Anesthesia Analgesia Oct_2010

Postoperative Sore Throat: More AnswersThan QuestionsPhillip E. Scuderi, MD

Postoperative sore throat (POST) is a common adverseevent after general anesthesia. Typically, the incidenceof POST is highest in patients who are tracheally

intubated; however, POST also occurs when a laryngeal maskairway (LMA) is used.1 Even patients who are managed witha facemask are not immune.1 Most of the measures that havebeen recommended for reducing this complication have beendirected at limiting the physical trauma that might result fromairway instrumentation and manipulation. Surprisingly fewinvestigations have evaluated pharmacologic interventionsas a means of reducing POST. Furthermore, no single drughas achieved widespread acceptance in the clinical commu-nity. In this issue of Anesthesia & Analgesia, 4 articlesdescribe simple prophylactic measures that seem to signifi-cantly reduce the incidence of POST.2–5 Two of thesearticles evaluate the effectiveness of topical benzydaminehydrochloride applied to the cuff of the endotracheal tube,directly to the pharyngeal mucosa, or both.2,3 A third articleevaluates the efficacy of inhaled fluticasone propionate.4

The fourth article evaluates Strepsils�, a nonprescriptionlozenge that contains 2 active ingredients, amylmetacresoland 2,4-dichlorobenzyl alcohol.5

When considered in aggregate, these 4 articles raise anumber of interesting questions. The expression “sorethroat” is obviously common to the vernacular of manydifferent cultures, yet it provides at best a parsimoniousdescription of the actual phenomena. Consequently, theexpression “postoperative sore throat” likely represents abroad constellation of signs and symptoms. For instance, inits simplest form, sore throat is a lay description of phar-yngitis, which in itself can have a variety of causes.However, sore throat may also include a variety of symp-toms including laryngitis, tracheitis, hoarseness, cough, ordysphagia. Postoperatively, it seems most plausible that thesymptoms are the result of mucosal injury with resultinginflammation caused by the process of airway instrumen-tation (i.e., laryngoscopy and suctioning) or the irritatingeffects of a foreign object (i.e., endotracheal tube, LMA, or

oral airway). The site or sites of mucosal injury wouldobviously vary depending on the airway device. For in-stance, endotracheal intubation can result in injury to anyportion of the pharynx as well as injury to the larynx andtrachea. Placement of an LMA can reasonably be expectedto cause injury to pharyngeal mucosa in the supraglotticregions only, whereas the use of a facemask with an oralairway should result in injury to only the oropharynx,assuming that no other injuries occurred because of suc-tioning or other airway maneuvers. It is therefore some-what surprising to note that the reported incidence of POSTafter LMA insertion is, at least in some studies, remarkablysimilar to that seen with endotracheal intubation.6,7 Al-though this might lead one to infer that the mechanism andlocation of injury must also be similar, a number of factsseem to contradict this assumption. For instance, reducingthe size of endotracheal tubes results in a significantdecrease in the incidence of POST.8 The design of tube cuffshas also been an area of intense research. The size,pressure/volume characteristics, and shape of cuff have allbeen implicated in tracheal mucosal injury and resultantPOST.9–12 Conversely, it has been suggested that cuffinflation pressure has less of a role in POST when an LMAis used.6 Both airway devices are clearly capable of induc-ing mucosal irritation and both can cause POST in patientsat rates that are not strikingly different. Yet, anatomically,the site or sites of injury cannot be the same.

There are several interesting observations that arisewhen one examines the data presented in the 4 articlespublished in this issue of Anesthesia & Analgesia. The datafrom Tazeh-kand et al.4 demonstrate that inhalation offluticasone propionate before the induction of anesthesiasignificantly reduces the incidence of POST at 1 hour and24 hours after surgery compared with a placebo control.This is not necessarily an unexpected result. Topical13,14

and systemic steroids15 have been demonstrated to reducethe incidence of POST presumably because of their sys-temic antiinflammatory effects. More puzzling are the datapresented by Ebneshahidi and Mohseni.5 Patients whoreceived a Strepsils lozenge before the induction of anes-thesia had a significantly lower incidence of POST andhoarseness both in the postanesthesia care unit and at 24hours after surgery than did the placebo control group thatreceived a flavored lozenge without the active ingredient.Unless one postulates a systemic effect from the activeingredients in the Strepsils lozenge (i.e., amylmetacresoland 2,4-dichlorobenzyl alcohol), the effect site must be the

From the Department of Anesthesiology, Wake Forest University School ofMedicine, Winston-Salem, North Carolina.

Accepted for publication June 14, 2010.

Disclosure: The author reports no conflicts of interest.

Address correspondence and reprint requests to Phillip E. Scuderi, MD,Department of Anesthesiology, Wake Forest University School of Medicine,Medical Center Blvd., Winston-Salem, NC 27157-1009. Address e-mail [email protected].

Copyright © 2010 International Anesthesia Research SocietyDOI: 10.1213/ANE.0b013e3181ee85c7

October 2010 • Volume 111 • Number 4 www.anesthesia-analgesia.org 831

EDITORIAL

pharyngeal mucosa. Whereas it is plausible to postulate areduction on pharyngeal irritation due to the lozenge, it isharder to postulate a mechanism for the reduction in“hoarseness” that was reported in this study. It is alsodifficult to understand how a preoperative lozenge couldreduce a reaction to injury to the larynx and trachea and theresultant laryngotracheitis that must have a role in POSTthat occurs after intubation.

Conversely, 2 of the articles2,3 describe a reduction inPOST with the application of benzydamine hydrochlorideto the endotracheal tube cuff alone compared with aplacebo control (normal saline and distilled water, respec-tively). It seems unlikely that the small dose of benzydam-ine hydrochloride (1.5 and 0.75 mg, respectively) usedwould have resulted in a systemic effect. Therefore, thereduction in POST that was observed in both of thesestudies must be assumed to have resulted from a localizeddecrease in mucosal injury and/or inflammatory response.The incidence of and reduction in POST is strikingly similarwhen Strepsils lozenges were used when compared with theapplication of benzydamine hydrochloride to the endotra-cheal tube cuff. There are unavoidable questions that must beasked: How can a lozenge that is administered orally result ina similar reduction in POST when compared with the topicalapplication of benzydamine hydrochloride to the endotra-cheal tube cuff? In addition, How can either an oral lozenge ortopical antiinflammatory agent applied to an endotrachealtube cuff yield reductions in POST that compare favorablywith a more widespread application of topical steroids to boththe pharyngeal and laryngotracheal mucosa?

POST is unquestionably a common adverse event aftergeneral anesthesia. A number of physical factors have beenimplicated as noted above. Most notable would seem to beendotracheal tube and cuff design and the approach to airwaymanagement (i.e., endotracheal tube, LMA, or mask anesthe-sia). In addition, female gender, younger patients, gynecologicsurgery, and the use of succinylcholine also seem to increasethe incidence.16 Of particular note, the use of topical lidocaineappears to confer no benefit and may in fact make POSTworse,17,18 a fact that seems to have been confirmed by Hunget al.2 However, what actually causes POST remains some-thing of a mystery. Hoarseness is a physical sign and cancertainly be evaluated objectively by a careful observer. Dys-phagia, although a symptom, is perhaps less likely to beinfluenced by intersubject variation in reporting, particularlyif questioning is done by a trained observer and is consistentlyapplied. Sore throat is more problematic. As noted above, thelocation of mucosal injury can vary widely but still result in asubject complaining of a “sore throat.” The 4 articles pre-sented in this issue of Anesthesia & Analgesia describe verydifferent strategies yet all achieved positive results. However,we are no closer to understanding the actual etiology of POSTthan we were before. A reasonable question to ask is “Does itmatter?” I believe that the answer is “yes.” If the preciseetiology (or etiologies) of pain after airway management canbe determined, it increases the likelihood that a specifictherapy or therapies could be recommended that woulddecrease symptoms and improve outcomes. POST is not themost important adverse event to avoid, at least from apatient’s perspective.19 Nevertheless, it is an adverse eventthat could easily be significantly decreased or even potentially

eliminated. The 4 studies presented here may provide theimpetus for a more careful evaluation of POST resulting inmore precisely targeted therapies.

AUTHOR CONTRIBUTIONSPES designed and conducted the study, analyzed the data, andwrote the manuscript. This author approved the final manuscript.

REFERENCES1. Higgins PP, Chung F, Mezei G. Postoperative sore throat after

ambulatory surgery. Br J Anaesth 2002;88:582–42. Hung NK, Wu CT, Chan SM, Lu CH, Hang YS, Yeh CC, Lee MS,

Cherng CH. The effect on postoperative sore throat of sprayingthe endotracheal tube cuff with benzydamine hydrochloride, 10%lidocaine, and 2% lidocaine. Anesth Analg 2010;111:882–6

3. Huang YS, Hung NK, Lee MS, Kuo CP, Yu JC, Huang GS,Cherng CH, Wong CS, Chu CH, Wu CT. The effectiveness ofbenzydamine hydrochloride spray on the endotracheal tubecuff or oral mucosa for postoperative sore throat. AnesthesiaAnalg 2010;111:887–91

4. Tazeh-kand NF, Eslami B, Mohammadian K. Inhaled flutica-sone propionate reduces postoperative sore throat, cough andhoarseness. Anesth Analg 2010;111:895–8

5. Ebneshahidi A, Mohseni M. Strepsils� tablets reduce sore throat andhoarseness after tracheal intubation. Anesth Analg 2010;111:892–4

6. Wakeling HG, Butler PJ, Baxter PJ. The laryngeal mask airway:a comparison between two insertion techniques. Anesth Analg1997;85:687–90

7. Joshi GP, Inagaki Y, White PF, Taylor-Kennedy L, Wat LI,Gevirtz C, McCraney JM, McCulloch DA. Use of the laryngealmask airway as an alternative to the tracheal tube duringambulatory anesthesia. Anesth Analg 1997;85:573–7

8. Stout DM, Bishop MJ, Dwersteg JF, Cullen BF. Correlation ofendotracheal tube size with sore throat and hoarseness follow-ing general anesthesia. Anesthesiology 1987;67:419–21

9. Loeser EA, Kaminsky A, Diaz A, Stanley TH, Pace NL. Theinfluence of endotracheal tube cuff design and cuff lubricationon postoperative sore throat. Anesthesiology 1983;58:376–9

10. Loeser EA, Bennett GM, Orr DL, Stanley TH. Reduction ofpostoperative sore throat with new endotracheal tube cuffs.Anesthesiology 1980;52:257–9

11. Loeser EA, Hodges M, Gliedman J, Stanley TH, Johansen RK,Yonetani D. Tracheal pathology following short-term intuba-tion with low- and high-pressure endotracheal tube cuffs.Anesth Analg 1978;57:577–9

12. Loeser EA, Orr DL II, Bennett GM, Stanley TH. Endotrachealtube cuff design and postoperative sore throat. Anesthesiology1976;45:684–7

13. Ayoub CM, Ghobashy A, Koch ME, McGrimley L, Pascale V,Qadir S, Ferneini EM, Silverman DG. Widespread applicationof topical steroids to decrease sore throat, hoarseness, andcough after tracheal intubation. Anesth Analg 1998;87:714–6

14. Sumathi PA, Shenoy T, Ambareesha M, Krishna HM. Con-trolled comparison between betamethasone gel and lidocainejelly applied over tracheal tube to reduce postoperative sorethroat, cough, and hoarseness of voice. Br J Anaesth 2008;100:215–8

15. Park SH, Han SH, Do SH, Kim JW, Rhee KY, Kim JH.Prophylactic dexamethasone decreases the incidence of sorethroat and hoarseness after tracheal extubation with a double-lumen endobronchial tube. Anesth Analg 2008;107:1814–8

16. McHardy FE, Chung F. Postoperative sore throat: cause, pre-vention and treatment. Anaesthesia 1999;54:444–53

17. Herlevsen P, Bredahl C, Hindsholm K, Kruhøffer PK. Prophy-lactic laryngo-tracheal aerosolized lidocaine against postopera-tive sore throat. Acta Anaesthesiol Scand 1992;36:505–7

18. Loeser EA, Stanley TH, Jordan W, Machin R. Postoperativesore throat: influence of tracheal tube lubrication versus cuffdesign. Can Anaesth Soc J 1980;27:156–8

19. Macario A, Weinger M, Carney S, Kim A. Which clinicalanesthesia outcomes are important to avoid? The perspectiveof patients. Anesth Analg 1999;89:652–8

EDITORIAL

832 www.anesthesia-analgesia.org ANESTHESIA & ANALGESIA

Activated Clotting Times, Heparin Responses, andAntithrombin: Have We Been Wrong All These Years?Jerrold H. Levy, MD, FAHA, and Roman M. Sniecinski, MD

It was recognized early in the development of moderncardiopulmonary bypass (CPB) that the fluidity ofblood needed to be maintained for extracorporeal cir-

culation, especially with the evolution of early oxygen-ators.1 It was not until the anticoagulant properties ofheparin were discovered that blood could circulate overnonendothelial surfaces without activating the coagulationcascade. Today, almost a century after its isolation byHowell and McLean in 1916, heparin remains the mainstayagent for cardiac surgery because of several advantages.2

Profound levels of anticoagulation can be quickly obtained(AntiXa levels of 3–5 U/mL) and an antidote, protamine, isreadily available. In addition, heparin has a relatively shortcontext-sensitive half-life, and can be used in patients withrenal dysfunction. However, an important limiting aspectof heparin is that it requires a cofactor, antithrombin (alsoreferred to as antithrombin III or ATIII), for its anticoagu-lation effect. Thus, heparin is still used extensively as ananticoagulant despite the development of newer agentsthat have been the focus of significant clinical interest andconsternation.3

The most common method for determining the adequacyof anticoagulation for cardiac surgery is the activatedclotting time (ACT), a modified whole blood Lee-Whiteclotting assay that can be affected by multiple factors.4

Although there is not universal agreement on the idealACT for CPB, when heparin does not produce thedesired ACT increase, “heparin resistance” is said tohave occurred.5 The term is misleading, however, be-cause it is really an alteration in the heparin doseresponse, possibly because of decreased antithrombin.“Heparin resistance” is perhaps more aptly termed “al-tered heparin responsiveness.”6

Antithrombin levels markedly decrease after CPB.7

These decreased levels may lead to a dangerous increase inthrombin activity.8 It would seem logical that this couldlead to postoperative thrombotic complications, suggestingthe need for antithrombin not only before CPB, but after it as

well. In this issue of the journal, 3 investigations from thesame group add further information and pose new questionsabout the interrelationship of antithrombin and heparin.

In the first study, Garvin et al.9 retrospectively evaluatedthe Hepcon HMS PLUS System (Medtronic Inc., Minneapo-lis, MN) for its accuracy in predicting heparin dose re-sponses for cardiac surgery with CPB. ACT-measuredheparin dose response and heparin concentrations wereevaluated in 3880 patients after a heparin dose calculated toachieve a target ACT. The result was wide variability inmeasurements. A target ACT of 300 seconds was notobtained in 7% of patients, and a target ACT of 350 secondswas not obtained in 17% of patients. The investigators alsofound that calculated and measured heparin dose re-sponses were not related at any heparin dose.

In the second study, the authors examined whetherthere was a direct association among preoperative anti-thrombin activity, heparin dose responses, and the heparinsensitivity index from 304 patients after CPB using HepconHMS PLUS Systems.10 The authors used multivariate linearregression to identify independent predictors of heparindose response. The baseline antithrombin activity wasnormal in this study and was not associated with eitherbaseline or postheparin ACT, heparin dose response, orheparin sensitivity index. Of note, only 16% of patients (49of 304) in the study presented with low baseline antithrom-bin levels as defined as �80%.

In the third study, the authors prospectively evaluatedwhether low levels of antithrombin were associated withpostoperative major adverse cardiac events in 1403 patientsundergoing coronary artery bypass graft.11 Major adversecardiac events were defined as postoperative death, reop-eration for coronary graft occlusion, myocardial infarction,stroke, pulmonary embolism, or cardiac arrest until firsthospital discharge. Antithrombin activity levels were mea-sured preoperatively, post-CPB, and on postoperative days1 to 5. Major adverse cardiac events occurred in 146patients (10.4%) and were independently associated withpostoperative antithrombin but not preoperative anti-thrombin levels.

What do these 3 studies tell us? For starters, they provethat an altered heparin response cannot be predicted bypreoperative antithrombin alone. Heparin’s effect on co-agulation is rather unpredictable, at least as measured byACT, an important perspective but not a new finding.12,13

For example, Metz and Keats14 gave 193 patients a singledose of heparin (300 U/kg). In 51 patients (26.4%), ACT

From the Department of Anesthesiology, Emory University School ofMedicine, Cardiothoracic Anesthesiology and Critical Care, Emory Health-care, Atlanta, Georgia.


Disclosure: The authors report no conflicts of interest.

Address correspondence and reprint requests to Jerrold H. Levy, MD, FAHA,Department of Anesthesiology, Emory University Hospital, 1364 Clifton Rd., NE,Atlanta, GA 30322. Address e-mail to [email protected].

Copyright © 2010 International Anesthesia Research SocietyDOI: 10.1213/ANE.0b013e3181f08a80


EDITORIAL

values were �400 seconds, including 4 patients �300seconds.

The target ACT values of 300 or 350 seconds for the 3studies may seem relatively low for many clinical practicesand the rate of failure to meet those targets may seemunusually high. Previous reports suggest there is a 5% to10% chance of a patient developing altered heparin respon-siveness. The current studies targeted ACT values �350,which may account for the authors finding a 17% incidenceof altered heparin responsiveness.14,15

Realistically, we still do not know the ideal ACT toinitiate CPB, which is reflected in the widespread variabil-ity in clinical practice.4,16 Several studies have demon-strated that thrombin is still activated at ACT values of 400to 480 seconds during CPB.17,18 As shown by the responseof patients with factor XII deficiency, a prolonged ACTdoes not guarantee anticoagulation.19

All of this calls into question the validity of using ACTsto assess adequate anticoagulation. Clinicians need to un-derstand the factors that influence both anticoagulation andits clinical measurement. The ACT is a relatively primordialtool, consisting only of a tube, some dirt or glass foractivation, and some heat and agitation to speed thereaction. As shown in Figure 1, there are a host of otherfactors besides heparin concentration and antithrombinlevels that affect ACT values. Unfortunately, there is no“gold standard” measurement to validate ACTs, complicat-ing assessment of the relationship between ACT and al-tered heparin responsiveness.

Based on the current studies, is there a role forantithrombin administration to improve heparin respon-siveness? Multiple studies report that antithrombinsupplementation improves intraoperative anticoagulation,increases ACT, and reduces biochemical markers of hemo-static activation.13,20–22 Although these 3 studies did notinvestigate antithrombin administration, the authors didfind a relationship between postoperative antithrombinlevels and major adverse cardiac events. This corroboratesthe observation by Ranucci et al.,23 that low levels ofantithrombin activity in the intensive care unit are associ-ated with a poor outcome in cardiac surgery. However,

there are still only a few supporting studies demonstratingthat antithrombin supplementation improves clinical out-comes, and these studies were not conducted in patientsundergoing cardiac surgery.24,25

In summary, the current 3 studies provide additionaldata about the complex issues of anticoagulation, heparinresponsiveness, and outcomes, including the role of anti-thrombin. Have we been wrong all these years regardingACTs, heparin responsiveness, and the role of antithrom-bin? More than 50 years after the development of CPB, weare still asking many of the questions that the earlypioneers confronted. We need better monitors and betterunderstanding of anticoagulation adequacy to treat alter-ations in heparin response and assess therapeutic efficacy.We hope it will not take another 50 years to find theseanswers.

REFERENCES1. Galletti PM. Cardiopulmonary bypass: a historical perspective.

Artif Organs 1993;17:675–862. Wardrop D, Keeling D. The story of the discovery of heparin

and warfarin. Br J Haematol 2008;141:757–633. Spyropoulos AC. Brave new world: the current and future use

of novel anticoagulants. Thromb Res 2008;123:S29–354. Despotis GJ, Gravlee G, Filos K, Levy J. Anticoagulation

monitoring during cardiac surgery: a review of current andemerging techniques. Anesthesiology 1999;91:1122–51

5. Staples MH, Dunton RF, Karlson KJ, Leonardi HK, Berger RL.Heparin resistance after preoperative heparin therapy or in-traaortic balloon pumping. Ann Thorac Surg 1994;57:1211–6

6. Levy JH. Heparin resistance and antithrombin: should it still becalled heparin resistance? J Cardiothorac Vasc Anesth 2004;18:129–30

7. Zaidan JR, Johnson S, Brynes R, Monroe S, Guffin AV. Rate ofprotamine administration: its effect on heparin reversal andantithrombin recovery after coronary artery surgery. AnesthAnalg 1986;65:377–80

8. Sniecinski R, Szlam F, Chen EP, Bader SO, Levy JH, TanakaKA. Antithrombin deficiency increases thrombin activityafter prolonged cardiopulmonary bypass. Anesth Analg2008;106:713– 8

9. Garvin S. Heparin concentration-based anticoagulation forcardiac surgery fails to reliably predict heparin bolus doserequirement. Anesth Analg 2010;111:849–55

Figure 1. The activated clotting time (ACT) is awhole blood point of care coagulation test usedextensively in cardiac surgery and in the cardiaccatheterization laboratory to monitor the anticoagu-lant effect of different agents including unfraction-ated heparin. In the ACT, blood is added to acartridge or tube that contains an activator, usuallycelite or kaolin, to speed the process by increasingcontact activation by the intrinsic coagulation cas-cade. Clot formation in the ACT represents theinteraction of plasma coagulation components(e.g., factors and fibrinogen), platelets, and redblood cells as this is a whole blood clotting assay.However, clot formation in the ACT is influenced bymultiple factors that include platelet count andplatelet function, factor deficiencies, fibrinogenlevels, pharmacologic agents (anticoagulants inplatelet inhibitors), temperature(especially hypo-thermia), and contact activation inhibitors (e.g.,aprotinin).

EDITORIAL


10. Garvin S. Heparin dose response is independent of perioper-ative antithrombin activity in patients undergoing coronaryartery bypass graft surgery using low heparin concentrations.Anesth Analg 2010;111:856–61

11. Garvin S. Postoperative activity, but not preoperative activity,of antithrombin is associated with major adverse cardiacevents after coronary artery bypass graft surgery. AnesthAnalg 2010;111:862–9

12. Gravlee GP, Case LD, Angert KC, Rogers AT, Miller GS.Variability of the activated coagulation time. Anesth Analg1988;67:469–72

13. Levy JH, Montes F, Szlam F, Hillyer CD. The in vitro effects ofantithrombin III on the activated coagulation time in patientson heparin therapy. Anesth Analg 2000;90:1076–9

14. Metz S, Keats AS. Low activated coagulation time duringcardiopulmonary bypass does not increase postoperativebleeding. Ann Thorac Surg 1990;49:440–4

15. Ranucci M, Isgro G, Cazzaniga A, Ditta A, Boncilli A, Cotza M,Carboni G, Brozzi S. Different patterns of heparin resistance:therapeutic implications. Perfusion 2002;17:199–204

16. Lobato RL, Despotis GJ, Levy JH, Shore-Lesserson LJ, CarlsonMO, Bennett-Guerrero E. Anticoagulation management duringcardiopulmonary bypass: a survey of 54 North Americaninstitutions. J Thorac Cardiovasc Surg 2010;139:1665–6

17. Brister SJ, Ofosu FA, Buchanan MR. Thrombin generationduring cardiac surgery: is heparin the ideal anticoagulant?Thromb Haemost 1993;70:259–62

18. Slaughter TF, LeBleu TH, Douglas JM Jr, Leslie JB, Parker JK,Greenberg CS. Characterization of prothrombin activationduring cardiac surgery by hemostatic molecular markers.Anesthesiology 1994;80:520–6

19. Salmenpera M, Rasi V, Mattila S. Cardiopulmonary bypassin a patient with factor XII deficiency. Anesthesiology 1991;75:539 – 41

20. Avidan MS, Levy JH, van Aken H, Feneck RO, Latimer RD, OttE, Martin E, Birnbaum DE, Bonfiglio LJ, Kajdasz DK, DespotisGJ. Recombinant human antithrombin III restores heparinresponsiveness and decreases activation of coagulation inheparin-resistant patients during cardiopulmonary bypass.J Thorac Cardiovasc Surg 2005;130:107–13

21. Levy JH, Despotis GJ, Szlam F, Olson P, Meeker D, Weis-inger A. Recombinant human transgenic antithrombin incardiac surgery: a dose-finding study. Anesthesiology 2002;96:1095–102

22. Williams MR, D’Ambra AB, Beck JR, Spanier TB, Morales DL,Helman DN, Oz MC. A randomized trial of antithrombinconcentrate for treatment of heparin resistance. Ann ThoracSurg 2000;70:873–7

23. Ranucci M, Frigiola A, Menicanti L, Ditta A, Boncilli A, BrozziS. Postoperative antithrombin levels and outcome in cardiacoperations. Crit Care Med 2005;33:355–60

24. Fourrier F, Chopin C, Goudemand J, Hendrycx S, Caron C,Rime A, Marey A, Lestavel P. Septic shock, multiple organfailure, and disseminated intravascular coagulation: comparedpatterns of antithrombin III, protein C, and protein S deficien-cies. Chest 1992;101:816–23

25. Eid A, Wiedermann CJ, Kinasewitz GT. Early administration ofhigh-dose antithrombin in severe sepsis: single center resultsfrom the KyberSept-trial. Anesth Analg 2008;107:1633–8

Activated Clotting Times, Heparin Responses, and Antithrombin


Bridging the Gap to Reduce Surgical Site InfectionsFrank J. Overdyk, MSEE, MD

Surgical site infections (SSI) remain one of the mostpersistent and costly preventable complications inhealth care. Although as anesthesiologists, we may

perceive our responsibility limited to the timely administra-tion of prophylactic antibiotics and sterile placement of inva-sive catheters, we feel the pain of our surgical colleagueswhen their service suffers a rash of SSI, and they resort tostrategies ranging from ones with sound scientific basis tomere superstition. Over the years, I have seen operatingrooms undergo lockdowns worthy of a maximum-securityprison, bathed in dance club–like ultraviolet light, or trans-formed to resemble a spacewalk rehearsal. Yet the problemof SSI persists, second only to urinary tract infectionsamong the most common types of nosocomial infections inhospitals.1 Approximately 1 in 50 surgical procedures arecomplicated by SSI, which translates to an estimated290,485 cases annually, of which 13,088 result in death. It isno surprise that SSI have made the 2010 Joint CommissionNational Patient Safety Goals for the second consecutiveyear.*

The importance of high tissue oxygen levels in im-proved surgical outcomes and reduced surgical infectionrates has been understood for many years. High tissueoxygen concentrations enhance the effects of leukocytesand antibiotics on microbes and are indicative of adequatetissue perfusion. Oxygen delivery to the tissues, a functionof both bloodflow and oxygen-carrying capacity, has beenthe underpinning of many of the nonpharmacologic strat-egies to prevent surgical wound infection.2 Traditionally,liberal intraoperative fluid management was thought toimprove outcomes by improving perfusion, yet more re-cent data suggest that tissue edema from overly aggressivehydration may be detrimental. We remain without a con-sensus on the optimal volume replacement strategy foroptimal surgical outcomes.3 Similarly, although the sym-pathectomy accompanying regional anesthesia improvestissue perfusion and certain outcomes (such as preservedgraft patency in vascular surgery and decreased incidenceof deep vein thrombosis), its impact on SSI is unclear.Avoiding intraoperative hypothermia helps maintain tissue

perfusion, reduces SSI, and shortens hospitalization.4 Mildintraoperative hypoventilation has been shown to increasetissue oxygen levels due to the vasodilating effect of carbondioxide in obese patients.5 More clarity may be brought tothe optimal intraoperative fluid, temperature, ventilation,and oxygenation strategies that minimize SSI if we couldeasily measure subcutaneous tissue oxygen partial pressure(Sto2). Arguably, Sto2 represents the common pathway ofcellular “contentment” during the “decisive period,” whichis defined as the hours surrounding surgical incision inwhich prophylactic antibiotics have been shown to have thegreatest impact on reducing SSI.6

In this issue of Anesthesia & Analgesia, Govinda et al. usea noninvasive infrared spectroscope (NIRS) to measureSto2 as a predictor of SSI that manifests as late as 30 dayspostoperatively.7 The authors used a receiver operatorcurve to optimize the sensitivity and specificity of upper-arm Sto2 measurements to differentiate patients who de-velop an SSI from those who do not. They found Sto2 below66% measured 75 minutes after colonic surgery on theupper arm to be predictive for SSI diagnosed, on average, 9days after surgery. This NIRS measurement was a betterpredictor for SSI than were 2 clinically accepted SSI riskassessment metrics, the National Nosocomial InfectionsSurveillance System and Study on the Efficacy of Nosoco-mial Infection Control.

As an introductory study on this technology, its limitationsare several and well delineated in the article. However, if thisresult is reproducible upon further investigation, we mayhave an important “early warning system” for SSI and a toolwith which to dissect the problem of SSI into 2 components.First, this noninvasive technique may allow us to investigatethe different intraoperative anesthetic strategies involvingfluid management, ventilation, oxygenation, hemodynamicindices, and temperature that optimize Sto2 during the imme-diate postoperative period. Second, once we optimize andcontrol for Sto2, we can explore additional postoperativestrategies to minimize SSI. The current debate on theefficacy of high intraoperative and postoperative Fio2 inreducing SSI demonstrates in which cases this Sto2 mea-surement may be helpful. A recent meta-analysis of 5randomized, double-blind clinical trials found a relativerisk reduction for SSI of 33% when hyperoxic gas mixtures(Fio2 � 80%) were used intraoperatively and immediatelypostoperatively for 2 to 6 hours versus a normoxic (Fio2

�30%) mixture.8 Yet, few anesthesiologists and surgeonshave adopted this approach, perhaps because a benefitrealized only once in every 33rd patient (number needed totreat) may not justify the inconvenience and risks of

*From http://www.jointcommission.org/PatientSafety/NationalPatientSafetyGoals/.

From Medical University of South Carolina, Charleston, South Carolina.


Address correspondence to Frank J. Overdyk, MSEE, MD, Medical Univer-sity of South Carolina, 167 Ashley Avenue, Suite 301, MSC 912 CharlestonSC 29425-9120. Address e-mail to [email protected].

Copyright © 2010 International Anesthesia Research SocietyDOI: 10.1213/ANE.0b013e3181ee85b1

836 www.anesthesia-analgesia.org October 2010 • Volume 111 • Number 4

EDITORIAL

patients wearing nonrebreather oxygen masks up to 6hours postoperatively. Atelectasis and increased pulmo-nary complications in the hyperoxic group, particularly inobese patients, remain a concern.9 Furthermore, these re-sults were contradicted by a randomized control trialpowered to detect an identical 33% relative risk reductionbetween an Fio2 of 80% and one of 30%, continued for only2 hours postoperatively.10 That study found no decreasedrisk of SSI with the higher inspired oxygen concentration.Was the shorter duration of postoperative hyperoxia deci-sive? Or were intraoperative fluid and temperature man-agement suboptimal for preventing SSI, as is suggested byHunt and Hopf in an accompanying editorial?11 Perhapsimmediate postoperative Sto2 measurements will give us a“halftime score,” which we can use to alter the game plan,and concentrate on postoperative interventions to reduceSSI, such as high supplemental oxygen concentrations,effective pain management, postoperative warming, andappropriate discontinuation of prophylactic antibioticswithin 24 hours of surgery end.12 However, if Sto2 mea-surements in the recovery room suggest that the “SSI die iscast,” it is clear that our efforts to reduce SSI should befocused on intraoperative strategies, including those notassociated with oxygenation and perfusion, such as timelyantibiotic prophylaxis within 1 hour before surgical inci-sion and tight control of blood glucose levels.

Additional work with Sto2 is warranted. Although dif-ferences in dependent variables such as intraoperativetemperature, volume replacement, hemodynamics, surgicalduration, and surgical technique between the patients withand without SSI were nonsignificant in this study, thepatients with SSI weighed more and had more pain, andFio2 during NIRS measurements was not controlled.Clearly, the impact of these confounding variables on theresults must be clarified. This study also contradictedprevious work in terms of the anatomic location of the Sto2

measurement, suggesting that measurement location is notmerely nuance, but may have important ramifications interms of the reproducibility of these results.

Lastly, the trade-off between sensitivity and specificityin determining a cutoff point (Sto2 �66%) needs to beconsidered in the context of the outcome. This threshold forSto2 was chosen by the authors from the receiver operatorcurve to optimize both sensitivity and specificity. In thecase of SSI, the annual cost to hospitals is approximately$7.5 billion, and the toll on patients and their familiesincalculable.13 It may be wise to sacrifice specificity for

better sensitivity, if most of the intra- and postoperativerisk mitigation strategies can be implemented cost-effectively. NIRS should help answer these questions.

As the above discussion illustrates, SSI is not merely asurgical issue, but one that anesthesiologists influencethrough intraoperative decisions and postoperative painmanagement strategies. This may be reflected at some pointby the inclusion of SSI in the National Anesthesia ClinicalOutcomes Registry.† Our involvement in investigating andreducing SSI will be appreciated by our administrators,surgeons, and our patients, and we will continue to berecognized as critical and invaluable members of the peri-operative team.

REFERENCES1. Klevens RM, Edwards J, Richards C, Horan T, Gaynes R,

Pollock D, Cardo D. Estimating health care–associated infec-tions and deaths in U.S. hospitals. Public Health Rep 2007;122:160–66

2. Sessler D. Non-pharmacologic prevention of surgical woundinfection. Anesthesiol Clin 2006;24:279–97

3. Chappel D, Jacob M, Hofmann-Kiefer K, Conzen P, Rehm M.Rational approach to perioperative fluid management. Anes-thesiology 2008;109:723–40

4. Kurz A, Sessler D, Lenhardt R. Perioperative normothermia toreduce the incidence of surgical wound infection and shortenhospitalization. NEJM 1996;334:1209–15

5. Hager H, Reddy D, Mandadi G, Pulley D, Eagon JC, Sessler D,Kurz A. Hypercapnia improves tissue oxygenation in mor-bidly obese surgical patients. Anesth Analg 2006;103:677–81

6. Classen DC, Evans RS, Pestotnik SL, Horn SD, Menlove RL,Burke JP. The timing of prophylactic administration of antibi-otics and the risk of surgical-wound infection. N Engl J Med1992;326(5):281–6

7. Govinda R, Kasuya Y, Bala E, Mahboobi R, Devarajan J, SesslerD, Akca O. Early postoperative subcutaneous tissue oxygenpredicts surgical site infection. Anesth Analg 2010;111:946–52

8. Qadan M, Akca O, Mahid S, Hornung C, Polk H. Perioperativesupplemental oxygen therapy and surgical site infection: ameta-analysis of randomized controlled trials. Arch Surg2009;144(4):359–66

9. Zoremba M, Dette F, Hunecke T, Braunecker S, Wulf H. Theinfluence of perioperative oxygen concentration on postopera-tive lung function in moderately obese adults. Eur J Anaesthe-siol 2010;27:501–7

10. Meyhoff C, Wetterslev J, Jorgensen L. For the PROXI random-ized clinical trial. JAMA 2009;302(14):1543–50

11. Hunt T, Hopf H. High inspired oxygen fraction and surgicalsite infection. JAMA 2009;302:1588–9

12. Bratzler DW, Houck PM. Antimicrobial prophylaxis for sur-gery: an advisory statement from the National Surgical Infec-tion Prevention Project. Amer J Surg 2005;189:395–404

13. Stone PW, Braccia D, Larson E. Systematic review of economicanalyses of health care–associated infections. Am J InfectControl 2005;33:501–9†From http://www.aqihq.org/. Accessed May 25, 2010.

Noninvasive Infrared Spectroscope and Surgical Site Infection


Labor Pain, Analgesia, and Chronobiology: WhatFactor Matters?Yvan Touitou, PhD,* Garance Dispersyn, PhD,† and Laure Pain, MD†

Alleviation and control of pain are of major impor-tance in medicine. Depending on the causes and onthe clinical situations, the patterns of pain may

vary during the day, with peak and trough times reportedat different times of day, although sometimes withcontradictory results.1 The therapeutic effects of localanesthetics and opioid analgesics also demonstrate cir-cadian variations.1

Two articles in this issue of Anesthesia & Analgesia reportinvestigations in parturients into the relationship betweentime of administration of intrathecal bupivacaine,2 or intra-thecal fentanyl versus systemic hydromorphone,3 and theduration of analgesia. In both studies, the duration ofanalgesia was defined as the time from the first adminis-tration of analgesia during labor until the second requestfor analgesia. The 2 groups of investigators approached theanalysis of the chronobiology of analgesic requirementsduring labor from somewhat different perspectives. Scav-one et al.3 addressed the problem from a clinical perspec-tive, with their main objective being to determine whethertime-related effects might have confounded the results of aprevious study by the group.4 Shafer et al.,2 on the otherhand, used their clinical data to examine the problems thatarise when developing models to analyze the often com-plex periodic data obtained from clinical situations.

Scavone et al.3 compared neuraxial versus systemicopioid analgesia in 692 healthy parturients early in labor.At the first request for analgesia, patients in the neuraxialgroup were given intrathecal fentanyl 25 �g, whereaspatients in the systemic analgesia group received 1 mg IVand 1 mg IM hydromorphone. Subjects given their initialanalgesia between 0701 hours and 2300 hours were consid-ered as the daytime group; those given their first analgesicbetween 2301 hours and 0700 hours, the nighttime group.The authors found no difference in the median duration ofeither neuraxial or systemic analgesia, irrespective of thetime of administration. A major advantage claimed by theauthors for their study was the large number of patients

investigated, considerably greater than that of previousstudies of this nature.

The paper by Shafer et al.2 represents the final productof a manuscript initially submitted to Anesthesia & Analge-sia, reporting the chronobiology of the duration of analgesiafollowing intrathecal bupivacaine. After peer review andextensive reanalysis of the original data, the manuscript hasevolved into a much more interesting exploration of themethods used in chronobiological analysis to detect theinfluence of external factors and in particular how one canuse these methods to detect possible artifacts in the data.This is an important illustration of the value of a well-conducted peer review process.

Shafer et al.2 make the important point in their paperthat very different conclusions can be drawn, depending onthe statistical methods used to analyze the data. They used3 different smoothing functions to explore circadianrhythms graphically. These revealed a bimodal pattern inthe duration of analgesia, with 1 peak at around 0630 hoursand a subsequent peak in the afternoon or evening. Incontrast, an analysis of variance (ANOVA) did not showany significant difference in the duration of analgesia,irrespective of the timing of the intrathecal injection ofbupivacaine. Fitting the data to a simple cosine function ofanalgesia duration versus time demonstrated a periodicwaveform with a period of 8 hours, but with peaks thatcorresponded with only 1 of the 3 peaks identified by thesmoother functions. The authors then used a bootstrapanalysis to show that 2 individual points were responsiblefor the statistical significance in their cosine fit. Removingthese points from the dataset also removed the previouslyobserved rhythmic effect of intrathecal bupivacaine onanalgesia. The authors concluded that these 2 points werelikely to be artifacts, though they acknowledge this mightbe uncertain, corresponding to the change in nursing andanesthesia shifts.

The literature contains conflicting reports on the time-dependent effects of neuraxial local anesthetics and opi-oids.1 These differences may be a consequence of the widevariety of factors that need to be considered when inter-preting the results from chronobiology experiments. Theyalso emphasize the importance of a very stringent experi-mental protocol, including methodology and statisticalmethods used to analyze the potential rhythmic effects of adrug. Many factors, both internal and external, can influ-ence the biological rhythms of a drug’s action. Theseinclude temporal variations in light–dark, rest–activity,

From *Unite de Chronobiologie, Fondation Ophtalmologique A. de Roths-child, Paris, France; and †INSERM U666 (GRERCA), CHU de Strasbourg,Faculte de medecine, Strasbourg, France.


Address correspondence to Yvan Touitou, Chronobiology Unit, FondationOphtalmologique A. de Rothschild, 29 rue Manin, 75019 Paris, France.Address e-mail to [email protected].

Copyright © 2010 International Anesthesia Research SocietyDOI: 10.1213/ANE.0b013e3181ee85d9


EDITORIAL

fasting–eating cycles, and other environmental factors suchas stress, alcohol consumption, tobacco use, and caffeineintake. All of these are able to alter the parameters charac-terizing a biological rhythm and should be considered ascofactors that can mask or unmask any circadian effects ofdrugs.5 When studying labor pain, specific obstetrical fac-tors should be controlled for, because they may haveconsiderable influence on pain intensity, e.g., parity, spon-taneous or pharmacologically induced labor, stage of cer-vical dilation, rupture of the membrane, labor duration,and pharmacological induction of uterine contractions(oxytocin administration).6–8

Ideally, although not always feasible, studies should con-sider subpopulations of patients, considering all of thesedifferent clinical situations. As an example, one may analyzethe duration of intrathecal analgesia in a population of partu-rients, nulliparous, with cervical dilation at 2 cm, with noadministration of oxytocin at all, and find circadian variationsthat could be masked in another study examining a differentsubgroup of population. This might explain the discrepanciesbetween the published studies. However, other factors canbias results. Although factors such as the timing of the shiftchanges in the work team can bias results, one needs to becautious about this type of explanation. For example, the timewhen the morning nursing shift begins could coincide withthe peak in cortisol, which is lowest in the evening andreaches its highest concentration during the early wakinghours. Which factor is then influencing the duration of anal-gesia, the change of nursing team or the endogenous circadianrhythm of cortisol?

Although biological rhythms are genetically deter-mined, they are continuously modulated (adjusted in time)by periodic events in the environment called synchroniz-ers.5 In mammals, the main synchronizer is light. Rhythmsynchronization of subjects is the “gold standard” of anychronobiological study protocol. This means that, in addi-tion to the factors discussed above, the subjects should keepas close as possible to their usual lifestyle, maintaining theenvironmental factors (synchronizers) that impose theirperiod on biological rhythms, such as sleep–wake time,rest–activity cycle, light–dark cycle, fasting–eating, etc.5

Thus, it is advisable to verify the synchronization of thesubjects’ circadian time organization by checking rhythmmarkers (body temperature, cortisol, melatonin, etc.) as aninitial step or as an integral aspect of the investigativeprotocol.9 The importance of the synchronizers that influ-ence endogenous rhythms varies according to the variableor the function under investigation. Some synchronizersare more predominant than others, e.g., the light–darkcycle and the sleep–wake activity in humans. In someexperimental situations, however, synchronizers that oth-erwise might be considered minor can assume predomi-nance. Changes in any one of these synchronizers may leadto changes in the temporal relationship between biologicalrhythms.5

Circadian rhythms can also be subject to seasonal modu-lation, and different circadian rhythm patterns have beendescribed in the same individuals but at different times ofthe year.10–12 Failure to consider this could explain someof the apparently discrepant results in the literature. All ofthose factors need be considered and standardized in any

study dealing with biological rhythms to avoid potentialbiases.

Many exogenous factors (masking agents) can influencecircadian patterns by masking the endogenous rhythms ofthe biological clock. A circadian rhythm can be masked byany environmental signal to which the organism is sensi-tive.13 To unmask the endogenous circadian structure, anexperimental protocol called “constant routine” can beused. This involves subjects staying awake but lying downfor 24 to 36 hours, in an environment of constant tempera-ture, humidity, and lighting, with identical and regularlyspaced meals. Though this protocol is considered by manyas the gold standard, it obviously presents several limita-tions,14 and for research in fields such as obstetrics isseldom practicable.

Subtracting the estimated exogenous or “masking” com-ponent of a rhythm from the observed rhythm may revealthe underlying endogenous component. Regression modelshave been developed to identify masking effects, but sepa-ration of the exogenous from the endogenous component isdifficult.5 When the data can be approximated by a sinu-soidal model, the cosinor function allows calculation ofrhythm parameters such as the mesor (the mean level thatis equal to the 24-hour average), amplitude (half of thepeak-to-trough difference of the fitted cosine function), andacrophase (the peak time of the rhythm given in degrees orhours and minutes). When the number of subjects is smallor the density of samplings low, the interindividual vari-ability should be smoothed. Nonparametric tests are usefulbecause they ensure that the absolute values do not influ-ence the results. Whatever methods are used, they shouldbe critically evaluated for their ability to test the hypothesesunder question and to assess the time series data of thevariable(s) being studied.

Finally, why are the studies by Scavone et al.3and Shaferet al.2 important? Both studies found that time of day didnot appear to influence the duration of analgesia producedby intrathecal local anesthetics or opioids. Knowledge ofthe time dependency of drugs is important because drugeffect can be optimized and toxicity minimized by basingdrug administration on the circadian patterns of drugactivity. Several studies have demonstrated circadian time–dependent changes in the toxicity and the pharmacokineticdisposition of local anesthetics.15 Shafer et al., in addition todemonstrating the importance of the peer review process,provide a set of clear guidelines that future investigators inthis field should find valuable, and hopefully help themavoid many of the potential pitfalls in chronobiologicalresearch.

ACKNOWLEDGMENTSThis editorial was solicited and handled by Dr. James Bovill,Guest Editor-in-Chief for Anesthesia & Analgesia.

REFERENCES1. Bruguerolle B, Labrecque G. Rhythmic pattern in pain and

their chronotherapy. Adv Drug Del Rev 2007;59:883–952. Shafer SL, Lemmer B, Boselli E, Boiste F, Bouvet L, Allaouch-

iche B, Chassard D. Pittfalls in chronobiology: a suggestedanalysis using intrathecal bupivacaine analgesia as an ex-ample. Anesth Analg 2010;111:980–5

Labor Pain, Analgesia, and Chronobiology


3. Scavone BM, McCarthy RJ, Wong CA, Sullivan JT. The influ-ence of time of day of administration on duration of opioidlabor analgesia. Anesth Analg 2010;111:986–91

4. Wong CA, Scavone BM, Peaceman AM, McCarthy RJ, SullivanJT, Diaz NT, Yaghmour E, Marcus RJ, Sherwani SS, SprovieroMT, Yilmaz M, Patel R, Robles C, Grouper S. The risk ofcesarean delivery with neuraxial analgesia given early versuslate in labor. N Engl J Med 2005;352:655–65

5. Touitou Y, Haus E, eds. Biological Rhythms in Clinical andLaboratory Medicine. Berlin: Springer, 1992

6. Sheiner E, Sheiner EK, Shoham-Vardi I. The relationship betweenparity and labor pain. Int J Gynecol Obstet 1998;63:287–8

7. Aya AGM, Vialles N, Mangin R, Robert C, Ferrer JM, Ripart J,de la Courssaye JE. Chronobiology of labour pain perception:an observational study. Br J Anaesth 2004;93:451–3

8. Veira WS, Hidalgo MPL, da Silva LT, Caumo W. Biologicalrhythms of spinal-epidural labor analgesia. Chronobiol Int2010;27:865–78

9. Selmaoui B, Touitou Y. Reproducibility of the circadianrhythms of serum cortisol and melatonin in healthy subjects. Astudy of three different 24-h cycles over six weeks. Life Sci2003;73:3339–49

10. Touitou Y, Lagoguey M, Bogdan A, Reinberg A, Beck H.Seasonal rhythms of plasma gonadotrophins: their persistencein elderly men and women. J Endocrinol 1983;96:15–21

11. Levi F, Canon C, Touitou Y, Reinberg A, Mathe G. Seasonalmodulation of the circadian time structure of circulating T andnatural killer lymphocyte subsets from healthy subjects. J ClinInvest 1988;81:407–13

12. Reinberg A, Touitou Y, Levi F, Nicolai A. Circadian andseasonal changes in ACTH-induced effects in healthy youngmen. Eur J Clin Pharmacol 1983;25:657–65

13. Mrosovsky N. Masking: history, definitions, and measure-ment. Chronobiol Int 1999;16:415–29

14. Waterhouse JM, Redfern P, Minors DS. An introduction tocircadian rhythms and their measurement in humans. In:Redfern P, ed. Chronotherapeutics. London: PharmaceuticalPress, 2003:1–30

15. Chassard D, Bruguerolle B. Chronobiology and anesthesia.Anesthesiology 2004;100:413–27

EDITORIAL


Can an Acute Pain Service Be Cost-Effective?Eric Sun, MD, PhD,*† Franklin Dexter, MD, PhD,‡ and Alex Macario, MD, MBA§

In many countries around the world, the anesthesiolo-gist is the primary physician responsible for pain con-trol in the first 24 hours after surgery. However, in the

United States of America (USA), postoperative analgesia istypically managed by the surgeon. This is because theirprofessional fee includes this responsibility while the pa-tient remains in the hospital and when the patient returnshome. At the other end of the spectrum is a dedicatedAcute Pain Service team with expertise and authority formanaging a patient’s surgical pain.

The well-done randomized clinical trial conducted byLee et al.1 and described in this issue of the journalestimates the cost-effectiveness of an anesthesiologist-led,nurse-based Acute Pain Service, mainly charged with man-aging IV patient-controlled analgesia. The control groupconsisted of patients receiving IM or IV boluses of opioidsas needed by nurses on the ward.1 It is not known howfrequently this technique is used in other parts of theworld, including the USA.

Our objectives for this editorial were to:

• Estimate the total annual number of surgical inpa-tients in the USA that might benefit from care pro-vided by an Acute Pain Service.

• Define possible structures and functions of an AcutePain Service because these may vary among institutions,and affect the economic viability of such a service.

• Outline research questions that deserve further studyto help hospitals make the needed investment in anAcute Pain Service.

HOW MANY PATIENTS ARE AT RISK FOR PAINAFTER INPATIENT SURGERY IN THE USA?To obtain an estimate of the number of inpatients havingsurgery, the Nationwide Inpatient Sample from the Health-care Cost and Utilization Project was used. It samples 20%of inpatient discharges from 1056 academic, community,and acute care hospitals in 42 states, and is intended to be

representative of 90% of all hospital discharges in the USA.For 2008, the year with the most recently available data,there were approximately 9.1 million inpatients havingsurgery in the USA. This constitutes approximately 23% ofall inpatient visits, and a 4% annual growth rate since 1998(Table 1). These data exclude federal and prison hospitals.

The data in Table 1 provide insight into the size andscope of the demand for pain management in the postop-erative setting. From a national health policy perspective,there are many different surgical procedures.2 However,cesarean deliveries, orthopedics, and general surgery pro-cedures should be a priority for optimizing clinical acutepostoperative pain care because they account for a largefraction of all inpatient surgery.

With public and private payers continuing to face tight-ening budgets, and with single bundled episode of carepayments for hospital and physician care looming, it is anopportunity for anesthesiologists to consider additionalvenues where the specialty can provide value. The dedi-cated Acute Pain Service seems to be a logical choice, giventhat it is a straightforward extension of the care anesthesi-ologists already provide in the operating room and chronicpain settings. However, given the expenses of such aservice, what are the incentives for hospitals and anesthe-siologists to participate? On the clinical side, controllingpain is of importance to patients, a goal made more difficultby the increasing number of patients who are admittedwith chronic pain syndromes including opioid prescrip-tions superimposed on their acute surgical pain. The po-tential clinical benefits of a dedicated Acute Pain Servicehave resulted in support from a variety of organizationssuch as the American Society of Anesthesiologists,3 theRoyal College of Surgeons, and the College of Anesthetists.

In addition to these clinical benefits, an Acute Pain Serviceprovides important visibility to the anesthesia group within ahospital. Patient satisfaction with pain control comprises animportant component of many measures of hospital quality,such as Press Ganey scores. Many facilities are providingresources to an Acute Pain Service to increase these scores. Infact, a patient’s Press Ganey response to, “How well your painwas controlled” was the second most important variablecorrelated to whether a patient recommended the hospital tosomeone else. By the way, the number one item was “Staffworked together to care for you.”4

POSSIBLE STRUCTURES AND FUNCTIONS OF ANACUTE PAIN SERVICEDespite these benefits, the use of a dedicated Acute PainService to manage postoperative pain did not gain wide-spread adoption in the 1990s in the USA5–7 or abroad.8,9

From the *RAND Corporation, Santa Monica; †the Department of HealthServices, University of California, Los Angeles, Los Angeles, California;‡University of Iowa, Iowa City, Iowa; and §Department of Anesthesia,Stanford University School of Medicine, Stanford, California.

Accepted for publication July 9, 2010.

Dr. Sun is funded by the Agency for Healthcare Research and Qualitythrough the UCLA/RAND training grant T32 HS 000046.


Address correspondence and reprint requests to Alex Macario, MD, MBA,Department of Anesthesia H3580, Stanford University School of Medicine,Stanford, CA 94305-5640. Address e-mail to [email protected].

Copyright © 2010 International Anesthesia Research SocietyDOI: 10.1213/ANE.0b013e3181f33533


EDITORIAL

Sometimes Acute Pain Services are services in name only.For example, 36% of 446 hospitals in Germany operatedan Acute Pain Service but half did not have specificpersonnel assigned to the service, policies for nights andweekends, written protocols for pain management, orregular assessment and documentation of pain scores atleast once a day.10 The challenges are in part financial,such as how coding/billing affects profitability, but alsoinclude organizational change barriers such as compet-ing managerial and clinical agendas, staff shortages,local politics, professional hierarchies, and workloadchanges.11

In Italy, although 42% of 163 public hospitals had anorganized Acute Pain Service, continuous IV analgesiausing elastomeric infusion systems was the most frequentlyused analgesic technique, and IV patient-controlled analge-sia and epidural techniques were used in fewer than 14% ofpatients.12 In Ireland, a 2007 study showed that 71%of teaching hospitals had a formalized Acute Pain Service,of which 85% were established after 1990.13

Moreover, where a dedicated Acute Pain Service exists,there is variability in its structure and the types of servicesit provides. To illustrate the variability in Acute PainService staffing and services provided at several USAhospitals, we created Table 2, which is based on informalcommunication with colleagues at those facilities.

The heterogeneity in structure and function of painservices across facilities makes it difficult to state unequivo-cally whether an Acute Pain Service is or is not cost-effective. To define the exact nature of an Acute PainService is a crucial component of its economic study.

Anesthesia groups in the USA currently are likely toprovide an Acute Pain Service for a variety of reasonssuch as:

• An anesthesia residency training program exists at thehospital. As of 2008, a month-long acute pain experi-ence for anesthesia trainees is required.

• A hospital’s practice is to place a large number ofperipheral nerve catheters, which then extends theanesthesiologist’s responsibility longer and leads to anAcute Pain Service.

• The anesthesia group is paid for its service by thirdparty payers such as the federal government or pri-vate insurance, and the marginal revenue exceeds themarginal costs.

• The anesthesia group’s contract and professionalservice agreement with the hospital stipulates itsexistence. The hospital may subsidize the AcutePain Service budget if there is a perceived competi-tive advantage to the hospital to attract morecases.

Table 1. Thirty Most Common Diagnosis Related Groups for Inpatients Having Surgery Requiring Anesthesia

DRG name

Discharges coded withoutcomplications and

comorbidities

Discharges coded withcomplications and

comorbiditiesTotal in theUSA in 2008

Cesarean 904,000 472,000 1,376,000Major joint replacement or reattachment of lower extremity 912,000 56,000 968,000Uterine and adnexa procedure for nonmalignancy 430,000 97,000 527,000Major small and large bowel procedures 95,000 275,000 370,000Laparoscopic cholecystectomy without common duct exploration 191,000 137,000 328,000Hip and femur procedures except major joint (e.g., hip fracture) 85,000 169,000 254,000Appendectomy 196,000 26,000 222,000Back and neck procedures except spinal fusion 157,000 51,000 208,000Spinal fusion except cervical 198,000 9100 207,100Lower extremity and humerus procedures except hip, foot, femur

(e.g., tibia fracture)149,000 54,000 203,000

Cervical spinal fusion 123,000 26,000 149,000Major cardiovascular procedures 84,000 53,000 137,000OR procedures for obesity 109,000 20,000 129,000Craniotomy and endovascular intracranial procedures 58,000 66,000 124,000Extensive OR procedure unrelated to principal diagnosis 24,000 99,000 123,000Major chest procedures 37,000 86,000 123,000Extracranial procedures 85,000 25,000 110,000Coronary bypass with cardiac catheterization 68,000 38,000 106,000Other respiratory system OR procedures 14,000 92,000 106,000Vaginal delivery with sterilization and/or D&C 101,000 None 101,000Major male pelvic procedures 79,000 17,000 96,000Infectious and parasitic diseases with OR procedure 1000 91,000 92,000Transurethral prostatectomy 44,000 45,000 89,000Peritoneal adhesiolysis 35,000 53,000 88,000Coronary bypass without cardiac catheterization 63,000 24,000 87,000Revision of hip or knee replacement 41,000 43,000 84,000Hernia procedures except inguinal and femoral 50,000 33,000 83,000Appendectomy with complicated principal diagnosis 48,000 24,000 72,000Cardiac valve and other major cardiothoracic procedures without

cardiac catheterization12,000 59,000 71,000

Total 4,393,000 2,240,100 6,633,100

Note: Because these are hospital discharges, a patient could have had more than 1 visit to the operating room for a surgical procedure during 1 hospitalization.Thus, the actual number of operations is likely higher than the numbers reported here.DRG � diagnosis related group; USA � United States of America; OR � operating room; D&C � dilatation and curettage.

EDITORIAL


• The surgeons’ desire to deflect the night phone callsand pain control questions to the anesthesia groupespecially if the anesthesia group is being paid fortaking calls (depending on the contract) whereas thesurgeons are not.

Financial considerations are often largely responsible forthe slow adoption of a dedicated Acute Pain Service.Although the study by Lee et al. in this issue of the journal,along with the studies that preceded it,14 represent impor-tant first steps, more research is needed to find ways toovercome organizational constraints and deliver the spe-cialized care of an Acute Pain Service in an economicallyviable manner.

RESEARCH QUESTIONS THAT DESERVEFURTHER STUDYThe literature leaves unanswered the question of what ser-vices a dedicated Acute Pain Service should provide and themost cost-effective method in which to provide them. Studiesto date have compared a wide variety of staffing models andservices to more conventional methods of postoperative painmanagement. However, no studies have explicitly comparedvarious Acute Pain Service models with each other. In alllikelihood, the most cost-effective way to implement an AcutePain Service will vary among hospitals because of localvariations in culture and personnel, for example. In addition,smaller hospitals that perform a range of surgical cases mayface more variability in the demand for expert postoperativepain management. This observation is based on the law oflarge numbers, which states that there is more variability insmaller samples. Such low volume hospitals will need to findways to manage this variability, such as pooling staff andresources across hospitals.

More work also needs to be done to quantify the efficacyof a dedicated Acute Pain Service. A typical outcome

considered is some measure of pain scores. The Quality-of-Recovery instrument, for instance, has 9 questions, at least4 of which pertain to pain or side effects from analgesics.15

However, a challenge for the health economist is that thesemeasures need to be translated into units more easilyunderstood by decision makers.

For example, efforts could be made to translate thesepain scores to quality-adjusted life years, the measure ofefficacy used by many public payers.16 However, the timehorizon in acute surgical care studies is typically measuredin days and weeks and not months or years as for treat-ments aimed at relieving chronic conditions. As a result, adedicated Acute Pain Service may have little effect on theconventional metric of quality-adjusted life years. There-fore, additional efforts should be made to define whatmeasures payers should consider, and present economicanalyses in a way that practicing clinicians and payers canunderstand and translate to their practice.

Future studies could also report alternative outcomes ofrelevance to hospital administrators. For example, a group ofpain service interventions increased Press Ganey measure-ments of patient satisfaction from the 87th to the 99th percen-tile.17 More understanding is needed regarding which of theseveral reported interventions had the largest impact.

Efforts to date to estimate the efficacy of a dedicated AcutePain Service have included consideration of the average effectin a study population. However, this approach may obscurethe utility of an Acute Pain Service for 2 reasons.18 The firstreason concerns differences in patient responses. If a dedicatedAcute Pain Service provides a large benefit for a small portionof the population, the average effect will be small, but clearlythe Acute Pain Service still has an important clinical role.

The second reason concerns dependence in responsesacross different alternatives to pain management. Put sim-ply, the issue here is whether patients who fail to respond

Table 2. Staffing and Care Provided by Acute Pain ServicesHospital

A B C D E F GAnesthesiologist (FTEs/d) 0.8 1 1 1 1 1 XHousestaff (FTEs/d) 1 1 1 No 1 1 NoRN (FTEs/d) No 1 0 No 1 No NoNP (FTEs/d) No No 1 No No No NoOther staff Fellow Pharmacist No No No No NoNo. of patients/d 12 18 12 (Monday)–30

(Friday)4 10–20 8–20 10–15

Services 0 � never provided; 10 � most frequently provided serviceManage routine IV PCA 0 0 0 2 1 0 10Manage complex IV PCA (e.g., patients

on chronic pain medications)10 8 3 7 3 8 2

Manage peripheral nerve catheter/infusions

10 10 10 10 10 3 0

Manage intraoperative spinal opioids 10 10 0 2 2 1 3Consult on nonsurgical acute pain (e.g.,

lumbar puncture, epidural cathetersfor rib fractures)

10 3 5 2 4 3 1

Manage epidural catheter 10 10 10 0 9 10 1Outpatient nerve block catheters No Yes No No Yes No No24-h coverage Yes Yes Yes Yes Yes Yes Yes

Note: Hospitals listed in this table are academic medical centers except hospitals D and G, which are community hospitals. Hospital D has no formal acute painservice in place but it is being proposed for funding. Hospital G has the post 2nd call person round on the patients in the morning, and the 1st call person answersthe calls at all other times.FTE � full time equivalent; RN � registered nurse; NP � nurse practitioner; PCA � patient-controlled analgesia.

Acute Pain Service Economics


to basic methods of pain management are likely to see largeimprovement with the expertise of an Acute Pain Service. Ifan Acute Pain Service is likely to provide large benefits forpatients who are not responding to other modalities, theservice also has an important clinical role, even if itsaverage effect (measured across a larger population) issmall. This latter possibility is particularly relevant in somehospitals where the anesthesiologist-staffed Acute PainService typically focuses on particularly difficult cases thathave been refractory to more conventional methods. Ulti-mately, these factors suggest that a dedicated Acute PainService can improve its cost-effectiveness by targeting itsefforts toward patients most likely to benefit. Unfortu-nately, our ability to identify these patients remains limitedand is another area for further study.19

Finally, in calculating the costs of an Acute Pain Service,the present study focuses on the costs of the service itself,measured as staff time and the costs of administeredmedications. Personnel costs will differ among countriesboth in the absolute cost of an anesthesiologist and nurse,and in the relative cost of one to the other. This markedlyaffects the optimal mix of providers for an Acute PainService.20 In addition, these measures of costs are incom-plete, because they do not account for the net costs, takinginto account the possibility that improved pain manage-ment may reduce costs by shortening hospital stay ordecreasing the probability of complications. As an example,total knee arthroplasty patients discharged home with acontinuous femoral nerve block had reduced hospitallength of stay and associated costs and charges, not includ-ing postdischarge resource utilization.21 Because stake-holders are likely to be interested in the net costs of runningan Acute Pain Service, future studies should investigatehow pain management affects total hospitalization anddownstream costs.

Providing an Acute Pain Service is not solely an anes-thesia issue. The hospital facility needs to invest in training,supplies, and policies. In fact, a dedicated Acute PainService provides the opportunity for anesthesiologists toextend their influence in perioperative medicine for themillions of inpatients having surgery. The tasks involvedalign well with the anesthesiologist’s existing skill set.However, the wide variety of practice models in the USAand abroad suggests there has not been agreement on howto provide these services in a financially viable way. Giventhe large number of surgical inpatients, optimizing painmanagement may yield large savings.

The truth is that whether a dedicated Acute Pain Servicecan be a cost-effective way of managing postoperative paindepends on multiple conditions, such as work culture,surgical case mix, and reimbursement systems.

AUTHOR CONTRIBUTIONSAll authors helped design and conduct the study, analyze thedata, and write the manuscript. All authors approved the finalmanuscript.

RECUSE NOTEFranklin Dexter is the Section Editor of Economics, Education,and Policy for the Journal. The manuscript was handled bySpencer Liu, Section Editor for pain Medicine. Dr. Dexter wasnot involved in any way with the editorial process or decision.

REFERENCES1. Lee A, Chang SKC, Chen PP, Gin T, Lau ASC, Chiu CH. The

costs and benefits of extending the role of the acute painservice on clinical outcomes after major elective surgery.Anesth Analg 2010;111:1042–50

2. Macario A. Truth in scheduling: is it possible to accuratelypredict how long a surgical case will last? Anesth Analg2009;108:681–5

3. American Society of Anesthesiologists Task Force on AcutePain Management. Practice guidelines for acute pain manage-ment in the perioperative setting: an updated report by theAmerican Society of Anesthesiologists Task Force on AcutePain Management. Anesthesiology 2004;100:1573–81

4. HospitalPulseReport2009.Availableat:http://www.pressganey.com/galleries/default-file/Hospital_Pulse_Report_2009.pdf. Ac-cessed June 15, 2010

5. Carr DB, Miaskowski C, Dedrick SC, Williams GR. Manage-ment of perioperative pain in hospitalized patients: a nationalsurvey. J Clin Anesth 1998;10:77–85

6. Ready LB. How many acute pain services are there in theUnited States, and who is managing patient-controlled analge-sia? Anesthesiology 1995;82:322

7. Warfield CA, Kahn CH. Acute pain management: programs inU.S. hospitals and experiences and attitudes among U.S.adults. Anesthesiology 1995;83:1090–4

8. Rawal N, Allvin R. Acute pain services in Europe: a 17-nationsurvey of 105 hospitals. The EuroPain Acute Pain WorkingParty. Eur J Anaesthesiol 1998;15:354–63

9. Powell AE, Davies HT, Bannister J, Macrae WA. Rhetoric andreality on acute pain services in the UK: a national postalquestionnaire survey. Br J Anaesth 2004;92:689–93

10. Stamer UM, Mpasios N, Stuber F, Maier C. A survey of acutepain services in Germany and a discussion of internationalsurvey data. Reg Anesth Pain Med 2002;27:125–31

11. Powell AE, Davies HT, Bannister J, Macrae WA. Challenge ofimproving postoperative pain management: case studies ofthree acute pain services in the UK National Health Service.Br J Anaesth 2009;102:824–31

12. Coluzzi F, Savoia G, Paoletti F, Costantini A, Mattia C.Postoperative pain survey in Italy (POPSI): a snapshot ofcurrent national practices. Minerva Anestesiol 2009;75:622–31

13. Hu P, Owens T, Harmon D. A survey of acute pain services inteaching hospitals in the Republic of Ireland. Ir J Med Sci2007;176:225–8

14. Lee A, Chan S, Chen PP, Gin T, Lau AS. Economic evaluationsof acute pain service programs: a systematic review. Clin J Pain2007;23:726–33

15. Myles PS, Hunt JO, Nightingale CE, Fletcher H, Beh T, Tanil D,Nagy A, Rubinstein A, Ponsford JL. Development and psycho-metric testing of a quality of recovery score after generalanesthesia and surgery in adults. Anesth Analg 1999;88:83–90

16. Tan JM, Macario A. How to evaluate whether a new technol-ogy in the operating room is cost-effective from society’sviewpoint. Anesthesiol Clin 2008;26:745–64

17. Philips BD, Liu SS, Wukovits B, Boettner F, Waldman S,Liguori G, Goldberg S, Goldstein L, Melia J, Hare M, Jasphey L,Tondel S. Creation of a novel recuperative pain medicineservice to optimize postoperative analgesia and enhance pa-tient satisfaction. HSS J 2010;6:61–5

18. Basu A, Philipson TJ. The impact of comparative effectivenessresearch on health and health care spending. NBER WorkingPaper No. 15633, 2010

19. Raja SN, Jensen TS. Predicting postoperative pain based onpreoperative pain perception: are we doing better than theweatherman? Anesthesiology 2010;112:1311–2

20. Macario A. What does one minute of operating room timecost? J Clin Anesth 2010;22:233–6

21. Ilfeld BM, Mariano ER, Williams BA, Woodard JN, Macario A.Hospitalization costs of total knee arthroplasty with a continu-ous femoral nerve block provided only in the hospital versuson an ambulatory basis: a retrospective, case-control, cost-minimization analysis. Reg Anesth Pain Med 2007;32:46–54

EDITORIAL


Society of Cardiovascular Anesthesiologists

Cardiovascular Anesthesiology Section Editor: Charles W. Hogue, Jr.

Perioperative Echocardiography and Cardiovascular Education Section Editor: Martin J. London

Hemostasis and Transfusion Medicine Section Editor: Jerrold H. Levy

Heparin Concentration–Based Anticoagulation forCardiac Surgery Fails to Reliably Predict HeparinBolus Dose RequirementsSean Garvin, MD,* Daniel C. FitzGerald, CCP,† George Despotis, MD,‡§� Prem Shekar, MD,†and Simon C. Body, MBChB, MPH*

BACKGROUND: Hemostasis management has evolved to include sophisticated point-of-caresystems that provide individualized dosing through heparin concentration–based anticoagulation. TheHepcon HMS Plus system (Medtronic, Minneapolis, MN) estimates heparin dose, activatedclotting time (ACT), and heparin dose response (HDR). However, the accuracy of this test has notbeen systematically evaluated in large cohorts.METHODS: We examined institutional databases for all patients who underwent cardiac surgerywith cardiopulmonary bypass (CPB) at our institution from February 2005 to July 2008. Duringthis period, the Hepcon HMS Plus was used exclusively for assessment of heparin dosing andcoagulation monitoring. Detailed demographic, surgical, laboratory, and heparin dosing datawere recorded. ACT, calculated and measured HDR, and heparin concentrations were recorded.Performance of the Hepcon HMS Plus was assessed by comparison of actual and target ACTvalues and calculated and measured HDR.RESULTS: In 3880 patients undergoing cardiac surgery, heparin bolus dosing to a target ACTresulted in wide variation in the postheparin ACT (r2 � 0.03). The postheparin ACT did not reachthe target ACT threshold in 7.4% (i.e., when target ACT was 300 s) and 16.9% (i.e., when targetACT was 350 s) of patients. Similarly, the target heparin level calculated from the HDR did notcorrelate with the postbolus heparin level, with 18.5% of samples differing by more than 2 levelsof the assay. Calculated and measured HDR were not linearly related at any heparin level.CONCLUSIONS: The Hepcon HMS Plus system poorly estimates heparin bolus requirements inthe pre-CPB period. Further prospective studies are needed to elucidate what constitutesadequate anticoagulation for CPB and how clinicians can reliably and practically assessanticoagulation in the operating room. (Anesth Analg 2010;111:849–55)

Administration of heparin, and monitoring of wholeblood activated clotting time (ACT), is still themainstay of anticoagulation for cardiopulmonary

bypass (CPB).1,2 Hemostasis management has evolved toinclude more sophisticated point-of-care systems that pro-vide individualized dosing of heparin and protamine,through heparin concentration–based estimation of hepa-rin dosage, and measurement of ACT.3,4 Individualizedheparin concentration–based anticoagulation managementhas been reported to cause less activation of thrombin,decreased fibrinolysis, reduced blood loss, fewer transfusions,and a decreased risk of reexploration for hemorrhage,5–8

although there is still a lack of consensus in this area.9,10

Previous studies of heparin-level monitoring devices, includ-ing the Hepcon HMS Plus, had varied results, with somestudies showing good agreement between whole blood hep-arin concentration and either ACT before CPB11 or plasmaanti-Xa levels before or during CPB,12 whereas others showeda lack of significant relationship between ACT and heparinconcentration13 or whole blood versus anti-Xa heparin con-centration.14 All of these studies were limited by their smallsample sizes. Importantly, there is previously observed inter-individual variation in measured heparin levels for a targetACT required for CPB.

Therefore, we examined the relationship between indi-vidualized in vitro estimation of the heparin level for aspecific target ACT, and the subsequently observed ACTand heparin level measured after bolus administration ofthe heparin dose response (HDR)-estimated heparindose. We hypothesized that: 1) there would be strongcorrelation between calculated and measured HDR forindividual patients; 2) residual variation in the predictedHDR would be due to patient factors that are accounted forin the calculation of the HDR; and 3) the HDR-estimatedheparin dose would achieve the desired ACT for CPB in allpatients.

From the *Department of Anesthesiology, Perioperative and Pain Medicine,and †Division of Cardiac Surgery, Brigham and Women’s Hospital, HarvardMedical School, Boston, Massachusetts; and ‡Departments of Pathology,§Immunology, and �Anesthesiology, Washington University School ofMedicine, St. Louis, Missouri.


Supported by departmental funds.

Address correspondence and reprint requests to Simon C. Body, MBChB,MPH, Brigham and Women’s Hospital, Harvard Medical School, 75 FrancisSt., Boston, MA 02215. Address e-mail to [email protected].

Copyright © 2010 International Anesthesia Research SocietyDOI: 10.1213/ANE.0b013e3181b79d09


METHODSThe Hepcon HMS Plus (Medtronic, Minneapolis, MN) wasintroduced in our institution in February 2004, in concertwith a change in heparin coagulation management. Use ofthis system was based on a belief that adequate anticoagula-tion for CPB using a heparin-coated circuit could be achievedwith lower ACTs as long as reasonable, albeit lower heparinlevels than had been historically used were maintained.Furthermore, there was a perception that reduced heparinadministration may result in reduced postoperative bleeding.

After IRB approval, we examined institutional databasesfor all patients who underwent CPB for cardiac surgery atour institution, from February 2005 to July 2008 (n � 4667).Patients with incomplete perioperative anticoagulationdata were excluded from analysis (n � 562). The majority ofpatients were excluded for having a target ACT of �350 sor subsequent administration of aprotinin (n � 157), norecord of postheparinization heparin level (n � 144), norecord of baseline ACT (n � 91), or no record of height orweight (n � 67). Furthermore, we excluded an additional225 patients who were not anticoagulated using the insti-tutional protocol described below, yielding 3880 analyzedpatients.

Anticoagulation ManagementThe Hepcon HMS Plus was used for anticoagulation man-agement according to the manufacturer’s recommenda-tions.15 The estimated blood volume for each patient wascalculated using the manufacturer’s instructions,15 accord-ing to the method described by Allen et al.14 After induc-tion of anesthesia, baseline kaolin ACT, predicted HDR,predicted heparin concentration, and heparin bolus calcu-lations were performed according to the manufacturer’sinstructions, using HDR cartridges encompassing wholeblood heparin concentration ranging from 0.4 to 3.4 U/mL.The Hepcon HMS Plus recommended CPB prime heparindose based on a 750- or 1000-mL prime volume was addedto the calculated heparin bolus and administered via acentral venous catheter. Heparin was not added to the CPBpump prime. Three minutes after United States Pharma-copeia porcine heparin administration and 10 min afteronset of CPB, heparin concentration and ACT were remea-sured. All patients received an �-aminocaproic acid load of7.5–10 g over 1 h, after the initial blood draw for baseline ACT,but before heparin administration and blood sampling formeasurement of postheparin ACT. �-Aminocaproic acid wassubsequently infused at a rate of 1.25–1.5 g/h.

Throughout the entire study period, an ACT of morethan 350 s and a minimum heparin concentration of 2U/mL were used in patients undergoing non–coronaryartery bypass graft (CABG) procedures or CABG with theuse of cardiotomy suction. Patients undergoing primaryCABG surgery without the use of cardiotomy suctionbefore March 2007 were anticoagulated using a protocolthat proscribed a minimum ACT of 300 s before the institu-tion of CPB. After February 2007, to standardize anticoagula-tion management, the minimum ACT before CPB was in-creased to 350 s with a minimum heparin concentration of 2.0U/mL for CABG surgery. These protocols were driven by adesire to reduce heparin and protamine doses to reduce blood

loss but were not supported by prior clinical evidence ofefficacy.

We also observed variability in anticoagulation practicethat was not proscribed by a protocol. Some perfusionistsadministered additional heparin when the measured hep-arin level was �1.4 or 2.0 U/mL, according to personalpreference. These patients were excluded from analysis, aspreviously described.

Data Collection and Statistical AnalysisDetailed demographic data, surgical indications, preopera-tive laboratory values, operative data, heparin bolus dose,ACT values, and measured heparin concentrations wereroutinely collected into a centralized database. The calcu-lated HDR was calculated as the target ACT � baselineACT divided by the target heparin level and reported ass � U�1 � mL�1. The calculated HDR differed from themethod used by the Hepcon HMS Plus,15 but prospectivecomparison of the Hepcon HMS Plus and the abovecalculation in 15 patients showed excellent correlation (r2 �0.991). This calculation uses the target heparin concentra-tion identified from the HDR slope generated by theHepcon HMS Plus, not the protocol-driven heparin level.Thus, the calculation of the HDR slope is independent ofthe clinical protocol and unaffected by the heparin leveldesired or obtained after heparin dosing.

The measured HDR was calculated as postheparinACT � baseline ACT divided by the measured heparinlevel and reported as s � U�1 � mL�1. By using the measuredpostheparinization ACT and measured heparin level, themeasured HDR is independent of the clinical protocolbecause the HDR slope has been demonstrated to be linearover the clinical range of heparin concentrations.11

The Hepcon HMS Plus heparin assay has limited fidel-ity, reporting only 6 categorical heparin levels, with inter-mediate values of heparin reported to the closest level.Accordingly, we could not directly compare the calculatedand measured HDR. Therefore, we report the calculatedHDR for each measured postbolus heparin level. Pearsonproduct-moment correlation was performed between cal-culated and measured values of HDR, and heparin concen-tration and ACT data. ACT data at each heparin level werecompared with Wilcoxon’s ranked sum test. Statisticalanalysis was performed with JMP version 7.02 (SAS, Cary,NC). All data are presented as mean � sd or median andinterquartile range, as appropriate. A 2-sided P � 0.05was considered as showing statistical significance.

RESULTSA total of 3880 patients were analyzed. Baseline demo-graphics and perioperative data are summarized in Table 1.Target ACTs of 300 and 350 s were used in 23.4% and 76.6%of patients, respectively. Heparin dose, ACT, and heparinlevel after the heparin bolus for each target ACT aredetailed in Table 2. The mean heparin dose calculated toachieve a target ACT of 300 s was 152 U/kg, and for atarget ACT of 350 s was 179 U/kg. After administration ofthe HDR-estimated heparin dose, the target ACT was notachieved in 7.4% of patients with a target ACT of 300 s, andit was not reached in 16.9% of patients with a target ACT of350 s (Fig. 1). Additional heparin was administered either

Anticoagulation for Cardiac Surgery


before or immediately after initiation of CPB to 91% ofpatients who failed to achieve their target ACT of 300 s andto 77% of patients who failed to achieve their target ACT of350 s.

There was wide variation among patients for the ACTseen at each heparin level (Fig. 2). Measured HDR variedsubstantially among patients (mean � sd, 104.7 � 24.3s � U�1 � mL�1; median [10%–90% interquartile range], 102[77–137]). Correlation (r2) between the calculated and mea-sured HDR was poor at all heparin levels (r2 � 0.02; Fig. 3).

Figure 1. Observed activated clotting time (ACT) at each target ACTlevel. Box plot represents the 5th, 25th, median, 75th, and 95thpercentiles of data.

Table 1. Demographics and Perioperative DataStudy cohort(N � 3880)

Age 65.2 � 13.4Male gender 2538 (65.4%)Height (cm) 171 � 10Weight (kg) 82.2 � 19.0BSA (m2) 1.94 � 0.24Indication for surgery

Coronary artery disease 2394 (63.3%)Valve disease 1932 (53.1%)Aortic dissection 41 (1.2%)

Medical historyPrevious myocardial infarction 989 (26.2%)Hypertension 2550 (67.5%)Insulin-dependent diabetes 343 (9.1%)Renal insufficiency (creatinine

�2.0 mg/dL)145 (4.0%)

Preoperative medicationsAspirin 2387 (63.1%)Adenosine diphosphate inhibitors 182 (4.8%)Intravenous heparin 249 (6.6%)Warfarin 34 (0.9%)

Preoperative laboratory valuesHematocrit (%) 38.8 � 4.9Platelet count (�109/L) 247 � 81INR 1.14 � 0.30INR �1.4 280 (7.8%)Partial thromboplastin time (s) 36.2 � 15.9Partial thromboplastin time �37 s 733 (20.9%)

Data are presented as mean � 1SD for demographics.The upper limit of the laboratory reference range for partial thromboplastintime is 37 s.BSA � body surface area; CABG � coronary artery bypass graft; INR �International normalized ratio.

Table 2. Intraoperative Anticoagulation ManagementMean � SD Median (10th–90th percentile)

Baseline ACT (s) 138 � 17 137 (118–159)Target ACT of 300 s (N � 908)

Initial heparin dose (U) 12,994 � 3909 12,000 (9000–18,000)Heparin dose (U/kg) 152 � 38 149 (108–203)Postbolus ACT (s) 387 � 73 377 (309–466)Postbolus ACT �300 s 67 (7.4%)CPB � 10 min ACT (s) 340 � 54 335 (274–409)

Postheparin bolus CPB � 10 minHeparin level (U/mL)

�1.4 16 (1.8%) 250 (27.9%)1.4 232 (25.6%) 498 (55.5%)2.0 460 (50.7%) 118 (13.2%)2.7 137 (15.1%) 21 (2.3%)3.4 60 (6.6%) 10 (1.1%)�3.4 3 (0.3%) 0 (0%)

Mean � SD Median (10th–90th percentile)Target ACT of 350 s (N � 2972)

Initial heparin dose (U) 14,313 � 4013 14,000 (10,000–20,000)Heparin dose (U/kg) 179 � 42 174 (132–232)Postbolus ACT (s) 418 � 83 406 (332–509)Postbolus ACT �350 s 501 (16.9%)CPB � 10 min ACT (s) 386 � 73 377 (313–466)

Postheparin bolus CPB � 10 minHeparin level (U/mL)

�1.4 20 (0.7%) 278 (9.4%)1.4 438 (14.7%) 1716 (58.2%)2.0 1449 (48.8%) 789 (26.8%)2.7 731 (24.6%) 125 (4.2%)3.4 315 (10.6%) 41 (1.4%)�3.4 19 (0.6%) 0 (0%)

ACT � activated clotting time; CPB � cardiopulmonary bypass.


To exclude an outlier effect, the analysis was repeated in asubgroup of patients within 1 sd of the mean postheparinACT, with similar results.

The predicted heparin level expected after heparin bolusadministration was compared with the heparin level mea-sured after heparin bolus. Levels that differed by more than0.3 U/mL differed by more than the sensitivity (1 channel)of the assay and were observed in 51.0% of samples (Fig. 4).Levels that differed by more than 0.7 U/mL differed bymore than 2 channels in the assay and were observed in18.5% of samples.

DISCUSSIONWe examined the performance of the Hepcon HMS Plus toguide anticoagulation in 3880 patients undergoing cardiacsurgical procedures. Data provided by the Hepcon HMSPlus were used to calculate heparin dosing throughout CPBand to direct the administration of protamine after CPB. Weobserved wide variability in heparin dose requirements toachieve an adequate ACT for CPB, as others have previ-ously described.1 We also observed poor correlation of thecalculated HDR with the measured HDR. This led to ACTvalues that were less than the target ACT values whenheparin dosing was guided by estimation of HDR, in7.4%–16.9% of patients. The imprecision for the calculatedHDR may explain the high frequency of failure to reach thetarget ACT.

Interpatient variability in HDR is well described, anduse of a dose-response plot recommended by Bull et al.3

provides the basis of Hepcon HMS Plus calculations. Ourinitial hypothesis that the HDR slope calculated by theHepcon HMS Plus would correlate with the measuredHDR slope is not supported by the data. The interindividualvariability in heparin response could be attributable to a

Figure 2. Observed activated clotting time at each measured post-bolus heparin level. Box plot represents the 5th, 25th, median, 75th,and 95th percentiles of data.

Figure 3. Correlation between calculated and measured heparindose response. Each panel contains data for individuals with aspecific measured heparin level after heparinization. Panel A repre-sents a measured heparin level of 1.4 U/mL (r2 � 0.01). Panel Brepresents a measured heparin level of 2.0 U/mL (r2 � 0.02). PanelC represents a measured heparin level of 2.7 U/mL (r2 � 0.01).Panel D represents a measured heparin level of 3.4 U/mL (r2 �0.03).



number of in vivo factors, including error in the estimation ofblood volume, effects of release of tissue factor pathwayinhibitor (TFPI) by heparin in vivo but not ex vivo, extravascu-lar sequestration of heparin, plasma protein binding, circulat-ing antithrombin, and platelet activation.16

Estimation of blood volume is a potential source of errorin heparin dose calculation. The methodology used byseveral previous studies5,6,12 is predominantly based on themethod described by Allen et al.14 in 1956. The magnitudeof error in estimation of blood volume in individualpatients undergoing cardiac surgery has not been system-atically evaluated. It is probable that the severity andnature of cardiac disease may be a substantial source ofvariation in blood volume estimates, thereby contributingto imprecision of point-of-care anticoagulation managementinstruments. Further investigation aimed at improving estima-tion of blood volume in patients undergoing cardiac surgerymay result in better performance of point-of-care heparinconcentration–based anticoagulation systems.

Potentially contributing to the imprecision of the Hep-con HMS Plus are the effects of release of TFPI by heparinin vivo but not ex vivo. TFPI in vivo is bound to endothelialcell surfaces with small amounts circulating in plasma andbound to platelets.17 TFPI is an endogenous serine proteaseinhibitor that exerts its action primarily through neutraliz-ing factor Xa by complexing with factor Xa and forming aTFPI-FXa complex, while also providing feedback inhibi-tion of the factor VIIa–tissue factor complex.18 Plasma TFPIis released from endothelial surface stores by administra-tion of unfractionated and low-molecular-weight hepa-rins,19 increasing plasma levels during CPB.20 Brodin etal.21 recently demonstrated significant synergy betweenantithrombin and TFPI in postheparin plasma with analmost equal contribution of each to the prolongation ofanticoagulation in a laboratory setting. In the absence of anendothelial source of TFPI, the contribution of TFPI to

heparin responsiveness may be underestimated by theinitial HDR calculation performed by the Hepcon HMSPlus. This source of error in the calculated HDR wouldseemingly lead to increased heparin dosages and higherpostheparin ACT, which is incongruous with our finding of7.4%–16.9% of patients failing to achieve their target ACT,but congruous with the mean ACT being higher than thetarget ACT in our overall population. Although the exactcontribution of TFPI to point-of-care measurements ofheparin responsiveness has not been established, strongcorrelation between the TFPI-responsive Heptest (Ameri-can Diagnostica, Stamford, CT) measurements of heparinlevel and ACT in the pre-CPB period22 has been observed,and release of TFPI correlates with the Heptest measure-ment of heparin response in normal volunteers receiving IVheparin.23 These observations reinforce the importance ofTFPI in heparin responsiveness and point to a likely impacton point-of-care assessments of HDR.

Earlier studies of these heparin-level monitoring de-vices, including the Hepcon HMS Plus, had varied resultswith some studies showing good agreement betweenwhole blood heparin concentration and either ACT beforeCPB11 or plasma anti-Xa levels,12 whereas others showed alack of significant relationship between ACT and heparinconcentration24 or whole blood versus anti-Xa heparinconcentration during CPB.13,14 One study12 revealed sub-stantial interpatient variability in HDR; however, the linearrelationship between heparin concentration and kaolinACT within individual patients was generally exceptional.Our retrospective analysis demonstrated significant vari-ability in the responsiveness of the kaolin ACT and heparinin vivo versus kaolin ACT and heparin ex vivo (i.e., afterIV administration of heparin), and calculated versusmeasured HDR, respectively (Fig. 3). This was paralleledby the finding that the Hepcon HMS Plus– guided hep-arin administration failed to result in adequate anti-coagulation for CPB in 16% of patients in this studywhen the target ACT was 350 s, even with the CPB primeheparin dose being administered as part of the initial IVbolus dose.

Early work performed by Bull et al.25 and Younget al.26 continues to form the basis for the ACT target usedto safely institute CPB, with many centers using an ACT ofmore than 480 s as a target. In 1981, Jobes et al.27 deter-mined that a heparin level of 2 U/mL was associated withan ACT �300 s in �95% of patients. These data were notthe basis for our use of a heparin level of 2 U/mL in thelatter portion of the study period, but are in line with ourobservation of wide variation in ACT values at specificheparin levels. The theoretical benefit of using heparinconcentration– based anticoagulation is that a stable hepa-rin level may be achieved, thereby minimizing hemostaticactivation7,28 and keeping the patient fully anticoagulatedthroughout the CPB period, which would lead to reducedblood loss and transfusion.5,6 We observed wide variabilityin ACT for any given heparin level (Fig. 2). Almost allstudies using heparin concentration have used this histori-cal reference as the target for defining adequate anticoagu-lation, but establishment of a “safe heparin level” has notbeen systematically studied.

Figure 4. Target and measured heparin level observed after bolusheparin administration. Box plot represents the 5th, 25th, median,75th, and 95th percentiles of data.


This study is limited by its retrospective nature. Wewere not able to capture all of the perioperative factors thatmay influence anticoagulation and an individual’s re-sponse to heparin administration. In addition, importantclinical outcomes are not part of this data set, includingperioperative complications that may relate to anticoagula-tion management, such as bleeding, transfusion, reexplora-tion, myocardial infarction, new or worsening renal insuf-ficiency, and cerebral vascular accident. Furthermore, ourcomparison of the calculated and measured HDR is limitedby the imprecision of the automated protamine titrationmethod, which measures whole blood heparin concentrationusing discrete cutoffs using only the 6 channels of the car-tridge. Thus, we are left with 4 likely sources of disparitybetween calculated and measured HDR: 1) inaccurate esti-mate of the patient’s blood volume; 2) lack of fidelity ofmeasured heparin concentration because only 6 categoriesof heparin level describe the linear range of heparin con-centration; 3) inherent inaccuracy of the device; and 4)differences between the anticoagulant activity of heparin exvivo compared with its several actions in vivo, notably uponthe release of TFPI. We were unable to subset these possiblecauses further, but they seem to be significant.

We conclude that the Hepcon HMS Plus fails to consis-tently provide the therapeutic heparin bolus dose uniformlyin all patients based on the wide discrepancy in calculatedversus measured HDR. This can lead to inadequate heparindoses to achieve a target ACT for CPB in as much as 16.9% ofpatients. However, the Hepcon HMS Plus was able to identifyan adequate heparin dose for the majority of the patients.Because this study, did not compare the Hepcon HMS Pluswith empiric dosing regimens, we are uncertain whetherempiric regimens can either under- or overperform whencompared to this system. Further prospective studies areneeded to elucidate what constitutes adequate anticoagula-tion for CPB and how clinicians can reliably and practicallyassess anticoagulation in the operating room.

REFERENCES1. Akl BF, Vargas GM, Neal J, Robillard J, Kelly P. Clinical

experience with the activated clotting time for the control ofheparin and protamine therapy during cardiopulmonary by-pass. J Thorac Cardiovasc Surg 1980;79:97–102

2. Hattersley PG. Activated coagulation time of whole blood.JAMA 1966;196:436–40

3. Bull BS, Huse WM, Brauer FS, Korpman RA. Heparin therapyduring extracorporeal circulation. II. The use of a dose-response curve to individualize heparin and protamine dos-age. J Thorac Cardiovasc Surg 1975;69:685–9

4. Raymond PD, Ray MJ, Callen SN, Marsh NA. Heparin moni-toring during cardiac surgery. Part 1: Validation of whole-blood heparin concentration and activated clotting time. Per-fusion 2003;18:269–76

5. Despotis GJ, Joist JH, Hogue CW Jr, Alsoufiev A, Kater K,Goodnough LT, Santoro SA, Spitznagel E, Rosenblum M,Lappas DG. The impact of heparin concentration and activatedclotting time monitoring on blood conservation. A prospective,randomized evaluation in patients undergoing cardiac opera-tion. J Thorac Cardiovasc Surg 1995;110:46–54

6. Jobes DR, Aitken GL, Shaffer GW. Increased accuracy andprecision of heparin and protamine dosing reduces blood lossand transfusion in patients undergoing primary cardiac opera-tions. J Thorac Cardiovasc Surg 1995;110:36–45

7. Despotis GJ, Joist JH, Hogue CW Jr, Alsoufiev A, Joiner-MaierD, Santoro SA, Spitznagel E, Weitz JI, Goodnough LT. Moreeffective suppression of hemostatic system activation in pa-tients undergoing cardiac surgery by heparin dosing based onheparin blood concentrations rather than ACT. Thromb Hae-most 1996;76:902–8

8. Bowie J, Kemma G. Automated management of heparin anti-coagulation in cardiovascular surgery. Proc Am Acad Cardio-vasc Perf 1985;6:1–10

9. Shore-Lesserson L, Reich DL, DePerio M. Heparin and prota-mine titration do not improve haemostasis in cardiac surgicalpatients. Can J Anaesth 1998;45:10–8

10. Slight RD, Buell R, Nzewi OC, McClelland DB, Mankad PS. Acomparison of activated coagulation time-based techniques foranticoagulation during cardiac surgery with cardiopulmonarybypass. J Cardiothorac Vasc Anesth 2008;22:47–52

11. Despotis GJ, Alsoufiev AL, Spitznagel E, Goodnough LT,Lappas DG. Response of kaolin ACT to heparin: evaluationwith an automated assay and higher heparin doses. AnnThorac Surg 1996;61:795–9

12. Despotis GJ, Summerfield AL, Joist JH, Goodnough LT, San-toro SA, Spitznagel E, Cox JL, Lappas DG. Comparison ofactivated coagulation time and whole blood heparin measure-ments with laboratory plasma anti-Xa heparin concentration inpatients having cardiac operations. J Thorac Cardiovasc Surg1994;108:1076–82

13. Hardy JF, Belisle S, Robitaille D, Perrault J, Roy M, Gagnon L.Measurement of heparin concentration in whole blood withthe Hepcon/HMS device does not agree with laboratorydetermination of plasma heparin concentration using a chro-mogenic substrate for activated factor X. J Thorac CardiovascSurg 1996;112:154–61

14. Allen TH, Peng MT, Chen KP, Huang TF, Chang C, Fang HS.Prediction of blood volume and adiposity in man from bodyweight and cube of height. Metabolism 1956;5:328–45

15. Medtronic Perfusion Systems. Hepcon HMS Plus operator’smanual. Minneapolis, MN, 2001


17. Bajaj MS, Kuppuswamy MN, Saito H, Spitzer SG, Bajaj SP.Cultured normal human hepatocytes do not synthesizelipoprotein-associated coagulation inhibitor: evidence that en-dothelium is the principal site of its synthesis. Proc Natl AcadSci USA 1990;87:8869–73

18. Broze GJ Jr, Miletich JP. Characterization of the inhibition oftissue factor in serum. Blood 1987;69:150–5

19. Hoppensteadt DA, Walenga JM, Fasanella A, Jeske W, FareedJ. TFPI antigen levels in normal human volunteers afterintravenous and subcutaneous administration of unfraction-ated heparin and a low molecular weight heparin. Thromb Res1995;77:175–85

20. Adams MJ, Cardigan RA, Marchant WA, Grocott MP, MythenMG, Mutch M, Purdy G, Mackie IJ, Machin SJ. Tissue factorpathway inhibitor antigen and activity in 96 patients receivingheparin for cardiopulmonary bypass. J Cardiothorac VascAnesth 2002;16:59–63

21. Brodin E, Appelbom H, Osterud B, Hilden I, Petersen LC,Hansen JB. Regulation of thrombin generation by TFPI inplasma without and with heparin. Transl Res 2009;153:124–31

22. Hellstern P, Bach J, Simon M, Saggau W. Heparin monitoringduring cardiopulmonary bypass surgery using the one-steppoint-of-care whole blood anti-factor-Xa clotting assay heptest-POC-Hi. J Extra Corpor Technol 2007;39:81–6

23. Hoffmann U, Harenberg J, Bauer K, Huhle G, Tolle AR,Feuring M, Christ M. Bioequivalence of subcutaneous andintravenous body-weight-independent high-dose low-molecular-weight heparin Certoparin on anti-Xa, Heptest, andtissue factor pathway inhibitor activity in volunteers. BloodCoagul Fibrinolysis 2002;13:289–96

24. Culliford AT, Gitel SN, Starr N, Thomas ST, Baumann FG,Wessler S, Spencer FC. Lack of correlation between activatedclotting time and plasma heparin during cardiopulmonarybypass. Ann Surg 1981;193:105–11



25. Bull BS, Korpman RA, Huse WM, Briggs BD. Heparin therapyduring extracorporeal circulation. I. Problems inherent inexisting heparin protocols. J Thorac Cardiovasc Surg1975;69:674–84

26. Young JA, Kisker CT, Doty DB. Adequate anticoagulationduring cardiopulmonary bypass determined by activated clot-ting time and the appearance of fibrin monomer. Ann ThoracSurg 1978;26:231–40

27. Jobes DR, Schwartz AJ, Ellison N, Andrews R, Ruffini RA,Ruffini JJ. Monitoring heparin anticoagulation and its neutral-ization. Ann Thorac Surg 1981;31:161–6

28. Koster A, Fischer T, Praus M, Haberzettl H, Kuebler WM,Hetzer R, Kuppe H. Hemostatic activation and inflammatoryresponse during cardiopulmonary bypass: impact of heparinmanagement. Anesthesiology 2002;97:837–41


Heparin Dose Response Is Independent ofPreoperative Antithrombin Activity in PatientsUndergoing Coronary Artery Bypass Graft SurgeryUsing Low Heparin ConcentrationsSean Garvin, MD,* Daniel Fitzgerald, CCP,† Jochen D. Muehlschlegel, MD,* Tjorvi E. Perry, MD,*Amanda A. Fox, MD,* Stanton K. Shernan, MD,* Charles D. Collard, MD,‡ Sary Aranki, MD,†and Simon C. Body, MBChB, MPH*

BACKGROUND: Unfractionated heparin’s primary mechanism of action is to enhance theenzymatic activity of antithrombin (AT). We hypothesized that there would be a direct associationbetween preoperative AT activity and both heparin dose response (HDR) and heparin sensitivityindex (HSI) in patients undergoing coronary artery bypass graft surgery.METHODS: Demographic and perioperative data were collected from 304 patients undergoingprimary coronary artery bypass graft surgery. AT activity was measured after induction of generalanesthesia using a colorimetric method (Siemens Healthcare Diagnostics, Tarrytown, NY).Activated coagulation time (ACT), HDR, and HSI were measured using the Hepcon HMS Plussystem (Medtronic, Minneapolis, MN). Heparin dose was calculated for a target ACT usingmeasured HDR by the same system. Multivariate linear regression was performed to identifyindependent predictors of HDR. Subgroup analysis of patients with low AT activity (�80% normal;�0.813 U/mL) who may be at risk for heparin resistance was also performed.RESULTS: Mean baseline ACT was 135 � 18 seconds. Mean calculated HDR was 98 � 21s/U/mL. Mean baseline AT activity was 0.93 � 0.13 U/mL. Baseline AT activity was notsignificantly associated with baseline or postheparin ACT, HDR, or HSI. Addition of AT activity tomultivariable linear regression models of both HDR and HSI did not significantly improve modelperformance. Subgroup analysis of 49 patients with baseline AT �80% of normal levels did notreveal a relationship between low AT activity and HDR or HSI. Preoperative AT activity, HDR, andHSI were not associated with cardiac troponin I levels on the first postoperative day, intensivecare unit duration, or hospital length of stay.CONCLUSION: Although enhancing AT activity is the primary mechanism by which heparinfacilitates cardiopulmonary bypass anticoagulation, low preoperative AT activity is not associatedwith impaired response to heparin or to clinical outcomes when using target ACTs of 300 to 350seconds. (Anesth Analg 2010;111:856–61)

More than 50 years after heparin’s discovery,1

Rosenberg and Damus2,3 identified its primarymechanism of action as enhancement of the en-

zymatic activity of antithrombin (AT). AT binds and inac-tivates serine proteases’ contact activation and commonpathways, principally factors IIa, Xa, IXa, and VIIa, bydecreasing their binding efficiency for substrate. AT’s inhi-bition of serine protease activity is increased several ordersof magnitude by heparin as a result of a conformationalchange induced by binding of a specific pentasaccharideunit of heparin with high affinity for AT.4

Heparin resistance has been reported in 4% to 22% ofpatients undergoing cardiopulmonary bypass (CPB).5–10

Potential risk factors include age older than 65 years,

platelet count �300,000 cells/mm3, recent heparin expo-sure, and AT deficiency.9 Preoperative AT activity of �80%of normal has been associated with reduced heparin re-sponse in adults undergoing cardiac surgery.11 Postopera-tive AT deficiency has also been associated with worsenedclinical outcomes, including increased intensive care unit(ICU) length of stay, risk of reexploration for bleeding, andthromboembolic events.12

Because the primary mechanism of action of heparin isto facilitate the enzymatic activity of AT, we surmised thatreduced AT activity would be associated with heparinresponse. Little evidence is available to support this hy-pothesis, with most data regarding AT activity and heparindose response (HDR) reported from studies examining resto-ration of heparin responsiveness in cardiac surgical patientsby administration of recombinant AT concentrate, usuallywithout measurement of the patient’s AT level.5–8,13,14 Weaimed to assess the relationship between preoperative ATactivity and heparin sensitivity in a cohort of primarycoronary artery bypass graft (CABG) patients and addi-tionally assess this association in a subgroup of subjectswith low preoperative AT activity. We further examinedwhether AT activity or measures of heparin sensitivitywere associated with severity of myocardial injury or

From the *Department of Anesthesiology, Perioperative and Pain Medicine,and †Division of Cardiac Surgery, Brigham and Women’s Hospital, HarvardMedical School, Boston, Massachusetts; and ‡Division of CardiovascularAnesthesia at the Texas Heart Institute, Baylor College of Medicine, SaintLuke’s Episcopal Hospital, Houston, Texas.

Accepted for publication November 18, 2009.

Address correspondence and reprint requests to Simon C. Body, MBChB,MPH, Brigham and Women’s Hospital, Harvard Medical School, 75 FrancisSt., Boston, MA 02215. Address e-mail to [email protected].

Copyright © 2010 International Anesthesia Research SocietyDOI: 10.1213/ANE.0b013e3181ce1ffa


duration of ICU and hospital stays, as surrogates of sever-ity of illness.

METHODSWith IRB approval and individual patient consent, 346patients undergoing primary CABG using CPB from Feb-ruary 2005 to December 2006 were enrolled into a parentstudy called the CABG Genomics Program with the aim ofdetermining genetic risks for adverse perioperative out-comes. Patients were excluded from the parent study ifthey were younger than 20 years of age; underwent repeator off-pump CABG, planned concomitant valve, or othercardiac surgery; had a preoperative hematocrit �25%; or ifthey had received leukocyte-rich blood products within 30days before surgery. For this secondary analysis, detaileddemographic data, preoperative laboratory values, opera-tive data, heparin bolus dose, anticoagulation data includ-ing activated coagulation time (ACT) values, and measuredheparin concentrations were collected from hospital records.Patients with missing laboratory or demographic data (n �27) and those receiving warfarin (n � 15) were furtherexcluded from analysis, yielding 304 analyzable patients.ICU length of stay until first discharge was recorded inhours. Hospital length of stay was recorded in days, withthe surgical day and last hospital day included as wholedays.

Baseline blood samples for ACT, HDR, AT, and otherassays were obtained after induction of anesthesia butbefore surgical incision (described as preoperative hereaf-ter). Free unbound AT activity was measured with acolorimetric method (Modular Analytics biochemistry ana-lyzer, Siemens Healthcare Diagnostics, Tarrytown, NY)performed by Charles River Laboratory (Montreal, Can-ada). The assay limit of quantitation was 21.6%. To reporthuman plasma AT activity results in IU/mL, the NationalInstitute for Biological Standards and Control Second In-ternational Reference Standard was used to determine aconversion factor of activity in IU/mL � AT activity in % �0.0102. AT content was measured using an immunoneph-elometric method using a BN-100 Prospec nephelometer(Dade Behring Diagnostics, Marburg, Germany). Coeffi-cient of variation for AT measurement was �5% withinassay and �10% across assays. The assay limit of quanti-tation was 0.00672 mg/mL. To report human plasma ATcontent results in IU/mL, National Institute for BiologicalStandards and Control Second International ReferenceStandard was used to determine a conversion factor ofcontent in IU/mL � content in g/L � 3.64. Both assaysmeasure free AT, rather than AT complexed with heparin.Plasminogen-activator inhibitor-1, tissue factor, d-dimer,protein C, and cardiac troponin I (cTnI) were measuredusing a sandwich immunoassay on a triage platform usingmonoclonal and polyclonal antibodies (Biosite, San Diego,CA). Hemoglobin, platelet, and white blood counts, alongwith prothrombin time, partial thromboplastin time, andinternational normalized ratio were measured by a centralhematology laboratory according to institutional protocols.Complete blood counts were performed using the Advia2120 Hematology System (Siemens Healthcare Diagnostics,Deerfield, IL), and coagulation studies were performed onSTA-R Evolution (Diagnostica Stago, Parsippany, NJ).

For anticoagulation management, the Hepcon HMS Plussystem (Medtronic, Minneapolis, MN) was used accordingto the manufacturer’s recommendations.15 The estimatedblood volume for each patient was calculated using themanufacturer’s instructions,15 according to the methoddescribed by Allen et al.16 After induction of anesthesia,baseline kaolin ACT, predicted HDR, predicted heparinconcentration, and heparin bolus calculations were per-formed according to the manufacturer’s instructions, usingheparin-protamine titration cartridges encompassing wholeblood heparin concentration ranging from 0.7 to 3.4 U/mLand kaolin as the activator. The recommended HepconHMS Plus CPB prime heparin dose based on a 750- or1000-mL prime volume was added to the calculated hepa-rin bolus and administered via a central venous catheter.Heparin was not added to the CPB pump prime. Through-out the entire study period, an ACT of �350 seconds wasused in patients undergoing surgery where cardiotomysuction was to be used. Patients undergoing primary CABGsurgery without the use of cardiotomy suction were anti-coagulated using a protocol that prescribed a minimal ACTof 300 seconds before the institution of CPB. Three minutesafter USP porcine heparin (APP Pharmaceuticals, Schaum-burg, IL) administration, heparin concentration and ACTwere remeasured. This heparin management protocolincluding use of heparin-coated circuits was adopted ina comprehensive institutional program to reduce the rateof reoperation for bleeding. All patients received an�-aminocaproic acid initial loading dose of 7.5 to 10 g over1 hour, after the initial blood draw for AT level and baselineACT, but before heparin administration and blood sam-pling for measurement of postheparin ACT.

HDR was measured as the difference in ACT betweentarget and baseline ACT measurements, divided by targetheparin level estimated from the Hepcon HMS Plus system.Because the Hepcon HMS Plus system has a limited fidelityin reporting whole blood heparin concentrations, in that itprovides values with discrete categories (i.e., 0.7, 1.4, 2.0,2.7, and 3.4 U/mL) rather than a continuous variable, theheparin sensitivity index (HSI) was also calculated fromchange in ACT between before and after heparin adminis-tration, divided by heparin dose, per kilogram of bodyweight.11 The same calculation was performed using hep-arin dose per liter of estimated blood volume withoutsignificantly changing the results; therefore, body weightwas used.

We estimated the accessible effect size for HDR, usingthe available sample size (n � 319), a 20% rate of heparinresistance based on prior studies, a mean HDR of 99s/U/mL and an sd of 22 s/U/mL, a type I error rate of 5%and a type II error rate of 20%. We estimated that we wouldbe able to observe differences in HDR of 9 s/U/mL, whichwe thought would be more sensitive than a clinicallyimportant difference.

Data are presented as mean � sd when normallydistributed, or median and 5th and 95th percentiles whennot normally distributed. Data that were nonnormallydistributed were compared using the Wilcoxon ranked sumtest. The Student t test was used to compare means ofnormally distributed data. Subgroup analysis of patientswith low AT activity (�80% of laboratory normal; AT


activity �0.813 U/mL) who might be at risk for heparinresistance was performed. Multivariable linear regres-sion modeling was performed to examine for clinical andlaboratory predictors of HDR and HSI; variables withunivariate P values �0.2 were entered into a combinedforward/backward stepwise linear regression model, withan exit P value of 0.05. A 2-sided P � 0.05 was consideredas showing statistical significance. Statistical analyses wereperformed using SAS version 9.1.3 and JMP 7.03 (SASInstitute, Cary, NC).

RESULTSBaseline demographics and perioperative data for 304patients are summarized in Table 1. Thirty-two patients(10.5%) required additional heparin administration beforethe institution of CPB for failing to achieve their respectivetarget ACT with administration of the heparin dose esti-mated by the HepCon HMS Plus system. No patientreceived �300 U/kg heparin. No patients were given freshfrozen plasma or AT supplementation to reach target ACTvalues.

Table 1. Demographics and Perioperative Data of Patients with Low Antithrombin (AT) Activity (<80% ofNormal; Defined as AT Activity <0.813 U/mL) and Normal AT Activity

Study cohort(N � 304)

AT activity <80% of normala

(N � 49)AT activity >80% of normala

(N � 255) PAge 65 � 9 70 � 9 64 � 9 �0.0001Male gender 247 (81.3%) 42 (85.7%) 205 (80.4%) 0.43Race (Caucasian) 272 (89.5%) 44 (89.8%) 228 (89.4%) 1.0Height (cm) 173 � 9 174 � 9 173 � 9 0.56Weight (kg) 88.3 � 18.6 87.0 � 16.6 88.4 � 19.0 0.62Medical history

Previous myocardial infarction 130 (42.7%) 27 (55.1%) 103 (40.4%) 0.06Myocardial infarction within prior 2 wk 56 (18.4%) 16 (32.7%) 40 (15.7%) 0.008Hypertension 224 (73.7%) 30 (61.2%) 194 (76.1%) 0.03Hypercholesterolemia 236 (77.6%) 33 (67.4%) 203 (79.6%) 0.06Insulin-dependent diabetes 31 (10.2%) 9 (18.4%) 22 (8.6%) 0.07Renal insufficiency (creatinine �2.0 mg/dL) 8 (2.6%) 2 (4.1%) 6 (2.4%) 0.62Liver failure 0 (0%) 0 (0%) 0 (0%) —

Preoperative medicationsAspirin 256 (84.2%) 38 (77.6%) 218 (85.4%) 0.20Platelet inhibitor (excluding aspirin) 71 (23.3%) 14 (28.6%) 57 (22.4%) 0.36Intravenous heparin in current admission 88 (28.9%) 33 (67.4%) 55 (21.6%) �0.0001Any statin 252 (82.8%) 41 (83.7%) 211 (82.8%) 1.0

Preoperative laboratory valuesHematocrit (%) 39.9 � 4.5 39.5 � 5.3 40.0 � 4.3 0.52Platelet count (109/mL) 237 (169–322) 245 (170–339) 236 (167–317) 0.55White cell count (103/mL) 8.0 (5.5–11.2) 8.6 (5.4–11.8) 7.8 (5.5–11.1) 0.27INR 1.0 (1.0–1.1) 1.0 (1.0–1.2) 1.0 (1.0–1.1) 0.02INR �1.4 0 (0%) 0 (0%) 0 (0%) —Partial thromboplastin time (s) 30.2 (26.5–63.2) 39 (28–94) 30 (26–40) �0.0001Antithrombin content (U/mL) 0.862 � 0.123 0.703 (0.582–0.790) 0.881 (0.772–1.04)Antithrombin content (mg/mL) 0.23 (0.196–0.282) 0.193 (0.160–0.217) 0.242 (0.212–0.286)Antithrombin activity (U/mL) 0.933 � 0.126 0.761 (0.649–0.811) 0.961 (0.855–1.11)Antithrombin activity (%) 91.1 (76.3–107.2) 74.8 (63.6–79.5) 93.8 (83.7–108.7)Protein C (�g/mL) 5.01 (2.27–6.98) 4.75 (3.32–6.63) 5.07 (3.53–6.96) 0.10Plasminogen activator inhibitor-1 (ng/mL) 1.53 (0.0–20.2) 1.18 (0.23–3.88) 1.58 (0.06–6.92) 0.16D-dimer (ng/mL) 183.3 (0.0–2569) 142.8 (6.8–1084) 186.6 (7.2–1381) 0.52Tissue factor (pg/mL) 16.4 (0.03–495.6) 12.6 (0.04–73.3) 16.5 (0.04–118.8) 0.49

Heparin administrationBaseline ACT (s) 135 � 18 138 (114–163) 136 (114–156) 0.12Target ACT

300 s 257 (84.5%) 41 (16.0%) 216 (84.0%)350 s 47 (15.5%) 8 (17.0%) 42 (83.0%) 0.86

Heparin dose (U/kg) 150 (86–246) 154 (112–198) 150 (111–211) 0.75Heparin dose �200 U/kg 59 (19.4%) 8 (16.3%) 51 (20.0%) 0.69ACT after heparin administration

300 s 366 (300–456) 357 (300–469) 366 (300–456) 0.53350 s 389 (324–460) 395 (366–425) 389 (319–470) 0.81

Calculated HDR slope (s/U/mL) 98 � 21 96 � 20 98 � 21 0.50Heparin sensitivity index (s/U/kg) 1.59 � 0.41 1.58 � 0.40 1.60 � 0.41 0.68

Clinical outcomesCardiac troponin I on POD 1 (�g/L) 0.99 (0.13–7.72) 1.17 (0.31–3.21) 0.97 (0.32–3.89) 0.48ICU duration (h) 45 (21–94) 48 (24–116) 44 (21–90) 0.13Postsurgical hospital stay (d) 6 (4–10) 6 (4–11) 6 (4–10) 0.17

Data presented as mean � SD or median (10th–90th percent confidence intervals).INR � International Normalized Ratio; ACT � activated clotting time; POD 1 � postoperative day 1; HDR � heparin dose response; ICU � intensive care unit.a Defined as 100% of the laboratory normal for AT activity, not corrected for the National Institute for Biological Standards and Control Second InternationalReference Standard (see Methods).

Antithrombin and Heparin Response


AT activity and content were highly correlated (r2 �0.801), therefore AT activity is reported. Higher AT levels(P � 0.05) were observed in females, younger individuals,current smokers, hypercholesterolemia, and individualswho had not had a myocardial infarction within the previ-ous 2 weeks. No association was observed between base-line AT activity and baseline ACT (r2 � 0.001; P � 0.10) orwith HDR (r2 � 0.001; P � 0.59) (Fig. 1), HSI (r2 � 0.001; P �0.73), or platelet count (r2 � 0.004; P � 0.24). Those patientswith recent preoperative heparin exposure had signifi-cantly lower AT activity (0.87 vs 0.96 U/mL; P � 0.001) butdid not show a significant difference in heparin require-ments or heparin responsiveness by any measure includingHDR or HSI.

Individuals with AT activity �80% normal (�0.813 U/mL)had higher partial thromboplastin time and were more likelyto have received heparin during the current admission beforesurgery, but otherwise showed no differences in coagulation

testing or management (Table 1). No relationship between ATactivity and HDR was observed in these patients (r2 � 0.001).Furthermore, there was no evidence of diminished heparinresponsiveness in this group, because 94.1% (48 of 51)achieved the target ACT after administration of the calculatedheparin bolus dose. Neither HDR nor HSI was significantlyrelated to AT activity in univariate relationship (Fig. 1), orafter accounting for other covariates using multivariable lin-ear regression (Tables 2 and 3).

Preoperative AT activity, HDR, and HSI were notassociated with cTnI levels on the first postoperative day(all r2 � 0.002; P � 0.5), ICU duration, or hospital lengthof stay.

DISCUSSIONWe observed no relationship between preoperative AT

activity and response to heparin, measured using eitherHDR or HSI, in these 304 patients undergoing primary

Figure 1. Preoperative antithrombin (AT)activity versus heparin dose response(HDR) (r2 � 0.001; P � 0.59). Confi-dence interval of HDR for a single mea-surement of AT activity is shaded. HDRslope (s/U/mL) � 94.3 � 5.31 � ATactivity (U/mL).

Table 2. Multivariable Predictors of Heparin-Dose Response (HDR)

Heparin dose response (s/U/mL)

Without AT activity With AT activity

r2 � 0.138 r2 � 0.140

Estimate P Estimate PFemale gender �1.82 � 1.66 0.27 �2.02 � 1.67 0.23Age (y) 0.16 � 0.13 0.23 0.176 � 0.131 0.18Caucasian race �3.81 � 1.87 0.042 �3.69 � 1.87 0.049Hypercholesterolemia 5.61 � 1.39 �0.0001 5.52 � 1.40 �0.0001Hematocrit 0.71 � 0.29 0.016 0.70 � 0.29 0.016Prothrombin time (s) 5.57 � 1.82 0.002 5.83 � 1.84 0.002White cell count (�106/mL) �1.52 � 0.51 0.003 �1.50 � 0.51 0.004AT activity (U/mL) 8.32 � 9.64 0.39

AT � antithrombin.


CABG surgery. Additional subgroup analysis in patientswith baseline AT activity �80% failed to identify ATactivity as a predictor of HDR or HSI. However, no patientexhibited a requirement for a heparin dose �300 U/kg toachieve target ACTs of either 300 or 350 seconds.

Acquired AT deficiency is common in patients withprevious heparin administration,17 critical illness, severehepatic dysfunction, and after major cardiovascular sur-gery.18,19 After cardiac surgery, lower levels of AT havebeen independently associated with prolonged ICU stayand a higher incidence of neurologic and thromboembolicevents.12 In this cohort of patients, we observed associa-tions between clinical markers of anticoagulation (pro-thrombin time and ACT) and factors that may relate to theseverity of illness including white blood cell count andhematocrit, and HDR. However, neither AT activity norheparin sensitivity was associated with severity of myocar-dial injury, ICU length of stay, or hospital length of stay, assurrogates of severity of illness.

Administration of unfractionated heparin remains themainstay of anticoagulation management for patients un-dergoing cardiac surgery requiring CPB, with the goal ofmaintaining therapeutic anticoagulation, thereby prevent-ing thromboembolic complications. Its unique properties ofrapidly providing systemic anticoagulation that can bemaintained for the duration of CPB and rapid completereversal with subsequent administration of protaminemake heparin the anticoagulant of choice for CPB. Interin-dividual variability in heparin response is well described,20

and contributing to this variation may be previously ob-served risk factors for diminished heparin responsivenessthat include low AT levels, platelet count �300,000cells/mm3, and heparin pretreatment.9 Although we ob-served substantial variability in heparin responsiveness,our data failed to show any relationship between heparinresponse and AT activity or heparin pretreatment.

AT is critical for maintenance of anticoagulation duringCPB and is consumed during the process.21 Despotis etal.14 demonstrated a strong association between in vitro ATand heparin responsiveness measured by ACT slope overthe range of AT levels of 0.2 to 1 U/mL. At AT levels above1 U/mL, there was no further increase in heparin respon-siveness. However, there was much weaker association in31 patients undergoing cardiac surgery with CPB.14 Simi-larly, Dietrich et al.11 observed a diminished HSI in adult

patients with preoperative AT activity �80% of normal,whereas no relationship was found in patients with normalAT levels. Our data contrast with this, in that we observedno relationship between AT activity and heparin respon-siveness in either group.

The ACT response to heparin is complex and affected bydifferent factors as previously noted. Although AT en-hancement is a primary mechanism of heparin’s action,other factors may influence HDR including tissue factorpathway inhibitor levels in vivo but not in vitro, extravas-cular sequestration of heparin, plasma protein binding,leukocyte lactoferrin, and activated platelets.7,14,22 Tissuefactor pathway inhibitor is an endothelium-derived endog-enous serine protease inhibitor with enhanced expressionafter heparin administration23 that may enhance anticoagu-lation in vivo.24 Other endogenous mechanisms mayfurther modify the activity of heparin in vivo. Highermolecular weight heparin (�13 kDa) is sequestered anddeactivated by endothelial endocytosis and depolymeriza-tion.25 In addition, neutrophil-derived lactoferrin can neu-tralize heparin by ionic binding.22 Heparin has shown to besimilarly neutralized by histidine-rich glycoproteins suchas vitronectin, fibronectin, and kininogen.7 Platelet factor 4is a potent heparin binding agent released from activatedplatelets that reverses the ACT,26,27 potentially contributingto heparin resistance. These endogenous mechanisms maybe important in modifying heparin’s anticoagulant activityin vivo, but difficult to quantify in vitro.

This study has important limitations. Our study did notidentify patients with severe heparin resistance, with only10.5% of patients failing to reach the target ACT, and nonerequired fresh frozen plasma or AT to initiate CPB. Perhapsthe relatively low ACT target contributed to the lack ofidentification of heparin-resistant patients. Historically,CPB has been initiated with higher ACTs to give a marginof safety28 and in many studies, a large number of patientsdiagnosed as heparin resistant had target ACTs �400seconds.5–8,13,14 Definitions of heparin resistance based ona dose of heparin to achieve a specific target ACT such asrequiring �500 IU/kg to achieve a target ACT of 480seconds assume linearity of the HDR.10 Although Despotiset al.29 observed a strong linear relationship between ACTand heparin concentration observed over a range of hepa-rin concentrations, to assume that our results could be

Table 3. Multivariable Predictors of Heparin Sensitivity Index (HSI)

Heparin sensitivity index (s/U/kg)

Without AT activity With AT activity

r2 � 0.136 r2 � 0.137

Estimate P Estimate PFemale gender 0.070 � 0.032 0.029 0.068 � 0.032 0.037Age (y) 0.001 � 0.003 0.79 0.001 � 0.003 0.72Caucasian race 0.019 � 0.037 0.61 0.021 � 0.037 0.58Platelet count (�109/mL) �0.0007 � 0.0003 0.013 �0.0007 � 0.0003 0.013PT (s) 0.10 � 0.04 0.006 0.10 � 0.04 0.005White cell count (�106/mL) �0.022 � 0.011 0.048 �0.022 � 0.011 0.048PAI-1 level (ng/mL) 0.032 � 0.009 0.0005 0.031 � 0.009 0.0007Tissue factor level (pg/mL) �0.0008 � 0.0003 0.015 �0.0008 � 0.0003 0.016AT activity (U/mL) 0.121 � 0.198 0.54

PAI-1 � plasminogen activator inhibitor level-1; AT � antithrombin; PT � prothrombin time.

Antithrombin and Heparin Response


extrapolated to higher target ACTs would not be appropri-ate. Furthermore, we did not measure circulating concen-trations of several important members of the coagulationpathway that may have affected these results.

In conclusion, we found that heparin responses wereindependent of preoperative plasma AT activity in patientsundergoing primary CABG using target ACTs of 300 to 350seconds. Preoperative heparin exposure was associatedwith diminished AT activity, but no change in HDR wasobserved. Perioperative AT activity, HDR, or HSI were notassociated with cTnI levels on the first postoperative day,ICU duration, or hospital length of stay.

DISCLOSUREDr. Garvin was the recipient of a research fellowship fundedby Talecris Biotherapeutics, which allowed time for generationof this and other articles. Dr. Body has received consulting feesfrom Talecris Biotherapeutics for another study.

Talecris Biotherapeutics paid for the costs of antithrombinanalyses performed by Charles River Laboratories in Ontario,Canada, but had no input into these analyses.

REFERENCES1. McLean J. The thromboplastin action of cephalin. Am J Physiol

1916;41:250–72. Rosenberg RD, Damus PS. The purification and mechanism of

action of human antithrombin-heparin cofactor. J Biol Chem1973;248:6490–505

3. Rosenberg RD. Actions and interactions of antithrombin andheparin. N Engl J Med 1975;292:146–51

4. Langdown J, Johnson DJ, Baglin TP, Huntington JA. Allostericactivation of antithrombin critically depends upon hinge re-gion extension. J Biol Chem 2004;279:47288–97

5. Avidan MS, Levy JH, Scholz J, Delphin E, Rosseel PM, HowieMB, Gratz I, Bush CR, Skubas N, Aldea GS, Licina M, BonfiglioLJ, Kajdasz DK, Ott E, Despotis GJ. A phase III, double-blind,placebo-controlled, multicenter study on the efficacy of recom-binant human antithrombin in heparin-resistant patientsscheduled to undergo cardiac surgery necessitating cardiopul-monary bypass. Anesthesiology 2005;102:276–84




9. Ranucci M, Isgro G, Cazzaniga A, Soro G, Menicanti L, FrigiolaA. Predictors for heparin resistance in patients undergoingcoronary artery bypass grafting. Perfusion 1999;14:437–42

10. Staples M, Dunton D, Karlson K, Leonardi H, Berger R.Heparin resistance after preoperative heparin therapy or in-traaortic ballon pumping. Ann Thorac Surg 1994;57:1211–6

11. Dietrich W, Braun S, Spannagl M, Richter JA. Low preopera-tive antithrombin activity causes reduced response to heparinin adult but not in infant cardiac-surgical patients. AnesthAnalg 2001;92:66–71


13. Lemmer JH Jr, Despotis GJ. Antithrombin III concentrate totreat heparin resistance in patients undergoing cardiac sur-gery. J Thorac Cardiovasc Surg 2002;123:213–7

14. Despotis GJ, Levine V, Joist JH, Joiner-Maier D, Spitznagel E.Antithrombin III during cardiac surgery: effect on response ofactivated clotting time to heparin and relationship to markersof hemostatic activation. Anesth Analg 1997;85:498–506

15. Medtronic Perfusion Systems. Hepcon HMS Plus Operator’sManual. Minneapolis, MN: Medtronic Perfusion Systems, 2001

16. Allen TH, Peng MT, Chen KP, Huang TF, Chang C, Fang HS.Prediction of blood volume and adiposity in man from bodyweight and cube of height. Metabolism 1956;5:328–45

17. Dietrich W, Spannagl M, Schramm W, Vogt W, Barankay A,Richter JA. The influence of preoperative anticoagulation onheparin response during cardiopulmonary bypass. J ThoracCardiovasc Surg 1991;102:505–14

18. Maclean PS, Tait RC. Hereditary and acquired antithrombindeficiency: epidemiology, pathogenesis and treatment options.Drugs 2007;67:1429–40

19. Levy JH, Despotis GJ, Szlam F, Olson P, Meeker D, Weisinger A.Recombinant human transgenic antithrombin in cardiac surgery:a dose-finding study. Anesthesiology 2002;96:1095–102

20. Bull BS, Korpman RA, Huse WM, Briggs BD. Heparin therapyduring extracorporeal circulation. I. Problems inherent in existingheparin protocols. J Thorac Cardiovasc Surg 1975;69:674–84

21. Hashimoto K, Yamagishi M, Sasaki T, Nakano M, Kurosawa H.Heparin and antithrombin III levels during cardiopulmonarybypass: correlation with subclinical plasma coagulation. AnnThorac Surg 1994;58:799–804; discussion 804–5

22. Wu HF, Lundblad RL, Church FC. Neutralization of heparinactivity by neutrophil lactoferrin. Blood 1995;85:421–8

23. Bajaj MS, Kuppuswamy MN, Saito H, Spitzer SG, Bajaj SP.Cultured normal human hepatocytes do not synthesizelipoprotein-associated coagulation inhibitor: evidence that en-dothelium is the principal site of its synthesis. Proc Natl AcadSci USA 1990;87:8869–73

24. Brodin E, Appelbom H, Osterud B, Hilden I, Petersen LC,Hansen JB. Regulation of thrombin generation by TFPI inplasma without and with heparin. Transl Res 2009;153:124–31

25. Barzu T, van Rijn JL, Petitou M, Tobelem G, Caen JP. Heparindegradation in the endothelial cells. Thromb Res 1987;47:601–9

26. Lane DA, Pejler G, Flynn AM, Thompson EA, Lindahl U.Neutralization of heparin-related saccharides by histidine-richglycoprotein and platelet factor 4. J Biol Chem 1986;261:3980–6

27. Levy JH, Cormack JG, Morales A. Heparin neutralization byrecombinant platelet factor 4 and protamine. Anesth Analg1995;81:35–7

28. Young JA, Kisker CT, Doty DB. Adequate anticoagulationduring cardiopulmonary bypass determined by activated clot-ting time and the appearance of fibrin monomer. Ann ThoracSurg 1978;26:231–40

29. Despotis GJ, Summerfield AL, Joist JH, Goodnough LT, San-toro SA, Spitznagel E, Cox JL, Lappas DG. Comparison ofactivated coagulation time and whole blood heparin measure-ments with laboratory plasma anti-Xa heparin concentration inpatients having cardiac operations. J Thorac Cardiovasc Surg1994;108:1076–82


Postoperative Activity, but Not Preoperative Activity,of Antithrombin Is Associated with Major AdverseCardiac Events After Coronary Artery BypassGraft SurgerySean Garvin, MD,* Jochen D. Muehlschlegel, MD,* Tjorvi E. Perry, MD,* Junliang Chen, PhD,†Kuang-Yu Liu, PhD,* Amanda A. Fox, MD,* Charles D. Collard, MD,‡ Sary F. Aranki, MD,§Stanton K. Shernan, MD,* and Simon C. Body, MB, ChB, MPH*

BACKGROUND: Low levels of antithrombin (AT) have been independently associated with prolongedintensive care unit stay and an increased incidence of neurologic and thromboembolic events aftercardiac surgery. We hypothesized that perioperative AT activity is independently associated withpostoperative major adverse cardiac events (MACEs) in patients undergoing coronary artery bypassgraft (CABG) surgery.METHODS: We prospectively studied 1403 patients undergoing primary CABG surgery with cardio-pulmonary bypass (CPB) (http://clinicaltrials.gov/show/NCT00281164). The primary clinical endpoint was occurrence of MACE, defined as a composite outcome of any one or more of the following:postoperative death, reoperation for coronary graft occlusion, myocardial infarction, stroke, pulmo-nary embolism, or cardiac arrest until first hospital discharge. Plasma AT activity was measuredbefore surgery, after post-CPB protamine, and on postoperative days (PODs) 1–5. Multivariate logisticregression modeling was performed to estimate the independent effect of perioperative AT activityupon MACE.RESULTS: MACE occurred in 146 patients (10.4%), consisting of postoperative mortality (n � 12),myocardial infarction (n � 108), stroke (n � 17), pulmonary embolism (n � 8), cardiac arrest (n �16), or a subsequent postoperative or catheter-based treatment for graft occlusion (n � 6). AT activityat baseline did not differ between patients with (0.91 � 0.13 IU/mL; n � 146) and without (0.92 �0.13 IU/mL; n � 1257) (P � 0.18) MACE. AT activity in both groups was markedly reducedimmediately after CPB and recovered to baseline values over the ensuing 5 PODs. Postoperative ATactivity was significantly lower in patients with MACE than those without MACE. After adjustment forclinical predictors of MACE, AT activity on PODs 2 and 3 was associated with MACE.CONCLUSIONS: Preoperative AT activity is not associated with MACE after CABG surgery. MACEis independently associated with postoperative AT activity but only at time points occurringpredominantly after the MACE. (Anesth Analg 2010;111:862–9)

Antithrombin (AT) is a serine protease inhibitor(serpin) and the principal inhibitor of the finalcommon pathway of the coagulation system by

inactivation of circulating thrombin (factor IIa) and factorXa, among other serine proteases. Heparin increases ATactivity 2000- to 4000-fold due to a conformational changein the quaternary structure of AT by heparin binding andthrough formation of a ternary complex of thrombin, AT,and heparin. Heparin-augmented AT activity is still theprincipal mechanism of anticoagulation for cardiopulmo-nary bypass (CPB).

AT levels are decreased after administration of heparindue to degradation of the ternary complex. Additionally,acquired AT deficiency is common in patients with criticalillness, severe hepatic dysfunction, and after major cardio-vascular surgery.1,2 The magnitude of reduction in AT aftercardiac surgery is similar to that in patients with heterozy-gous AT deficiency, which is associated with increased riskof thromboembolic events.3,4 After cardiac surgery, lowerlevels of AT have been independently associated withprolonged intensive care unit (ICU) stay and a higherincidence of neurologic and thromboembolic events.5 Wetherefore examined for independent association betweenperioperative AT activity and the frequency of postopera-tive major adverse cardiac events (MACEs) in patientsundergoing coronary artery bypass graft (CABG) surgery.

METHODSThe study cohort was obtained from a continuing prospectivelongitudinal parent study of 1447 patients undergoing primaryCABG surgery with CPB between August 2001 and May 2006 at2 United States academic medical centers (CABG GenomicsProgram; http://clinicaltrials.gov/show/NCT00281164). WithIRB approval, written informed consent was obtained from eachpatient. Patients were excluded from the parent study if theywere younger than 20 yr old, underwent repeat or off-pump

From the *Department of Anesthesiology, Perioperative and Pain Medicine,Brigham and Women’s Hospital, Harvard Medical School, Boston, Massa-chusetts; †Talecris Biotherapeutics, Research Triangle Park, Durham, NorthCarolina; ‡Baylor College of Medicine Division of Cardiovascular Anesthe-sia at the Texas Heart Institute, Saint Luke’s Episcopal Hospital, Houston,Texas; and §Division of Cardiac Surgery, Brigham and Women’s Hospital,Harvard Medical School, Boston, Massachusetts.


The authors had full access to the data and take responsibility for itsintegrity. All authors have read and agree to the manuscript as written.

Address correspondence and reprint requests to Simon C. Body, MB, ChB,MPH, Department of Anesthesiology, Perioperative and Pain Medicine,Brigham and Women’s Hospital, 75 Francis St., Boston, MA 02115. Addresse-mail to [email protected].

Copyright © 2010 International Anesthesia Research SocietyDOI: 10.1213/ANE.0b013e3181b7908c


CABG, had a preoperative hematocrit �25%, or if they hadreceived leukocyte-rich blood products within 30 days beforesurgery. All patients enrolled in the parent study were included.Patients without postoperative cardiac troponin I (cTnI) levels(n � 43) were excluded from further analysis. Demographicdata, medical and surgical history, medications, and out-comes were recorded by trained research staff using definedprotocols in a purpose-built case report form. The examina-tion of the relationship between AT levels and MACE was notprespecified in the original parent study.

Perioperative anticoagulation protocols differed be-tween institutions. At Brigham and Women’s Hospital,patients received 300 U per kg body weight of porcineheparin to achieve an activated clotting time (ACT) of �400s, until February 2004. From February 2004, patients re-ceived a Hepcon HMS Plus (Medtronic, Minneapolis, MN)calculated dose of porcine heparin to achieve an ACT ofeither 300 or 350 s. At Texas Heart Institute, 300 U per kgbody weight of either bovine or porcine heparin was givento achieve an ACT of �400 s.

Clinical End Points for the Test CohortThe primary clinical end point was prespecified as theoccurrence of a MACE, defined as a composite outcome ofany one or more of the following: postoperative mortality(defined as all deaths occurring within 30 days of theoperation or occurring during the primary hospitalization),reoperation for coronary graft occlusion, myocardial infarc-tion (MI) (predefined as peak postoperative cTnI concen-tration �12 ng/mL, being the upper 8th percentile), cardiacarrest (defined as a postoperative event requiring cardio-pulmonary resuscitation) until first hospital discharge,thromboembolic event consisting of stroke (defined as aclinical diagnosis of focal or global neurological deficit), orpulmonary embolism (diagnosed by ventilation perfusionscan of moderate to high probability or by a positivepulmonary angiogram).

Cardiac Biomarker AssayBlood samples were obtained before surgery, 5 min afteradministration of post-CPB protamine, and on the morningsof postoperative days (PODs) 1–5. Citrated plasma was storedin vapor-phase liquid nitrogen until analysis for cTnI with asandwich immunoassay on a Triage® platform using mono-clonal and polyclonal antibodies (Biosite, San Diego, CA) ata single core facility. Patient caregivers were not aware ofthe results of the assays as they were performed afterpatient discharge.

Antithrombin AssaysAT activity was measured with a colorimetric method using aModular Analytics biochemistry analyzer (Siemens Health-care Diagnostics, Tarrytown, NY). The assay limit of quanti-tation was 21.6%. To report human plasma AT activity resultsin IU/mL, the National Institute for Biological Standards andControl Second International Reference Standard was used todetermine a conversion factor of activity in IU/mL � activityin % � 0.0102. AT content was measured using an immuno-ephelometric method using a BN-100 ProSpec nephelometer(Dade Behring Diagnostics, Marburg, Germany). The assaylimit of quantitation was 0.00672 mg/mL. To report human

plasma AT content results in IU/mL, the National Institutefor Biological Standards and Control Second InternationalReference Standard was used to determine a conversionfactor of content in IU/mL � content in g/L � 3.64. Bothassays measure free AT rather than AT complexed withheparin. Assays were performed by Charles River Labora-tory in Montreal, Canada by personnel blinded to outcomestatus. Subsequent comparison of paired AT activity andcontent data revealed high correlation (r2 � 0.878), so onlythe activity is reported.

Statistical MethodsStatistical analyses were performed using SAS, version9.1.3, and JMP 7.0 (SAS Institute, Cary, NC). AT activitywas normally distributed at all time points, so was nottransformed. Data are presented as mean (sd) and medianwith 10%–90% interquantile range, unless otherwise stated.Continuous variables were compared using analysis of vari-ance or Wilcoxon Mann-Whitney ranked sum test whenappropriate. Categorical variables were compared with �2 orFisher’s exact test.

Multivariate logistic regression modeling was per-formed to identify and account for MACE risk factors thatmight confound any association between low AT activityand MACE. The multivariate analysis used a forwardstepwise technique to identify independent risk factors forMACE, whereby clinically relevant demographic variablesand variables with a two-tailed univariate P � 0.2 wereentered into the model and P � 0.2 was necessary to remainin the model. Age, gender, race, body mass index, and institu-tion were forced into the model. Nagelkerke generalized r2 andlikelihood ratio test were used to determine the additionalpredictive value of AT upon MACE. F tests were used tocompare generalized r2. Odds ratios and 95% confidenceintervals for a 0.1 IU/mL decrease in AT activity wereestimated. A two-sided P � 0.05 was considered significant.

RESULTSThe cohort comprised 1403 patients undergoing CABGsurgery whose characteristics are described in Table 1.MACE occurred in 146 patients (10.4%), consisting ofpostoperative mortality (n � 12), MI (n � 108), stroke (n �17), pulmonary embolism (n � 8), cardiac arrest (n � 16), ora subsequent postoperative or catheter-based treatment forgraft occlusion (n � 6). Nineteen patients had 2 or 3 events,usually MI, with either subsequent death or stroke. Mostadverse events occurred before or on POD 2. Of 12 patientswith operative mortality, 2 patients died at POD 0 and 10patients died on or after POD 5. Of 17 patients with stroke,6 patients had a stroke on or before POD 2. Of 108 patientswith MI, 89 patients had a diagnosis of MI first occurringon POD 1. MACE frequency did not differ between insti-tutions (Table 1).

AT activity and content were measured at the 7 timepoints (Table 2). AT activity at baseline did not differbetween patients with MACE (0.91 � 0.13 IU/mL; n � 146)and those without MACE (0.92 � 0.13 IU/mL; n � 1257)(P � 0.18). AT activity was significantly reduced at thepost-CPB measurement compared with the preoperativetime point (P � 0.0001) and returned to baseline levels overthe ensuing 5-day period in patients with and without MACE.


Table 1. Demographic and Clinical Characteristics of the Cohort as a Whole and Stratified by theOccurrence of Major Adverse Cardiac Events (MACE)

Entire cohort(n � 1403)

Patients with MACE(n � 146)

Patients without MACE(n � 1257) P

Age�55 yr 240 (17.1) 22 (15.1) 218 (17.3)55 to �65 yr 494 (35.2) 51 (34.9) 443 (35.2)65 to �75 yr 422 (30.1) 42 (28.8) 380 (30.2)75 to �85 yr 231 (16.5) 29 (19.9) 202 (16.1)At least 85 yr 16 (1.1) 2 (1.4) 14 (1.1) 0.75

Male 1136 (81.0) 109 (74.7) 1027 (81.7) 0.045Caucasian race 1189 (84.6) 116 (79.5) 1073 (85.4) 0.068Body mass index (kg/m2) 29.4 (5.5) 30.2 (5.5) 29.3 (5.4) 0.063Institution

BWH 1061 (75.6) 109 (10.3) 952 (89.7) 0.76THI 342 (24.4) 37 (10.8) 305 (89.2)

Medical historyDiabetes (drug treated; %) 466 (33.2) 52 (35.6) 414 (32.9) 0.52Pulmonary disease (%) 224 (16.0) 17 (11.6) 207 (16.5) 0.13Creatinine (mg/dL) 1.10 (0.334) 1.13 (0.322) 1.10 (0.335) 0.27Hematocrit (%) 40.11 (4.724) 39.48 (4.825) 40.2 (4.708) 0.099Hypertension (%) 1051 (74.9) 117 (80.1) 934 (74.3) 0.12Hypercholesterolemia (%) 1045 (74.5) 106 (72.6) 939 (74.7) 0.58Previous MI 620 (44.2) 83 (56.9) 537 (42.7) 0.002Time since last MI

�2 wk 256 (18.3) 44 (30.1) 212 (16.9)2–13 wk 47 (3.4) 6 (4.1) 41 (3.3)�13 wk 266 (19.0) 31 (21.2) 235 (18.7)Never 834 (59.4) 65 (44.5) 769 (61.2) 0.0002

Previous thrombolysis 71 (5.1) 13 (8.9) 58 (4.6) 0.025IABP placed preoperatively 38 (2.7) 11 (7.5) 27 (2.2) 0.001Arrhythmia requiring therapy 148 (10.6) 19 (13.0) 129 (10.3) 0.31Peripheral vascular disease 130 (9.3) 20 (13.7) 110 (8.8) 0.051Prior PVD procedure 39 (2.8) 10 (6.9) 29 (2.3) 0.0049Prior stroke 63 (4.5) 10 (6.9) 53 (4.2) 0.15LVEF preoperative �40% 182 (13.0) 29 (19.9) 153 (12.2) 0.013

Medications—preoperativeACE inhibitor 648 (46.2) 66 (45.2) 582 (46.3) 0.86Beta-blocker 1094 (78.0) 116 (79.5) 978 (77.8) 0.75Ca�� antagonist 195 (13.9) 24 (16.4) 171 (13.6) 0.35Aspirin 1072 (76.4) 113 (77.4) 959 (76.3) 0.84HMG CoA reductase inhibitor 1082 (77.1) 107 (73.3) 975 (77.6) 0.24Heparin (intravenous) 351 (25.0) 47 (32.2) 304 (24.2) 0.035Platelet inhibitor (not aspirin) 304 (21.7) 32 (21.9) 272 (21.6) 0.94

Preoperative laboratory dataHemoglobin (g/dL) 13.7 (1.7) 13.5 (1.7) 13.7 (1.6) 0.12Creatinine (mg/dL) 1.10 (0.33) 1.13 (0.32) 1.10 (0.34) 0.27Platelet count (109/mL) 240 (72) 234 (70) 241 (72) 0.32cTnI (ng/mL) 0.4 (2.5) 1.9 (7.4) 0.2 (0.7) 0.006

Intraoperative managementNumber of grafts

1 28 (2.0) 2 (1.4) 26 (2.1)2 188 (13.4) 27 (18.5) 161 (12.8)3 627 (44.8) 63 (43.2) 564 (44.9)�4 558 (39.8) 54 (37.0) 504 (40.2) 0.28

CPB duration (min) 99.2 (42.21) 119.63 (57.821) 96.78 (39.353) �0.0001Aortic cross-clamp duration (min) 71.5 (35.03) 81.92 (45.125) 70.27 (33.471) 0.003IABP placed intraoperatively 63 (4.5) 23 (15.8) 40 (3.2) �0.0001Heparin administration (mg)

BWH before February 2004 206 (83.7) 208 (76.0) 206 (85.0) 0.8319BWH after February 2004 194 (61.0) 187 (54.7) 195 (61.5) 0.4608THI 306 (92.0) 300 (80.6) 306 (93.4) 0.6998

Other surgical procedureConcurrent mitral valve 35 (2.5) 10 (6.9) 25 (2.0) 0.0020Concurrent aortic valve 22 (1.6) 3 (2.1) 19 (1.5) 0.49Concurrent other valve 2 (0.1) 2 (1.4) 0 0.01Other cardiac surgery 108 (7.7) 24 (16.4) 84 (6.7) �0.0001Other noncardiac surgery 17 (1.2) 1 (0.7) 16 (1.3) 1.0

(Continued)

Antithrombin and Adverse Events


Postoperative AT activity was significantly lower in patientswith MACE than those without MACE (Table 2).

Decreased preoperative AT activity was independentlypredicted by older age, male gender, and prior heparin usewithin the same hospitalization (Table 3). Other clinical

variables that may possibly be indicative of recent heparinuse at prior recent hospitalization, such as recent MI, werealso independently predictive. Decreased preoperative ATactivity was also independently associated with lowerplatelet count and increased partial thromboplastin time,

Table 1. ContinuedEntire cohort(n � 1403)

Patients with MACE(n � 146)

Patients without MACE(n � 1257) P

Intraoperative inotropesEpinephrine 345 (24.6) 59 (40.4) 286 (22.8) �0.0001Norepinephrine 169 (12.1) 20 (13.7) 149 (11.9) 0.52Phenylephrine 668 (47.6) 63 (43.2) 605 (48.1) 0.25Dopamine �5�g � kg�1 � min�1 38 (2.7) 3 (2.1) 35 (2.8) 0.79Dobutamine 18 (1.3) 1 (0.7) 17 (1.4) 1.0Milrinone 29 (2.1) 6 (4.1) 23 (1.8) 0.11Vasopressin 17 (1.2) 3 (2.1) 14 (1.1) 0.41

Postoperative inotropesEpinephrine 230 (16.4) 51 (34.9) 179 (14.2) �0.0001Norepinephrine 143 (10.2) 31 (21.2) 112 (8.9) �0.0001Phenylephrine 113 (8.1) 11 (7.5) 102 (8.1) 0.81Dopamine �5�g � kg�1 � min�1 30 (2.1) 8 (5.5) 22 (1.8) 0.009Dobutamine 12 (0.9) 3 (2.1) 9 (0.7) 0.12Milrinone 32 (2.3) 9 (6.2) 23 (1.8) 0.004Vasopressin 91 (6.5) 26 (17.8) 65 (5.2) �0.0001

Intraoperative and postoperativetransfusiona (median, 10–90percentile)

Red blood cell transfusion (units) 1 (0–5) 2 (0–8) 1 (0–5) �0.0001Coagulation factor transfusion (units) 0 (0–2) 0 (0–4) 0 (0–2) 0.0055

Postoperative eventsHLOS (d) �12 d 130 (9.3) 43 (29.5) 87 (6.9) �0.0001ICU LOS (d) �4 d 136 (9.7) 44 (30.1) 92 (7.3) �0.0001Peak postoperative cTnI �12 ng/mL 108 (8.0) 108 (76.6) 0 �0.0001Peak postoperative cTnI 4.06 (8.6) 21.9 (17.8) 2.0 (2.1) �0.0001

a All blood products administered during the hospital stay. Red blood cell transfusion includes both packed red blood cells and whole blood. Coagulation factortransfusion includes fresh frozen plasma, cryoprecipitate, and platelet transfusion.BWH � Brigham and Women’s Hospital; THI � Texas Heart Institute; MI � myocardial infarction; cTnI � cardiac troponin I; IABP � intraaortic balloon pump;LVEF � left ventricular ejection fraction; ACE � angiotensin converting enzyme; CPB � cardiopulmonary bypass; HLOS � hospital length of stay; ICU LOS � intensivecare unit length of stay; PVD � peripheral vascular disease.

Table 2. Postoperative Changes in Antithrombin Activity in Patients with Major Adverse Cardiac Events(MACE) (n � 146) and Without MACE (n � 1257): Results of Univariate Analysis at Each Time Point

Baseline Post-CPB

Postoperative day

1 2 3 4 5Antithrombin activity

(IU/mL)Patients with MACE 0.91 � 0.13 0.59* � 0.12 0.66* � 0.12 0.72* � 0.12 0.78* � 0.13 0.85* � 0.15 0.90† � 0.16Patients without

MACE0.92 � 0.13 0.62 � 0.11 0.71 � 0.13 0.77 � 0.12 0.83 � 0.12 0.90 � 0.13 0.95 � 0.14

Change inantithrombinactivity frombaseline (IU/mL)

Patients with MACE — �0.32 � 0.13 �0.24† � 0.13 �0.19† � 0.13 �0.13† � 0.14 �0.06* � 0.15 �0.00† � 0.16Patients without

MACE— �0.30 � 0.12 �0.22 � 0.13 �0.16 � 0.13 �0.09 � 0.13 �0.02 � 0.13 0.03 � 0.13

Percentage change inantithrombinactivity frombaseline (%)

Patients with MACE — �34.7† � 12.8 �26.2† � 12.3 �20.4* � 12.7 �13.5† � 15.0 �5.9† � 16.4 0.7† � 17.9Patients without

MACE— �32.0 � 11.2 �22.7 � 13.5 �16.1 � 13.3 �9.2 � 14.7 �1.1 � 14.5 4.6 � 15.5

Data are reported as mean � standard deviation.CPB � cardiopulmonary bypass.Significance is reported between MACE groups at each time point: *P � 0.001 and †P � 0.05 by Student’s t-test.


independent of recent heparin use, perhaps indicating adose effect of heparin administration upon decreased ATactivity. In the 1205 patients who had complete data for allvariables in the model, decreased post-CPB AT activity wasindependently predicted by lower preoperative AT activityand clinical variables that indicate greater hemodilution,such as lower body weight and height and increasedtransfusion incidence or a prolonged procedure (Table 4).

A clinical model predicting MACE was developed(Table 5; adjusted r2 � 0.156) for 1403 patients who hadcomplete data for all variables in the clinical model. Vari-ables that have been associated with MACE in priorstudies, notably recent MI, longer perfusion time, a require-ment for intraaortic counterpulsation, and red blood celltransfusion were also associated with MACE in this study.AT activities at baseline, post-CPB, PODs 1–5, and thechange from baseline at these time points for each patientwere added to the clinical model, one time point at a time.Preoperative, post-CPB, and POD 1 AT activity were notindependently predictive of MACE, whereas AT activity on

PODs 2 and 3 was independently predictive of MACE. MACEwas independently associated with change in AT activityfrom baseline on PODs 2–5 (Table 5). The receiver operat-ing characteristic of the model was significantly improvedby the addition of AT activity on PODs 2–4 to the model(Fig. 1).

DISCUSSIONThis prospective, observational, cohort study confirms earlierfindings that preoperative AT activity is reduced in patientswith recent exposure to heparin or recent MI.5,6 Furthermore,AT activity is reduced after CPB, likely due to consumptionby heparin administration6 and dilution. AT activity remainedsignificantly reduced from baseline levels over the postoperativeperiod, recovering to baseline levels within 5 days, on average.

In this population of patients, MACE was associatedwith factors that have been previously associated with MIor mortality including recent MI, peripheral vascular dis-ease, longer perfusion time, use of intraaortic counterpul-sation, and red blood cell transfusion.7–10 The clinical

Table 3. Multivariable Predictors of Preoperative Antithrombin Activity

Predictor

Predictors of preoperative AT activity (U/mL)(n � 1194a; r2 � 0.168)

Univariateestimate

Multivariateestimate

Standard error ofmultivariate estimate P

Age (1 yr increment) �0.0020 �0.0022 0.0004 �0.0001Gender (female) 0.0288 0.0177 0.0059 0.0028Race (Caucasian) �0.0078 �0.0026 0.0054 0.624Weight (1 kg increment) �0.0003 �0.0003 0.0002 0.233Height (1 cm increment) �0.0007 �0.0031 0.0005 0.568Institution 0.0183 0.0081 0.0053 0.129Previous MI (Y) �0.0273 �0.0104 0.0036 0.004Preoperative platelet count (10 �

109/mL increment)0.0023 0.0013 0.0005 0.006

Preoperative PTT (1 s increment) �0.0014 �0.0008 0.0002 �0.0001Preoperative heparin use �0.0863 �0.0352 0.0042 �0.0001Preoperative diuretic use 0.0275 0.0147 0.0042 0.0005a 209 subjects were missing one or more of the model’s predictor variables and are not included in this analysis.AT � antithrombin; MI � myocardial infarction; PTT � partial thromboplastin time.

Table 4. Multivariable Predictors of Post-CPB Antithrombin (AT) Activity

Predictor

Predictors of post-CPB AT activity (U/mL)(n � 1205a; r2 � 0.504)

Univariateestimate

Multivariateestimate

Standard error ofmultivariate estimate P

Age (1 yr increment) �0.0030 �0.0006 0.0002 0.0110Gender (female) 0.0356 0.0023 0.0039 0.5581Race (Caucasian) �0.0068 �0.0022 0.0033 0.4979Height (1 cm increment) 0.0023 0.0010 0.0003 0.0045Weight (1 kg increment) 0.0015 0.0006 0.0001 �0.0001Institution �0.0492 �0.0236 0.0038 �0.0001Preoperative AT activity (U/mL) 0.4870 0.4698 0.0174 �0.0001CPB duration (10 min increment) �0.006 �0.004 0.001 �0.0001Lowest venous temperature during CPB

(1°C increment)0.0030 0.0044 0.0011 �0.0001

Intraoperative packed red blood celltransfusion (1 unit increment)

�0.0095 �0.0085 0.0021 �0.0001

Intraoperative cryoprecipitate transfusion(1 pooled unit increment)

0.0497 0.0491 0.0150 0.0011

Post-CPB hemoglobin (1 g/dL increment) 0.0177 0.0184 0.0018 �0.0001a 198 subjects were missing one or more of the model’s predictor variables and are not included in this analysis.CPB � cardiopulmonary bypass.



model of MACE generated from this cohort had modestpredictive value, similar to prior clinical models.9 Theaddition of AT activity on POD 2 and beyond to the modelmodestly improved model performance. However, theclinical value of AT activity as a predictor of MACE islimited by the majority of adverse events that comprisedMACE occurring before POD 2. Specifically, the majority ofMACE events were MI, which was typically manifested asthe peak cTnI level occurring on POD 1. Rather, it may bethat lower postoperative AT activity may be a consequenceof more extensive surgery and other factors that are asso-ciated with increased incidence of MACE, although suchassertion cannot be proven in this observational cohort. Wecannot exclude the possibility that lower AT levels mayhave enhanced postoperative MACE.

Increased risk of adverse cardiovascular events has beenassociated with lower levels of AT in other critically illpopulations, such as patients with severe sepsis.11–13 Theseobservations have prompted clinical trials of supplementalAT administration.14–17 Although several trials have

shown survival benefit, a meta-analysis of 20 trials encom-passing 3458 patients failed to show a survival benefit ofadministration of AT to critically ill patients.18

There are limited data regarding AT activity and ad-verse outcomes after cardiac surgery. The majority ofstudies describe use of AT for “heparin resistance” duringCPB and lack postoperative clinical outcome data.19–25 Asingle well-conducted observational study of 647 patientsevaluated the association between preoperative and imme-diate postoperative AT levels and outcomes in cardiacsurgery patients. Low levels of AT activity upon ICUarrival were associated with prolonged ICU stay, higherrate of reexploration for bleeding, thromboembolism, andadverse neurologic sequelae.5 Our study examined a longertime period that encompassed the period of MACE occur-rence and the recovery of AT levels, thus providing addi-tional insights. Importantly, we replicated the finding thatAT level after CPB is associated with clinical factors indica-tive of longer, more extensive procedures, perhaps withmore hemodilution.

Table 5. Multivariable Predictors of Major Adverse Cardiac Events (MACE)

Predictor

Without AT activity information in model(n � 1403; r2 � 0.156)

Odds ratio95% confidence

interval PAge

�55 yr 1.00 —55–64 yr 1.26 0.72–2.2165–74 yr 1.26 0.70–2.2875–84 tears 1.66 0.86–3.20�85 yr 0.70 0.11–4.36 0.58

Gender (female) 1.13 0.72–1.78 0.60Race (Caucasian) 0.59 0.36–0.95 0.03Body mass index

�20 0.67 0.15–3.0420–24.9 0.52 0.28–0.9625–34.9 0.48 0.30–0.77�35 1.00 — 0.03

Institution 0.73 0.44–1.21 0.22Myocardial infarction �2 wk prior 1.70 1.11–2.59 0.014Prior peripheral vascular procedure 3.38 1.53–7.46 0.003Perfusion time (10 min increment) 1.09 1.05–1.14 �0.0001Concurrent other cardiac procedure 1.47 0.82–2.67 0.20Preoperative or intraoperative IABP 4.02 2.18–7.42 �0.0001RBC transfusion during hospital stay (per unit) 1.10 1.04–1.16 0.0004

Additional Predictive Value of AT Activity at Each Time Point Added to the Clinical Model

AT activity (0.1 IU/mLdecrease)

OddsRatio

95% Confidenceinterval

P value of ATvariable

P value of improvementin overall model

Preoperative AT activity 0.98 0.85–1.13 0.7820 0.7822Postoperative AT activity 1.10 0.91–1.31 0.3193 0.3179POD 1 AT activity 1.13 0.96–1.33 0.1565 0.1543POD 2 AT activity 1.26 1.07–1.48 0.0056 0.0053POD 3 AT activity 1.19 1.01–1.41 0.0412 0.0399POD 4 AT activity 1.17 1.00–1.37 0.0506 0.0493POD 5 AT activity 1.17 1.00–1.37 0.0553 0.0546Change in AT activity Post-CPBa 1.15 0.93–1.41 0.1888 0.4058Change in AT activity Day 1a 1.20 1.00–1.45 0.0558 0.1394Change in AT activity Day 2a 1.36 1.13–1.62 0.0009 0.0035Change in AT activity Day 3a 1.26 1.04–1.52 0.0163 0.0525Change in AT activity Day 4a 1.26 1.06–1.51 0.0104 0.0294Change in AT activity Day 5a 1.20 1.01–1.44 0.0407 0.1201a Refers to change in AT activity from the baseline level.AT � antithrombin; RBC � red blood cell; POD � postoperative day.


Although our study contributes to our understanding ofthe role AT plays in the perioperative period, we areconcerned by the absence of a temporal correlation betweenlower AT activity and MACE in our cohort, likely indicat-ing that AT level is not an important determinant ofthrombotic complications such as MI, stroke, and graftocclusion in the cardiac surgical setting. Thus, our obser-vational cohort cannot directly address any putative bio-logical mechanism for the role of AT in MACE generation.

CONCLUSIONIn a large cohort of patients undergoing CABG surgery,postoperative AT activity was independently associated withMACE. Because this occurs at time points predominantly afterthe MACE event, the clinical utility of AT as a biomarker ofrisk remains unknown.

DISCLOSUREThere are four potential conflicts of interest. Dr. Garvin was therecipient of a Research Fellowship funded by Talecris Biothera-peutics, which allowed time for generation of this and otherarticles. Dr. Chen is an employee of Talecris Biotherapeuticsand performed the majority of the analyses. To prevent theappearance or substance of conflict, Simon Body personallydirected the conduct of all analyses and confirmed the conductof the analyses by reviewing the code and output of allanalyses. In addition, Simon Body personally reran the impor-tant components of the analyses to confirm the findings andcan attest that there was no potential for conflict in the analysisphase. Dr. Body received a total of �$5000 to allow his time fortravel to Talecris in North Carolina to coordinate the analyses.Talecris also paid for the costs of antithrombin analysesperformed by Charles River Laboratories in Ontario, Canada,but had no opportunity to intervene in these analyses. In brief,

we believe there is no conflict of interest in the conduct of thisstudy.

REFERENCES1. Maclean PS, Tait RC. Hereditary and acquired antithrombin

deficiency: epidemiology, pathogenesis and treatment options.Drugs 2007;67:1429–40

2. Levy JH, Despotis GJ, Szlam F, Olson P, Meeker D, Weisinger A.Recombinant human transgenic antithrombin in cardiac surgery:a dose-finding study. Anesthesiology 2002;96:1095–102

3. Sanson BJ, Simioni P, Tormene D, Moia M, Friederich PW,Huisman MV, Prandoni P, Bura A, Rejto L, Wells P, MannucciPM, Girolami A, Buller HR, Prins MH. The incidence of venousthromboembolism in asymptomatic carriers of a deficiency ofantithrombin, protein C, or protein S: a prospective cohortstudy. Blood 1999;94:3702–6

4. Simioni P, Sanson BJ, Prandoni P, Tormene D, Friederich PW,Girolami B, Gavasso S, Huisman MV, Buller HR, Wouter tenCate J, Girolami A, Prins MH. Incidence of venous thrombo-embolism in families with inherited thrombophilia. ThrombHaemost 1999;81:198–202


6. Ranucci M, Ditta A, Boncilli A, Cotza M, Carboni G, Brozzi S,Bonifazi C, Tiezzi A. Determinants of antithrombin consump-tion in cardiac operations requiring cardiopulmonary bypass.Perfusion 2004;19:47–52

7. Bradshaw PJ, Jamrozik K, Le M, Gilfillan I, Thompson PL.Mortality and recurrent cardiac events after coronary arterybypass graft: long term outcomes in a population study. Heart2002;88:488–94

8. Charlesworth DC, Likosky DS, Marrin CA, Maloney CT,Quinton HB, Morton JR, Leavitt BJ, Clough RA, O’Connor GT.Development and validation of a prediction model for strokesafter coronary artery bypass grafting. Ann Thorac Surg2003;76:436–43

9. Koch CG, Weng YS, Zhou SX, Savino JS, Mathew JP, Hsu PH,Saidman LJ, Mangano DT. Prevalence of risk factors, and notgender per se, determines short- and long-term survival aftercoronary artery bypass surgery. J Cardiothorac Vasc Anesth2003;17:585–93

10. van Brussel BL, Plokker HW, Voors AA, Ernst JM, Ernst NM,Knaepen PJ, Koomen EM, Tijssen JG, Vermeulen FE. Multivar-iate risk factor analysis of clinical outcome 15 years aftervenous coronary artery bypass graft surgery. Eur Heart J 1995;16:1200–6

11. Martinez MA, Pena JM, Fernandez A, Jimenez M, Juarez S,Madero R, Vazquez JJ. Time course and prognostic significanceof hemostatic changes in sepsis: relation to tumor necrosisfactor-alpha. Crit Care Med 1999;27:1303–8

12. Sakr Y, Reinhart K, Hagel S, Kientopf M, Brunkhorst F.Antithrombin levels, morbidity, and mortality in a surgicalintensive care unit. Anesth Analg 2007;105:715–23

13. Wilson RF, Mammen EF, Tyburski JG, Warsow KM, KubinecSM. Antithrombin levels related to infections and outcome.J Trauma 1996;40:384–7

14. Baudo F, Caimi TM, de Cataldo F, Ravizza A, Arlati S, CasellaG, Carugo D, Palareti G, Legnani C, Ridolfi L, Rossi R,D’Angelo A, Crippa L, Giudici D, Gallioli G, Wolfler A, CaloriG. Antithrombin III (ATIII) replacement therapy in patientswith sepsis and/or postsurgical complications: a controlleddouble-blind, randomized, multicenter study. Intensive CareMed 1998;24:336–42

15. du Cheyron D, Bouchet B, Bruel C, Daubin C, Ramakers M,Charbonneau P. Antithrombin supplementation for anticoagu-lation during continuous hemofiltration in critically ill patientswith septic shock: a case-control study. Crit Care 2006;10:R45

16. Eid A, Wiedermann CJ, Kinasewitz GT. Early administration ofhigh-dose antithrombin in severe sepsis: single center resultsfrom the KyberSept-trial. Anesth Analg 2008;107:1633–8

Figure 1. Receiver operating characteristics of the addition ofantithrombin (AT) activity on postoperative days (PODs) 2–4 to theclinical model.



17. Kienast J, Juers M, Wiedermann CJ, Hoffmann JN, OstermannH, Strauss R, Keinecke HO, Warren BL, Opal SM. Treatmenteffects of high-dose antithrombin without concomitant heparinin patients with severe sepsis with or without disseminatedintravascular coagulation. J Thromb Haemost 2006;4:90–7

18. Afshari A, Wetterslev J, Brok J, Moller A. Antithrombin III incritically ill patients: systematic review with meta-analysis andtrial sequential analysis. BMJ 2007;335:1248–51


20. Conley JC, Plunkett PF. Antithrombin III in cardiac surgery: anoutcome study. J Extra Corpor Technol 1998;30:178–83

21. Despotis GJ, Levine V, Joist JH, Joiner-Maier D, Spitznagel E.Antithrombin III during cardiac surgery: effect on response ofactivated clotting time to heparin and relationship to markersof hemostatic activation. Anesth Analg 1997;85:498–506

22. Kanbak M. The treatment of heparin resistance with anti-thrombin III in cardiac surgery. Can J Anaesth 1999;46:581–5

23. Lemmer JH Jr, Despotis GJ. Antithrombin III concentrate totreat heparin resistance in patients undergoing cardiac sur-gery. J Thorac Cardiovasc Surg 2002;123:213–7

24. Levy JH, Montes F, Szlam F, Hillyer CD. The in vitro effects ofantithrombin III on the activated coagulation time in patientson heparin therapy. Anesth Analg 2000;90:1076–9



�2-Adrenergic Receptor-Coupled Phosphoinositide3-Kinase Constrains cAMP-Dependent Increases inCardiac Inotropy Through Phosphodiesterase4 ActivationChristopher J. Gregg,* Jochen Steppan, MD,* Daniel R. Gonzalez, PhD,† Hunter C. Champion, MD, PhD,‡Alexander C. Phan,* Daniel Nyhan, MD,* Artin A. Shoukas, PhD,* Joshua M. Hare, MD,†Lili A. Barouch, MD,* and Dan E. Berkowitz, MD*

BACKGROUND: Emerging evidence suggests that phosphoinositide 3-kinase (PI3K) may modu-late cardiac inotropy; however, the underlying mechanism remains elusive. We hypothesized that�2-adrenergic receptor (AR)-coupled PI3K constrains increases in cardiac inotropy through cyclicadenosine monophosphate (cAMP)-dependent phosphodiesterase (PDE) activation.METHODS: We tested the effects of PI3K and PDE4 inhibition on myocardial contractility by usingisolated murine cardiac myocytes to study physiologic functions (sarcomere shortening [SS] andintracellular Ca� transients), as well as cAMP and PDE activity.RESULTS: PI3K inhibition with the reversible inhibitor LY294002 (LY) resulted in a significantincrease in SS and Ca2� handling, indicating enhanced contractility. This response depended onGi� protein activity, because incubation with pertussis toxin (an irreversible Gi� inhibitor)abolished the LY-induced hypercontractility. In addition, PI3K inhibition had no greater effect onSS than both a PDE3,4 inhibitor (milrinone) and LY combined. Furthermore, LY decreased PDE4activity in a concentration-dependent manner (58.0% of PDE4 activity at LY concentrations of 10�M). Notably, PI3K� coimmunoprecipitated with PDE4D. The �2-AR inverse agonist, ICI 118,551(ICI), abolished induced increases in contractility.CONCLUSIONS: PI3K modulates myocardial contractility by a cAMP-dependent mechanismthrough the regulation of the catalytic activity of PDE4. Furthermore, basal agonist-independentactivity of the �2-AR and its resultant cAMP production and enhancement of the catalytic activityof PDE4 through PI3K represents an example of integrative cellular signaling, which controlscAMP dynamics and thereby contractility in the cardiac myocyte. These results help to explain themechanism by which milrinone is able to increase myocardial contractility in the absence ofdirect �-adrenergic stimulation and why it can further augment contractility in the presence ofmaximal �-adrenergic stimulation. (Anesth Analg 2010;111:870–7)

The �-adrenergic receptor (�-AR) isoforms character-ized in the heart have distinct but overlapping signaltransduction mechanisms that regulate many aspects

of myocardial pump function. Augmentation of cardiacinotropy through �-ARs is mediated through a well-studied system that increases intracellular cyclic adenosinemonophosphate (cAMP) leading to increases in cAMP-dependent protein kinase A activity.1 The �1-AR is coupledto the heterotrimeric Gs protein and is considered theprimary means for catecholamine-induced increases incAMP concentrations and thus myocardial contractility.The role of the �2-AR in modulation of contractility is more

complex, as this receptor is coupled to both Gs and Gi, thelatter of which is recognized to attenuate inotropy throughthe cGMP dependent effects of nitric oxide produced fromendothelial nitric oxide synthase 3. �2-AR signaling is likelycompartmentalized in light of the dual and seeminglyopposing effects of these signaling cascades.2 Those cas-cades are critically dependent on cAMP concentrations andspatial localization.3 Thus, changes in the balance betweencAMP production through adenylyl cyclase activation andcAMP breakdown through cAMP-dependent phosphodies-terases (PDE) will modulate protein kinase A activity andmyocardial contractility.

The roles of phosphoinositide 3-kinase (PI3K), a fam-ily of lipid kinases whose downstream targets includebioactive lipids and proteins, in signaling has emergedover the past few years.4 Specifically, it has been dem-onstrated that the �-AR kinase–1 (�-ARK1) and PI3Kinteract and that �-ARK1 recruitment of PI3K is crucialfor mediating �2-AR internalization as a component ofthe receptor/internalization scaffold complex.1 In addi-tion to its role in receptor internalization, the direct roleof PI3K in modulating myocardial contractility has beeninvestigated in an elegant study observing the role ofPI3K in progression of myocardial hypertrophy. Specifi-cally, mice deficient in the catalytic subunit of the PI3K�

From the *Johns Hopkins Medical Institutions (current affiliation: Universityof California, San Diego), †Johns Hopkins Medical Institutions (currentaffiliation: University of Miami), ‡Johns Hopkins Medical Institutions,Baltimore, Maryland (current affiliation: University of Pittsburgh).


This work was supported in part by a grant from the National SpaceBiomedical Research Institute (CA00405) through National Aeronautics &Space Administration (AS) and a National Institutes of Health grant (R01AG 021,523) (DEB).

Disclosure: The authors report no conflict of interest.

Address correspondence to Dan E. Berkowitz, MD, Johns Hopkins MedicalInstitutions, 600 N Wolfe Street, CCM Tower 711, Baltimore, MD 21287.Address e-mail to [email protected].

Copyright © 2010 International Anesthesia Research SocietyDOI: 10.1213/ANE.0b013e3181ee8312


isoform (p110��/�) demonstrate increased basal myo-cardial contractility in a cAMP-dependent manner, andinhibition of PI3K augments isoproterenol-induced con-tractile responses.5 The cellular mechanism of this effect,however, remains incompletely understood.

Thus, we hypothesized that PI3K is a negative upstreamregulator of myocardial contractility by modulating cAMP-dependent PDE activity.

METHODSAnimalsAnimal treatment and care was approved and provided inaccordance with the institutional animal care and use

committee of the Johns Hopkins University School ofMedicine, as well as with the National Institutes of Healthand American Physiological Society guidelines. Three5-month-old C57/BL6J mice were purchased from JacksonLaboratories (Bar Harbor, ME).

Cardiac Myocyte IsolationMyocytes were isolated by enzymatic digestion with colla-genase type 2 (1 g/L; Worthington Biochemical, Lakewood,NJ) and protease type XIV (0.1 g/L), as previously de-scribed.6,7 Cell suspension was obtained by mechanicallydisrupting digested ventricles, filtering, centrifugation, and

Figure 1. Phosphoinositide 3-kinase constrains contractility in cardiac myocytes. (A,B) Effects of the reversible phosphoinositide 3-kinase(PI3K) inhibitor LY (10 �M) on sarcomere shortening (SS) (A, n � 14, *** P � 0.001 vs baseline) and systolic [Ca2�]i (B, n � 12, *** P �0.001 vs baseline). Washout confirms the reversibility for both SS (n � 9, n.s. vs baseline) and [Ca2�]i (n � 8, n.s. vs baseline). Sampletransients are shown above the graphic panels. (C) LY (0.1 to 100 �M) dose response showing a concentration-dependent increase in SS withan EC50 of 2.2 �M. (D) Effects of wortmannin (5 nM) on SS (n � 4, ** P � 0.01 vs baseline). (E) Monoexponential time constant, � (indexof lusitropy), fit to diastolic portion of averaged Ca2� fluorescence transients (n � 15, * P � 0.05 vs baseline fit) and washout.


resuspension in Tyrode solution (1.0 mM Ca2�). The myo-cyte cell suspension was incubated with 5 �mol/L fura-2/am (Molecular Probes, Eugene, OR), then transferred toan inverted microscope (TE 200; Nikon), continuouslysuperfused with Tyrode solution and stimulated at 1 Hz.Change in average sarcomere length was determined byfast Fourier transform of the Z-line density trace to thefrequency domain. Calcium concentrations were measuredusing a dual-excitation spectrofluorometer (IonOptix, Mil-ton, MA), as previously described.6,7 The experiments wereconducted in the presence or absence of the reversible PI3Kinhibitor LY294002 (LY, 10 �M), the irreversible PI3Kinhibitor wortmanin (5 nM), the Gi� protein inhibitorpertussis toxin (PTX, 1.5 �g/mL), the �1-and �2-adrenoreceptor agonist isoproterenol (50 nM), the PDE3,4inhibitor milrinone (0.3, 3, 30, 300 �M), the PDE4 inhibitorrolipram (1, 10, 100 �M), and ICI 118,551 (ICI, 100 nM).

Phosphodiesterase Activity AnalysisTotal low Km cAMP-dependent PDE activity was assayedby fluorescence polarization (Molecular Devices, Sunny-vale, CA) and read on a microplate reader (Perkin Elmer,Wellesley, MA), as described by the manufacturer in thepresence or absence of LY (1, 3, 10 �mol/L), isoproterenol(100 nmol/L), milrinone (30 �mol/L), and ICI (100nmol/L).

Cyclic Nucleotide AssayMyocytes were used in a cAMP enzyme immunoassay(Amersham Pharmacia Biotech, Piscataway, NJ), as de-scribed by the manufacturer, in the presence or absence ofisoproterenol (50 nM), LY (10 �M), milrinone (Mil) (30 �M),and ICI (100 nM).

CoimmunoprecipitationPI3K (p110�) or control (bovine serum albumin) antibodies(both: Cell Signaling, Danvers, MA) were cross-linked toprotein A/G beads (Pierce Biotechnology, Rockford, IL)and used to immunoprecipitate proteins from mouse heartcell lysates. Western blots were then performed with aPDE4D antibody.

Western BlotsThe entire coimmunoprecipitation was denatured, re-duced (150 mmol/L dithiothreitol), and separated byTris-Glycine/sodium dodecyl sulfate-polyacrylamide gelelectrophoresis (Invitrogen, Carlsbad, CA). The proteinswere transferred to polyvinylidene fluoride membranes(Billerica, MA) with Towbin transfer buffer, and themembranes were blocked with 5% nonfat dry milk/trisbuffered saline (NFDM/TBS). Primary antibodies (PI3-kinase, Cell Signaling; PDE4D, Fabgennix, Frisco, TX)were diluted in 5% NFDM/TBS. The immunoblots werewashed and incubated with the appropriate secondaryantibodies (Santa Cruz Biotechnology, Santa Cruz CA).The membranes were washed again and developed withSuperSignal Pico and Femto substrate (Pierce Biotechnol-ogy, Rockford, IL).

Data Analysis and Statistical ProceduresData are presented as mean � SEM, with the exception ofphysiologic data that are presented as change from base-line � SEM. Statistical significance was considered for P

value �0.05 after t tests for pairs or nonparametric rankingand one-way ANOVA.

RESULTSPI3K Constrains Contractility in IsolatedCardiac MyocytesFigure 1 illustrates that the dose-dependent inhibition ofPI3K results in an increased sarcomere shortening (SS) andsystolic [Ca2�]i. In detail, inhibition of PI3K with thespecific reversible inhibitor LY (10 �M) resulted in a 2.21 �0.12-fold increase in SS (Fig. 1A) and a 1.68 � 0.10-foldincrease in systolic [Ca2�]i (Fig. 1B). This increase incontractility was reversible and concentration dependent(Fig. 1C). Similar results were observed with the irrevers-ible PI3K inhibitor wortmannin, which produced a 2.31 �0.04-fold increase in SS (Fig. 1D). Furthermore, the timeconstant of Ca2� fluorescence decay decreased from 0.16 �0.01 to 0.11 � 0.01 s upon PI3K inhibition (Fig. 1E).

PI3K Is Regulated Through Gi� SignalingFigure 2 shows the relation between PI3K and Gi� signal-ing. The enhanced contractility mediated by PI3K inhibi-tion was abolished by preincubation with the Gi� inhibitor,PTX. LY increased SS 2.52 � 0.75-fold, while in the presenceof PTX, SS did not change significantly (Fig. 2).

PI3K Inhibition Enhances �-AR-StimulatedContractility in MyocytesFigure 3 demonstrates the additive effect of PI3K inhibitionand �-AR stimulation. The combination of PI3K inhibition

Figure 2. Enhancement of myocardial contractility by PI3K inhibitionis sensitive to Gi activity. Effects of the irreversible Gi� inhibitorpertussis toxin (PTX) on basal and phosphoinositide-3 kinase (PI3K)inhibited myocardial contractility. LY significantly increased contrac-tility in non-PTX treated myocytes (n � 6, ** P � 0.01 vs baseline).In PTX-treated myocytes, LY response was significantly blunted,compared with LY only group (n � 6, ** P � 0.01, LY vs LY � PTX)and was not significantly different than baseline (n � 6, P � 0.05 vsbaseline).

PI3K Constrains Cardiac Contractility


with LY and �-AR stimulation with isoproterenol aug-mented SS 1.97 � 0.27-fold and systolic [Ca2�]i 1.50 �0.13-fold over isoproterenol alone (Fig. 3).

PI3K Inhibition Increases Myocyte cAMP LevelsAs shown in Figure 4, increasing cAMP levels are linked toPI3K inhibition and �-AR stimulation. Isoproterenol in-creased cAMP levels from 5.20 � 1.20 to 18.48 � 2.21fmol/�g protein, and LY increased cAMP levels to 24.32 �2.73 fmol/�g protein. The combination of LY and isopro-terenol had a synergistic effect and augmented cAMP levelsfurther to 37.44 � 3.78 fmol/�g protein (Fig. 4).

PI3K Is Coupled to cAMP-DependentPDE ActivityFigure 5 and 6 demonstrate the mechanism by which PI3Kregulates contractility in a cAMP-dependent manner. LY

decreased PDE4 activity in a concentration-dependentmanner, which paralleled the results from the PDE4 spe-cific inhibitor CB524717 (Fig. 5A). The PDE3,4 inhibitor Mil(30 �M) and LY increased cAMP levels from 4.70 � 0.77 to14.59 � 2.12 and 19.85 � 3.42 fmol/�g protein, respectively(Fig. 5B). There was no difference when incubated with LYand Mil alone or a combination of both (18.28 � 4.31fmol/�g protein).

Mil and the PDE4 specific inhibitor rolipram resulted ina concentration-dependent increase in SS and systolic[Ca2�]i (Fig. 6A, B). At EC50 concentrations, Mil (10 �M)produced a 1.35 � 0.08-fold increase in SS, and LY pro-duced a 2.27 � 0.25-fold increase. The combination of Miland LY yielded no synergistic effect (2.25 � 0.18-foldincrease), while the addition of isoproterenol (100 nM)yielded an additional enhancement in contractility (3.41 �0.32-fold change in SS, Fig. 6C).

Agonist Independent �2-AR Coupling ThroughPI3K Maintains Contractile ToneThe results depicted in Figure 7 illustrate that decreasingbasal agonist-independent cAMP production (with ICI), thecatalytic activity of PDE isoforms (regulated by PI3K) ceaseto have an effect on cardiac contractility. Myocytes incu-bated with the �2-AR inverse agonist ICI demonstrated atime and concentration-dependent decrease in SS (data notshown) and exhibited a decrease in SS from baseline (Fig.7A). Furthermore, ICI completely reversed LY and Mil-induced contractile responses, cAMP levels, and systolic[Ca2�]i (Fig. 7, B–D).

Interaction of PDE4 and PI3K�PDE4D was detected in HeLA and U937 cells (control) andin homogenates immunoprecipitated with PI3K� (p110�)but not in homogenates immunoprecipitated with bovineserum albumin. As expected, PI3K� (p110�) was detected

Figure 3. Synergistic effect of isoproterenol and LY on myocyte contractility. Cumulative data showing significant increases in sarcomereshortening (SS) (A) with adrenergic stimulation alone (isoproterenol: n � 9, * P � 0.05 vs baseline) and the addition of phosphoinositide-3kinase (PI3K) inhibition (isoproterenol � LY: n � 9, ** P � 0.01 vs baseline, # P � 0.01 isoproterenol versus isoproterenol � LY). Systolic[Ca2�]i (B) also showed significant enhancement with adrenergic stimulation alone (isoproterenol: n � 9, * P � 0.05 vs baseline) and theaddition of PI3K inhibition (isoproterenol � LY: n � 9, ** P � 0.01 vs baseline, # P � 0.01 isoproterenol versus isoproterenol � LY). Sampletransients from these experiments are shown above the graphic panels.

Figure 4. PI3K inhibition increases myocyte cAMP production andaugments adrenergically stimulated cAMP. Effect of isoproterenol(50 nM), LY (10 �M) and their combination on myocyte cAMPlevels (fmol/�g protein, n � 6, * P � 0.05). Both LY andisoproterenol increased cAMP levels independently, however, thecombination of isoproterenol and LY increased cAMP levels abovelevels seen with isoproterenol and LY alone, (n � 8, *** P �0.001, baseline versus isoproterenol � LY, # P � 0.01 isopro-terenol versus isoproterenol � LY).


in lysates immunoprecipitated with the same antibodies(Fig. 8).

DISCUSSIONIn the present study, we have demonstrated that PI3Knegatively modulates myocardial contractility by a Gi�-dependent mechanism. Furthermore, PI3K negatively regu-lates adrenergic stimulation of myocardial contractilitysince inhibition significantly enhances �-AR–mediated con-tractility. �2-AR–coupled PI3K decreases cAMP by amechanism that involves activation of PDE4, as PI3K

inhibition decreases PDE4 activity and increases cAMPlevels. Finally, this mechanism is constitutively activatedgiven that the inverse �2-AR agonist ICI completely re-verses both PI3K and PDE-mediated increases in contrac-tility and cAMP levels (Fig. 9).

We demonstrated that inhibition of PI3K enhances ad-renergic stimulation of cardiac myocytes, which is consis-tent with the findings of Jo et al.8 and Crackower et al.5

However, it does differ significantly from Jo et al.8 withregard to the effect of LY on basal contractility. While theydemonstrated no effect of PI3K inhibition with LY on basal

Figure 5. PI3K inhibition decreases the activity of cAMP-dependent PDE4 to increase myocardial cAMP levels. Concentration-dependentdepression of LY (A) and CB524717 (panel inlay) on phosphodiesterase (PDE) 4 activity (** P � 0.01, n � 5 vs baseline activity). (B) Effectof LY (10�61,517), milrinone (30 �M) alone, and in combination on myocyte cAMP levels (fmol/�g protein). LY and milrinone increasedmyocyte cAMP levels above baseline (* P � 0.05, n � 4 vs baseline). There is no significant increase in cAMP levels if the drugs are combined(n � 4, p � n.s. LY versus milrinone, milrinone versus LY � milrinone, LY versus LY � milrinone).

Figure 6. PI3K regulates myocardial contractility by cAMP-dependent PDE activation. Concentration-dependent enhancement of myocytesarcomere shortening (SS) by milrinone (A, n � 4) and rolipram (B, n � 10). (C) Effect of milrinone (30 �M), LY (5 �M), and isoproterenol (10nM) alone and in combination on myocyte SS and systolic [Ca2�]i. Milrinone and LY alone produced significant increases in SS and systolic[Ca2�]i. However, administration of a combination of both LY and milrinone produced no added effect over LY alone (n � 13, p � n.s. LY versusLY, milrinone). Addition of isoproterenol to milrinone and LY produced a further increase in SS (n � 13, ** P � 0.01, milrinone � LY �isoproterenol versus milrinone � LY).



cellular contractility,8 we demonstrated a concentration-dependent increase in both SS and systolic [Ca2�]i. This isconsistent with the findings of Crackower et al.5 whodemonstrated significantly enhanced basal contractility inisolated myocytes from mice in which the catalytic subunitof the � isoform of PI3K (p110�) has been knocked out. Weagree with Crackower et al.5 that myocyte cAMP levels aremodulated by the activity of PI3K. This is in contrast to theconclusion made by Jo et al. 8 that the �2-AR–mediatedenhancement in contractility is not mediated by an increasein cAMP, because they demonstrate no difference in myo-cyte cAMP levels in response to a �2-AR selective agonist inthe presence or absence of LY. However, these experiments

were performed in the presence of the nonspecific PDEinhibitor isobutylmethylxanthine. This difference in theassays performed could explain their different conclusion,while indirectly supporting our hypothesis that inhibitionof PI3K enhances contractility through a pathway involvingPDE inhibition.

We show that PI3K is coupled to cAMP-dependentPDE4 activity, which is in contrast to Patrucco et al.9 whodemonstrated that mice expressing a PI3Kg kinase deadmutant (PI3Kg

KD) have no overt alteration in myocardial

Figure 7. Agonist independent coupling of the �2-AR to cAMP production and the role of PI3K. (A) Depression of myocyte sarcomere shortening(SS) by the �2-AR inverse agonist ICI 118,551 (100 nM) (n � 6, * P � 0.05 vs baseline). (B) ICI abolishes LY and milrinone-induced increasesin SS (n � 10, ** P � 0.01 LY or milrinone versus baseline; p � n.s. LY or milrinone � ICI versus baseline) and systolic [Ca2�]i (D, n � 11,* P � 0.05 LY versus baseline; ** P � 0.01 milrinone versus baseline; p � n.s. LY or milrinone � ICI versus baseline). (C) Both LY (n � 6,*** P � 0.001 vs baseline) and milrinone (n � 6, ** P � 0.01 vs baseline) increased cAMP levels that were completely abolished by ICIcoincubation. Sample transients are shown above their respective graphic panels.

Figure 8. Molecular interaction between PI3K and PDE4D. Phospho-diesterase (PDE) 4D was detected in lysates from HeLa and U937cells (control). In mouse heart homogenates immunoprecipitatedwith phosphoinositide 3-kinase (PI3K) (p110�), PDE4D was detectedbut not in homogenates immunoprecipitated with A/G-linked bovineserum albumin Ab. Figure 9. Schematic of the proposed PI3K interaction with the �-AR.

�2-AR � beta-2 adrenoreceptor; AC � adenylate cyclase; Gs �G-protein S; Gi � G-protein i; PI3K � phosphoinositide 3-kinase;PDE � phosphodiesterase; cAMP � cyclic adenosine monophos-phate; PKA � protein kinase A.


contractility. They interpreted their results to indicate thatthe kinase domain of the PI3K was not necessary formodulating PDE and constraining cAMP-dependent alter-ations in myocardial contractility. Therefore, Patrucco etal.9 suggested that protein-protein interaction betweenPI3Kg and the PDE3B isoform is sufficient for PDE3Bactivation. Our data, on the other hand, suggest that thePI3K catalytic domain may indeed be important in activa-tion of a downstream PDE, in this case PDE4D. LY294002,a specific reversible inhibitor of PI3K, does so by competingwith adenosine triphosphate for the active site of thecatalytic subunit for PI3K�, p110�.10 Our results withLY294002 and wortmannin support the notion that kinaseactivity is critically important for the modulation of cAMP-dependent inotropic changes. We cannot refute the ideathat LY294002, in binding and inhibiting the kinase activityof PI3K, may, in addition, alter the binding of a PDE top110�. An alternative explanation is that PI3K may actthrough an intermediate to result in regulation of a PDE.This is hypothetically supported by the observation thatPDE has regulatory phosphorylation sites for Akt, thekinase downstream of PI3K.11,12 Thus, PI3K may not onlyact as a scaffold for PDE9 but may also be phosphorylatedby the enzyme or its downstream effector Akt.

PDE3 and PDE4 are the major isoforms that degradecAMP in cardiac myocytes.13 Recent data suggest that theseisoforms localize to distinct compartments in the cell andhave different functional roles.3,13–15 Specifically, PDE4appears to be functionally coupled to the pool of cAMPstimulated by �-ARs as a result of spatial confinement by alocalization signal, while PDE3 appears to regulate thegeneral pool of cAMP stimulated by forskolin.13 This isconsistent with the observations of Xiang et al.16 whodemonstrated that PDE4 rather than PDE3 is required for�2-AR signaling in neonatal mouse myocytes. They demon-strated an enhanced response to isoproterenol in the presenceof selective PDE4 inhibitors and not in cells treated with aselective PDE3 inhibitor. Furthermore, PDE4 knockout micehave augmented responses to isoproterenol, while myocytesfrom �2-AR knockout mice have no augmented response toPDE4 inhibition.16 These findings are consistent with ourfindings and support the idea that PDE4 is essential for �2-ARsignaling. Moreover, we have extended the observation thatthis phenomenon is dependent on PI3K.

There are several shortcomings of the present study thatneed to be considered. First, all physiologic experimentswere done in a rodent model of isolated myocyte and notverified in vivo or in human tissues. There are considerabledifferences between species, and it is not possible togeneralize all results and translate them to a differentspecies. Also, there could be a direct biochemical effect ofPI3K inhibitors on PDE activity. None, however, have yetbeen described in the literature. Moreover, even though weshowed that PI3K acts through PDE4, the cAMP-dependentPDE isoform remains controversial.

Milrinone, a well-known PDE 3,4 inhibitor, is frequentlyused to treat advanced heart failure, especially in the contextof cardiac surgery. Our novel finding that inhibition of PI3Kthrough Gi� enhances myocardial contractility by a PDE4-dependent mechanism that is independent of direct �-ARagonist activation further demonstrates the exquisite balance

of cAMP production and cAMP breakdown in the cardiacmyocyte and sheds a light on Milrinone’s mechanism ofaction (Fig. 9). In another way, cAMP is a convergence pointof two opposing signaling pathways. More specifically, cAMPlevels are the integrated output of two temporally and spa-tially sensitive signaling inputs that serve to regulate myocar-dial inotropic responses. This signaling convergence of cAMPproduction and breakdown at cAMP levels indicates tighterregulation of cAMP availability than has been previouslyidentified. This could potentially prevent a destructive, run-away cAMP event on the second timescale or could controlthe gain of the �2-AR over the lifetime of the organism byrelatively regulating transcription of both Gs and Gi proteins.Furthermore, by indirectly inhibiting the breakdown ofcAMP, Milrinone is able to increase myocardial contrac-tility in the absence of direct �-adrenergic stimulation andfurther augments contractility even in the presence ofmaximal �-adrenergic stimulation, as seen clinically. Thus,an understanding of the importance of this physiologicpathway in the regulation of myocyte contractility is im-portant in defining the molecular mechanism underlyingMilrinone’s action as a critical inotrope in our limitedarmamentarium.

REFERENCES1. Rockman HA, Koch WJ, Lefkowitz RJ. Seven-transmembrane-

spanning receptors and heart function. Nature 2002;415:206–122. Devic E, Xiang Y, Gould D, Kobilka B. Beta-adrenergic receptor

subtype-specific signaling in cardiac myocytes from beta(1)and beta(2) adrenoceptor knockout mice. Mol Pharmacol2001;60:577–83

3. Baillie GS, Houslay MD. Arrestin times for compartmentalisedcAMP signalling and phosphodiesterase-4 enzymes. CurrOpin Cell Biol 2005;17:129–34

4. Oudit GY, Sun H, Kerfant BG, Crackower MA, Penninger JM,Backx PH. The role of phosphoinositide-3 kinase and PTEN incardiovascular physiology and disease. J Mol Cell Cardiol2004;37:449–71

5. Crackower MA, Oudit GY, Kozieradzki I, Sarao R, Sun H, SasakiT, Hirsch E, Suzuki A, Shioi T, Irie-Sasaki J, Sah R, Cheng HY,Rybin VO, Lembo G, Fratta L, Oliveira-dos-Santos AJ, Benovic JL,Kahn CR, Izumo S, Steinberg SF, Wymann MP, Backx PH,Penninger JM. Regulation of myocardial contractility and cell sizeby distinct PI3K-PTEN signaling pathways. Cell 2002;110:737–49

6. Barouch LA, Harrison RW, Skaf MW, Rosas GO, Cappola TP,Kobeissi ZA, Hobai IA, Lemmon CA, Burnett AL, O’Rourke B,Rodriguez ER, Huang PL, Lima JA, Berkowitz DE, Hare JM.Nitric oxide regulates the heart by spatial confinement of nitricoxide synthase isoforms. Nature 2002;416:337–9

7. Khan SA, Lee K, Minhas KM, Gonzalez DR, Raju SV, TejaniAD, Li D, Berkowitz DE, Hare JM. Neuronal nitric oxidesynthase negatively regulates xanthine oxidoreductase inhibi-tion of cardiac excitation-contraction coupling. Proc Natl AcadSci U S A 2004;101:15944–8

8. Jo SH, Leblais V, Wang PH, Crow MT, Xiao RP. Phosphatidyl-inositol 3-kinase functionally compartmentalizes the concur-rent G(s) signaling during beta2-adrenergic stimulation. CircRes 2002;91:46–53

9. Patrucco E, Notte A, Barberis L, Selvetella G, Maffei A,Brancaccio M, Marengo S, Russo G, Azzolino O, Rybalkin SD,Silengo L, Altruda F, Wetzker R, Wymann MP, Lembo G,Hirsch E. PI3Kgamma modulates the cardiac response tochronic pressure overload by distinct kinase-dependent and-independent effects. Cell 2004;118:375–87

10. Walker EH, Pacold ME, Perisic O, Stephens L, Hawkins PT,Wymann MP, Williams RL. Structural determinants of phos-phoinositide 3-kinase inhibition by wortmannin, LY294002,quercetin, myricetin, and staurosporine. Mol Cell 2000;6:909–19



11. Kenan Y, Murata T, Shakur Y, Degerman E, ManganielloVC. Functions of the N-terminal region of cyclic nucleotidephosphodiesterase 3 (PDE 3) isoforms. J Biol Chem 2000;275:12331– 8

12. Wijkander J, Landstrom TR, Manganiello V, Belfrage P, Deger-man E. Insulin-induced phosphorylation and activation ofphosphodiesterase 3B in rat adipocytes: possible role forprotein kinase B but not mitogen-activated protein kinase orp70 S6 kinase. Endocrinology 1998;139:219–27

13. Mongillo M, McSorley T, Evellin S, Sood A, Lissandron V,Terrin A, Huston E, Hannawacker A, Lohse MJ, Pozzan T,Houslay MD, Zaccolo M. Fluorescence resonance energytransfer-based analysis of cAMP dynamics in live neonatal ratcardiac myocytes reveals distinct functions of compartmental-ized phosphodiesterases. Circ Res 2004;95:67–75

14. Baillie GS, Sood A, McPhee I, Gall I, Perry SJ, Lefkowitz RJ,Houslay MD. beta-Arrestin-mediated PDE4 cAMP phosphodi-esterase recruitment regulates beta-adrenoceptor switchingfrom Gs to Gi. Proc Natl Acad Sci U S A 2003;100:940–5

15. Perry SJ, Baillie GS, Kohout TA, McPhee I, Magiera MM, AngKL, Miller WE, McLean AJ, Conti M, Houslay MD, LefkowitzRJ. Targeting of cyclic AMP degradation to beta 2-adrenergicreceptors by beta-arrestins. Science 2002;298:834–6

16. Xiang Y, Naro F, Zoudilova M, Jin SL, Conti M, Kobilka B.Phosphodiesterase 4D is required for beta2 adrenoceptorsubtype-specific signaling in cardiac myocytes. Proc Natl AcadSci U S A 2005;102:909–14


ECHO ROUNDS

Intraoperative Transesophageal Echocardiography-Guided Patent Ductus Arteriosus Ligation in anAsymptomatic Nonbacterial Endocarditis PatientHaibo Song, MD, Fei Liu, MD, Ke Dian, MD, and Jin Liu, MD

On routine examination an asymptomatic 16-year-old girl was found to have a continuous (machin-ery like) murmur in the second intercostal space. A

transthoracic echocardiogram (TTE) demonstrated a patentductus arteriosus (PDA) with continuous left-to-rightshunting and a mobile hyperechogenic vegetation (40mm � 4 mm) attached to the wall of the pulmonary artery,with normal-sized right heart and pulmonary artery pres-sure. Three consecutive blood cultures were negative forbacterial growth. A chest radiograph was normal. Thepatient was diagnosed with PDA and noninfective endo-carditis. Consent for publication of this case was obtainedfrom the patient and her parents.

Intraoperative transesophageal echocardiography (TEE)(Philip IE33 � 7 to 3) during PDA ligation showed aslightly enlarged left atrium and ventricle, mild mitralvalve regurgitation, and a normal triscuspid valve. Mod-erate aortic regurgitation and a 6-mm vegetation (Fig. 1A) (Video 1; see Supplemental Digital Content 1,http://links.lww.com/AA/A177; see the Appendix forvideo legends) was noted in the midesophageal aortic valvelong-axis view. Using live three-dimensional TEE imaging,we noted the vegetation to be present on the right coronarycusp (Figure 1B) (Video 2; see Supplemental Digital Con-tent 2, http://links.lww.com/AA/A178; see the Appendixfor video legends). Imaging in the midesophageal rightventricular inflow–outflow view revealed a dilated mainpulmonary artery (MPA) without pulmonic valve regurgi-tation. A PDA that measured 18 mm long and 8 mm widewas detected in the modified upper esophageal aortic archshort-axis view. A mobile vegetation was seen in thepulmonary artery on the downstream side of the PDA (Fig.2A). Color Doppler imaging demonstrated flow into theMPA from descending aorta (Figure 2B). Continuous waveDoppler showed a continuous spectrum; the peak velocityof the systolic shunt flow was 4.5 m/s, and diastolic flowwas 3 m/s.

After institution of cardiopulmonary bypass, TEE-guided ligation of the PDA was first performed (Video 2,clip 2). During the procedure, TEE monitoring of theanterior wall of the pulmonary artery documented thepresence of the vegetation. Intraoperative surgical findingsincluded a dilated and atherosclerotic PDA with small

calcified vegetations across the lumen and at the pulmo-nary artery end. Vegetations were removed from the pul-monic and aortic valves. Finally, aortic valvuloplasty wasperformed. TEE confirmed the absence of flow through thePDA, and the absence of vegetations and residual regurgi-tation through the aortic valve when the operation wasfinished. Pathologic examination of the vegetations provedto be thrombi, and cultures were negative.

TEE has been shown to be more sensitive than TTE foridentifying small (�1 cm) valvular vegetations.1 In this caseit provided confirmatory information as well as detectingan aortic valve vegetation and valvular insufficiency notappreciated on preoperative TTE, which prevents infectiveendocarditis (IE) and other complications after surgery.

We believe that this is the first case report of anasymptomatic patient with nonbacterial thrombotic endo-carditis in a patient with PDA. The fetal ductus arteriosusarises from the aorta opposite the origin of the left subcla-vian artery to the bifurcation of the MPA at the origin of theleft pulmonary artery. PDA normally closes in the weeksafter birth, but persists in 10% of cases of adult congenitalheart disease.2 PDA-related IE is a rare but very dangerousevent, which could cause septic shock or even pulmonarythrombolization.3,4 Turbulent bloodflow produced by PDAis likely a causative factor in the development of nonbac-terial endocarditis and IE. Continuous turbulent flowcauses microscopic trauma to the endothelium of thepulmonary artery and aortic valve, which provides anenvironment conducive to platelet and fibrin depositionand progression to IE if septicemia sets in. Before theintroduction of antibiotic therapy and surgical closure,PDA-related IE had a 45% mortality rate. With recentadvancement in diagnostic tools, especially echocardiogra-phy, preoperative antimicrobial therapies, and surgery, themortality rate of patients with PDA-related IE seems tohave declined.3

PDA is usually demonstrated in the upper esophagealascending aortic short-axis view with color Doppler flowmapping.2 However, we made a small change to enhancevisualization of the PDA. We first get the upper esophagealaortic arch short-axis view. We advanced the probe 1 to 2cm from this view and rotated it from 0° to 60° to obtain aview of the descending aorta and MPA (Figure 2). Color-flow Doppler was used to detect flow through the PDAfrom 0° to 60°, from which we could visualize left-to-rightcontinuous shunt flow originating from the descendingaorta. The systolic velocity was 4.5 m/s, yielding a peakgradient between the descending aortic artery and pulmo-nary artery of 81 mm Hg, and a diastolic velocity (3 m/s)yielding the lowest gradient of 49 mm Hg.

TEE visualization of the PDA is potentially problematicbecause of interposition of the trachea between the esoph-agus and aortic arch (Figure 3 A). We used the techniquedescribed by Li et al.,5 in which a saline-filled balloon is

From the West China Hospital, Sichuan University, Chengdu, Sichuan,China.

Accepted for publication May 27, 2010.

Supplemental digital content is available for this article. Direct URL citationsappear in the printed text and are provided in the HTML and PDF versionsof this article on the journal’s Web site (www.anesthesia-analgesia.org).

Address correspondence to Fei Liu, MD, West China Hospital, SichuanUniversity, Chengdu, Sichuan 610041, China. Address e-mail [email protected].

Copyright © 2010 International Anesthesia Research SocietyDOI: 10.1213/ANE.0b013e3181effffb


advanced through the endotracheal tube after initiation ofcardiopulmonary bypass but before aortic cross-clampingto attenuate the ultrasound scattering by the tracheal aircolumn (Fig. 3 B). This resulted in substantial enhancementof the PDA in this case (Fig. 3C).

In summary, in this case, TEE monitoring played animportant role in PDA ligation, facilitating assessment ofresidual bloodflow. It also allowed monitoring of the statusof the vegetations in the pulmonary artery and on the aorticvalve.

Figure 1. A, Two-dimensional transesophagealechocardiographic examination of the aortic valvelong axis view. Arrow indicates a vegetation overly-ing the aortic valve. B, Live 3-dimensional viewshows the vegetation overlying the right coronaryartery cusp. DAO � descending aorta; AO � aorta;RV � right ventricle; LV � left ventricle; AVV �aortic valve vegetation; RCS � right coronary cusp;LCS � left coronary cusp; NCS � non-coronarycusp.

Figure 2. A, Two-dimensional transesophagealechocardiographic examination of the upper esoph-ageal aortic arch short-axis sectional view at 30°.Arrows indicate the vegetation in the pulmonaryartery. B, Color doppler image of shunted flowthrough the patent ductus arteriosus in the upperesophageal aortic arch short-axis view. MPA �main pulmonary artery; PV � pulmonary vegetation;DAO � descending aorta; PDA � patent ductusarteriosus.

Figure 3. A, Desending aorta and a blurred image ofthe main pulmonary artery. B, Relationship betweenthe esophagus, left main bronchus, and patentductus arteriosus (PDA). We used a saline balloonin the left main bronchus to improve the image ofthe PDA. C, Clear pulmonary artery after we used asaline balloon. DAO � descending aorta, MPA �main pulmonary artery, AAO �; AO arch � aorticarch, LPA � left pulmonary artery, LB � left mainbronchus; TEE � transesophageal echocardiogra-phy; RPA � right pulmonary artery; LPA � leftpulmonary artery.

Comprehensive TEE for PDA Ligation


APPENDIX: VIDEO LEGENDSVideo 1. Real-time monitoring of persistent patent ductus arteriosus(PDA) during surgical ligation. Initial bloodflow from the descendingaorta to the pulmonary artery ceases by the end of the videorecording. White arrow shows colorful bloodflow ceases after PDAligation. MPA � main pulmonary artery; DAO � descending aorta.Video 2. In the first part of the video is the live 2-dimensionaltransesophageal echocardiography (TEE) view of aortic valve vegeta-tion in the aortic valve long-axis view. White arrow indicates avegetation overlying the aortic valve. In the second part of the videois the live 3-dimensional TEE view of aortic valve vegetation. Whitearrow indicates a vegetation overlying the aortic valve. LV � leftventricle; LA � left atrium; AO � aorta.

REFERENCES1. Shapiro SM, Young E, De Guzman S, Ward J, Chiu CY, Ginzton

LE, Bayer AS. Transesophageal echocardiography in diagnosisof infective endocarditis. Chest 1994;105:377–82

2. Neuman MD, Fox JD, Muehlschlegel JD. Incidental discovery ofa large patent ductus arteriosus in an adult during aorticreconstruction: echocardiographic findings and diagnostic di-lemmas. Anesth Analg 2007;105:1227–8

3. Ozkokeli M, Ates M, Uslu N, Akcar M. Pulmonary and aorticvalve endocarditis in an adult patient with silent patent ductusarteriosus. Jpn Heart J 2004;45:1057–61

4. Kouris NT, Sifaki MD, Kontogianni DD, Zaharos I, KalkandiEM, Grassos HE, Babalis DK. Patent ductus arteriosus endarte-ritis in a 40-year-old woman, diagnosed with transesophagealechocardiography. A case report and a brief review of theliterature. Cardiovasc Ultrasound 2003;1:2

5. Li YL, Wong DT, Wei W, Liu J. A new method for detecting theproximal aortic arch and innominate artery by transesophagealechocardiography. Anesthesiology 2006;105:226–7

Clinician’s Key Teaching Points By Martin M. Stechert, MD, Roman M. Sniecinski, MD,and Martin J. London, MD

• Patent ductus arteriosus (PDA) is the persistence of a normal fetal connection between the left pulmonary artery andthe descending aorta. Clinical manifestations are determined by the size of the PDA and the degree of left-to-rightshunting. Although large PDAs cause left ventricular overload and are usually diagnosed in the newborn, a small lesionmay remain undetected into adulthood. Turbulent flow through the PDA can lead to endothelial injury, complicated bynonbacterial thrombotic endocarditis, bacterial seeding, and infective endocarditis.

• Transesophageal echocardiography (TEE) can demonstrate a PDA with the upper esophageal aortic arch views, whichshow a dilated main pulmonary artery and, when color-flow Doppler is applied, turbulent flow into the left pulmonaryartery from the descending aorta. However, because of the interposition of the air-filled trachea over the aortic arch,the left pulmonary artery is not always well visualized. Therefore, transthoracic echocardiography using thesuprasternal and parasternal windows is the preferred method of diagnosis.

• In this case, the authors initially obtained the upper esophageal aortic short-axis view, advanced the TEE probe 1 to2 cm, then adjusted the omniplane angle to 60° to successfully visualize the PDA. The image was further improvedby inserting a saline-filled balloon through the endotracheal tube while on bypass, which decreased scattering of theultrasound by endotracheal air.

• TEE may be useful for perioperative PDA assessment, provided an appropriate echo window can be found. In selectpatients, visualization might be improved by filling the endotracheal cuff with normal saline to provide such a window.

ECHO ROUNDS


Society for Ambulatory Anesthesiology

Section Editor: Peter S. A. Glass

Effect on Postoperative Sore Throat of Spraying theEndotracheal Tube Cuff with BenzydamineHydrochloride, 10% Lidocaine, and 2% LidocaineNan-Kai Hung, MD,* Ching-Tang Wu, MD,* Shun-Ming Chan, MD,* Chueng-He Lu, MD,*Yuan-Shiou Huang, MD,* Chun-Chang Yeh, MD,* Meei-Shyuan Lee, DPH,†and Chen-Hwan Cherng, MD, DMSc*

BACKGROUND: Postoperative sore throat (POST) is a common complication after endotrachealintubation. We compared the effectiveness on POST of spraying the endotracheal tube (ETT) cuffwith benzydamine hydrochloride, 10% lidocaine, and 2% lidocaine.METHODS: Three hundred seventy-two patients were randomly allocated into 4 groups. The ETTcuffs in each group were sprayed with benzydamine hydrochloride, 10% lidocaine hydrochloride,2% lidocaine hydrochloride, or normal saline before endotracheal intubation. After insertion, thecuffs were inflated to an airway leak pressure of 20 cm H2O. Anesthesia was maintained withpropofol. The patients were examined for sore throat (none, mild, moderate, or severe) at 1, 6,12, and 24 hours after extubation.RESULTS: The highest incidence of POST occurred at 6 hours after extubation in all groups.There was a significantly lower incidence of POST in the benzydamine group than 10% lidocaine,2% lidocaine, and normal saline groups (P � 0.05) at each observation time point. At 6 hoursafter extubation, the incidence of POST was significantly lower in the benzydamine group (17.0%)compared with 10% lidocaine (53.7%), 2% lidocaine (37.0%), and normal saline (40.8%) groups(P � 0.05). The benzydamine group had significantly decreased severity of POST compared withthe 10% lidocaine, 2% lidocaine, and normal saline groups (P � 0.05) at each observation timepoint. Compared with the 2% lidocaine and normal saline groups, the 10% lidocaine group hadsignificantly increased severity of POST at 1, 6, and 12 hours after extubation. There were nosignificant differences among groups in local or systemic side effects.CONCLUSIONS: Spraying benzydamine hydrochloride on the ETT cuff is a simple and effectivemethod to reduce the incidence and severity of POST. (Anesth Analg 2010;111:882–6)

Postoperative sore throat (POST) after general anes-thesia with an endotracheal tube (ETT) is an unde-sirable outcome1 with an incidence varying from

40% to 100%.2–5 Although the symptoms resolve spontane-ously without any treatment, prophylactic management fordecreasing its frequency and severity is still recommendedto improve the quality of postanesthesia care.6–8

Several pharmacological methods have been suggestedto reduce POST including inhaling beclomethasone; apply-ing lidocaine spray or lidocaine to the ETT; administeringaspirin, ketamine, or benzydamine hydrochloride; or gar-gling with azulene sulfonate.6,9–12 Benzydamine hydro-chloride is a topical nonsteroidal antiinflammatory drugthat also has analgesic, antipyretic, and antimicrobial

properties.13–16 It has been reported that moderate tosevere sore throat may be resolved with gargling benzy-damine hydrochloride.17 In addition, it is widely used inradiation-induced oral mucositis,18 for arthritis as a gelointment preparation applied to the skin,13 and for symp-tomatic treatment of acute sore throat.17,19 Preventive topicalbenzydamine hydrochloride applied to the oropharyngealcavity before endotracheal intubation or before endotrachealintubation and continuously for 48 hours postoperatively hasbeen reported to decrease the incidence and severity of POSTafter ETT insertion and laryngeal mask airway inser-tion.19–21 However, patients may sustain numbness or thesensation of tingling in the tissues in the oral cavity, drymouth, thirst, and nausea because of application of benzy-damine hydrochloride by gargling or oropharyngealspray.21 Combes et al.22 demonstrated that mucosal dam-age occurring at the cuff level is thought to be an importantcausative factor for tracheal morbidity. Therefore, applica-tion of benzydamine hydrochloride to the ETT rather thangargling may provide an alternative and effective methodto reduce the incidence and severity of POST.

The hypothesis of this study was that simply sprayingbenzydamine hydrochloride on the endotracheal cuff willprovide better prevention of POST after extubation than10% lidocaine hydrochloride, 2% lidocaine hydrochloride,or normal saline sprayed over the cuff of the ETT.

From the *Department of Anesthesiology, Tri-Service General Hospital andNational Defense Medical Center; and †School of Public Health, NationalDefense Medical Center, Taipei, Taiwan, Republic of China.

Accepted for publication January 11, 2010.

Supported by grants from Tri-Service General Hospital (TSGH-C98-125) ofTaiwan, Republic of China.

Address correspondence and reprint requests to Dr. Chen-Hwan Cherng,Department of Anesthesiology, Tri-Service General Hospital and NationalDefense Medical Center, #325, Section 2, Chenggung Rd., Neihu 114, Taipei,Taiwan, ROC. Address e-mail to [email protected].

Copyright © 2010 International Anesthesia Research SocietyDOI: 10.1213/ANE.0b013e3181d4854e


mohamed

Highlight

mohamed

Note

benzydamine = tantum verde

METHODSAfter obtaining hospital ethics committee approval andwritten informed consent, 420 patients (n � 105 per group),ASA physical status I to III, aged 18 to 84 years, wereincluded in this prospective, randomized, and double-blindstudy. All patients were scheduled for surgical proceduresunder general anesthesia with orotracheal intubation in thesupine position. The exclusion criteria included patientsundergoing oral cavity surgery, cervical spine surgery,thyroid surgery, patients with a difficult airway (after �1orotracheal intubation attempt), requiring a nasogastrictube, or a history of perioperative sore throat.

Patients were randomly allocated into 4 groups bychoosing blinded envelopes. All study solutions and nor-mal saline were contained in similarly configured bottles.The grouping and medication were as follows:

1. The benzydamine (Comfflam, United Biomedical, Asia)(containing ethanol, glycerin, and menthol as additives)group: 10 puffs of benzydamine hydrochloride weresprayed on the ETT cuffs, which contain approximately1.5 mg benzydamine hydrochloride.

2. The 10% lidocaine (Xylocaine Spray 10%, AstraZeneca,Sweden) (containing ethanol, polyethylene glycol 400,menthol, saccharin, and essence of banana as addi-tives) group: 10 puffs of 10% lidocaine hydrochloridewere sprayed on the ETT cuffs, which contain ap-proximately 100 mg lidocaine hydrochloride.

3. The 2% lidocaine (Xylocaine 2%, AstraZeneca) (con-taining sodium chloride as an additive) group: 10puffs of 2% lidocaine hydrochloride were sprayed onthe ETT cuffs, which contain approximately 20 mglidocaine hydrochloride.

4. The normal saline group: 10 puffs of normal salinewere sprayed on the ETT cuffs, which contain ap-proximately 1 mL normal saline.

Sterile ETTs (ILM Endotracheal Tube, Euromedical, Ke-dah, Malaysia) with high-volume, low-pressure cuffs wereused. Application of substances to the ETT cuff was per-formed 5 minutes before the induction of anesthesia. Maleand female patients received 7.5- and 7.0-mm inner diam-eter ETTs, respectively. Before intubation, oxygen wasadministered, adequate IV access was established, andstandard American Society of Anesthesiologists clinicalmonitoring was applied. After administration of 2 to 3�g/kg fentanyl, induction was accomplished with 2 to 2.5mg/kg propofol and 0.6 mg/kg rocuronium. Endotrachealintubation was performed by residents with at least 2 yearsof experience and attending physicians, who were blindedto group allocation. The ETT cuffs were inflated with roomair to achieve a seal at 20 cm H2O of peak airway pressure.Anesthesia was maintained by a target-controlled infusionsystem (Fresenius Base Primea�, Brezins, France) withpropofol, intermittent fentanyl, and cisatracurium/rocuroniumadministered as required. Nitrous oxide was not used. Atthe end of surgery, neuromuscular blockade was antago-nized by neostigmine and atropine. After full recovery andawakening, the ETT was removed after gentle suctioning oforal secretions. Patients were then transferred to the post-anesthesia care unit.

At 1, 6, 12, and 24 hours after extubation, the patientswere asked about sore throat and hoarseness by a singleinvestigator who was blinded to the group allocation. POSTwas graded by a modified 4-point scale (1–4): 1, no sorethroat; 2, mild sore throat (complains of sore throat only onasking); 3, moderate sore throat (complains of sore throaton his or her own); and 4, severe sore throat (change ofvoice or hoarseness, associated with throat pain).10,11

We recorded the patients’ age, sex, weight, height,duration of surgery, total fentanyl consumption, type ofsurgery, and postoperative analgesia methods. Potentialside effects associated with tracheal intubation or thestudy drugs, such as nausea, vomiting, cough, throatnumbness or stinging, dry mouth, and hoarseness, werealso recorded.

Before initiation of the study, power analysis was per-formed. A minimum of 91 patients was required in eachgroup to detect a decrease of the incidence of POST from40% to 20% with a power of 80% and a significance level of95%. Statistical analysis was performed using the SPSS forWindows version 15 (SPSS, Chicago, IL). Results are ex-pressed as mean (SD), median with range, or percentage.Patient age, height, weight, and duration of surgery werecompared among groups and tested statistically by analysisof variance. The incidence of POST and side effects amonggroups were tested by �2 tests. To avoid a type I error, forthose significant variables in the �2 test, we recalculated allpossible six 2 � 2 �2 tests by applying the Bonferroniinequality to adjust the � level (i.e., P[�2 � 6.97] � 0.05/6 �0.0083 for 1 degree of freedom). Differences in the severityof symptoms among groups were evaluated by Kruskal-Wallis tests. Furthermore, the Dunn procedure was appliedto compare the difference among groups. The area underthe curve (AUC) by adding 3 trapezoid areas was gener-ated from the level of POST and time (within 24 hours).Both differences in the severity of symptoms and AUCwere tested by Kruskal-Wallis analysis of variance andfollowed by the Dunn procedure. Values of P � 0.05 wereconsidered statistically significant.

RESULTSCharacteristics of the study groups are shown in Table 1.There were 11, 12, 13, and 12 patients withdrawn from thebenzydamine, 10% lidocaine, 2% lidocaine, and normalsaline groups, respectively, because of �1 attempt at oro-tracheal intubation or nasogastric tube insertion during theoperation. Therefore, there were 372 patients enrolled inthis study.

In the 24-hour evaluation period, the highest incidencesof POST occurred at the sixth hour time interval afterextubation, with incidences (95% confidence interval)of 17.0% (9.4%–24.6%), 53.7% (43.6%– 63.9%), 37.0%(27.1%–46.8%), and 40.8% (30.7%–50.9%) in the benzydam-ine, 10% lidocaine, 2% lidocaine, and normal saline groups,respectively. There was a significantly lower incidence ofPOST in the benzydamine group than the 10% lidocaine,2% lidocaine, and normal saline groups (P � 0.05) at eachobservation time point (Table 2). Similarly, the benzydam-ine group had significantly decreased severity of POST


compared with the 10% lidocaine, 2% lidocaine, and nor-mal saline groups (P � 0.05) at each observation time point(Table 2). Compared with 2% lidocaine and normal salinegroups, the 10% lidocaine group had significantly in-creased severity of POST at 1, 6, and 12 hours afterextubation (Table 2). Of the mean AUCs generated from thelevel of POST and time (within 24 hours), the benzydaminegroup (25.6) had a significantly smaller area than the 10%lidocaine (41.4) and normal saline groups (33.0) (P � 0.05),but not the 2% lidocaine group (37.5). There were nosignificant differences in AUC among any other groups(data not shown).

There were no significant differences among groups forthe potential side effects relevant to tracheal intubation orstudy drugs.

DISCUSSIONThis study demonstrated that spraying benzydamine hy-drochloride on an ETT cuff may reduce the incidence andseverity of POST compared with applying 10% lidocaine,2% lidocaine, and normal saline. An unexpected finding

was that applying 10% lidocaine on the ETT cuff increasedthe severity of POST.

POST is one of the common side effects associatedwith endotracheal intubation.2–5 Tracheal mucosa lesionsafter intubation and an overinflated ETT cuff have beenproposed to be a possible cause of POST.22 These com-plications can occur after even a “smooth” intubation.Immediate POST may be primarily due to the action ofextubation, and late POST may be related to trachealmucosa trauma.23 According to the results of this study,the highest incidence of POST occurred at the sixth hourafter extubation, but not the first hour. Sore throat at thefirst hour after extubation might be masked by residualanalgesic effects after general anesthesia.

Benzydamine hydrochloride is indicated for the relief ofpainful conditions of the mouth and throat such as tonsil-litis, sore throat, radiation mucositis, and postorosurgicaland periodontal procedures. Previous studies demon-strated that topical application of benzydamine hydrochlo-ride to the pharynx before laryngeal mask airway or ETTinsertion decreased the incidence of POST.19–21 The side

Table 1. Demographic Data of the Patients and Data Related to the Surgery

Group

Benzydaminehydrochloride

(n � 94)

10% lidocainehydrochloride

(n � 93)

2% lidocainehydrochloride

(n � 92)Normal saline

(n � 93)Gender (M/F) 45/49 48/45 46/46 48/45Age (y) 48.5 (16) 47.8 (15) 45.3 (17) 46.3 (17)Height (cm) 162.9 (9.2) 162.5 (8.8) 162.7 (9.3) 161.6 (9.5)Weight (kg) 62.2 (11.4) 63.8 (13.4) 64.3 (12.5) 62.3 (12.1)Duration of surgery (min) 183.7 (110) 181.6 (91) 183.8 (114) 176.8 (100)Total fentanyl consumption (�g) 173 (73) 173 (59) 173 (61) 171 (61)Postoperative analgesia method (PCA/meperidine IM) 78/16 76/17 75/17 78/15Types of surgery

Colon rectal surgery 6 4 6 4General surgery 24 18 20 16Genitourinary surgery 3 5 5 4Gynecologic surgery 20 16 18 24Ophthalmologic surgery 12 17 16 15Orthopedic surgery 23 18 16 21Plastic surgery 6 15 11 9

Values are presented as mean (SD) or number of patients.PCA � patient-controlled analgesia.

Table 2. Incidence (n, %) and Severity (Median, Range) of Postoperative Sore Throat

Evaluation timeBenzydamine

(n � 94)10% lidocaine

(n � 93)2% lidocaine

(n � 92)Normal saline

(n � 93) P*1 h after extubation

Incidence 10 (10.6%)* 30 (33.2%) 21 (22.8%) 22 (23.7%) 0.005Severity 1 (1–4)* 1 (1–4)†‡ 1 (1–4) 1 (1–4) 0.004

6 h after extubationIncidence 16 (17.0%)* 50 (53.7%) 34 (37.0%) 38 (40.8%) �0.001Severity 1 (1–4)* 2 (1–4)†‡ 1 (1–4) 1 (1–4) �0.001

12 h after extubationIncidence 5 (5.3%)* 37 (39.8%) 22 (23.9%) 30 (32.2%) �0.001Severity 1 (1–4)* 1 (1–4)†‡ 1 (1–4) 1 (1–4) �0.001

24 h after extubationIncidence 2 (2.1%)* 25 (26.9%) 16 (17.4%) 19 (20.4%) �0.001Severity 1 (1–4)* 1 (1–4)† 1 (1–4) 1 (1–4) �0.001

Values are presented as number of patients (%) or median (range).* P � 0.05, benzydamine group versus 10% lidocaine, 2% lidocaine, and normal saline groups.† P � 0.05, 10% lidocaine group versus 2% lidocaine group.‡ P � 0.05, 10% lidocaine group versus normal saline.

Benzydamine Hydrochloride Reduces Postoperative Sore Throat


effects of topical use of benzydamine hydrochloride in-clude local numbness or burning, stinging sensation, nau-sea or vomiting, cough, dry mouth, throat discomfort,drowsiness, and headache,24 which may be evident beforeinduction of anesthesia. To avoid these adverse effects,we applied benzydamine hydrochloride on the ETT cuffinstead of perioperative topical application to the oralpharyngeal cavity. We found that this maneuver pro-vided excellent prevention of POST and reduced itsincidence from placebo or 2% lidocaine spray by �50%.Therefore, the application of benzydamine hydrochlo-ride on the ETT cuffs may provide a simple and effectivemethod to attenuate the incidence and severity of POSTafter tracheal intubation.

Application of lidocaine spray to the oral pharyngealcavity before intubation seems to increase the incidence ofsore throat.5,25,26 In this study, we also found that spraying10% lidocaine on the ETT cuff also increased the severity ofPOST compared with 2% lidocaine or placebo. Ten percentlidocaine solution contains ethanol, polyethylene glycol400, menthol, saccharin, and macrogolum as additives inthe solvent, whereas the 2% lidocaine solution we usedcontained sodium chloride as an additive. In fact, bothmenthol and ethanol can irritate tracheal mucosa, poten-tially causing tracheal mucosa damage, thus leading toincreased severity of POST. However, Soltani and Agha-davoudi27 reported that using intracuff lidocaine (ETT cuffsprefilled with 7 to 8 mL of 2% lidocaine for 90 minutesbefore intubation and refilled with enough 2% lidocaineafter intubation) was superior to spraying topical 10%lidocaine on laryngopharyngeal structures or on the distalend of the ETT for decreasing the incidence of POST.Theoretically, chemical irritation from the additives may beavoided by using intracuff lidocaine. We also found that 2%lidocaine spray did not attenuate the incidence and severityof POST compared with normal saline. The duration of theanalgesic effect of lidocaine spray applied to oral mucosa is�15 minutes.28 In this study, at the end of surgery (aver-aging 180 minutes after tracheal intubation), the analgesiceffect of lidocaine spray might have already disappeared.This probably explains why we found the incidence ofPOST to be no different between the 2% lidocaine andnormal saline groups.

One limitation of our study is that there was no recordof coughing or bucking at the time of extubation. Althoughthe extubation protocol was the same in all groups, we didnot evaluate the correlation between the frequency ofcoughing or bucking at the time of extubation and theincidence of POST. The second limitation is that benzy-damine hydrochloride is available under different tradenames in different countries, its formulations are quitedifferent in each country, and the additives might also vary.The drug is not available in the United States, Canada, andmost of Western Europe (the United Kingdom being anexception). For this reason, the results of this study mightnot be widely applicable. Moreover, the safety and dosageof benzydamine hydrochloride applied to the trachea needfurther investigation, even though we did not find anyadverse effects in our patients. The third limitation is thatthe additives to 2% and 10% lidocaine solution are differ-ent, which may have influenced the result.

In conclusion, application of benzydamine hydrochlo-ride on the ETT cuff effectively attenuates the incidenceand severity of POST. Application of 10% lidocaine sprayshould be avoided because of worsening of POST, andspraying 2% lidocaine on the ETT cuff does not preventPOST.

REFERENCES1. Macario A, Weinger M, Carney S, Kim A. Which clinical

anesthesia outcomes are important to avoid? The perspectiveof patients. Anesth Analg 1999;89:652–8

2. Biro P, Seifert B, Pasch T. Complaints of sore throat aftertracheal intubation: a prospective evaluation. Eur J Anaesthe-siol 2005;22:307–11

3. Al-Qahtani AS, Messahel FM. Quality improvement in anes-thetic practice—incidence of sore throat after using smalltracheal tube. Middle East J Anesthesiol 2005;18:179–83

4. Monroe MC, Gravenstein N, Saga-Rumley S. Postoperativesore throat: effect of oropharyngeal airway in orotracheallyintubated patients. Anesth Analg 1990;70:512–6


6. el Hakim M. Beclomethasone prevents postoperative sorethroat. Acta Anaesthesiol Scand 1993;37:250–2

7. Monem A, Kamal RS. Postoperative sore throat. J Coll Physi-cians Surg Pak 2007;17:509–14



10. Canbay O, Celebi N, Sahin A, Celiker V, Ozgen S, Aypar U.Ketamine gargle for attenuating postoperative sore throat. Br JAnaesth 2008;100:490–3

11. Agarwal A, Nath SS, Goswami D, Gupta D, Dhiraaj S, SinghPK. An evaluation of the efficacy of aspirin and benzydaminehydrochloride gargle for attenuating postoperative sore throat:a prospective, randomized, single-blind study. Anesth Analg2006;103:1001–3

12. Ogata J, Minami K, Horishita T, Shiraishi M, Okamoto T,Terada T, Sata T. Gargling with sodium azulene sulfonatereduces the postoperative sore throat after intubation of thetrachea. Anesth Analg 2005;101:290–3

13. Quane PA, Graham GG, Ziegler JB. Pharmacology of benzy-damine. Inflammopharmacology 1998;6:95–107

14. Baldock GA, Brodie RR, Chasseaud LF, Taylor T, WalmsleyLM, Catanese B. Pharmacokinetics of benzydamine after intra-venous, oral, and topical doses to human subjects. BiopharmDrug Dispos 1991;12:481–92

15. Guglielmotti A, Aquilini L, Rosignoli MT, Landolfi C, Soldo L,Coletta I, Pinza M. Benzydamine protection in a mouse modelof endotoxemia. Inflamm Res 1997;46:332–5

16. Modeer T, Yucel-Lindberg T. Benzydamine reduces prosta-glandin production in human gingival fibroblasts challengedwith interleukin-1 beta or tumor necrosis factor alpha. ActaOdontol Scand 1999;57:40–5

17. Turnbull RS. Benzydamine Hydrochloride (Tantum) in themanagement of oral inflammatory conditions. J Can DentAssoc 1995;61:127–34

18. Epstein JB, Silverman S Jr, Paggiarino DA, Crockett S, SchubertMM, Senzer NN, Lockhart PB, Gallagher MJ, Peterson DE,Leveque FG. Benzydamine HCl for prophylaxis of radiation-induced oral mucositis: results from a multicenter, random-ized, double-blind, placebo-controlled clinical trial. Cancer2001;92:875–85

19. Gulhas N, Canpolat H, Cicek M, Yologlu S, Togal T, DurmusM, Ozcan Ersoy M. Dexpanthenol pastille and benzydaminehydrochloride spray for the prevention of post-operative sorethroat. Acta Anaesthesiol Scand 2007;51:239–43


20. Mazzarella B, Macarone Palmieri A, Mastronardi P, Spatola R,Lamarca S, De Rosa G, Mastella A. Benzydamine for theprevention of pharyngo-laryngeal pathology following tra-cheal intubation. Int J Tissue React 1987;9:121–9

21. Kati I, Tekin M, Silay E, Huseyinoglu UA, Yildiz H. Doesbenzydamine hydrochloride applied preemptively reduce sorethroat due to laryngeal mask airway? Anesth Analg 2004;99:710–2

22. Combes X, Schauvliege F, Peyrouset O, Motamed C, Kirov K,Dhonneur G, Duvaldestin P. Intracuff pressure and trachealmorbidity: influence of filling with saline during nitrous oxideanesthesia. Anesthesiology 2001;95:1120–4

23. Keane WM, Denneny JC, Rowe LD, Atkins JP Jr. Complicationsof intubation. Ann Otol Rhinol Laryngol 1982;91:584–7

24. Passali D, Volonte M, Passali GC, Damiani V, Bellussi L.Efficacy and safety of ketoprofen lysine salt mouthwash versusbenzydamine hydrochloride mouthwash in acute pharyngealinflammation: a randomized, single-blind study. Clin Ther2001;23:1508–8

25. Maruyama K, Sakai H, Miyazawa H, Iijima K, Toda N,Kawahara S, Hara K. Laryngotracheal application of lidocainespray increases the incidence of postoperative sore throat aftertotal intravenous anesthesia. J Anesth 2004;18:237–40

26. Hara K, Maruyama K. Effect of additives in lidocaine spray onpostoperative sore throat, hoarseness and dysphagia after totalintravenous anaesthesia. Acta Anaesthesiol Scand 2005;49:463–7

27. Soltani HA, Aghadavoudi O. The effect of different lidocaineapplication methods on postoperative cough and sore throat.J Clin Anesth 2002;14:15–8

28. Schonemann NK, van der Burght M, Arendt-Nielsen L, Bjer-ring P. Onset and duration of hypoalgesia of lidocaine sprayapplied to oral mucosa—a dose response study. Acta Anaes-thesiol Scand 1992;36:733–5

Benzydamine Hydrochloride Reduces Postoperative Sore Throat


The Effectiveness of Benzydamine HydrochlorideSpraying on the Endotracheal Tube Cuff or OralMucosa for Postoperative Sore ThroatYuan-Shiou Huang, MD,* Nan-Kai Hung, MD,* Meei-Shyuan Lee, MPH,† Chang-Po Kuo, MD,*Jyh-Cherng Yu, MD,‡ Go-Shine Huang, MD,* Chen-Hwan Cherng, MD, DMSC,*Chih-Shung Wong, MD, PhD,* Chi-Hong Chu, MD, PhD,‡ and Ching-Tang Wu, MD*

BACKGROUND: The etiology of postoperative sore throat (POST) is considered to be the resultof laryngoscopy, intubation damage, or inflated cuff compression of the tracheal mucosa. In thisstudy, we compared the effectiveness in alleviating POST using different approaches tobenzydamine hydrochloride (BH) administration by spraying the endotracheal tube (ET) cuff or theoropharyngeal cavity, or both.METHODS: Three hundred eighty patients were included in this prospective and double-blindstudy, which was randomized into 4 groups: group A, oropharyngeal cavity spray of BH, anddistilled water on the ET cuff; group B, both the oropharyngeal cavity and the ET cuff received BHspray; group C, the ET cuff received BH spray, and the oropharyngeal cavity received distilledwater; and group D, distilled water sprayed on both the ET tube and into the oropharyngeal cavity.The patients were examined for sore throat (none, mild, moderate, severe) at 0, 2, 4, and 24hours postextubation.RESULTS: The incidence of POST was 23.2%, 13.8%, 14.7%, and 40.4% in groups A, B, C, andD, respectively. POST occurred significantly less frequently in groups B and C compared withgroup D (odds ratio: 0.36; 95% confidence interval: 0.21–0.60; P � 0.05). However, there wasno significant difference between groups A and D (odds ratio: 0.62; 95% confidence interval:0.38–1.01). Moreover, there was no significant interaction between spraying BH over theoropharyngeal cavity and the ET cuff on the incidence of POST (P � 0.088). The severity of POSTwas significantly more intense in group D compared with groups B and C (P � 0.001). Group Bhad a significantly higher incidence of local numbness, burning, and/or stinging sensationcompared with patients in group D (P � 0.05).CONCLUSIONS: This study indicates that spraying BH on the ET cuff decreases the incidenceand severity of POST without increased BH-related adverse effects. (Anesth Analg 2010;111:887–91)

Postoperative sore throat (POST) after intubated gen-eral anesthesia (GA) is a troublesome complicationand is recognized as one of the undesirable outcomes

in the postoperative period.1 The incidence of POST rangesfrom 21% to 66%2,3 in accordance with different surgicaland anesthetic manipulations.4,5 Therapeutic managementfor decreasing its frequency and severity is still advised toimprove the quality of postanesthesia care even though itwill resolve without treatment.4,6 Various methods havebeen reported to alleviate POST, such as using a smallersized endotracheal tube (ET), lubricating the ET cuff, andavoiding excessive intracuff pressure.2 In addition, severalanalgesic and antiinflammatory drugs have also been used,including IV dexamethasone,4 inhaled beclomethasone,and gargling with azulene sulfonate,6 aspirin,7 or benzy-damine hydrochloride (BH).8,9 BH is a topical nonsteroidal

antiinflammatory drug and has analgesic, antipyretic, an-timicrobial, and antiinflammatory effects.10 It has beenreported that moderate to severe oral inflammatory condi-tions may be resolved by gargling BH.11 In addition, it iswidely used in radiation-induced oral mucositis,12 forarthritis as a gel ointment preparation applied to the skin,10

and for symptomatic treatment of acute sore throat.3,11 Theaim of this study was to compare the effectiveness of 0.15%BH in alleviating POST using different approaches: spray-ing on the ET cuff, into the oropharyngeal cavity, or both.

METHODSThis study was approved by our institutional ethics com-mittee, and written informed consent was obtained from allparticipants. In this prospective, randomized, and double-blind study, we enrolled 380 patients. All patients werescheduled for elective surgery under GA with ET intuba-tion. Patients who had head and neck surgery, cervicalspine surgery, or thyroid surgery, and had a history ofpreoperative sore throat were excluded. Patients with firstattempt failed laryngoscopy, postoperative endotrachealintubation or reintubation, and nasogastric tube insertionwere also excluded from further analysis.

All patients were randomly assigned into 4 groups (95patients in each group). The grouping and medication wereas follows: group A, 5 puffs of 0.15% BH (total 0.75 mg,Comfflam; United Biomedical Asia, Taiwan) were sprayed

From the *Department of Anesthesiology, †School of Public Health, and‡Division of General Surgery, Tri-Service General Hospital and NationalDefense Medical Center, Taipei, Taiwan, Republic of China.

Accepted for publication April 25, 2010.


Address correspondence and reprint requests to Ching-Tang Wu, MD,Department of Anesthesiology, Tri-Service General Hospital and Na-tional Defense Medical Center, #325, Section 2, Chenggung Rd., Neihu114, Taipei, Taiwan. Address e-mail to [email protected] [email protected].

Copyright © 2010 International Anesthesia Research SocietyDOI: 10.1213/ANE.0b013e3181e6d82a


into the oropharyngeal cavity and 5 puffs of distilled waterwere sprayed on the ET cuff (5 puffs containing approxi-mately 0.5 mL distilled water); group B, the oropharyngealcavity and the ET cuff were both sprayed with 5 puffs ofBH; group C, the ET cuff was sprayed with 5 puffs of BHand the oropharyngeal cavity was sprayed with 5 puffs ofdistilled water; and group D, the oropharyngeal cavity andthe ET cuff were both sprayed with 5 puffs of distilledwater. All medications were sprayed 5 minutes beforeinduction of anesthesia by a nurse anesthetist blinded to thetreatment. The treatments were prepared by a pharmacistin our pharmacy department blinded to the medication sothat the different treatments had the same external appear-ance. No patient received premedication, and standardmonitors were applied in the operating room. GA wasinduced with 2 to 3 �g/kg fentanyl, 1 to 1.5 mg/kglidocaine, and 2 to 2.5 mg/kg propofol. Laryngoscopy andET intubation were facilitated by administration of 0.6mg/kg rocuronium. Patients were tracheally intubatedwith a 7.0- and 6.5-mm inner diameter ET with a high-volume and low-pressure cuff (ILM Endotracheal Tube;Euromedical, Kedah, Malaysia), respectively. Intubationswere performed by residents with at least 2 years ofexperience or attending physicians who were blinded to thetreatment. The ET cuffs were inflated with room air to acuff pressure of 20 to 25 cm H2O. We measured the cuffpressure immediately after ET intubation using a manom-eter (VBM, Sulz, Germany) that was connected to the pilotballoon of the ET cuff via a 3-way stopcock, and the cuffpressure was measured once in each patient at 60 minutesafter intubation. Anesthesia was maintained with 8% to12% desflurane in a total flow of 300 mL/min oxygenunder a closed system without nitrous oxide. Fentanyl wasadministered depending on the surgical stimulus and thehemodynamic response; the dosage was approximately 1�g/kg/h after intubation. Neuromuscular transmissionwas monitored using train-of-four supramaximal stimula-tions (2 Hz, 50 mA) (TOF Watch SX; Organon, Dublin,Ireland). Intravenous boluses of cisatracurium 2 mg wereadministered when more than 2 responses were detected intrain-of-four stimulation until closure of the peritoneumwas commenced. At the end of surgery, residual neuro-muscular blockade was antagonized by neostigmine andatropine. When the T4/T1 ratio reached �90% and patientscould follow simple commands, the patients were trache-ally extubated after gentle oropharyngeal suction and thentransferred to the postanesthesia care unit.

When arriving at the postanesthesia care unit (0 hour)and thereafter at 2, 4, and 24 hours, patients were assessedfor the incidence and severity of sore throat by anotheranesthesiologist blinded to the treatment. POST wasgraded on a 4-point scale (0–3): 0, no sore throat; 1, mildsore throat (complaints of sore throat only whenprompted); 2, moderate sore throat (complaints of sorethroat volunteered by the patient without prompting); and3, severe sore throat (change of voice or hoarseness, asso-ciated with throat pain).11 Total POST cases were thosepatients who reported any degree of sore throat over the24-hour evaluation period. We recorded the patient’s age,sex, smoking history, weight, height, duration of surgery,

hoarseness, nausea and vomiting, and the type of postop-erative analgesia. Potential side effects associated with thestudy drugs, such as local numbness, burning, and/orstinging sensation, cough, and dry mouth, were also re-corded during spraying and postoperatively by partici-pants’ self report.

Before initiation of the study, a power calculation wasperformed to determine the required sample size based onour institute’s previous data. A minimum of 91 patientswas required in each group to detect a decreased incidenceof POST from 40% to 20% with a power of 80% and asignificance level of 0.05. To compensate for potentialdropouts, we enrolled 95 patients in each group. Statisticalanalysis was performed using SPSS for Windows version14 (SPSS, Chicago, IL). Data are expressed as mean (SD),and median with range or percentage. We used a 1-wayanalysis of variance to compare patients’ ages, heights,weights, and durations of surgery among groups. Theoverall incidence of POST and side effects among groupswere tested using �2 tests. For those significant variables inthe �2 test, we recalculated all possible six 2 � 2 �2 tests byapplying the Bonferroni inequality to adjust � level [i.e.,P(�2 � 6.97) � 0.05/6 � 0.0083 for 1 degree of freedom] toavoid type I error. A Kruskal-Wallis test followed by theDunn procedure was applied to compare the differences inthe severity of POST among groups. We used univariableand multivariable logistic regressions for evaluating theinteraction between 2 treatments (mucosa and cuff) and therelative risk, as odds ratio (OR) and 95% confidence inter-val (CI) of BH on POST. Probability values �0.05 wereconsidered statistically significant.

RESULTSThe characteristics of the study groups are shown in Table1; there were no significant differences among the 4 groupsin age, sex, height, weight, body mass index, and theduration of anesthesia. Two patients in groups B and Dreceived unanticipated nasogastric tubes, and the statisticalanalysis was performed without those patients.

Table 2 lists the incidence and severity of POST for thestudy groups at 0, 2, 4, and 24 hours postoperatively.Within the 24-hour period of evaluation, the overall inci-dence of POST (patients with any POST during 24-hourevaluation/patient numbers) in groups A, B, C, and D was23.2%, 13.8%, 14.7%, and 40.4%, respectively, mostly at thesecond hour after extubation. There were significantlyfewer incidences in groups B and C, but not in group Acompared with group D, for the total incidence of POST.The severity of POST was significantly higher in group Dcompared with groups B and C (P � 0.001).

No statistically significant differences of POST incidencewere found among patients who received BH spray only onthe ET cuff, only in the oropharyngeal cavity, or both,which indicated no additive or synergistic effects of appli-cation of BH (P � 0.088). Patients who received BH on theET cuff had a significantly lower risk of POST (OR: 0.36;95% CI: 0.21–0.60) compared with the placebo group.Those who received BH in the oropharyngeal cavity alsohad less risk of POST (OR: 0.62; 95% CI: 0.38–1.01) com-pared with the placebo group, although not significant.After adjusting for potential confounders (age, gender,

Topical Benzydamine Hydrochloride Attenuates Postoperative Sore Throat


smoking, and duration of anesthesia), the association be-tween treatments for the risk of developing POST did notchange substantially (Table 3).

The incidence of local numbness, burning, or stingingsensation in groups A, B, C, and D was 8.42%, 10.6%, 3.19%,and 1.06%, respectively. Only group B had a significantlyhigher incidence of local numbness, burning, or stingingsensation than group D (P � 0.05). There were 4, 4, 3, and5 patients who had hoarseness in groups A, B, C, and D,respectively. There was no significant difference in the

potential side effects relevant to BH, such as nausea orvomiting, cough, and mouth dryness (Table 4).

DISCUSSIONOur major finding was that spraying BH on the ET cuff mayreduce the incidence and severity of POST up to 24 hourspostoperatively compared with the application of distilledwater. Moreover, spraying BH in the oropharyngeal cavitydid not reduce the incidence and severity of POST. Indeed,there was no additional advantage from spraying BH on the

Table 1. Demographic Data and Duration of SurgeryGroup

A (n � 95) B (n � 94) C (n � 95) D (n � 94) P value

Gender (male/female) 49/46 44/50 45/50 44/50 0.919Age (y) 49.2 (18) 47.1 (15) 48.7 (16) 45.5 (18) 0.446Height (cm) 163.7 (8.0) 162.1 (8.8) 162.9 (9.4) 162.4 (9.6) 0.631Weight (kg) 63.2 (12.1) 62.9 (10.8) 62.0 (11.5) 62.7 (11.7) 0.897BMI (kg/m2) 23.5 (3.5) 23.9 (3.6) 23.3 (3.4) 23.7 (3.9) 0.655Duration of anesthesia (min) 175 (99) 186 (91) 176 (84) 175 (83) 0.917Intraoperative fentanyl use 318 (65) 312 (57) 310 (60) 306 (60) 0.706Smoking habit 35 40 38 42 0.964Postoperative analgesia (PCA with

fentanyl/demerol, prn)80/15 78/16 77/18 75/19 0.862

Type of surgeryColon rectal surgery 8 6 5 6 0.850General surgery 21 20 21 18 0.954Genitourinary surgery 4 5 7 5 0.815Gynecologic surgery 22 16 18 19 0.758Ophthalmologic surgery 12 16 16 18 0.673Orthopedic surgery 22 20 20 18 0.928Plastic surgery 6 11 8 10 0.585

BMI � body mass index; PCA � patient-controlled analgesia; prn � pro re nata (when necessary).Values are presented as mean (SD) or number.

Table 2. The Incidence and Severity of Postoperative Sore Throat for the Study GroupsEvaluation time Group A (n � 95) Group B (n � 94) Group C (n � 95) Group D (n � 94) P value

0 h 6 (6.3%)* 4 (4.2%)* 5 (5.2%)* 20 (20.6%) �0.001Grading of discomfort 0.001

Mild 4 2 3 12Moderate 2 2 2 8Severe 0 0 0 0

2 h 22 (23.2%) 13 (13.8%)* 14 (14.7%)* 38 (40.4%) �0.001Grading of discomfort �0.001


4 h 12 (12.6%)* 8 (8.3%)* 10 (10.4%)* 30 (30.9%) �0.001Grading of discomfort 0.001


24 h 10 (10.5%) 7 (7.3%)* 8 (8.3%)* 22 (22.7%) 0.003Grading of discomfort 0.039


Total incidence of POST 22 (23.2%) 13 (13.8%)* 14 (14.7%)* 38 (40.4%) �0.001Severity of POST 0 (0–3) 0 (0–3)* 0 (0–3)* 0 (0–3) �0.001

POST � postoperative sore throat.Values are presented as number of subjects (%) or median (range).* P � 0.05, compared with group D. The incidence of POST was done by �2 test by applying the Bonferroni inequality to adjust � level �i.e., P(�2 � 6.97) �0.05/6 � 0.0083 for 1 degree of freedom� to avoid type I error. The severity of POST was determined by Kruskal-Wallis test and pairwise posteriori comparisonswere done by Dunn procedure.


cuff and into the oropharyngeal cavity simultaneously forPOST prevention, even though there was a decreased trend ofPOST incidence when spraying BH on the ET cuff comparedwith into the oropharyngeal cavity. Our results indirectlydemonstrated that mucosal irritation occurring at the level ofthe ET cuff is the most important causative factor for trachealmorbidity, and this outcome was consistent with previousstudies.13,14

Local spraying of BH can resolve moderate to severe sorethroat after ET intubation.11,15 Preventive topical applicationof BH into the oropharyngeal cavity before intubation andcontinuous use for 48 hours after the operation effectivelydecreased the incidence and severity of POST after ET andalso laryngeal mask airway insertion.3,8 In this study, weshowed that spraying BH into the oropharyngeal cavity didnot reduce the incidence and severity of POST. This might bebecause of the small dosage of BH used in the oropharyngealcavity (0.75 mg) in our study compared with previous studies(which ranged from 2.16 to 22.5 mg).3,7

There are many side effects (64%) of topical use of BH,including local numbness, burning or stinging sensation,nausea or vomiting, cough, dry mouth, thirst, throat dis-comfort, drowsiness, and headache.15 Kati et al.8 andAgarwal et al.7 reported local numbness, dysgeusia, andthroat irritation in BH-pretreated patients. In our currentstudy, we found that 8% to 10% of patients experiencedthese side effects after topical application of BH beforeanesthesia induction. To avoid these drawbacks, we sug-gest that the application of BH onto the ET cuff instead of

into the oropharyngeal cavity before induction of anesthe-sia may ameliorate the local irritation of BH.

The first limitation in our study is the pharmacokineticsand safety of BH on the tracheal mucosa. BH is well absorbedthrough the oral mucosa, and its effects last for 1.84 hours.15 Inour study, BH may be well absorbed after application to thetrachea because of its thinner epithelial layer. However,concern for damage to the tracheal structure should beconsidered, even though we did not find any adverse event inour limited data. Silvestrini et al.16 had reported that grossand histological structure of the major organs such as lung,liver, kidney, and spleen were not altered by benzydamine inan animal study. However, no relevant human study regard-ing a histological examination of the trachea has been re-ported. A second limitation is that we did not ensure that eachgroup did in fact have an equal amount of BH, although welimited the dose to 5 sprays each on either the ET or into theoropharyngeal cavity, and an equal dosage was prescribed. Athird limitation of our randomized study is that postoperativepain management can influence POST, although there was nosignificant difference in postoperative pain management. Atthe same time, dry oxygen and IV lidocaine would affectPOST; however, we used these routinely and standardizedthe dosage for each patient.

In conclusion, BH topical spray on the ET cuff beforeintubation and GA can reduce the incidence and severity ofPOST without increased BH-related adverse effects.

REFERENCES1. Macario A, Weinger M, Carney S, Kim A. Which clinical

anesthesia outcomes are important to avoid? The perspectiveof patients. Anesth Analg 1999;89:652–8

2. Al-Qahtani AS, Messahel FM. Quality improvement in anes-thetic practice: incidence of sore throat after using smalltracheal tube. Middle East J Anesthesiol 2005;18:179–83

3. Gulhas N, Canpolat H, Cicek M, Yologlu S, Togal T, DurmusM, Ozcan Ersoy M. Dexpanthenol pastille and benzydaminehydrochloride spray for the prevention of post-operative sorethroat. Acta Anaesthesiol Scand 2007;51:239–43


5. Sumathi PA, Shenoy T, Ambareesha M, Krishna HM. Controlledcomparison between betamethasone gel and lidocaine jelly ap-plied over tracheal tube to reduce postoperative sore throat,cough, and hoarseness of voice. Br J Anaesth 2008;100:215–8

6. Ogata J, Minami K, Horishita T, Shiraishi M, Okamoto T,Terada T, Sata T. Gargling with sodium azulene sulfonatereduces the postoperative sore throat after intubation of thetrachea. Anesth Analg 2005;101:290–3

7. Agarwal A, Nath SS, Goswami D, Gupta D, Dhiraaj S, SinghPK. An evaluation of the efficacy of aspirin and benzydaminehydrochloride gargle for attenuating postoperative sore throat:a prospective, randomized, single-blind study. Anesth Analg2006;103:1001–3

8. Kati I, Tekin M, Silay E, Huseyinoglu UA, Yildiz H. Does benzy-damine hydrochloride applied preemptively reduce sore throatdue to laryngeal mask airway? Anesth Analg 2004;99:710–2

9. Hung NK, Wu CT, Chan SM, Lu CH, Huang YS, Yeh CC, LeeMS, Cherng CH. Effect on postoperative sore throat of spray-ing the endotracheal tube cuff with benzydamine hydrochlo-ride, 10% lidocaine, and 2% lidocaine. Anesth Analg 2010 Mar19 [Epub ahead of print]

10. Quane PA, Graham GG, Ziegler JB. Pharmacology of benzy-damine. Inflammopharmacology 1998;6:95–107

Table 3. Odds Ratios and 95% ConfidenceIntervals of Benzydamine Hydrochloride onPostoperative Sore Throat at 2 HoursAfter Surgery

No. of sorethroats/no.of patients

UnivariableOR (95% CI)

Multivariablea

OR (95% CI)BH on mucosa

Yes 36/189 0.63 (0.39–1.03) 0.65 (0.40–1.06)No 51/189 1.00 (reference) 1.00 (reference)

BH on cuffYes 27/189 0.36 (0.22–0.60)* 0.36 (0.22–0.60)*No 60/189 1.00 (reference) 1.00 (reference)

OR � odds ratio; CI � confidence interval; BH � benzydamine hydrochloride.a Adjusted for gender, age (in years), smoking, and anesthetic time.* P � 0.05 compared with the reference group.

Table 4. The Local and Systemic Side Effects ofBenzydamine Hydrochloride in the Study Groups

Group A(n � 95)

Group B(n � 94)

Group C(n � 95)

Group D(n � 94)

Nausea 20 25 22 25Vomiting 18 16 24 20Cough 19 22 20 18Local stinging or

numbness of thethroat and mouth

8 10* 3 1

Dry mouth 42 46 45 48Hoarseness 4 4 3 5

Values are presented as number.* P � 0.05 compared with group D. �2 test and applying the Bonferroniinequality to adjust � level �i.e., P(�2 � 6.97) � 0.05/6 � 0.0083 for 1degree of freedom� to avoid type I error.

Topical Benzydamine Hydrochloride Attenuates Postoperative Sore Throat


11. Turnbull RS. Benzydamine hydrochloride (Tantum) in themanagement of oral inflammatory conditions. J Can DentAssoc 1995;61:127–34

12. Epstein JB, Silverman S Jr, Paggiarino DA, Crockett S, SchubertMM, Senzer NN, Lockhart PB, Gallagher MJ, Peterson DE,Leveque FG. Benzydamine HCl for prophylaxis of radiation-induced oral mucositis: results from a multicenter, random-ized, double-blind, placebo-controlled clinical trial. Cancer2001;92:875–85


14. Peppard SB, Dickens JH. Laryngeal injury followingshort-term intubation. Ann Otol Rhinol Laryngol 1983;92:327–30

15. Passali D, Volonte M, Passali GC, Damiani V, Bellussi L.Efficacy and safety of ketoprofen lysine salt mouthwash versusbenzydamine hydrochloride mouthwash in acute pharyngealinflammation: a randomized, single-blind study. Clin Ther2001;23:1508–18

16. Silvestrini B, Barcellona PS, Garau A, Catanese B. Toxicology ofbenzydamine. Toxicol Appl Pharmacol 1967;10:148–59


Strepsils� Tablets Reduce Sore Throat andHoarseness After Tracheal IntubationAmin Ebneshahidi, MD, and Masood Mohseni, MD

BACKGROUND: Amyl-m-cresol (Strepsils�) has been successfully used in the prophylaxis andtreatment of oral inflammations, but its effects on postintubation sore throat and hoarseness areunknown. We conducted this study to evaluate the effects of Strepsils in reducing postintubationsore throat and hoarseness.METHODS: One hundred fifty patients, ASA physical status I to II, scheduled to undergo generalanesthesia and elective orthopedic or gynecologic surgery were enrolled. Participants wererandomly allocated to receive either Strepsils or identical-looking placebo tablets immediatelybefore arrival to the operating room. The incidence and severity of postoperative sore throat andhoarseness were evaluated immediately and 24 hours after surgery.RESULTS: The incidence of early postoperative sore throat was 13.7% and 33.3% andhoarseness was 12.3% and 26.4% in the Strepsils and placebo groups, respectively (P � 0.05).One day after surgery, the incidence of sore throat decreased to 6.8% and 18.1% in the Strepsilsand control groups, respectively. The incidence of hoarseness 1 day after the operationdecreased to 8.2% in the Strepsils group and 19.4% in the placebo group, but the differenceremained statistically significant (P � 0.05).CONCLUSION: Perioperative use of Strepsils tablets reduces postoperative sore throat andhoarseness of voice. (Anesth Analg 2010;111:892–4)

Postoperative sore throat and hoarseness are minor butfrequent complications of endotracheal intubation thatoccur in up to 50% of patients.1,2 These complications

negatively influence patient satisfaction and occasionally re-quire treatment with supplementary analgesics. Several drugssuch as clonidine,3 betamethasone gel,4 and chamomile-extract spray5 have been used to reduce postoperative sorethroat and hoarseness with varying efficacy. A simple, safe,and inexpensive perioperative intervention to prevent post-operative sore throat and hoarseness would be useful.

Amyl-m-cresol (Strepsils�) has been successfully used in thetreatment of oral inflammation and for the prevention of painand inflammation after oral surgery,6,7 but the effects on postin-tubation sore throat and hoarseness are unknown. This studywas conducted to evaluate whether the perioperative use ofStrepsils lozenges would reduce the incidence of postoperativesore throat and hoarseness compared with placebo.

METHODSPatientsThis prospective study was approved by the ethics com-mittee of Sadi Hospital, and written informed consent wasobtained from all patients. One hundred fifty ASA physicalstatus I to II patients, aged 19 to 63 years, undergoingelective orthopedic or gynecologic surgery under generalanesthesia who were expected to remain in the hospital for�24 hours were enrolled. Patients with significant sorethroat for any reason or obvious hoarseness were notincluded. Participants were randomly allocated to receive

either Strepsils or identical-looking placebo tablets imme-diately before arrival to the operating room (45 minutesbefore induction of anesthesia on average). A blinded nurseadministered the tablets to the patients.

Strepsils honey and lemon lozenge (Boots, Nottingham, UK)contained the active ingredients 2,4-dichlorobenzyl alcohol 1.2mg, amylmetacresol 0.6 mg, sucrose, glucose syrup, honey,tartaric acid, peppermint oil, terpeneless lemon oil, and quinolineyellow (E104). The placebo tablets (Anata, Tabriz, Iran) containedsugar, glucose, citric acid, lemon flavor, and color (E104). Thelozenges were not easily distinguishable from placebo tabletswith regard to color, taste, and smell. The randomization wasperformed by the hospital pharmacy using a table of randomnumbers, and the patients, nurses, and anesthetic technician whowere involved in the patients’ care and data recording wereblinded to the nature of the assignment.

Method of AnesthesiaAll patients were premedicated with oral oxazepam 10 mgand ranitidine 150 mg 2 hours before surgery. Fentanyl 3 to4 �g � kg�1 and IV lidocaine 1.5 mg � kg�1 were adminis-tered 3 to 5 minutes before tracheal intubation. After theadministration of 100% oxygen at 5 L min�1 for severalminutes, anesthesia was induced with propofol 1.5 to2 mg � kg�1 and atracurium (0.5 mg � kg�1). Laryngoscopywas performed 3 to 5 minutes after atracurium injectionwhen adequate neuromuscular blockade was confirmedusing train-of-four nerve stimulation. A train-of-four �10%was used as an indication of satisfactory neuromuscularblockade for intubation. Difficulty in laryngoscopy wasgraded as 1, no difficulty; 2, only posterior extremity ofglottis visible; 3, only epiglottis seen; and 4, no recognizablestructures. Endotracheal intubation was performed using apolyvinyl chloride endotracheal tube (Supa, Tehran, Iran)with 7- to 8-mm internal diameter for women and 7.5- to8.5-mm internal diameter for men. The high-volume, low-pressure cuff was inflated until no air leak could be heardwith peak airway pressure at 20 cm H2O. Cricoid pressure

From the Department of Anesthesiology, Sadi Hospital, Isfahan, Iran.

Accepted for publication November 29, 2009.

Supported by departmental funds.

Address correspondence and reprint requests to Amin Ebneshahidi, MD,Department of Anesthesiology, Sadi Hospital, Khodaverdi Alley, Sadi Blv,Isfahan, Iran. Address e-mail to [email protected].

Copyright © 2010 International Anesthesia Research SocietyDOI: 10.1213/ANE.0b013e3181d00c60


was not applied for intubation. Patients requiring nasogas-tric tube placement or those in whom laryngoscopy wasattempted more than once were withdrawn.

General anesthesia was maintained with propofol, fentanyl,and atracurium. Controlled mechanical ventilation with an initialtidal volume of 10 mL � kg�1 and respiratory frequency of 10breaths � min�1 was adjusted to maintain normocapnia.

Neuromuscular blockade was reversed with neostig-mine 0.05 mg � kg�1 and atropine 0.02 mg � kg�1.

When spontaneous ventilation was adequate and thepatients were able to follow verbal commands, trachealextubation was performed immediately after suctioning ofthe oropharynx at the discretion of the responsible anesthe-siologist. An oral airway was inserted before extubationand emergence in all patients.

Outcome MeasuresPostoperative sore throat and hoarseness were evaluatedwithin 20 minutes in the recovery room and 24 hours aftersurgery. At the time of the first evaluation, patients with aRamsay Sedation Score8 of 2 (cooperative, oriented, andtranquil) or 3 (responding to commands only) were in-cluded. The severity of sore throat was graded as follows: 0,no sore throat; 1, minimal; 2, moderate; 3 severe; and forhoarseness, the grading was as follows: 0, no hoarseness; 1,slight hoarseness; 2, severe hoarseness; 3, cannot speak be-cause of hoarseness. Additional analgesics were not adminis-tered until completion of the first evaluation. All scales werecompleted by a nurse blinded to the study groups.

Demographic data, perioperative steroid use, type ofsurgery, the time needed for laryngoscopy (time fromopening the mouth to placement of the endotracheal tube),intubation period (time from intubation to extubation), andgrades of laryngoscopic view as well as the incidence ofbucking were recorded by a blinded anesthetic technician.Patients’ discomfort after lozenge administration includingsubjective symptoms of allergy to lozenges, bad taste orsmell, and nausea were also recorded by the same techni-cian before induction of anesthesia.

Statistical AnalysisResults are presented as either mean (SD) or percentages, asappropriate. The incidences of sore throat and hoarsenesswere compared between the 2 groups using the Fisher exacttest. The scores of sore throat and hoarseness as well as gradeof laryngoscopic view were compared between the 2 groupswith the Mann-Whitney U test. The duration of laryngoscopyand intubation period were compared between groups by anindependent 2-sample t test. A nomogram based on estimatedstandardized difference was used to calculate the requiredsample size. With a power of 80% and � level of 0.05 for2-tailed statistical analysis, and estimated incidence of sorethroat of 0.351,9 and 0.10 in the 2 groups, a sample size of 70patients for each group was calculated as being appropriate.Correlation coefficient between grade of laryngoscopic viewand sore throat/hoarseness score in either group was assessedwith Spearman analysis. Statistical analysis was performedwith SPSS version 11.0 software (SPSS, Chicago, IL).

RESULTSOf 150 included participants, 5 patients were excluded be-cause of unfavorable Ramsay Sedation Scores in the recovery

room. The distribution of surgical procedures and positionsare shown in Table 1. Patient characteristics, grade of diffi-culty of laryngoscopy, and the duration of laryngoscopy(13.7% vs 33.3%, P � 0.81) and intubation (12.3% vs 26.4%,P � 0.14) were comparable between the 2 groups (Table 1).

None of the patients complained of any kind of discomfortafter using either Strepsils or placebo. The incidence of earlypostoperative sore throat in patients who received Strepsilswas about one-third compared with the placebo group (P �0.003). Likewise, hoarseness was reported less frequently inthe Strepsils than placebo group (P � 0.04). As shown inFigure 1, the incidence of both sore throat and hoarsenessdecreased at the 24-hour assessment after surgery, but thedifferences between the 2 groups remained statistically sig-nificant (P � 0.04). Further analysis showed that the mean �SD severity scores of early sore throat (0.52 � 0.85 vs 0.20 �0.57), late sore throat (0.22 � 0.50 vs 0.08 � 0.32), earlyhoarseness (0.36 � 0.65 vs 0.16 � 0.47), and late hoarseness(0.22 � 0.45 vs 0.08 � 0.27) in the placebo group weresignificantly higher than in the Strepsils group (P � 0.05).

None of the patients complained of grade 3 hoarseness(unable to speak because of hoarseness). Only 1 patient inthe Strepsils group and 3 patients in the placebo group had

Figure 1. Incidence of sore throat and hoarseness at recovery fromanesthesia and 24 hours postoperatively in the Strepsils (white) andplacebo (gray) groups. Data shown as percentages; P � 0.05 with �2 test.

Table 1. Group Baseline Characteristicsand Covariates

VariablesStrepsils(n � 73)

Placebo(n � 72)

Age (y)a 29 (9) 31 (10)Male sexb 43 (58.9) 40 (55.5)Weight (kg)a 58 (9) 57 (11)Duration of laryngoscopy (s)a 13.7 (3.1) 14.3 (3.6)Duration of intubation (min)a 149 (44) 158 (51)Grade of laryngoscopic viewc 2 (1,2) 2 (1,2)Type of surgeryb

Myomectomy/hysterectomy 9 (12.3) 7 (9.7)Ovarectomy 4 (5.5) 5 (6.9)Diagnostic laparoscopy 10 (13.7) 12 (16.6)Bone tumor resection 7 (9.5) 8 (11.1)Total hip replacement 4 (5.5) 3 (4.2)Traumatic fractures 31 (42.4) 29 (40.2)Shoulder surgery/arthroscopy 4 (5.5) 3 (4.2)Bone grafting 4 (5.5) 5 (6.9)

Positionb

Supine 55 (75.3) 52 (72.2)Lateral decubitus 14 (19.2) 17 (23.6)Sitting 4 (5.5) 3 (4.2)

a Data are presented as mean (SD).b Data are presented as n (%).c Data are presented as median (25%, 75%).


severe sore throat. Pearson correlation coefficient analysesbetween grade of laryngoscopic view and early sore throat(r � 0.13, P � 0.05), 24-hour sore throat (r � 0.10, P � 0.05),early hoarseness (r � 0.16, P � 0.05), and 24-hour hoarse-ness (r � 0.10, P � 0.05) showed no statistically significantresults. Reanalysis of the data after the exclusion of 6patients with perioperative steroid use yielded similarresults in all sets of analysis (data not shown).

DISCUSSIONSore throat and hoarseness are minor but common postop-erative complaints with an estimated incidence of 14.4% to90% for sore throat1,9 and 10% to 50.1% for hoarseness.1,10

In this study, the overall incidence of sore throat andhoarseness in the immediate postoperative period was20.8% and 18.6%, respectively. Both complications wereshown to be significantly reduced with the use of Strepsils.Likewise, 1-day follow-up of patients confirmed the effi-cacy of Strepsils tablets for the prevention of postoperativesore throat and hoarseness.

Several causal factors for sore throat and hoarsenessafter intubation have been reported, including sex, largetracheal tube size, cuff design,11,12 increased intracuff pres-sure by nitrous oxide,13 use of succinylcholine, and pro-longed laryngoscopy.14 In this study, patient characteristicsand duration of laryngoscopy and surgery were compa-rable between the 2 groups. We excluded the potentialeffects of difficult intubation on postoperative throat com-plications in our series and had a standardized protocol forinduction, tracheal intubation, and extubation of patients.Thus, it seems reasonable to conclude that the observeddifference in the outcome measures between the 2 groupscan be safely attributed to the favorable effects of Strepsils.

Prior studies have demonstrated that mucosal damagecan occur even after uneventful intubation for routinesurgery.12 Because of the potential role of inflammation inthe causation of tracheal morbidities, earlier investigatorshave suggested the use of inhaled and topical steroids.15–17

Gaspar et al.6 used Strepsils for the treatment of 22 patientswith oral inflammatory diseases and preoperatively in 20oral surgery cases. In this preliminary study, they reportedthat Strepsils may be effective in the prophylaxis andtreatment of oral inflammation. In a complementary survey of272 patients with either oral inflammatory diseases or treatedprophylactically for possible oral inflammation or infection,Gaspar et al.7 found that healing was shortened by 30% andpain and functional assessment were improved by 30% inpatients treated with Strepsils compared with control. Strep-sils tablets were well tolerated by all patients. Taken together,these data suggest that Strepsils is an effective antiinflamma-tory choice for mucosal damage in the orotracheal cavity. Inthis study, we demonstrated that postoperative sore throatand hoarseness were reduced by using Strepsils lozenges.Although we did not evaluate the effect of Strepsils on thepotential role of the inflammatory process in the generation ofthese adverse effects, it is likely that this is its mechanism ofaction based on the above literature.

Study LimitationsResponsiveness of patients in the immediate postoperativeperiod may be questioned. Although we excluded patients

with inappropriate sedation scores, fentanyl or propofol stillcirculating in the bloodstream immediately after surgery couldhave affected their ability to accurately perceive their throatsymptoms. However, standardized protocols for intubation andextubation of patients along with similar results in the later(24-hour) evaluation of patients support our initial findings.

One limitation of this study is that we did not evaluate patientsatisfaction, which could be an important indicator of the efficacyof our intervention. Finally, we did not evaluate the best timingfor the administration of Strepsils and the effects of repeated use,especially during the day after surgery.

In conclusion, the perioperative use of Strepsils lozengesmay help eliminate sore throat and hoarseness after inpa-tient surgery. The best timing of lozenge administrationand the efficacy of repeated use in patients with remainingsymptoms should be evaluated in further studies.

REFERENCES1. Maruyama K, Sakai H, Miyazawa H, Toda N, Iinuma Y,

Mochizuki N, Hara K, Otagiri T. Sore throat and hoarsenessafter total intravenous anaesthesia. Br J Anaesth 2004;92:541–3

2. Higgins PP, Chung F, Mezei G. Postoperative sore throat afterambulatory surgery. Br J Anaesth 2002;58:582–4

3. Maruyama K, Yamada T, Hara K. Effect of clonidine premedi-cation on postoperative sore throat and hoarseness after totalintravenous anesthesia. J Anesth 2006;20:327–30

4. Kazemi A, Amini A. The effect of betamethasone gel inreducing sore throat, cough, and hoarseness after laryngo-tracheal intubation. Middle East J Anesthesiol 2007;19:197–204

5. Kyokong O, Charuluxananan S, Muangmingsuk V, RodanantO, Subornsug K, Punyasang W. Efficacy of chamomile-extractspray for prevention of post-operative sore throat. J Med AssocThai 2002;85(suppl 1):S180–5

6. Gaspar L, Turi J, Toth BZ, Suri C, Vago P. [The use of benzyl alcoholand amyl-m-cresol (Strepsils) in the oral cavity: review of theliterature and first clinical experiences]. Fogorv Sz 1998;91:143–50

7. Gaspar L, Szmrtyka A, Turi J, Toth BZ, Suri C, Vago P, Sefer A, alHaj C. [Clinical experience with the use of benzylalcohol andamyl-m-cresol (Strepsils) in stomatological diseases]. Fogorv Sz2000;93:83–90

8. Ramsay MA, Savege TM, Simpson BR, Goodwin R. Controlledsedation with alphaxalone-alphadolone. Br Med J 1974;22:656–65

9. Lev R, Rosen P. Prophylactic lidocaine use preintubation: areview. J Emerg Med 1994;4:499–506

10. Christensen AM, Willemoes-Larsen H, Lundby L, JakobsenKB. Postoperative throat complaints after tracheal intubation.Br J Anaesth 1994;73:786–7

11. Jensen PJ, Hommelgaard P, Sondergaard P, Eriksen S. Sorethroat after operation: influence of tracheal intubation, intra-cuff pressure and type of cuff. Br J Anaesth 1982;54:453–7




15. Sumathi PA, Shenoy T, Ambareesha M, Krishna HM. Controlledcomparison between betamethasone gel and lidocaine jelly ap-plied over tracheal tube to reduce postoperative sore throat,cough, and hoarseness of voice. Br J Anaesth 2008;100:215–8

16. Ayoub MC, Ghobashy A, McGrimley L, Koch ME, Qadir S,Silverman DG. Wide spread application of topical steroids todecrease sore throat, hoarseness and cough after trachealintubation. Anesth Analg 1998;87:714–6

17. el-Hakim M. Beclomethasone prevents postoperative sorethroat. Acta Anaesthesiol Scand 1993;37:250–2

Strepsils� Reduces Sore Throat and Hoarseness


Inhaled Fluticasone Propionate ReducesPostoperative Sore Throat, Cough, and HoarsenessNasrin Faridi Tazeh-kand, MD, Bita Eslami, MPH, and Khadijeh Mohammadian, RN

BACKGROUND: Sore throat is a common complication after surgery. Postoperative cough andhoarseness can also be distressing to patients. We sought to determine the effect of an inhalersteroid on sore throat, cough, and hoarseness during the first 24 hours of the postoperativeperiod.METHODS: We enrolled 120 women with ASA physical status I or II and term singleton pregnancywho were scheduled for elective cesarean delivery under general anesthesia. Patients wererandomized into 2 groups: in the sitting position, group F patients received 500 �g inhaledfluticasone propionate via a spacer device during 2 deep inspirations, after arrival in the operatingroom, and group C had no treatment. The patients were interviewed by a blinded investigator forpostoperative sore throat, cough, and hoarseness at 1 and 24 hours after surgery.RESULTS: There were no significant differences in age, height, weight, body mass index,duration of surgery, intubation, and grade of laryngeal exposure between the 2 groups. Theincidence of sore throat, cough, and hoarseness was significantly lower in group F (3.33%,3.33%, and 3.33%) compared with the control group (36.67%, 18.33%, and 35%) (P � 0.05 forall comparisons), not only in the first postoperative hour but also 24 hours after surgery (13.33%,13.33%, and 25% in group F vs 40%, 41.67%, and 50% in the control group). The incidence ofmoderate and severe hoarseness in group F at the first hour was significantly less than thecontrol group (P � 0.05).CONCLUSIONS: Inhaled fluticasone propionate decreases the incidence and severity of postop-erative sore throat, cough, and hoarseness in patients undergoing cesarean delivery undergeneral anesthesia. (Anesth Analg 2010;111:895–8)

Sore throat is a common complication after surgery.Postoperative cough and hoarseness also can be dis-tressing to patients and affect patient satisfaction.

Many factors can contribute to postoperative sore throat,and the incidence has been found to vary with the methodof airway management.1 The incidence is the highest aftertracheal intubation (45.4%), whereas after laryngeal maskairway use the incidence is lower (17.5%) and much less(3.3%) when a facemask is used for the maintenance ofanesthesia.1–3 Female sex increases the incidence and sever-ity of postoperative sore throat.4

We sought to determine the effect of an inhaler steroidon sore throat, cough, and hoarseness during the first 24postoperative hours.

METHODSAfter receiving hospital ethics committee approval andinformed consent from subjects, we enrolled 120 womenwith ASA physical status I or II and term singleton preg-nancies who were scheduled for elective cesarean deliveryunder general anesthesia. The study was conducted in aprospective, randomized, and single-blinded manner, fromAugust to December 2008. Patients with a history ofperioperative sore throat and asthma, Mallampati grade�2, recent nonsteroidal antiinflammatory drug use, weight

�115 kg, diabetes mellitus, pregnancy-induced hyperten-sion, �2 attempts at intubation, and patients receiving steroidtherapy were excluded from the study. Patients were random-ized into the 2 groups with the help of a computer-generatedtable of random numbers.

In the sitting position, group F received 500 �g inhaledfluticasone propionate (Flixotide™ Evohaler™ 250 �g, GlaxoWellcome Production, Evreux, France) via a spacer deviceduring 2 deep inspirations, after their arrival in the oper-ating room, and group control (C) received no treatment. Ingroup F, the inhaler was given by an anesthesia nurse. Allpatients were given the impression they would receivemedication to reduce the incidence of sore throat. Theywere told that some of them would be randomized toreceive the medication by inhalation (F group) and some byinjection (C group). Those randomized to group C did notreceive any active drug.

Standard monitoring included noninvasive arterialblood pressure, pulse oximetry, electrocardiogram, andend-tidal carbon dioxide. Uterine displacement wasachieved by tilting the operating table to the left. Weinserted a wide-bore IV catheter into a forearm vein andstarted a slow infusion of Ringer solution. In all patients,after 4 minutes of administration of oxygen, rapid sequenceinduction with cricoid pressure was achieved using thio-pental 5 mg/kg and succinylcholine 1.5 mg/kg. The tra-chea was intubated with a soft seal cuffed sterile polyvinylchloride endotracheal tube with a standard cuff (SupaMedical Devices, Tehran, Iran) and an internal diameter of7 mm. The tracheal tube cuff was inflated until no airleakage could be heard with a peak airway pressure at 20cm H2O. The degree of laryngeal exposure was rankedfrom complete visualization of the vocal cords (grade I) to

From the Department of Anesthesiology, Roointan-Arash Hospital, TehranUniversity of Medical Sciences, Tehran, Iran.

Accepted for publication October 26, 2009.

Address correspondence and reprint requests to Nasrin Faridi Tazeh-kand,MD, Roointan-Arash Hospital, Rashid Ave., Tehranpars, Tehran, Iran.Address e-mail to [email protected].

Copyright © 2010 International Anesthesia Research SocietyDOI: 10.1213/ANE.0b013e3181c8a5a2


a partial view of the vocal cords (grade II), epiglottis only(grade III), and the inability to view even the epiglottis(grade IV).

Anesthesia was maintained with 50% nitrous oxide inoxygen and 0.5 minimum alveolar concentration of halo-thane. Atracurium was given as required for further musclerelaxation. After delivery and clamping of the umbilicalcord, we administered 2 �g/kg fentanyl and 0.02 mg/kgmidazolam and 10 IU oxytocin IV. At the end of surgery,oxygen 100% was administered, and residual neuromuscu-lar block was antagonized using neostigmine and atropine.Oral suctioning was done just before extubation only. Thetrachea was extubated after deflating the cuff when thepatient was fully awake and was following commands. Allpatients received oxygen by a facemask after surgery. Theanesthesiologist intubating and providing care did notknow whether a patient had been allocated to either the For C group. The patients were interviewed by a blindedinvestigator for postoperative sore throat, cough, andhoarseness at 1 and 24 hours after surgery, using thequestionnaire based on the scoring system in Table 1.

Based on the results of a pilot study with 30 patients ineach group that showed an incidence of problems (sorethroat, cough, and hoarseness) of 56% in group F and 83%in group C, we calculated that 60 patients would berequired in each group to detect a difference in the inci-dence with a power of 90% and � � 0.05 by using the EpiInfo Web site (www.cdc.gov/epiinfo/). Statistical analysiswas performed with JMP software (version 4, SAS Institute,Cary, NC). Statistical significance for differences was testedby Student t test and �2 test when appropriate. A P value�0.05 was considered statistically significant.

RESULTSAs shown in Table 2, there were no significant differencesin age, height, weight, body mass index, duration ofsurgery, intubation, and grade of laryngeal exposure be-tween the 2 groups. The total incidence of sore throat,cough, and hoarseness was significantly lower in group Fcompared with group C (P � 0.05) not only in the firstpostoperative hour but also 24 hours after surgery (Table3). During the first postoperative hour, none of the patients

in group F had moderate or severe (grades 2 and 3) sorethroat, cough, or hoarseness. However, in group C, mod-erate and severe symptoms were reported in 3 patients forsore throat, 3 for cough, and 5 for hoarseness. The incidenceof combined moderate and severe hoarseness in group F inthe first hour was significantly lower than in group C (P �0.05). Meanwhile, the evaluation of complications 24 hoursafter surgery showed that in group F, only 2 patients hadmoderate or severe sore throat, and 2 patients had moder-ate or severe hoarseness. None of the patients in group Fhad moderate or severe cough after 24 hours. However,moderate and severe complications were common in groupC and when combined, moderate and severe complicationswere significantly more frequent than in group F (P � 0.05).

DISCUSSIONWe found that the incidence of postoperative sore throat,cough, and hoarseness was significantly less when inhaledfluticasone (500 �g) was administered compared with notreatment. Some studies found the incidence of postopera-tive sore throat, cough, and hoarseness to be as high as 6.6%to 90%. The results of our study showed that the incidenceof these problems in group C was higher than in group F(40% vs 13.3% for cough; 41.67% vs 13.3% for sore throat;and 50% vs 25% for hoarseness) after 24 hours. Manyfactors including airway management, female sex, youngerpatients, gynecological procedure, and succinylcholine ad-ministration predict postoperative sore throat.3 All patientsin this trial were young females and were candidates forcesarean delivery. Therefore, age and sex were eliminatedas possible confounding factors.

Researchers recognizing the potential role of inflammationin these postoperative airway sequelae have described the useof inhaled and topical steroids.5–7 Stride8 concluded that 1%hydrocortisone water-soluble cream was ineffective forreducing the incidence of postoperative sore throat.Sumathi et al.9 showed that the widespread application ofbetamethasone gel on the tracheal tube decreased theincidence and severity of postoperative sore throat, cough,and hoarseness. The differences in the findings betweenthese 2 studies may be attributable to the fact that Stride

Table 1. Scoring System for Sore Throat, Cough,and HoarsenessScore

Sore throat0 No sore throat1 Mild (less than a common cold)2 Moderate (similar to a common cold)3 Severe (more than a common cold)

Cough0 No cough1 Mild (less than a common cold)2 Moderate (similar to a common cold)3 Severe (more than a common cold)

Hoarseness0 No hoarseness1 Mild (no hoarseness at the time of interview but had

it previously)2 Moderate (is only perceived by the patient)3 Severe (recognizable at the time of interview)

Table 2. Characteristics of Study PopulationFluticasonepropionate

group(n � 60)

Controlgroup

(n � 60) PAge (y) 26.32 � 5.18 26.70 � 5.26 NSHeight (cm) 162.03 � 4.90 161.39 � 4.99 NSWeight (kg) 79.18 � 10.91 76.79 � 12.16 NSBody mass index

(kg/m2)30.09 � 3.41 29.42 � 4.13 NS

Duration of surgery(min)

39.92 � 8.36 40.67 � 11.55 NS

Duration of trachealintubation (min)

53.83 � 8.65 54 � 11.96 NS

Grade of laryngealexposure

I 43 (71.67) 43 (71.67)II 15 (25) 17 (28.33)III 2 (3.33) 0 (0)V 0 (0) 0 (0) NS

NS � not significant.

Postoperative Sore Throat and Fluticasone


lubricated the tube only to the 5-cm mark, whereas Sumathiet al. lubricated the tube to the 15-cm mark. Thus, becauseof more widespread application of steroid gel to the tube,more gel came in contact with the posterior pharyngealwall, vocal cords, and trachea and was not just confined tothe tip and cuff of the tracheal tube.

Our study showed that inhaled steroid could be simi-larly effective in decreasing postoperative sore throat,cough, and hoarseness, and is similar to the study bySumathi et al.9 In the 2 studies, the patient populations andthe type of surgery were different. Their patients wereeither sex, aged between 18 and 50 years, and undergoingelective surgery.

Inhaled fluticasone delivers the drug in smaller dosesand in a shorter time to the patient’s airway compared withwidespread lubrication of the tube with betamethasone,which may increase the dose of drug that comes in contactwith the mucosa of the oropharynx, larynx, and trachea,resulting in higher systemic absorption and a possibleaggravation of local subtle infection, especially in pregnantpatients.

Coughing, wheezing, and shortness of breath are symp-toms of asthma. Asthma treatment includes inhaled bron-chodilators, which reverse airflow obstruction, and inhaledcorticosteroids to prevent asthma exacerbations by damp-ing the inflammatory processes that underlie asthma at-tacks. Inhaled fluticasone propionate is a relatively newinhaled corticosteroid for the treatment of asthma.

In many studies, fluticasone propionate was used forimproved outcomes in children or adults who were at riskof asthma. Treatment with fluticasone propionate or acombination of this drug with a long-acting �-2 agonistsuch as salmeterol was significantly effective in asthmaticpatients.10–12 A study of the safety of intranasal corticoste-roids such as fluticasone propionate did not identify anysystemic adverse events, which suggests that this drug canbe safely administered.13

Some studies evaluated the effect of treatment withfluticasone propionate nasal spray in rhinitis and inhaledfluticasone during pregnancy. The authors did not reportany adverse effects on maternal and fetal health.14–16 It is

therefore unlikely that a single dose of fluticasone willaffect the fetus. In this study, we used fluticasone propi-onate in healthy women and did not observe any sideeffects.

A limitation of our study was the infrequency of datacollection, which we could also have performed at 6 and 12hours. Also, we did not use an inert inhaler in the controlgroup, leading to potential bias.

In conclusion, inhaled fluticasone propionate (500 �g)decreases the incidence and severity of postoperative sorethroat, cough, and hoarseness in patients undergoing cesar-ean delivery under general anesthesia.

REFERENCES1. McHardy FE, Chung F. Postoperative sore throat: cause, pre-

vention and treatment. Anaesthesia 1999;54:444–532. Joshi GP, Inagaki Y, White PF, Taylor-Kennedy L, Wat LI,

Gevirtz C, McCraney JM, McCulloch DA. Use of laryngealmask airway as an alternative to the tracheal tube duringambulatory anesthesia. Anesth Analg 1997;85:573–7


4. Maruyama K, Sakai H, Miyazawa H, Tida N, Iinuma Y,Mochizuki N, Hara K, Otagiri T. Sore throat and hoarsenessafter total intravenous anaesthesia. Br J Anaesth 2004;92:541–3

5. Ayoub MC, Ghobashy A, McGrimely L, Koch ME, Gadir S,Silverman DG. Wide spread application of topical steroids todecrease sore throat, hoarseness and cough after trachealintubation. Anesth Analg 1998;87:714–6

6. Selvaraj T, Dhanpal R. Evaluation of the application of topicalsteroids on the endotracheal tube in decreasing postoperativesore throat. J Anaesthesiol Clin Pharmacol 2002;18:167–70

7. El-Hakim M. Beclomethasone prevents postoperative sorethroat. Acta Anaesthesiol Scand 1993;37:250–2

8. Stride PC. Postoperative sore throat: topical hydrocortisone.Anaesthesia 1990;45:968–71


10. Bacharier LB, Guilbert TW, Zeiger RS, Strunk RC, Morgan WJ,Lemanske RF Jr, Moss M, Szefler SJ, Krawiec M, Boehmer S,Mauger D, Taussig LM, Martinez FD. Patient characteristicsassociated with improved outcomes with use of an inhaledcorticosteroid in preschool children at risk for asthma. JAllergy Clin Immunol 2009;123:1077–82

Table 3. Postoperative Complications by Grade After 1 and 24 hAfter 1 h After 24 h

Fluticasonepropionate group

Controlgroup P

Fluticasonepropionate group

Controlgroup P

Sore throat1 2 (3.33) 19 (31.67) 6 (10) 15 (25)2 0 (0) 2 (3.33) 2 (3.33) 7 (11.67)3 0 (0) 1 (1.67) 0 (0) 2 (3.33)

Total 2 (3.33) 22 (36.67) �0.0001 8 (13.33) 24 (40) 0.001Cough

1 2 (3.33) 8 (13.33) 8 (13.33) 14 (23.33)2 0 (0) 3 (5) 0 (0) 9 (15)3 0 (0) 0 (0) 0 (0) 2 (3.33)

Total 2 (3.33) 11 (18.33) 0.008 8 (13.33) 25 (41.67) 0.005Hoarseness

1 2 (3.33) 16 (26.67) 13 (21.67) 23 (38.33)2 0 (0) 4 (6.67) 2 (3.33) 4 (6.67)3 0 (0) 1 (1.67) 0 (0) 3 (5)

Total 2 (3.33) 21(35) �0.0001 15 (25) 30 (50) 0.005


11. de Blic J, Ogorodova L, Klink R, Sidorenko I, Valiulis A, HofmanJ, Bennedbaek O, Anderton S, Attali V, Desfougeres JL, PoterreM. Salmeterol/fluticasone propionate vs. double dose flutica-sone propionate on lung function and asthma control inchildren. Pediatr Allergy Immunol. 2009;20:763–71

12. Markham A, Jarvis B. Inhaled salmeterol/fluticasone propi-onate combination: a review of its use in persistent asthma.Drugs 2000;60:1207–33

13. Demoly P. Safety of intranasal corticosteroids in acute rhino-sinusitis. Am J Otolaryngol 2008;29:403–13

14. Choi S, Han JY, Kim MY, Velazques-Armenta EY, Nava-OcampoAA. Pregnancy outcomes in women using inhaled fluticasoneduring pregnancy: a case series. Allergol Immunopathol 2007;35:239–42

15. Rahimi R, Nikfar S, Abdollahi M. Meta-analysis finds use ofinhaled corticosteroids during pregnancy safe: a systematicmeta-analysis review. Hum Exp Toxicol 2006;25:447–52

16. Ellegård EK, Hellgren M, Karlsson NG. Fluticasone propionateaqueous nasal spray in pregnancy rhinitis. Clin OtolaryngolAllied Sci 2001;26:394–400

Postoperative Sore Throat and Fluticasone


International Society for Anaesthetic Pharmacology

Anesthetic Pharmacology and Preclinical Pharmacology Section Editor: Marcel E. Durieux

Clinical Pharmacology Section Editor: Tony Gin

Repinotan, a Selective 5-HT1A-R-Agonist, AntagonizesMorphine-Induced Ventilatory Depression inAnesthetized RatsU. Guenther, MD,* H. Wrigge, PhD,* N. Theuerkauf, MD,* M. F. Boettcher, MD,† G. Wensing, MD,†J. Zinserling, PhD,* C. Putensen, PhD,* and A. Hoeft, PhD*

BACKGROUND: Spontaneous breathing during mechanical ventilation improves arterial oxygen-ation and cardiovascular function, but is depressed by opioids during critical care. Opioid-inducedventilatory depression was shown to be counteracted in anesthetized rats by serotonin(1A)-receptor (5-HT1A-R)-agonist 8-OH-DPAT, which cannot be applied to humans. Repinotan hydro-chloride is a selective 5-HT1A-R-agonist already investigated in humans, but the effects onventilation and nociception are unknown. In this study, we sought to establish (a) the effects ofrepinotan on spontaneous breathing and nociception, and (b) the interaction with the standardopiate morphine.METHODS: The dose-dependent effects of repinotan, given alone or in combination withmorphine, on spontaneous minute ventilation (MV) and nociceptive tail-flick reflex latencies(TFLs) were measured simultaneously in spontaneously breathing anesthetized rats. An addi-tional series with NaCl 0.9% and the 5-HT1A-R-antagonist WAY 100 135 served as controls.RESULTS: (a) Repinotan dose-dependently activated spontaneous breathing (MV, mean [95%confidence interval]; 53% [29%–77%]) of pretreatment level) and suppressed nociception (TLF,91% maximum possible effect [68%–114%]) with higher doses of repinotan (2–200 �g/kg). Onthe contrary, nociception was enhanced with a small dose of repinotan (0.2 �g/kg; TFL, �47%maximum possible effect [�95% to 2%]). Effects were prevented by 5-HT1A-antagonist WAY 100135. (B) Morphine-induced depression of ventilation (MV, �72% [�100% to �44%]) wasreversed by repinotan (20 �g/kg), which returned spontaneous ventilation to pretreatment levels(MV, 18% [�40% to 77%]). The morphine-induced complete depression of nociception wassustained throughout repinotan and NaCl 0.9% administration. Despite a mild decrease in meanarterial blood pressure, there were no serious cardiovascular side effects from repinotan.CONCLUSIONS: The 5-HT1A-R-agonist repinotan activates spontaneous breathing in anesthetizedrats even in morphine-induced ventilatory depression. The potency of 5-HT1A-R-agonists to stimulatespontaneous breathing and their antinociceptive effects should be researched further. (Anesth Analg2010;111:901–7)

Opioids are potent analgesics, but their clinical admin-istration is limited by the intrinsic risk of fatal apnea.Hence, pain therapy involving opioids must be bal-

anced against respiratory depression.1 The serotonin(1A)-receptor (5-HT1A-R)-agonist buspirone has been shown tostimulate spontaneous breathing in cats2 and to overcomeneurogenic breathing disturbances in humans.3,4 Another5-HT1A-R-agonist, 8-hydroxy-2-(di-n-propylamino)tetralin(8-OH-DPAT), a substance not approved for use in humans,

has been demonstrated to overcome opioid-induced ventila-tory depression in anesthetized rats.5

Nociception, another target of 5-HT1A-R-agonists, wasreported either to be depressed6–8 or enhanced by 5-HT1A-R-agonists.9,10 Later, 5-HT1A-R-agonist F13640 was foundto exert a dual effect, hyperalgesic and analgesic, depend-ing on plasma and brain concentrations.11 Recently,enhancement of nociceptive reflexes by small doses of8-OH-DPAT and suppression by higher doses were con-firmed in 2 different experimental models.12

Repinotan (R-(�)-2-{4-[(chroman-2-ylmethyl)-amino]-butyl}-1,1-dioxobenzo[d]-isothiazolone hydrochloride) is ahighly effective, selective, full 5-HT1A-R-agonist.13,14 Unlikeother 5-HT1A-R-agonists, repinotan is approved for IV use inhumans and has already undergone a series of clinical inves-tigations into the effects of neuroprotection after traumaticbrain injury and stroke.15–18 However, its effects on nocicep-tion and ventilation are not yet established. The aim of thisstudy was to verify the effects of repinotan on spontaneousbreathing and nociception simultaneously, and to determinethe interaction with the standard opiate morphine on sponta-neous breathing and nociception in anesthetized rats. Two

From the *University Hospital of Bonn, Clinic of Anaesthesiology andIntensive Care Medicine, Bonn; and †Department of Pharmacological Re-search, Bayer Schering Pharma AG, Wuppertal, Germany.


Supported by Bayer Schering Pharma AG, Germany, and departmentalfunding.

Address correspondence and reprint requests to Ulf Guenther, MD, Univer-sity Hospital of Bonn, Clinic of Anaesthesiology and Intensive Care Medi-cine, Sigmund-Freud-Strasse 25, 53105 Bonn, Germany. Address e-mail [email protected].

Copyright © 2010 International Anesthesia Research SocietyDOI: 10.1213/ANE.0b013e3181eac011


hypotheses were tested: (1) repinotan at a dose to stimulatespontaneous breathing does not enhance a nociceptive reflex,and (2) repinotan antagonizes morphine-induced depressionof spontaneous breathing.

METHODSAnimalsThis study was performed with approval from the localInstitutional Animal Review Board for animal research and inaccordance with the “Guide for the Care and Use of Labora-tory Animals.” Animals were housed in standard laboratoryconditions with a 12-hour light/dark schedule and free accessto food and water. Thirty-nine male Sprague-Dawley ratsweighing 260 g (244–277 g) (mean, 95% confidence interval[CI]) were deeply anesthetized with sodium-pentobarbitone(60 mg/kg) intraperitoneally and placed supine on a heatingpad to maintain rectal temperature constantly at 37°C �0.5°C. The right inguinal vessels were cannulated via asmall surgical incision for continuous monitoring ofarterial blood pressure and systemic administration ofstudy drugs. Anesthesia was maintained with sevoflu-rane, leveled at an inspiratory concentration of 1.5 to 2.5vol% to ensure immobility, stable spontaneous breath-ing, and detectable tail-flick reflex (TFR).

MeasurementsAnimals breathed spontaneously via a tracheotomy tube(inner diameter, 1.2 mm). The expired air was led throughthe flowhead (order number, MLT1L; ADInstrumentsGmbH, Spechbach, Germany) of a spirometer (ML141)connected to an A/D-interface (PowerLab 4/25®; all de-vices from ADInstruments GmbH) to record respiratoryrate (RR) and tidal volume (Vt) by integration of ventila-tory airflow over time. Minute ventilation (MV) was calcu-lated as MV [mL/min] � RR [1/min] � Vt [mL].

The TFR was evoked by a 100-W light beam sourcemounted 15 mm over the base of the tail to reachmaximum temperature within a second. The latency ofthe reflex response (TFR latency, TFL) was recorded witha strain gauge attached to the tail distal to the heatingspot. A shortened TFL indicates enhanced nociceptiveresponsiveness; an elongated TFL indicates depressednociception. Heating was stopped when the tail flicked orafter a maximum heating time of 15 seconds (TFLoffset) toprevent damage to the tail. TFLs were calculated as changein percent of the maximum possible effect [% MPE] accord-ing to the formula8: % MPE � 100 � [TFLtreatment �TFLpretreatment] � [TFLoffset � TFLpretreatment]

�1). A 100%MPE means complete suppression of nociception. Threesweeps were recorded and averaged. A blood pressuretransducer, temperature probe, and strain-gauge trans-ducer were also connected to the same A/D-interface suchas the spirometer (PowerLab 4/25®; ADInstrumentsGmbH).

Drug Administration ProtocolsTwo different sets of experiments were performed: (a) thefirst set was aimed at determining the effects of repinotanon spontaneous breathing, and (b) the second set assessedinteractions of repinotan with the opiate morphine (see Fig.1 for schematic overview).

RepinotanRepinotan was injected IV every 15 minutes with dosesranging from 0.02 through 200 �g/kg (Fig. 1A). The doseswere chosen because it was concluded from preliminarydose-finding experiments that the 20 �g/kg dose was themost efficient to counteract opioid-induced ventilatorydepression. The wide range of dosage was necessary toverify whether repinotan also possesses dose-dependentpro- and antinociceptive effects similar to the standard5-HT1A-R-agonist, 8-OH-DPAT. The number of experi-ments involving repinotan (n � 8) was chosen based on ourprevious experience with 8-OH-DPAT, in which the small-est effective dose increased spontaneous MV by 46% with astandard deviation of 35%. With an � set at 0.05, the powerwas calculated as 0.93 with n � 8 experiments in thedouble-sided power analysis.

Before the first drug administration, a series of 3 TFLsweeps was averaged and taken as the pretreatment level.Subsequent TFLs were taken 10 minutes after each drugadministration. For control experiments, NaCl 0.9% wasinjected every 15 minutes instead of study drugs (n � 8). Inanother series of 4 experiments, the selective 5-HT1A-R-

NaCl 0.9%

NaCl 0.9%

NaCl 0.9%

NaCl 0.9%

NaCl 0.9%

Repinotan

Repinotan

Repinotan

Repinotan

Repinotan

WAY 100135

Repinotan

Repinotan

Repinotan

Repinotan

0.2µg/kg

2µg/kg

20µg/kg

200µg/kg

1µg/kg0.02µg/kg

0.2µg/kg

2µg/kg

20µg/kg

200µg/kg

n=8 n=4 n=8

Morphine12[6-18]mg/kg

Repinotan

Repinotan

Repinotan

Repinotan

0.2µg/kg

2µg/kg

20µg/kg

200µg/kg

n=8

0.02µg/kg

60µg/kg

Repinotan

Repinotan

Repinotan


n=5


NaCl 0.9%

NaCl 0.9%

NaCl 0.9%

NaCl 0.9%

n=6

A Repinotan

B Repinotan / Morphine

6µg/kg

Figure 1. A, Serotonin(1A)-receptor (5-HT1A-R)-agonist repinotan wasinjected every 15 minutes at increasing doses. In a second series,the selective 5-HT1A-R-antagonist WAY 100 135 was given beforerepinotan. A third series involving only NaCl 0.9% served as controls.B, Morphine was given at increments of 5 mg/kg until a targetdepression of respiratory rate of �50% was established. The meanrequired morphine dosing is given as mean (95% confidence inter-val). Repinotan was given in 2 distinct series to cover all dosesrequired to delineate the top of the bell-shaped dose-response curveand to maintain experiments at comparable length.

Repinotan Antagonizes Ventilatory Depression


antagonist WAY 100 135 (1 mg/kg) was injected beforeadministration of repinotan (n � 4).

Repinotan/Morphine CoadministrationMorphine was injected at increments of 5 mg/kg untilrespiratory frequency was depressed to at least 50% of thepretreatment level. Thereafter, repinotan was added cumu-latively at the same doses as in the first set (Fig. 1B). Controlexperiments were again performed by injection of 200 �L ofNaCl 0.9% (n � 6). After completion of this series, theventilatory dose-response curve had a bell shape. To delin-eate the top of the bell shape more precisely, 5 additionalexperiments were performed with repinotan concentra-tions that were between the initially intended measuringpoints (Fig. 1). To verify integrity of the TFR at the end ofexperiments, naloxone (1 mg/kg) was given (not shown).

Statistical AnalysisAll data were tested for normal distribution (Kolmogorov-Smirnov test). Pretreatment levels of matched groups werecompared with the Student t test. All ventilatory variableswere calculated as change in percent of pretreatment level(% change). Results of experiments without morphine werecompared with the values before the first administration ofstudy drugs (pretreatment level). Results of experimentsinvolving morphine/repinotan coadministration were com-pared with the variables obtained after morphine administra-tion. The results of the ventilatory experiments were analyzedby 1-way repeated-measures analysis of variance. Each repi-notan concentration was compared with the pretreatmentlevel (i.e., 0% change) by Dunnett multiple comparison posttest.19 TFLs were calculated as change in % MPE as statedabove and also analyzed by comparison of each drug concen-tration to pretreatment levels by Dunnett multiple compari-son test. Data were processed with the Chart 4.0 and Scope 4.0software package (ADInstruments GmbH); statistical analyseswere performed using Prism4® software package for Macin-tosh (GraphPad Software Inc., San Diego, CA). Power analy-ses were done with the Simple Interactive Statistical Analysis(SISA) online software package (http://www.quantitativeskills.com/sisa/calculations/power.htm).

DrugsRepinotan was provided by the manufacturer, BayerHealthcare AG (Wuppertal, Germany). Morphine-sulfatewas purchased from Merck KG (Darmstadt, Germany),with permission from the German Institute for Pharmacyand Medical Products. WAY 100 135 was obtained fromTocris (Bristol, UK). Naloxone-HCl was purchased fromRatiopharm (Ulm, Germany). All compounds were dilutedin isotonic saline at the respective concentrations; injectionvolumes were 200 �L.

RESULTSEffects of RepinotanSpontaneous VentilationThe mean pretreatment RR was 58 breaths/min (95% CI,46–70 breaths/min) in the repinotan group versus 61breaths/min (54–69 breaths/min) in the control group, andthe mean arterial blood pressure (MAP) was 111 mm Hg(100–122 mm Hg) in the repinotan group versus 110 mm

Hg (100–120 mm Hg) in the control group. The mean TFLwas 7 seconds (5–9 seconds) in both groups. All data werenormally distributed, and pretreatment values did notdiffer statistically. Repinotan dose dependently increasedspontaneous MV (Fig. 2A), reaching the maximum effect(MV; 53% [29%–77%]) with the 200 �g/kg dose.

Tail-Flick Reflex LatencyThe repinotan effects on nociception were dose dependent:a small dose of repinotan (0.2 �g/kg) shortened TFL,whereas a high dose of repinotan (200 �g/kg) elongatedTFL (Fig. 2B), meaning that nociception was enhanced withsmall doses and suppressed with higher doses.

ControlsThe 5-HT1A-antagonist WAY 100 135 (1 mg/kg, n � 4) didnot significantly alter MV itself, but prevented repinotanfrom activating spontaneous MV (data not shown). Injec-tions of NaCl 0.9% had neither detectable effects on venti-lation nor nociception.

A

B

0.02 0.2 2 20 200Repinotan (µg/kg)

RepinotanNaCl 0.9%

0.02 0.2 2 20 200Repinotan (µg/kg)

RepinotanNaCl 0.9%

Figure 2. Effects of the serotonin(1A)-receptor (5-HT1A-R)-agonistrepinotan (n � 8) on spontaneous minute ventilation (MV) andtail-flick reflex (TFR) latencies (TFLs). Control experiments wereperformed with NaCl 0.9% (n � 8). Values of MV are given as percentchange from pretreatment level (% change), results of TFLs areshown as percent of maximum possible effect (% MPE; mean [95%confidence interval]). *P � 0.05, ***P � 0.001, compared withpretreatment level, 1-way repeated-measures analysis of variance.A, Repinotan dose dependently increased MV to a maximum of 53%(10%–29%) above the pretreatment level with the 200 �g/kg dose.B, Repinotan effects on nociception were dose dependent. Initialshortening of TFL (�47 [�95 to 2], indicating enhanced nociceptiveresponsiveness) with small doses (0.2 �g/kg) was followed byelongation of TFL to 91% MPE (68%–114% MPE) with the highestdose (200 �g/kg), meaning that TFR was profoundly suppressed.


Repinotan/Morphine CoadministrationFigure 3 shows a representative experiment on the antago-nization of morphine-induced ventilatory depression, dur-ing which nociceptive TFR remains completely depressed by

the sustained action of morphine. Morphine depressed spon-taneous MV to �72% (�100% to �44%) and always abolishedthe TFR with the first dosing increment (Fig. 4). Morphineconsumption did not differ between groups (Fig. 1).

Spontaneous BreathingRepinotan (after morphine) dose dependently activated spon-taneous breathing (Fig. 4), resulting in a maximum MV of 18%(�40% to 77%) above the pretreatment level with the 20�g/kg dose (P � 0.01, compared with morphine level).Further increases to doses �20 �g/kg returned MV to lowerlevels, giving the dose-response curve a bell shape.

NociceptionTFR was abolished with the first bolus of morphine (TFL,100% MPE), and remained completely suppressed through-out administrations of repinotan (n � 13) and control drugs(n � 6; P � 0.001, compared with pretreatment levels).Naloxone-HCl at the end of experiments verified the integ-rity of the TFR, because it returned with a TFL of 8% MPE(�18% to 33% MPE) (not shown).

Cardiovascular Side EffectsRepinotan depressed MAP with higher doses (Table 1, n �8), but this did not have deleterious effects on the experi-ments. Likewise, although morphine (12 mg/kg [8–16mg/kg], n � 13) markedly depressed MAP, repinotan didnot further aggravate arterial hypotension, and neither didNaCl 0.9% (Table 1). There were no serious cardiovascularcomplications.

DISCUSSIONThis study was performed to clarify whether the 5-HT1A-R-agonist repinotan also antagonizes opioid-induced ven-tilatory depression similar to other 5-HT1A-R-agonists. Itwas verified that (a) higher doses (2–200 �g/kg) of repino-tan stimulated spontaneous breathing, small doses (0.2�g/kg) enhanced nociception, and the highest dose (200�g/kg) depressed nociception; and (b) morphine-induceddepression of spontaneous breathing was antagonized byhigher doses of repinotan, whereas depression of nocicep-tion persisted. Despite a mild depression of MAP, repino-tan did not produce serious cardiovascular complications.

These findings confirm previous work in which the5-HT1A-R-agonist 8-OH-DPAT was shown to antagonizean opioid-induced ventilatory depression without impair-ing antinociception.12 8-OH-DPAT, however, is not ap-proved for human use. Buspirone, the only commerciallyavailable 5-HT1A-R-agonist for use in humans, has beenshown to stabilize apneustic breathing disturbances.3 Bu-spirone, being only a partial 5-HT1A-R-agonist, failed tocounteract a morphine-induced ventilatory depression inhealthy volunteers,20 nor did it cause antinociceptive ef-fects in healthy volunteers.21 In a direct comparison with8-OH-DPAT, only a weak ventilatory stimulation in coad-ministration with fentanyl was found for buspirone in an insitu perfused brainstem–spinal cord preparation, whereas8-OH-DPAT proved to be an effective ventilatory stimulant.22

Both enhancement and depression of nociception by5-HT1A-R-agonists, given alone or in combination with

Morphine 10 mg/kg

0

10

mL/s

10 s2 s

heat on

Pre-treatment

0

10

mL/s

10 s

Morphine 10 mg/kg, Repinotan 20 µg/kg

10 s

0

10

mL/s

heat on

TFR

heat off

2 s

heat on

2 s

A

B

C

Figure 3. Left panel, Expiratory airflow (mL/s). Animal was breathingspontaneously through a tracheostomy tube. Right panel, Tracings ofthe tail-flick reflex (TFR). Arrows indicate onset and offset of heat tothe tail, and the circle marks the artifact evoked by the tail flick. A,Pretreatment level. Left panel, Spontaneous respiratory frequency,51 breaths/min. Right panel, Heat to the tail evoked a tail flick witha latency (TFL) of 3.9 seconds. B, Morphine (10 mg/kg) completelydepressed spontaneous ventilation; the last 3 breaths before apneaare shown (left panel). The TFR was abolished (right panel). C,Repinotan (20 �g/kg) reestablished spontaneous ventilation (leftpanel), whereas TFR remained suppressed (right panel).

0.02 0.2 2 20 200Repinotan (µg/kg)

MO

Morphine/ RepinotanMorphine/ NaCl 0.9%

MV

(%ch

ange

[95%

CI])

n = 13 5 8 8 88 5 5

Figure 4. Effects of morphine (MO) and subsequent serotonin(1A)-receptor (5-HT1A-R)-agonist repinotan on spontaneous minute venti-lation (MV). NaCl 0.9% served as controls (n � 6). Values are givenas mean (95% confidence interval) and percent change from pre-treatment level (% change). *P � 0.05, **P � 0.01, compared withmorphine levels, 1-way repeated-measures analysis of variance.Morphine (11 mg/kg [8–16 mg/kg]) depressed MV to �72%(�100% to �44%) of pretreatment level. Repinotan dose depen-dently activated MV to a maximum of 18% (�40% to 77%). Repino-tan dosage �20 �g/kg re-decreased MV, giving the dose-responsecurve a bell shape (see Discussion).



opioids, have been variously reported.8,9,23–26 More re-cently, the highly selective 5-HT1A-R-agonist F13640 wasreported to induce both hyperalgesia and/or analgesiadepending on the blood and brain concentration timecourse.11 Most notably, F13640 was also shown to alleviateopioid-induced hyperallodynia and neuropathic pain inrats.27,28 The dose-dependent pro- and antinociceptive ef-fects of repinotan found in this study contribute to recon-ciling the past contradictory findings, which were at least inpart attributable to different experimental models, drugadministration routes, and dosing ranges.

The dose-response curve of 5-HT1A-R stimulation of spon-taneous breathing after morphine-induced ventilatory depres-sion is inversely U shaped or “bell shaped,” meaning thatstimulatory effects subsided with high concentrations.29 Also,repinotan produced a combination of low-dose stimulation ofthe TFR followed by high-dose inhibition. This dose-responsecharacteristic is generally referred to as “hormesis.”30 Morethan 30 receptor systems, including opioid and adrenergicreceptors, were identified to have hormetic dose responses,and the serotonin (5-HT) receptor system is among them.31

The neuroprotective effects of 5-HT1A-R-agonists have beenshown to have bell-shaped dose responses.32 It is proposedthat the basic biological principle behind this is that a mildstress may promote function or action, and extreme stressmay promote depressive or toxic action.30

Although hormesis can be observed in a wide range ofreceptor systems and agents, there is no one-for-all molecu-lar mechanism. The 5-HT1A-R are variously located andinvolved at different levels in the modulation of opioider-gic effects on nociceptive pathways.9,33 Activation of cen-tral 5-HT1A-R, for instance, has been shown to enhanceopioidergic inhibition of spinal reflexes,33 whereas systemic(intraperitoneal, IV) administration of 5-HT1A-R-agonistsproduced both pro- and antinociceptive effects.7,11 Directlyapplied onto the spinal cord, activation of 5-HT1A-R inhib-ited nociceptive neural responses only with the higheststudied dose of 8-OH-DPAT.8,34 We speculate that IVrepinotan overpowered possible pronociceptive effects(mediated by spinal 5-HT1A-R) by actions via central

5-HT1A-R, once the administered dose was high enough toestablish sufficient brain tissue concentrations. The under-lying mechanisms of hormetic dose responses clearly de-serve further research.30,35

The cardiovascular depression after repinotan adminis-tration was much less severe than that by 8-OH-DPAT. Inprevious work, we reported severe, occasionally fatal,cardiocirculatory depression with the highest dose of8-OH-DPAT (100 �g/kg) in anesthetized rats.12 Others sawthat 8-OH-DPAT prevented arterial hypotension inducedby the short-acting opioid remifentanil in conscious, non-anesthetized rats.36 Unlike repinotan,13 8-OH-DPAT alsostimulates 5-HT7-R,37 which are critical for activation ofcardiac vagal input.38 For instance, blockade of central5-HT7-R attenuates the bradycardia and pressor responseto both chemoreflex activation (induced by intracisternalinjection of potassium cyanide) and baroreflex activation(induced by IV phenylephrine).39 Activation of 5-HT7-R inturn might add to the depression of MAP seen in this studyafter morphine administration (Table 1), which was likelyinduced by peripheral vasodilation.40 Furthermore, it hasbeen shown in anesthetized animals that the 5-HT1A-R-agonist F13640 markedly reduced the intraoperative re-quirement of the volatile anesthetic.36 The concentration ofthe anesthetic was maintained constant in this study ac-cording to our protocol, which certainly contributed toarterial hypotension caused by increasing Pco2 as theconsequence of hypoventilation.

Some limitations of this study warrant comment. First,repinotan, a 5-HT1A-R-agonist, was developed as an antide-pressant, and was also found to exert neuroprotective effectson in vivo rats.15 Despite promising clinical data in humans,16

multicenter studies failed to show favorable effects on neuro-logical outcomes in patients with stroke and traumatic braininjury.17,18 Specific serotonergic complications of repinotan,such as headache, nausea and vomiting, flush, tachycardia,and agitation, in humans were reported.17,41 These symptomsmay even be aggravated in coadministration with mor-phine.20 Specific serotonergic side effects were not seen in thisstudy because of the experimental setup, but they could,

Table 1. Effects of Morphine and Repinotan on Mean Arterial Blood PressureRepinotan NaCl 0.9%

Mean (95% CI) P value Mean (95% CI) P valueRepinotan (% change of pretreatment)

Repinotan 0.02 �g/kg 10 (4–17) �0.05 0 (�13 to 13) NSRepinotan 0.2 �g/kg 4 (�6 to 14) NS �3 (�21 to 16) NSRepinotan 2 �g/kg 1 (�7 to 8) NS �5 (�25 to 15) NSRepinotan 20 �g/kg �17 (�29 to �4) �0.001 �10 (�25 to 10) NSRepinotan 200 �g/kg �19 (�29 to �9) �0.001 �7 (�21 to 10) NS

Morphine, repinotan (% change of pretreatment)Morphine �30 (�46 to �14) �0.001 �21 (�54 to �13) �0.05Repinotan 0.02 �g/kg �16 (�31 to �2) NS �21 (�54 to �11) NSRepinotan 0.2 �g/kg �24 (�40 to �8) NS �22 (�50 to �6) NSRepinotan 2 �g/kg �40 (�55 to �25) NS �21 (�53 to �11) NSRepinotan 6 �g/kg �28 (�50 to �5) NSRepinotan 20 �g/kg �30 (�46 to �14) NS �19 (�37 to 0) NSRepinotan 60 �g/kg �21 (�60 to �17) NSRepinotan 200 �g/kg �25 (�64 to 15) NS �19 (�49 to �11) NS

CI � confidence interval; NS � not significant; MAP � mean arterial blood pressure.MAP is given as mean (95% CI). Repinotan depressed MAP with higher doses (20, 200 �g/kg, n � 8, repeated-measures analysis of variance, P � 0.001).Morphine (12 mg/kg �8–16 mg/kg�) markedly depressed MAP (n � 13, P � 0.001, compared with pretreatment level). Subsequent repinotan did not furtheraggravate hypotension. There were no serious cardiovascular complications.


however, limit the clinical applicability of repinotan at least inconscious patients.

Second, it should be highlighted that the morphineconcentrations in this investigation were much higher thanin studies aiming solely at nociception.42,43 This happenedbecause morphine dosage was targeted to produce venti-latory depression, requiring higher dosing. The TFR wasalways abolished with the first bolus of morphine andalways before ventilatory depression occurred. The pos-sible attenuation of morphine-induced antinociception bysmall doses of repinotan was presumably overpowered bythe strong morphine effect. We did not investigate todetermine whether small pronociceptive doses of repinotanwould interfere with more moderate doses of morphine.This should be considered for further research.

Third, the TFR is an acute, polysynaptic nociceptivespinal reflex.44 Although pronociceptive effects were seenonly with very small doses of repinotan, but not within thedosing range to stimulate breathing, it is conceivable thatsmall pronociceptive doses of repinotan could alleviate mor-phine antinociception. It was shown by others that 5-HT1A-Rinfluence nociceptive processing differently, according to thetype of noxious stimulus.45 Thus, nociceptive modalities otherthan the one investigated here may be activated by smalldoses of 5-HT1A-R-agonists, which may not be treated withopioids. This will be clarified by further investigations.

In conclusion, this work confirmed that the 5-HT1A-R-agonist repinotan activates spontaneous breathing andsuppresses nociception with higher doses, and that itantagonizes morphine-induced ventilatory depression inanesthetized rats. Selective 5-HT1A-R-agonists thus arepromising candidates for research into the stabilization ofspontaneous breathing and pain therapy.

AUTHOR CONTRIBUTIONSUG, MFB, GW, HW, CP, and AH helped with study design;UG, NT, JZ, and GW helped with study conduction; UG, NT,and JZ helped with data collection; UG, NT, and JZ helpedwith data analysis; and UG, HW, MFB, NT, CP, and AH helpedwith manuscript preparation. All authors read and approvedthe final manuscript. UG and MFB reviewed the original studydata and data analysis. UG maintains the study records.

DISCLOSUREBayer Schering Pharma AG, Germany, provided the studydrug and funded part of this study. MFB and GW areemployees of Bayer Schering Pharma AG.

REFERENCES1. Couzin J. Medicine: a sigh of relief for painkillers. Science

2003;301:1502. Garner SJ, Eldridge FL, Wagner PG, Dowell RT. Buspirone, an

anxiolytic drug that stimulates respiration. Am Rev Respir Dis1989;139:946–50

3. Wilken B, Lalley P, Bischoff AM, Christen HJ, Behnke J,Hanefeld F, Richter DW. Treatment of apneustic respiratorydisturbance with a serotonin-receptor agonist. J Pediatr1997;130:89–94

4. El-Khatib MF, Kiwan RA, Jamaleddine GW. Buspirone treat-ment for apneustic breathing in brain stem infarct. Respir Care2003;48:956–8

5. Sahibzada N, Ferreira M, Wasserman AM, Taveira-DaSilva AM,Gillis RA. Reversal of morphine-induced apnea in the anesthe-tized rat by drugs that activate 5-hydroxytryptamine(1A) recep-tors. J Pharmacol Exp Ther 2000;292:704–13

6. Bardin L, Tarayre JP, Malfetes N, Koek W, Colpaert FC.Profound, non-opioid analgesia produced by the high-efficacy5-HT(1A) agonist F 13640 in the formalin model of tonicnociceptive pain. Pharmacology 2003;67:182–94

7. Bardin L, Tarayre JP, Koek W, Colpaert FC. In the formalinmodel of tonic nociceptive pain, 8-OH-DPAT produces5-HT1A receptor-mediated, behaviorally specific analgesia.Eur J Pharmacol 2001;421:109–14

8. Nadeson R, Goodchild CS. Antinociceptive role of 5-HT1Areceptors in rat spinal cord. Br J Anaesth 2002;88:679–84

9. Clarke RW, Ogilvie J, Houghton AK. Enhancement and depressionof spinal reflexes by 8-hydroxy-2-(di-n-propylamino) tetralin in thedecerebrated and spinalized rabbit: involvement of 5-HT1A-and non-5-HT1A-receptors. Br J Pharmacol 1997;122:631–8

10. Zhang YQ, Gao X, Ji GC, Huang YL, Wu GC, Zhao ZQ.Expression of 5-HT1A receptor mRNA in rat lumbar spinaldorsal horn neurons after peripheral inflammation. Pain2002;98:287–95

11. Bardin L, Assie MB, Pelissou M, Royer-Urios I, Newman-Tancredi A, Ribet JP, Sautel F, Koek W, Colpaert FC. Dual,hyperalgesic, and analgesic effects of the high-efficacy5-hydroxytryptamine 1A (5-HT1A) agonist F 13640 [(3-chloro-4-fluoro-phenyl)-[4-fluoro-4-{[(5-methyl-pyridin-2-ylmethyl)-amino]-methyl}piperidin-1-yl]methanone, fumaric acid salt]:relationship with 5-HT1A receptor occupancy and kinetic param-eters. J Pharmacol Exp Ther 2005;312:1034–42

12. Guenther U, Manzke T, Wrigge H, Dutschmann M, ZinserlingJ, Putensen C, Hoeft A. The counteraction of opioid-inducedventilatory depression by the serotonin 1A-agonist 8-OH-DPAT does not antagonize antinociception in rats in situ andin vivo. Anesth Analg 2009;108:1169–76

13. De Vry J, Schohe-Loop R, Heine HG, Greuel JM, Mauler F,Schmidt B, Sommermeyer H, Glaser T. Characterization of theaminomethylchroman derivative BAY � 3702 as a highlypotent 5-hydroxytryptamine1A receptor agonist. J PharmacolExp Ther 1998;284:1082–94

14. Schwarz T, Beckermann B, Buehner K, Mauler F, SchuhmacherJ, Seidel D, Steinke W, Weinz C, Zimmerd D. Pharmacokineticsof repinotan in healthy and brain injured animals. BiopharmDrug Dispos 2005;26:259–68

15. Harkany T, Mulder J, Horvath KM, Keijser J, van derMeeberg EK, Nyakas C, Luiten PG. Oral post-lesion admin-istration of 5-HT(1A) receptor agonist repinotan hydrochlo-ride (BAY � 3702) attenuates NMDA-induced delayedneuronal death in rat magnocellular nucleus basalis. Neu-roscience 2001;108:629 – 42

16. Ohman J, Braakman R, Legout V. Repinotan (BAY � 3702): a5HT1A agonist in traumatically brain injured patients. J Neuro-trauma 2001;18:1313–21

17. Teal P, Silver FL, Simard D. The BRAINS study: safety,tolerability, and dose-finding of repinotan in acute stroke. CanJ Neurol Sci 2005;32:61–7

18. Teal P, Davis S, Hacke W, Kaste M, Lyden PD, Fierus M. Arandomized, double-blind, placebo-controlled trial to eval-uate the efficacy, safety, tolerability, and pharmacokinetic/pharmacodynamic effects of a targeted exposure of intrave-nous repinotan in patients with acute ischemic stroke: modi-fied Randomized Exposure Controlled Trial (mRECT). Stroke2009;40:3518–25

19. Motulsky HJ. Analyzing Data with GraphPad Prism. SanDiego: GraphPad Software Inc., 1999

20. Oertel BG, Schneider A, Rohrbacher M, Schmidt H, Tegeder I,Geisslinger G, Lotsch J. The partial 5-hydroxytryptamine1Areceptor agonist buspirone does not antagonize morphine-induced respiratory depression in humans. Clin PharmacolTher 2007;81:59–68

21. Pavlakovic G, Tigges J, Crozier TA. Effect of buspirone onthermal sensory and pain thresholds in human volunteers.BMC Clin Pharmacol 2009;9:12

22. Guenther U, Bischoff A, Kettler D, Richter DW. 5-HT1A-agnoists protect against opioidergic depression of respiration.In: Urban BW, Barann M, eds. Molecular und basic mecha-nisms of anesthesia D-49525 Lengerich: Papst Science Publish-ers, 2001:400–3



23. Colpaert FC, Wu WP, Hao JX, Royer I, Sautel F, Wiesenfeld-Hallin Z, Xu XJ. High-efficacy 5-HT1A receptor activationcauses a curative-like action on allodynia in rats with spinalcord injury. Eur J Pharmacol 2004;497:29–33

24. Galeotti N, Ghelardini C, Bartolini A. 5-HT1A agonists inducecentral cholinergic antinociception. Pharmacol Biochem Behav1997;57:835–41

25. Bardin L, Colpaert FC. Role of spinal 5-HT(1A) receptors inmorphine analgesia and tolerance in rats. Eur J Pain2004;8:253–61

26. Graeff FG. Serotonin, the periaqueductal gray and panic.Neurosci Biobehav Rev 2004;28:239–59

27. Deseure K, Breand S, Colpaert FC. Curative-like analgesia in aneuropathic pain model: parametric analysis of the dose andthe duration of treatment with a high-efficacy 5-HT(1A) recep-tor agonist. Eur J Pharmacol 2007;568:134–41

28. Colpaert FC, Deseure K, Stinus L, Adriaensen H. High-efficacy5-hydroxytryptamine 1A receptor activation counteracts opi-oid hyperallodynia and affective conditioning. J PharmacolExp Ther 2006;316:892–9

29. Calabrese EJ, Baldwin LA. U-shaped dose-responses in biol-ogy, toxicology, and public health. Annu Rev Public Health2001;22:15–33

30. Calabrese EJ. Hormesis: basic, generalizable, central to toxicol-ogy and a method to improve the risk-assessment process. IntJ Occup Environ Health 2004;10:466–7

31. Calabrese EJ, Baldwin LA. Hormesis: the dose-response revo-lution. Annu Rev Pharmacol Toxicol 2003;43:175–97

32. Mauler F, Horvath E. Neuroprotective efficacy of repinotanHCl, a 5-HT1A receptor agonist, in animal models of strokeand traumatic brain injury. J Cereb Blood Flow Metab2005;25:451–9

33. Clarke RW, Ward RE. The role of 5-HT(1A)-receptors infentanyl-induced bulbospinal inhibition of a spinal withdrawalreflex in the rabbit. Pain 2000;85:239–45

34. Gjerstad J, Tjolsen A, Hole K. The effect of 5-HT1A receptorstimulation on nociceptive dorsal horn neurones in rats. EurJ Pharmacol 1996;318:315–21

35. Manzke T, Dutschmann M, Schlaf G, Morschel M, Koch UR,Ponimaskin E, Bidon O, Lalley PM, Richter DW. Serotonin targetsinhibitory synapses to induce modulation of network functions.Philos Trans R Soc Lond B Biol Sci 2009;364:2589–602

36. Dutschmann M, Waki H, Manzke T, Simms AE, Pickering AE,Richter DW, Paton JF. The potency of different serotonergicagonists in counteracting opioid evoked cardiorespiratory dis-turbances. Philos Trans R Soc Lond B Biol Sci 2009;364:2611–23

37. Eriksson TM, Golkar A, Ekstrom JC, Svenningsson P, OgrenSO. 5-HT7 receptor stimulation by 8-OH-DPAT counteracts theimpairing effect of 5-HT(1A) receptor stimulation on contex-tual learning in mice. Eur J Pharmacol 2008;596:107–10

38. Kellett DO, Ramage AG, Jordan D. Central 5-HT7 receptors arecritical for reflex activation of cardiac vagal drive in anaesthe-tized rats. J Physiol 2005;563:319–31

39. Damaso EL, Bonagamba LG, Kellett DO, Jordan D, RamageAG, Machado BH. Involvement of central 5-HT7 receptors inmodulation of cardiovascular reflexes in awake rats. Brain Res2007;1144:82–90

40. White DA, Reitan JA, Kien ND, Thorup SJ. Decrease invascular resistance in the isolated canine hindlimb after gradeddoses of alfentanil, fentanyl, and sufentanil. Anesth Analg1990;71:29–34

41. Lutsep HL. Repinotan, a 5-HT1A agonist, in the treatment ofacute ischemic stroke. Curr Drug Targets CNS Neurol Disord2005;4:119–20

42. Hutto CW Jr, Crowder WF. Dosage choices of rats for mor-phine, for heroin, and between morphine and heroin. Pharma-col Biochem Behav 1997;58:133–40

43. Eisenach JC, Gebhart GF. Intrathecal amitriptyline: antino-ciceptive interactions with intravenous morphine and intra-thecal clonidine, neostigmine, and carbamylcholine in rats.Anesthesiology 1995;83:1036 – 45

44. Sandkuhler J, Gebhart GF. Characterization of inhibition of aspinal nociceptive reflex by stimulation medially and laterallyin the midbrain and medulla in the pentobarbital-anesthetizedrat. Brain Res 1984;305:67–76

45. Bonnefont J, Chapuy E, Clottes E, Alloui A, Eschalier A. Spinal5-HT1A receptors differentially influence nociceptive process-ing according to the nature of the noxious stimulus in rats:effect of WAY-100635 on the antinociceptive activities ofparacetamol, venlafaxine and 5-HT. Pain 2005;114:482–90


Society for Technology in Anesthesia

Section Editor: Dwayne Westenskow

Goal-Directed Fluid Management Based on the PulseOximeter–Derived Pleth Variability Index ReducesLactate Levels and Improves Fluid ManagementPatrice Forget, MD,* Fernande Lois, MD,* and Marc de Kock, MD, PhD*

BACKGROUND: Dynamic variables predict fluid responsiveness and may improve fluid manage-ment during surgery. We investigated whether displaying the variability in the pulse oximeterplethysmogram (pleth variability index; PVI) would guide intraoperative fluid management andimprove circulation as assessed by lactate levels.METHODS: Eighty-two patients scheduled for major abdominal surgery were randomized into 2groups to compare intraoperative PVI-directed fluid management (PVI group) versus standardcare (control group). After the induction of general anesthesia, the PVI group received a 500-mLcrystalloid bolus and a crystalloid infusion of 2 mL � kg�1 � h�1. Colloids of 250 mL wereadministered if the PVI was �13%. Vasoactive drug support was given to maintain the meanarterial blood pressure above 65 mm Hg. In the control group, an infusion of 500 mL ofcrystalloids was followed by fluid management on the basis of fluid challenges and their effectson mean arterial blood and central venous pressure. Perioperative lactate levels, hemodynamicdata, and postoperative complications were recorded prospectively.RESULTS: Intraoperative crystalloids and total volume infused were significantly lower in thegoal-directed PVI group. Lactate levels were significantly lower in the PVI group during surgery and48 hours after surgery (P � 0.05).CONCLUSIONS: PVI-based goal-directed fluid management reduced the volume of intraoperativefluid infused and reduced intraoperative and postoperative lactate levels. (Anesth Analg 2010;111:910–4)

Hypovolemia occurs frequently in the operatingroom. Its diagnosis remains difficult, but assessmentof the adequacy of intravascular volume is of prime

importance to maintain cardiac output and thus avoid tissuehypoxia. In 1 meta-analysis the authors observed that periop-erative hemodynamic optimization reduced mortality.1

For many years, cardiac filling pressures were used toguide intravascular volume therapy. This, however, is not areliable predictor of fluid responsiveness.2 Dynamic vari-ables (indices evaluating the response to a cyclic preloadvariation) provide a better prediction of fluid responsive-ness.3 Among these, the arterial pulse pressure variationinduced by mechanical ventilation has been demonstratedas one of the best tools to guide volume therapy.3 Lopes etal. showed an improvement in postoperative outcome afterhigh-risk surgery when the pulse pressure variation wasused to guide intraoperative fluid therapy.4 Natalini et al.and Cannesson et al. demonstrated that respiratory varia-tions in the amplitude of the pulse oximeter plethysmo-graphic waveform and in the pulse pressure both predictfluid responsiveness.5–8 Zimmerman et al. showed thatpleth variability index (PVI) predicts fluid responsiveness

as accurately as does stroke volume variation.9 Neverthe-less, it remains unknown whether the optimization of theplethysmogram variability that occurs intraoperatively im-proves fluid management and circulation. To investigatethis, we used a pulse oximeter to continuously monitor thePVI.7 We measured the impact of PVI-based goal-directedfluid management on perioperative lactate levels.

METHODSAfter approval of the Ethics Committee of St.-Luc Hos-pital (Brussels, Belgium) (www.clinicaltrials.gov, no.NCT00816153), and after obtaining written informedconsent, a pilot study including 20 patients (10 per group)was conducted for the power analysis. Results showed animprovement of 20% of the primary outcome (whole bloodlactate levels) with the use of the PVI. A sample size of 37patients per group was calculated for a 0.05 difference(2-sided) with a power of 80%.

Between May and September 2008, we obtained writteninformed consent from 86 patients who met the inclusioncriteria: older than 18 years and the absence of cardiacarrhythmias, ultrasonographic cardiac ejection fraction�30%, lung pathology prohibiting mechanical ventilationwith tidal volumes larger than 6 mL � kg�1, and kidneydialysis. They were scheduled for esophagectomy, gastricresection/suture, hepatectomy, pancreatectomy, or intesti-nal and colorectal surgeries. Patients were randomized toeither the PVI group or the control group.

Heart rate, arterial blood pressure, oxygen saturation,inhaled gas concentrations, and temperature were measured

From the *Department of Anesthesiology, Universite catholique de Louvain,St.-Luc Hospital, Brussels, Belgium.


Address correspondence and reprint requests to Patrice Forget, Departmentof Anesthesiology, St.-Luc Hospital, av. Hippocrate 10–1821, 1200 Brussels,Belgium. Address e-mail to [email protected].

Copyright © 2010 International Anesthesia Research SocietyDOI: 10.1213/ANE.0b013e3181eb624f


continuously by a Datex S/5 monitor (Datex Ohmeda®, GEHealthcare). A Masimo Set version V7.1.1.5 pulse oximeter(Masimo Co., Irvine, California) was placed on the patient’sfinger for the continuous monitoring of PVI. A 20-G radialarterial catheter and a central venous access catheter wereinserted at the end of the induction phase. A thoracic epiduralcatheter was placed before the induction of the generalanesthesia. Anesthesia was induced with propofol 2 to 4 mg �kg�1 and atracurium or rocuronium 0.4 to 0.6 mg � kg�1 andmaintained with sevoflurane or desflurane. The lungs wereventilated with 6 to 8 mL � kg�1 of tidal volume, I:E � 1:2. Thefrequency was set to maintain normocapnia (Paco2 target �40 � 3 mm Hg).

In the PVI group, 500 mL of crystalloids (NaCl 0.9% orP-Lyte®, Baxter) was infused during induction, followed bya 2 mL � kg�1 � h�1 continuous infusion. If PVI was higherthan 13% for �5 minutes, we gave a 250-mL bolus ofcolloid (hydroxyethyl starch 6%, Voluven®, FreseniusKabi). The dose was repeated every 5 minutes if PVI wasstill higher than 13%. Norepinephrine was given as neededto maintain a mean arterial blood pressure �65 mm Hg.

In the control group, 500 mL of crystalloids was infusedduring induction, followed by a continuous infusion ofcrystalloids (4 to 8 mL � kg�1 � h�1). A bolus of colloids wasgiven if acute blood loss of �50 mL occurred, if the meanarterial blood pressure decreased below 65 mm Hg, or if thecentral venous pressure decreased below 6 mm Hg. Arepeat bolus was given after waiting 5 minutes if any one ofthe criteria was met. If the mean arterial blood pressuredecreased below 65 mm Hg and remained unresponsive tofluids, norepinephrine was given to maintain the meanarterial blood pressure above 65 mm Hg.

Arterial blood samples were taken at the time of skinincision, each hour during surgery and 6, 12, 18, 24, 36, and48 hours after the end of surgery. The lactate concentrationwas measured using an ABL 620 analyzer (Radiometer,Copenhagen, Denmark). Serum creatinine concentrationswere measured 24 and 48 hours after surgery. The anesthe-siologist identified and recorded instances of intraoperativehypotension (systolic blood pressure 20% below the valuemeasured the day before surgery, while the patient wasresting quietly for at least 15 minutes) and oliguria (urineoutput �0.5 mL � kg�1 for �2 hours).

During the first 30 days after surgery, a blinded postop-erative care team member identified, collected, and re-corded instances of postoperative infection, pulmonaryembolism, acute myocardial infarction, acute lunginjury/acute respiratory distress syndrome, pulmonaryedema, arrhythmia, stroke, cardiac arrest, coagulopathy(platelets �100,000 �L�1, international normalized ratio�2), hepatic dysfunction, nausea or vomiting necessitatingtreatment, upper digestive hemorrhage, leakage of anasto-mosis, and mortality.

Statistical AnalysisData were analyzed by comparing the patients in the PVIgroup with those in the control group using a modifiedintention-to-treat analysis (4 patients were excluded afterthe randomization for intraoperative arrhythmia or cancel-lation of the surgery, 2 per group; Fig. 1). The remaining 82

patients completed the protocol and were analyzed. Nopatient met abandon criteria.

Student’s t test was used to compare normally distrib-uted continuous variables and �2 for categorical variables.Homogeneity of variances was verified by the Levene’stest. A P value �0.05 was considered statistically signifi-cant. Data are expressed as mean (�sd), mean [95% confi-dence interval], or number (percentage). STATISTICA (dataanalysis software system) version 7 (Statsoft, Inc., 2004) wasused for all analyses.

RESULTSTable 1 lists the patients’ history and surgery. There wereno preoperative differences between the goal-directed fluid

Figure 1. Trial profile. PVI group: pleth variability index–guided fluidmanagement.

Table 1. Preoperative Characteristics, Incidenceof Chronic Diseases, Type and Duration of Surgeryand Anesthesia, Use of Epidural Analgesia in thePleth Variability Index (PVI) Group (PVI-GuidedFluid Management) and Control Group

PVI group(N � 41)

Control group(N � 41)

Age (years) 59 � 14 61 � 12Weight (kg) 71 � 15 68 � 16Height (cm) 169 � 9 170 � 9Sex (female/male) 16/25 (39/61) 16/25 (39/61)ASA score

2 22 (54) 22 (54)3 19 (46) 19 (46)

Chronic diseasesCirrhosis 3 (7) 0 (0)Chronic obstructive pulmonary

disease2 (5) 2 (5)

Hypertension 18 (44) 13 (32)Peripheral vascular disease 7 (17) 7 (17)Coronary artery disease 5 (12) 2 (5)Other cardiomyopathy 2 (5) 4 (10)Diabetes mellitus 4 (10) 2 (5)

Preoperative biological valuesHemoglobin (g � dL�1) 12.5 � 2 12.7 � 2Serum creatinine (mg � dL�1) 0.96 � 0.2 0.97 � 0.3

Type of surgeryUpper gastrointestinal 7 (17) 5 (12)Hepato-biliary 11 (27) 15 (37)Lower gastrointestinal 24 (59) 22 (54)

Laparoscopic approach 5 (12) 5 (12)Duration of surgery (minutes) 295 � 125 301 � 154Duration of anesthesia

(minutes)346 � 125 356 � 158

Epidural analgesia 33 (81) 29 (71)

One patient per group had two types of surgery. Data are presented asmean � SD or number (%). P � 0.05 for all the data.


management group and the control group. Table 2 showsthat during surgery, patients in the PVI-directed fluidmanagement group were given less total fluid and lesscrystalloid intraoperatively than was the control group.There were no differences postoperatively. Figure 2 shows

that lactate levels were lower in the PVI group during andafter surgery. There were no statistically significant differ-ences in the incidence of hypotension, cardiovascular res-cue, or renal dysfunction. Two patients in the PVI groupdied from septic shock 20 days and 33 days after surgerybecause of a failed anastomosis (Table 3).

DISCUSSIONWe found that PVI-guided fluid management resulted inless crystalloid administered perioperatively and reducedlactate levels during and after major abdominal surgery.Lactate levels provide an indirect but sensitive measure oforgan perfusion. Lactate is clearly correlated with theadequacy of intravascular volume, tissue hypoxia, andenergy failure due to bloodflow redistribution.10 Lactatelevels can be improved by the optimization of the fluidstatus and cardiac preload.2,4

Our results confirm the conclusion of Lopes et al. Theuse of the noninvasive PVI, or the invasively obtainedpulse pressure variation, improves perioperative fluidmanagement.4 In Lopes et al.’s study, the averageamount of fluids was larger in the group guided by pulsepressure variation, in contrast with our results. Thedifference in results may be explained by the presence ofhypovolemia in some patients and hypervolemia inothers. These results therefore argue the superiority of

Figure 2. Lactate levels during and after surgery in the plethvariability index (PVI)–guided group (PVI-guided fluid management)and in the control group. Intraoperative: maximum intraoperativevalue. Data are presented as mean � SEM. *P � 0.05.

Table 2. Fluids Administered, Blood Loss, Hemodynamic Status, Physiologic Status, and Renal FunctionDuring and After Surgery in the Pleth Variability Index (PVI) Group (PVI-Guided Fluid Management) and inthe Control Group

PVI group(N � 41)

Control group(N � 41) P value

Intraoperative fluids (mL)Crystalloids 1363 �1185–1540� 1815 �1568–2064� 0.004Colloids 890 �709–1072� 1003 �779–1227� 0.43Blood products 141 �53–230� 99 �20–179� 0.48Total of intraoperative fluids 2394 �2097–2692� 2918 �2478–3358� 0.049Blood losses 349 �230–468� 440 �242–637� 0.43

Postoperative fluids (24 hours)Crystalloids 3107 �2760–3454� 3516 �3009–4024� 0.17Colloids 268 �126–409� 358 �175–540� 0.43Blood products 8 ��8–25� 44 ��45–133� 0.41

Lactate levels (mMol � L�1)Maximum intraoperative 1.2 �1–1.4� 1.6 �1.2–2� 0.04At 24 hours 1.4 �1.3–1.5� 1.8 �1.5–2.1� 0.02At 48 hours 1.2 �1–1.3� 1.4 �1.2–1.5� 0.03

Lactate levels �1.7 mMol � L�1

Intraoperatively 7 (17) 4 (10) 0.33At 24 hours 2 (5) 28 (68) <0.0001At 48 hours 0 8 (20) 0.003

Lactate levels �5 mMol � L�1

Intraoperatively 0 1 (2) 0.31At 24 hours 0 1 (2) 0.31At 48 hours 0 1 (2) 0.31Intraoperative hypotension 22 (54) 28 (68) 0.17

Continuous infusion of norepinephrineIntraoperative 9 (22) 9 (22) 1.0At 24 hours 3 (7) 1 (2) 0.31

Renal function diuresisIntraoperative oliguria 13 (32) 17 (42) 0.34Postoperative oliguria (24 hours) 3 (8) 3 (8) 0.97

Serum creatinine (mg � dL�1)At 24 hours 1.01 �0.9–1.1� 1.12 �0.9–1.3� 0.32At 48 hours 0.91 �0.8–1� 1.09 �0.9–1.3� 0.11

Initiation of dialysis 1 (2) 0 (0) 0.32

Lactate levels: normal value 0.9–1.7 mMol � L�1. Oliguria was defined as a urinary output �0.5 mL � kg�1 for more than 2 hours. Data are presented asmean �95% confidence interval� or number (%).P � 0.05 was considered as statistically significant (boldface numerical entries).

Pleth Variability Index and Fluid Management


goal-directed fluid management over simplistic restrictiveor liberal approaches for fluid management, avoiding hy-povolemia and hypervolemia.11,12

Unlike Lopes et al., we did not find an improvementin terms of the number of complications. The muchhigher incidence of hypovolemia reported by Lopes et al.may account for this difference. The clinical significance oflower lactate levels in our relatively small study may bequestioned. Additionally, fluid management in the controlgroup was different by design, favoring greater fluidcrystalloid administration (2 mL � kg�1 � h�1 in the PVIgroup vs. 4 to 8 mL � kg�1 � h�1 in the control group), andit was possibly influenced by the fact that the control grouphad a greater blood loss (440 [242 to 637] mL vs. 349 [230 to468] mL) (although this was not statistically significant).When mean arterial blood pressure decreased to �65mm Hg, the PVI group received norepinephrine, whereasthe control group received norepinephirne and a bolus ofcrystalloid. In our study a “learning contamination bias”may have blunted the differences between the groups. Thisbias occurs when a team member gains experience withpulse pressure variation and begins, intuitively, to userespiratory variations of the arterial pressure curve to treatpatients in the control group. However, small variations aredifficult to see without using a device that makes thecalculations from the curve.

The PVI was calculated by the new Masimo Set pulseoximeter (Masimo Co., Irvine, California) from the respira-tory variations in the perfusion index (PI). The PI is thepercentage amplitude difference between the pulsatile in-frared signal and the nonpulsatile infrared signal. The PVIis calculated by measuring changes in the PI during therespiratory cycle: PVI � [(PImax � PImin)/PImax] � 100.Cannesson et al. have demonstrated that the PVI predictsfluid responsiveness in the operating room. They showedthat the cutoff value to distinguish responders from non-responders to intravascular volume expansion (in terms ofan increase of cardiac index) was a PVI �14%.7 Weconfirmed their results in a preliminary study (data not

shown) and consequently chose 14% as the threshold forfluid loading.

Whereas the PVI may be useful in most patients, ourexclusion criteria limit the application of our results insome patients. To maintain homogeneity between the 2study groups, we did not include patients with severecardiac insufficiency (ejection fraction �30%) or chronicdialysis. Moreover, the dynamic variables must not becalculated in the presence of arrhythmia. One patient pergroup was excluded because of an intraoperative arrhyth-mia. Additionally, these results cannot be extrapolated toother devices that calculate the respiratory variation of theplethysmographic curve. The algorithm used to processthe signal may explain the poor accuracy observed byothers.13 Moreover, we did not measure the possibleimpact of the use of epidural analgesia and thoracotomyin some patients.

In conclusion, the use of PVI-guided fluid managementwas associated with lower lactate levels during majorabdominal surgery. Patients in the PVI-guided group weregiven less crystalloid. Reduced lactate levels in PVI-guidedpatients suggests that PVI-guided fluid management maylead to fluid administration that is tailored to each indi-vidual patient’s needs.

ACKNOWLEDGMENTSMasimo Corporation graciously provided devices during thestudy protocol.

REFERENCES1. Poeze M, Greve JWM, Ramsay G. Meta-analysis of hemody-

namic optimisation: relationship to methodological quality.Crit Care 2005;9:R771–9

2. Cavallaro F, Sandroni C, Antonelli M. Functional hemody-namic monitoring and dynamic indices of fluid responsive-ness. Minerva Anestesiol 2008;74:123–35

3. Michard F, Teboul JL. Predicting fluid reponsiveness in ICUpatients. A critical analysis of the evidence. Chest 2002;121(6):2000 – 8

Table 3. Postoperative Complications and Intensive Care Unit/Hospital Stay in the Pleth Variability Index(PVI) Group (PVI-Guided Fluid Management) and in the Control Group

PVI group(N � 41)

Control group(N � 41) P Value

Postoperative complicationsInfection of surgery site 8 (20) 8 (20) 1.0Other infections (pulmonary, line-related, other abdominal) 6 (15) 7 (17) 0.77Cardiovascular complications (acute myocardial infarction,

acute lung injury/acute respiratory distresssyndrome, pulmonary edema, arrhythmia)

4 (10) 8 (20) 0.26

Coagulopathy 5 (12) 6 (15) 0.75Nausea and/or vomiting 0 (0) 4 (16) 0.08Upper digestive hemorrhage 4 (10) 3 (7) 0.78Leakage of anastomosis 5 (12) 5 (12) 1.0

Morbidity (event per patient) 1.2 � 1.8 1.5 � 2.2 0.46Mortality 2 (5) 0 (0) 0.16Length of stay

Postoperative mechanical ventilation 1 (2) 3 (7) 0.31Intensive care unit (days) 2.2 � 5.7 1.8 � 7.2 0.71Hospital (days) 15.1 � 14.3 16.0 � 17.8 0.78

Data are presented as mean � SD or number (%).


4. Lopes MR, Oliveira MA, Pereira VO, Lemos IP, Auler JO Jr,Michard F. Goal-directed fluid management based on pulsepressure variation monitoring during high-risk surgery: a pilotrandomized controlled trial. Crit Care 2007;11(5):R100

5. Desebbe O, Cannesson M. Using ventilation-induced plethys-mographic variations to optimize patient fluid status. CurrOpin Anaesthesiol 2008;21:772–8

6. Natalini M, Rosano A, Taranto M, Faggian B, Vittorielli E,Bernardini A. Arterial versus plethysmographic dynamicindices to test responsiveness for testing fluid administra-tion in hypotensive patients: a clinical trial. Anesth Analg2006;103(6):1478 – 84

7. Cannesson M, Desebbe O, Rosamel P, Delannoy B, Robin J,Bastien O, Lehot JJ. Pleth variability index to monitor therespiratory variations in the pulse oximeter plethysmographicwaveform amplitude and predict fluid responsiveness in theoperating theatre. Br J Anaesth 2008;101(2):200–6

8. Cannesson M, Attof Y, Rosamel P, Desebbe O, Joseph P,Metton O, Bastien O, Lehot JJ. Respiratory variations inpulse oximetry plethysmographic waveform amplitude topredict fluid responsiveness in the operating room. Anes-thesiology 2007;106(6):1105–11

9. Zimmermann M, Feibicke T, Keyl C, Prasser C, Moritz S, GrafBM, Wiesenack C. Accuracy of stroke volume variation com-pared with pleth variability index to predict fluid responsive-ness in mechanically ventilated patients undergoing majorsurgery. Eur J Anaesthesiol 2009; [Epub ahead of print]

10. Valenza F, Aletti G, Fossali T, Chevallard G, Sacconi F, Irace M,Gattinoni L. Lactate as a marker of energy failure in critically illpatients: hypothesis. Crit Care 2005;9(6):588–93

11. Bundgaard-Nielsen M, Holte K, Secher NH, Kehlet H.Monitoring of peri-operative fluid administration by indi-vidualized goal-directed therapy. Acta Anaesthesiol Scand2007;51(3):331– 40

12. Bundgaard-Nielsen M, Ruhnau B, Secher NH, Kehlet H. Flow-related techniques for preoperative goal-directed fluid optimi-sation. Br J Anaesth 2007;98(1):38–44

13. Landsverk SA, Hoiseth LO, Kvandal P, Hisdal J, Skare O,Kirkeboen KA. Poor agreement between respiratory variationsin pulse oximetry photoplethysmographic waveform ampli-tude and pulse pressure in intensive care unit. Anesthesiology2008;109(5):849–55

Pleth Variability Index and Fluid Management


A Long-Term Clinical Evaluation of AutoFlow DuringAssist-Controlled Ventilation: A RandomizedControlled TrialSigismond Lasocki, MD, PhD, Francoise Labat, MD, Gaetan Plantefeve, MD, Mathieu Desmard, MD,and Herve Mentec, MD

BACKGROUND: Many new mechanical ventilation modes are proposed without any clinicalevaluation. “Dual-controlled” modes, such as AutoFlow™, are supposed to improve patient–ventilator interfacing and could lead to fewer alarms. We performed a long-term clinicalevaluation of the efficacy and safety of AutoFlow during assist-controlled ventilation, focusing onventilator alarms.METHODS: Forty-two adult patients, receiving mechanical ventilation for more than 2 days witha Drager Evita 4 ventilator were randomized to conventional (n � 21) or AutoFlow (n � 21)assist-controlled ventilation. Sedation was given using a nurse-driven protocol. Ventilator-generated alarms were exhaustively recorded from the ventilator logbook with a computer. Dailyblood gases and ventilation outcome were recorded.RESULTS: A total of 403 days of mechanical ventilation were studied and 45,022 alarms wererecorded over a period of 8074 hours. The course of respiratory rate, minute ventilation, FIO2,positive end-expiratory pressure, PaO2/FIO2, PaCO2, and pH and doses and duration of sedationdid not differ between the 2 groups. Outcome (duration of mechanical ventilation, ventilator-associated pneumonia, course of Sequential Organ Failure Assessment score, or death) was notdifferent between the 2 groups. The number of alarms per hour was lower with AutoFlowassist-controlled ventilation: 3.3 [1.5 to 17] versus 9.1 [5 to 19], P � 0.0001 (median [quartilerange]). In multivariate analysis, a low alarm rate was associated with activation of AutoFlow anda higher midazolam dose.CONCLUSIONS: This first long-term clinical evaluation of the AutoFlow mode demonstrated itssafety with regard to gas exchange and patient outcome. AutoFlow also allowed a very markedreduction in the number of ventilator alarms. (Anesth Analg 2010;111:915–21)

“Dual-controlled” ventilation modes1,2 are reportedto combine the advantages of both volume and

pressure-controlled ventilation modes. The onlystudies comparing dual modes to conventional assist-controlled ventilation (ACV) focused on short-term effects(several hours). They report a reduction of inspiratorypressure.3,4 Drager’s AutoFlow™ (AF) is one of thesedual-controlled ventilation modes that also allows sponta-neous breathing throughout the respiratory cycle. Dragerclaims that AF is expected to improve patient–ventilatorinterfacing and could decrease the number of times patientsfight the ventilator. AF could consequently reduce thenumber of ventilator alarms. Alarms are partly responsiblefor the high noise level in intensive care units (ICU),5–7 anda reduction of ventilator alarms would therefore be benefi-cial to both patients and ICU staff. ICU staff also fail tocorrectly identify many of these alarms.8 Reducing the

number of alarms may therefore improve alarm efficiency.9

We hypothesize that activation of the AF would reduce thenumber of ventilation-generated alarms, without impairingthe patient’s ventilation. Because no clinical evaluation ofthis ventilation mode is available, this study was alsodesigned to clinically evaluate AF activation.

The aim of this randomized controlled study was toclinically evaluate AF activation during ACV, with regardto ventilator-generated alarms (primary aim) and to gasexchange and patient outcome (secondary aims).

METHODSThe study protocol was approved by the Comite Consul-tatif de Protection des Personnes dans la RechercheBiomedicale (independent ethics committee) of Saint-Germain-en-Laye, France (ClinicalTrial.gov identifier:NCT0092774). Written informed consent was obtained fromthe patient or next of kin.

PatientsAdult patients admitted to the ICU of Victor DupouyHospital, Argenteuil, France, were eligible when they re-quired ACV with an Evita4 ventilator (Drager Medical,Antony, France) for an expected duration of �2 days.Patients were not included in the case of coma, ventilationfor �12 hours before inclusion, pregnancy, or inclusion inanother study.

AutoFlowAF is a dual-control mechanical ventilation mode associ-ated with ACV.1,10 All breaths are pressure-controlled, with

From the Reanimation Polyvalente, Centre Hospitalier Victor Dupouy,Argenteuil, France.


Sigismond Lasocki, MD, PhD, is currently affiliated with ReanimationChirurgicale, APHP, CHU Bichat Claude Bernard, Paris, France.

The Hospital Victor Dupouy, Argenteuil, France, supported this study.Dragger SA provided only technical assistance for the recording of aventilator logbook, but had no access to the data and was not involved inthe preparation of this manuscript.

Address correspondence to Sigismond Lasocki, MD, PhD, ReanimationChirurgicale, CHU Bichat, 46 rue Henri Huchard, 75018 Paris, France.Address e-mail to [email protected].

Copyright © 2010 International Anesthesia Research SocietyDOI: 10.1213/ANE.0b013e3181f00015


a delivered level of pressure support that varies frombreath to breath to deliver the set tidal volume (Vt). AFuses a feedback loop that regulates inspiratory flow. Dy-namic compliance is measured breath by breath, and therequired � pressure for the next breath is calculated bydividing the desired Vt by dynamic compliance. Changesof inspiratory pressure from breath to breath are limited to3 mbar. When inspiratory pressure reaches the upperpressure limit minus 5 mbar, inspiratory time is increasedwithin the limits defined by the set respiratory rate.

Mechanical VentilationPatients were randomly assigned to the control group(AF�) or the AF group (AF�) by opening a sealed enve-lope. Attending physicians chose all respiratory settings,without intervention by the study investigators. The upperlimit of inspiratory pressure alarm was initially set at 50 cmH2O. Other alarm limits were set at the manufacturer’sdefault values, which varied according to the patient’sbody weight. Attending physicians were allowed to changeany alarm limit and ventilator mode when clinically indi-cated, except that AF was always used with ACV in theAF� group, and AF was never used with ACV in the AF�group. Discontinuation of mechanical ventilation was per-formed according to our ICU standard protocol, by thegradual reduction of pressure support. Blood gases wereobtained at least once daily during the first days. Morningvalues of ventilator settings, highest Fio2, highest positiveend-expiratory pressure, highest Paco2, lowest Pao2/Fio2

ratio, and lowest pH were recorded daily.

SedationPatients were sedated by continuous infusions of midazo-lam and fentanyl according to a nurse-driven protocol (Fig.1). The sedation goal was a Ramsay score11 of 2 or 3.

Ventilator-Generated AlarmsVentilators were periodically connected to a personalcomputer to download the number of alarms, changes inventilator settings, alarm limit settings, and the numberof times alarms were manually silenced (silence knobactivation).

OutcomeOrgan failure was assessed daily using the SequentialOrgan Failure Assessment (SOFA) score.12 Duration ofmechanical ventilation, ICU and hospital survival, theincidence of pneumothorax, and the incidence of ventilator-associated pneumonia were recorded.

Statistical AnalysisResults are expressed as median[Q1–Q3] or mean � sd.Comparisons were performed with a Mann–Whitney testor Student’s t test, a �2 test with Yates’ correction, or ananalysis of variance (ANOVA) for repeated measures asappropriate. Analysis was performed by either intention totreat on the whole study period or per protocol (for theperiod during which patients were on ACV, with orwithout AF), depending on the variable considered. Thuswe report the alarm rate for the different ventilation mode(ACV or non-ACV) in both groups. Obviously, in the AF�group, AF was not activated during non-ACV ventilation(because it was not available).

To estimate the sample size of the study, we assumedthat the total alarm rate in an ICU would be 36.5alarms/hr,13 and that ventilator alarms would account for38% of all alarms.14 The alarm rate during ACV wasassumed to be 14 alarms/hr. A 50% reduction when AFwas added to ACV was considered to be clinically relevant.Twenty-one patients in each group were needed to achievea 90% power, with an � risk of 5%. To assess factors

Figure 1. Sedation algorithm used by nurses to conduct patient sedation (VAS � visual analgesia scale; Ramsay � sedation score accordingto the Ramsay scale11).

Evaluation of AutoFlow Controlled Ventilation


influencing the alarm rate, we divided patients into 2groups—higher and lower than the median alarmrate—and we performed a logistic regression, including allthe parameters that had a P value of �0.1 in the univariateanalysis (i.e., indexed midazolam, indexed fentanyl and AFgroup, with indexed midazolam and indexed fentanylbeing the total dose of midazolam or fentanyl indexed tothe patient’s body weight and the duration of drug infu-sion). Statistical analysis was performed with SPSS 15.0.P � 0.05 was considered statistically significant.

RESULTSPatientsForty-two patients were included. No statistically signifi-cant differences in demographic data were observed be-tween the 2 groups (Table 1). Indications for mechanicalventilation were not statistically different between groups(P � 0.31).

Ventilation and Gas ExchangeBaseline ventilator settings and blood gases were notdifferent between the 2 groups (Table 2). An ANOVA forrepeated measures from day 1 to day 5 showed no signifi-cant difference between the 2 groups for either ventilatorsettings or blood gases (Fig. 2).

SedationSedation was not different between the 2 groups (Table 3).Total doses of midazolam (114 [0 to 163] vs. 150 [0 to 316]mg, P � 0.5) and fentanyl (6600 [0 to 55,480] vs. 8000 [0 to21,500] �g, P � 0.8) were not different between the 2 groups(AF� and AF� groups, respectively).

Ventilator Alarms and InterventionsA total of 403 days (8074 hours) of mechanical ventilationwere studied. ACV was used for 3997 hours. AF was usedfor 2133 hours. A total of 45,022 alarms were recorded;nearly 1 alarm every 10 minutes. During ACV, 7060 alarmswere recorded in the AF� group, and 16,817 alarms wererecorded in the AF� group. Figure 3 shows that theventilator alarm rate was lower when AF was used inconjunction with ACV (3.3 [1.5 to 17] alarms/hr with AF vs.9.1 [5.2 to 19] without AF [P � 0.0001]). In the AF� group,the alarm rate was lower for ventilation modes other thanACV (mainly pressure support) than they were for ACV(without AF), but no difference was observed betweenthese ventilation modes and ACV for the AF� group (Fig.3). The number of alarm setting modifications per hour wasnot different between groups (0.07 [0.02 to 0.23] vs. 0.09 [0.02to 0.23] modifications per hour for AF� and AF� patients,respectively; P � 0.85). Setting the high-pressure limit above50 cmH2O was less frequent in the AF� group (0 [0 to 2] vs.2 [0 to 8] times per patient; P � 0.0007).

The type of alarm differed between the 2 groups. ACVplus AF generated fewer pressure alarms than did ACValone (P � 0.0001) (Fig. 4). The silence knob activation ratewas lower in the AF� group during ACV (0.26 [0.1 to 1.1]vs. 0.72 [0.26 to 2] activation per hour; P � 0.0013).

The median alarm rate during ACV was 6.37 alarms/hr.Patients with alarm rates lower than 6.37 alarms/hr weremore frequently randomized to the AF� group (P �0.0002) and had a higher dosage of sedative drugs (fenta-nyl, P � 0.02; midazolam, P � 0.07) (Table 4). In multivar-iate analysis, an alarm rate lower than 6.37 alarms/hr wasassociated with activation of AF (OR [95% CI], 90 [5 to1570], P � 0.002) and a higher midazolam dosage (OR 1801[3 to 1.1 106] per mg/d/kg, P � 0.02).

OutcomeNo patient suffered from pneumothorax in the AF� groupin comparison with 2 patients in the AF� group (P � 0.48).Four cases of ventilator-associated pneumonia were ob-served in the AF� group in comparison with 8 in the AF�group (P � 0.16). The median duration of ventilation was 6[2 to 25] versus 9 [2 to 36] days in AF� and AF� groups,respectively (P � 0.33), and the median number of days freeof mechanical ventilation at day 28 were 15 [0 to 26] and 13[0 to 26], respectively (P � 0.55). An ANOVA for repeated

Table 1. Patient CharacteristicsAF� (n � 21) AF� (n � 21) P

Age (years) 71 �25–90� 65 �32–88� 0.70Body weight (kg) 73 �53–101� 64 �38–150� 0.75Height (cm) 168 �150–191� 167 �150–185� 0.49Gender M/F 14/7 14/7 0.99COPD (%) 2 (10) 6 (29) 0.24Type of patient 0.99

Medical 15 15Urgent surgery 6 6

Indication ofmechanicalventilation

0.31

ARF 15 12Ventilation

withoutARFa

4 3

ARF on COPD 1 3Postoperative

ventilation1 3

AdmissionSAPS II

60 �25–101� 51 �19–105� 0.33

Admission SOFA 9 �2–17� 9 �4–19� 0.57

Values are expressed as median �min–max� or numbers (%).AF� � assist-controlled ventilation with AutoFlow activation; AF� �assist-controlled ventilation without AutoFlow activation; COPD � chronicobstructive pulmonary disease; ARF � acute respiratory failure; SAPS II �simplified acute physiologic score (34); SOFA � Sequential Organ FailureAssessment (12).a Shock and PaO2/FIO2 �250 mm Hg.

Table 2. Ventilator Settings and GasExchange Variables

AF� (n � 21) AF� (n � 21) PRR (cycles/min) 18 �14–22� 20 �15–22� 0.87VT (mL) 500 �400–650� 500 �350–600� 0.40PEEP (cm H2O) 5 �0–15� 5 �0–13� 0.97FIO2 1 �0.35–1� 0.8 �0.40–1� 0.76pH 7.31 �6.98–7.60� 7.28 �7.11–7.57� 0.76PaCO2 (mm Hg) 39 �31–61� 42 �29–110� 0.51PaO2 (mm Hg) 141 �62–492� 152 �49–286� 0.87PaO2/FIO2 (mm Hg) 192 �62–532� 242 �49–356� 0.98

Values are expressed as median �min–max�.AF� � assist-controlled ventilation with AutoFlow activation; AF� � assist-controlled ventilation without AutoFlow activation; RR � respiratory rate; VT �tidal volume; PEEP � positive end-expiratory pressure; FIO2 � inspiratoryoxygen fraction.


measures from day 1 to day 5 showed no significantdifference between the 2 groups for SOFA score (Fig. 2). Sixpatients in the AF� group and 9 patients in the AF� groupdied in the ICU (P � 0.39), whereas 8 and 10 patients diedin the hospital, respectively (P � 0.62).

DISCUSSIONThis study is the first long-term clinical evaluation of AFduring ACV. Clinical outcome and blood gas variableswere not different with and without AF. There were fewerventilator alarms when AF was used.

Figure 2. Time course of ventilatory settings, gas exchange, and Sequential Organ Failure Assessment (SOFA) score12 during the first 5 days(D1 to D5) of mechanical ventilation. Solid squares represent the Drager’s AutoFlow (AF)� group, and open circles represent the control (AF�)group. Mean � SEM. An analysis of variance (ANOVA) for repeated measures was not significant. PEEP � positive end-expiratory pressure.



AF is based on an attractive principle: to guarantee a setVt and minute ventilation while maintaining the advan-tages of pressure-controlled ventilation.1,2 Despite thispotential advantage, clinical evaluations have not beenperformed. Even clinical efficiency in comparison withconventional ACV has not been formally demonstrated.This is the first report of around-the-clock observationand long-term bedside evaluation of this ventilationmodality in the context of standard care. We found nodifferences with or without AF for gas exchange (P/Fratio or Paco2) during ACV. No other studies are avail-able for comparison. Previous studies have only reportedthe physiological advantages associated with pressure-regulated ACV in comparison with volume-controlled

Figure 3. Hourly alarm rates. Rectangular boxes represent themedian hourly alarm rate during assist-controlled ventilation (ACV)and diabolo boxes represent the median hourly alarm rate during theother mechanical ventilation modalities, mainly during pressuresupport (non-ACV). Upper and lower edges represent the 75th and25th percentiles. Dark gray shapes represent rates observed forpatients in the AutoFlow group (AF�), and light gray shapes repre-sent rates observed in the conventional ACV group (AF�). Thedashed line represents the overall median hourly alarm rate, whichwas almost 1 alarm every 10 minutes. The lowest alarm rate wasobserved during ACV using AutoFlow, and the highest alarm rate wasobserved during ACV without AutoFlow.

Table 3. Sedation VariablesAF� (n � 21) AF� (n � 21) P

Sedation duration (days) 3 �1–10� 5 �0–12� 0.44Indexed midazolam

(mg/d/kg)a0.38 �0–1.13� 0.44 �0–0.78� 0.73

Indexed fentanyl (�g/d/kg)b 28 �0–93� 25 �10–52� 0.46Sedation modifications

(per day)1.5 �0–3.3� 1.4 �0–3.8� 0.66

Ramsay score of 2 or 3(per day)c

2.9 �0.5–6.3� 2.2 �0–5.5� 0.29

Values are expressed as median �min–max�.AF� � assist-controlled ventilation with AutoFlow activation; AF� � assist-controlled ventilation without AutoFlow activation.a Indexed midazolam: total dose of midazolam indexed to the patient’s bodyweight and the duration of drug infusion.b Indexed fentanyl: total dose of fentanyl indexed to the patient’s body weightand the duration of drug infusion.c Ramsay score (11) of 2 or 3, the mean number of Ramsay scoring not equalto 2 or 3 per day.

Figure 4. Distribution of alarm types in the 2 groups during assist-controlled ventilation (ACV). AutoFlow ACV group (AF�) is repre-sented with dark gray bars, and conventional ACV group (AF�) isrepresented with light gray bars. The left side of the chart shows thetotal number of all volume alarms generated during ACV (i.e., low orhigh tidal volume (VT) and low or high minute ventilation alarms) andthe total “partially delivered VT” alarm (this alarm is generated whenthe pressure limit is reached, leading to a stopping of inspiration),and the right side shows the total number of high and low airwaypressure alarms. ACV without AF generated many more pressurealarms than did ACV with AutoFlow. Mean � SEM.

Table 4. Univariate Analysis of Factors InfluencingAlarm Rate

Low alarmratea (n � 21)

High alarmratea (n � 21) P

Age (years) 66 �26–90� 66 �25–88� 0.89Gender (M/F) 15/6 13/8 0.11COPD (%) 2 (10) 5 (24) 0.41SAPS II (points) 59 �23–101� 54 �29–105� 0.27Tracheal aspirate

volume(points)

1.5 �0.7–2.6� 1.9 �1.0–2.8� 0.27

Nebulizations(total number)

0 �0–36� 0 �0–22� 0.25

Indexedmidazolam(mg/d/kg)b

0.5 �0.2–1.1� 0.3 �0–0.8� 0.07

Indexed fentanyl(�g/d/kg)c

30 �7–93� 21 �0–53� 0.02

Sedationduration(days)

4 �2–10� 4 �1–12� 0.72

ACV duration(hours)

85 �15–284� 54 �11–316� 0.16

Silence knobactivation (perhour)

0.26 �0.1–1.1� 0.80 �0.26–2� 0.001

Study group(AF�/ AF�)

17/4 4/17 0.0002

Values are expressed as median �min–max� or number.COPD � chronic obstructive pulmonary disease; SAPS II � Simplified AcutePhysiology Score II (34); ACV � assist-controlled ventilation; AF� � assist-controlled ventilation with AutoFlow activation; AF� � assist-controlledventilation without AutoFlow activation.a The patients were divided into 2 groups according to whether their ventilatoralarm rate was lower or higher than was the median alarm rate of the overallpopulation (6.37 alarms/hr).b Indexed midazolam, total dose of midazolam indexed to the patient’s bodyweight and the duration of drug infusion.c Indexed fentanyl, total dose of fentanyl indexed to the patient’s body weightand the duration of drug infusion.


ventilation, such as lower inspiratory pressures3,4,15 andlower Paco2.3,15

One concern regarding AF is that the level of support(i.e., the level of pressure delivered) could theoreticallydecrease as patient demand increases. Indeed, the work ofbreathing increases when pressure support decreases.16

This has been recently confirmed in a lung simulator.17

However, the absence of altered gas exchange in the AF�group and the trend towards a shorter duration of ventila-tion do not support this hypothesis, in vivo. In fact, for a setVt, pressure-controlled ventilation reduces the work ofbreathing in comparison with volume-controlled ventila-tion.18 In the present study, 2133 hours of AF ACV wererecorded, and no patient experienced any signs of respira-tory distress. However, our study, with a sample size of 42patients, is not powered to definitively assess these clinicalaspects. Indeed, it was powered to compare the ventilation-related alarm rate with and without AF during ACV.

An original tool was used to record all ventilator-generated alarms. A higher alarm rate was observed thanwas previously reported by Chambrin et al. (0.6 alarm/hrin14). Gabor et al. found a higher rate of sound increase(37 � 20 to 72 � 13 times per hour of sleep),13 but they didnot identify the source of each sound. The high alarm rateobserved in the present study is certainly due to themethod used, but could also be due to the target sedationlevel (Ramsay score of 2 to 3). A patient with less sedationmay fight the ventilator. However, this low target is widelyrecommended and validated.19–24 The lowest alarm ratewas observed during ACV with AF, and the highest ratewas observed during conventional ACV. It is not surprisingthat a pressure-controlled mode, such as AF, is associatedwith a reduction in pressure alarms. This is in accordancewith studies demonstrating a reduction of peak inspiratorypressure during pressure-controlled ventilation.3,4,15 Mul-tivariate analysis found only 2 factors associated with alower alarm rate: midazolam doses and AF activation.

It may be beneficial to reduce the number of alarms in anICU, because alarms are partly responsible for the highnoise level in an ICU.5–7 The first consequence of alarmnoise could be sleep disruption.25,26 In addition, an excessof false positive alarms may decrease alarm efficiency: Only26.8% of ventilator-generated alarms lead to an action.14

Only one half of critical alarms are correctly identified byICU staff.8 Reducing the total number of alarms shouldreduce noise in the ICU and improve alarm efficiency.9 Wealso showed that AF activation was associated with areduction in a surrogate marker of care interruption, acti-vation of the silence knob, which could help reduce cross-infections, because the ICU environment can be a reservoirfor pathogens.27,28 Further studies will be needed to con-firm the alarm reduction observed during AF ACV.

The major limitations of our study are the absence offixed alarm limits and the unblinded design. The attendingphysician was allowed to modify alarm limits as usual,because no clear recommendations have been published.29

Physicians are unlikely to have set high alarm limits inview of the very high alarm rate observed in this study.Furthermore, no difference was observed for changes ofalarm limits, apart from the upper-pressure alarm limit setabove 50 cmH2O (more frequent in the AF� group), which

should have reduced the alarms rate in that group. The trialcould not be blinded because the ventilator screen displaysventilator settings. However, the end points (gas exchange,alarms, and outcome) were assessed objectively. Lastly, nomodification of clinical outcome such as decreased dura-tion of mechanical ventilation or mortality was observed. Itis not surprising because no new mechanical ventilationmode has improved clinical outcome.10 In particular, alarge-scale study comparing pressure-controlled ventila-tion with volume-controlled ventilation did not find anydirect benefit.30 In the present study, sedation was con-trolled by a nurse-driven protocol, but it can be assumedthat if physicians had prescribed sedation, they would haveincreased sedation during conventional ACV to decreasethe alarm rate, which could have accentuated the trendsobserved in this study.

CONCLUSIONSACV with AF appears to be safe in terms of gas exchangeand clinical outcome in this first long-term around-the-clock clinical evaluation. AF was associated with a markeddecrease in ventilator-generated alarms. The beneficial ef-fect of such a reduction of alarm rate on the comfort ofpatients and ICU staff (quality of sleep and stress) deservesfurther evaluation.

REFERENCES1. Branson RD, Davis K Jr. Dual control modes: combining

volume and pressure breaths. Respir Care Clin N Am2001;7:397–408

2. Campbell RS, Davis BR. Pressure-controlled versus volume-controlled ventilation: does it matter? Respir Care 2002;47:416 –24

3. Alvarez A, Subirana M, Benito S. Decelerating flow ventilationeffects in acute respiratory failure. J Crit Care 1998;13:21–5

4. Guldager H, Nielsen SL, Carl P, Soerensen MB. A comparisonof volume control and pressure-regulated volume controlventilation in acute respiratory failure. Crit Care (London)1997;1:75–7

5. Freedman NS, Gazendam J, Levan L, Pack AI, Schwab RJ.Abnormal sleep/wake cycles and the effect of environmentalnoise on sleep disruption in the intensive care unit. Am JRespir Crit Care Med 2001;163:451–7

6. Meyer TJ, Eveloff SE, Bauer MS, Schwartz WA, Hill NS,Millman RP. Adverse environmental conditions in the respira-tory and medical ICU settings. Chest 1994;105:1211–6

7. Walder B, Francioli D, Meyer JJ, Lancon M, Romand JA. Effectsof guidelines implementation in a surgical intensive care unitto control nighttime light and noise levels. Crit Care Med2000;28:2242–7

8. Cropp AJ, Woods LA, Raney D, Bredle DL. Name that tone.The proliferation of alarms in the intensive care unit. Chest1994;105:1217–20

9. Siebig S, Sieben W, Kollmann F, Imhoff M, Bruennler T,Rockmann F, Gather U, Wrede CE. Users’ opinions on inten-sive care unit alarms—a survey of German intensive care units.Anaesth Intens Care 2009;37:112–6

10. Branson RD, Johannigman JA. What is the evidence base forthe newer ventilation modes? Respir Care 2004;49:742–60

11. Ramsay MA, Savege TM, Simpson BR, Goodwin R. Controlledsedation with alphaxalone-alphadolone. Br MedJ 1974;2:656–9

12. Vincent JL, de Mendonca A, Cantraine F, Moreno R, Takala J,Suter PM, Sprung CL, Colardyn F, Blecher S. Use of the SOFAscore to assess the incidence of organ dysfunction/failure inintensive care units: results of a multicenter, prospective study.Working group on “sepsis-related problems” of the EuropeanSociety of Intensive Care Medicine. Crit Care Med 1998;26:1793–800



13. Gabor JY, Cooper AB, Crombach SA, Lee B, Kadikar N,Bettger HE, Hanly PJ. Contribution of the intensive care unitenvironment to sleep disruption in mechanically ventilatedpatients and healthy subjects. Am J Respir Crit Care Med2003;167:708 –15

14. Chambrin MC, Ravaux P, Calvelo-Aros D, Jaborska A, ChopinC, Boniface B. Multicentric study of monitoring alarms in theadult intensive care unit (ICU): a descriptive analysis. IntensCare Med 1999;25:1360–6

15. Edibam C, Rutten AJ, Collins DV, Bersten AD. Effect ofinspiratory flow pattern and inspiratory to expiratory ratio onnonlinear elastic behavior in patients with acute lung injury.Am J Respir Crit Care Med 2003;167:702–7

16. Kreit JW, Capper MW, Eschenbacher WL. Patient work ofbreathing during pressure support and volume-cycled me-chanical ventilation. Am J Respir Crit Care Med 1994;149:1085–91

17. Mireles-Cabodevila E, Chatburn RL. Work of breathing inadaptive pressure control continuous mandatory ventilation.Respir Care 2009;54:1467–72

18. Cinnella G, Conti G, Lofaso F, Lorino H, Harf A, Lemaire F,Brochard L. Effects of assisted ventilation on the work ofbreathing: volume-controlled versus pressure-controlled ven-tilation. Am J Respir Crit Care Med 1996;153:1025–33

19. Brook AD, Ahrens TS, Schaiff R, Prentice D, Sherman G,Shannon W, Kollef MH. Effect of a nursing-implementedsedation protocol on the duration of mechanical ventilation.Crit Care Med 1999;27:2609–15

20. Heffner JE. A wake-up call in the intensive care unit. New EnglJ Med 2000;342:1520–2

21. Jacobi J, Fraser GL, Coursin DB, Riker RR, Fontaine D, Witt-brodt ET, Chalfin DB, Masica MF, Bjerke HS, Coplin WM,Crippen DW, Fuchs BD, Kelleher RM, Marik PE, Nasraway SAJr., Murray MJ, Peruzzi WT, Lumb PD. Clinical practiceguidelines for the sustained use of sedatives and analgesics inthe critically ill adult. Crit Care Med 2002;30:119–41

22. Kollef MH, Levy NT, Ahrens TS, Schaiff R, Prentice D,Sherman G. The use of continuous i.v. sedation is associatedwith prolongation of mechanical ventilation. Chest 1998;114:541– 8

23. Kress JP, Pohlman AS, Hall JB. Sedation and analgesia in theintensive care unit. Am J Respir Crit Care Med 2002;166:1024–8

24. Kress JP, Pohlman AS, O’Connor MF, Hall JB. Daily interrup-tion of sedative infusions in critically ill patients undergoingmechanical ventilation. New Engl J Med 2000;342:1471–7

25. Freedman NS, Kotzer N, Schwab RJ. Patient perception ofsleep quality and etiology of sleep disruption in the intensivecare unit. Am J Respir Crit Care Med 1999;159:1155–62

26. Parthasarathy S, Tobin MJ. Sleep in the intensive care unit.Intens Care Med 2004;30:197–206

27. Bures S, Fishbain JT, Uyehara CF, Parker JM, Berg BW.Computer keyboards and faucet handles as reservoirs ofnosocomial pathogens in the intensive care unit. Am J InfectControl 2000;28:465–71

28. Eggimann P, Pittet D. Infection control in the ICU. Chest2001;120:2059–93

29. Les recommandations des experts de la Societe de Reanimationde Langue Francaise: monitorage de la ventilation mecanique.Reanim Urgences 2000;9:407–12

30. Esteban A, Alia I, Gordo F, de Pablo R, Suarez J, Gonzalez G,Blanco J. Prospective randomized trial comparing pressure-controlled ventilation and volume-controlled ventilation inARDS. For the Spanish Lung Failure Collaborative Group.Chest 2000;117:1690–6


Oxygen Delivery During Transtracheal Oxygenation:A Comparison of Two Manual DevicesFrancois Lenfant, MD, PhD,* Didier Pean, MD,† Laurent Brisard, MD,† Marc Freysz, MD, PhD,*and Corinne Lejus, MD, PhD†

BACKGROUND: The Manujet™ and the ENK Oxygen Flow Modulator™ (ENK) deliver oxygenduring transtracheal oxygenation. We sought to describe the ventilation characteristics of these2 devices.METHODS: The study was conducted in an artificial lung model consisting of a 15-cm ringedtube, simulating the trachea, connected via a flow analyzer and an artificial lung. A 15-gaugetranstracheal wire reinforced catheter was used for transtracheal oxygenation. The ENK andManujet were studied for 3 minutes at respiratory rates of 0, 4, and 12 breaths/min, with andwithout the artificial lung, in a totally and a partially occluded airway. Statistical analysis wasperformed using analysis of variance followed by a Fisher exact test; P � 0.05 was consideredsignificant.RESULTS: Gas flow and tidal volume were 3 times greater with the Manujet than the ENK(approximately 37 vs 14 L � min�1 and 700 vs 250 mL, respectively) and were not dependent onthe respiratory rate. In the absence of ventilation, the ENK delivered a 0.6 � 0.1 L � min�1

constant gas flow. In the totally occluded airway, lung pressures increased to 136 cm H2O after3 insufflations with the Manujet, whereas the ENK, which has a pressure release vent, generatedacceptable pressures at a low respiratory rate (4 breaths/min) (peak pressure at 27.7 � 0.7 andend-expiratory pressure at 18.8 � 3.8 cm H2O). When used at a respiratory rate of 12breaths/min, the ENK generated higher pressures (peak pressure at 95.9 � 21.2 andend-expiratory pressure at 51.4 � 21.4 cm H2O). In the partially occluded airway, lung pressureswere significantly greater with the Manujet compared with the ENK, and pressures increased withthe respiratory rate with both devices. Finally, the gas flow and tidal volume generated by theManujet varied proportionally with the driving pressure.DISCUSSION: This study confirms the absolute necessity of allowing gas exhalation between 2insufflations and maintaining low respiratory rates during transtracheal oxygenation. In the caseof total airway obstruction, the ENK may be less deleterious because it has a pressure releasevent. Using a Manujet at lower driving pressures may decrease the risk of barotrauma and allowthe safe use of higher respiratory rates. (Anesth Analg 2010;111:922–4)

Transtracheal oxygenation may be lifesaving in “can-not intubate/cannot ventilate” patients. The FrenchSociety of Anesthesiologists has recently published

updated guidelines for difficult airway management andrecommends the use of dedicated devices for transtrachealoxygen delivery.1 Two devices are available, the ENKOxygen Flow Modulator™ (ENK) and the Manujet™. Bothhave been studied in a pig model.2 Both maintain oxygen-ation, but the ENK seemed to achieve better ventilationbecause of a continuous flow that provides CO2 washoutbetween insufflations. Very little is known concerning thelung pressures generated with these 2 devices.3 Because itis technically impossible to assess these variables in clinicalsituations, we conducted this study in an artificial lung to

characterize these 2 devices in terms of oxygen flow, tidalvolumes, and airway pressures.

METHODSThe Manujet III™ jet ventilator (VBM�; Vitrolles, France)was connected to a 4-bar oxygen source and the drivingpressure was set at 3 bar. Manual activation of the triggerdelivered oxygen to a transtracheal catheter. The ENKOxygen Flow Modulator™ (COOK�, Charenton, France)was connected to an oxygen wall flow regulator set at 15L/min. Oxygen was delivered when the 5 holes in theENK’s perforated tube were manually occluded.

The 2 devices were evaluated by connecting them to a15-gauge (2-mm internal diameter, 75-mm length) wirereinforced transtracheal catheter (COOK). The catheter wasinserted into the simulated trachea, a 15-cm ringed tube(Intersurgical, Fontenay sous Bois, France) occluded at oneend by a plastic cork. The other end of the tube wasconnected to an IMT Medical Flow Analyzer PF-300™ (IMTMedical�, Buchs, Switzerland) to measure flow rate, tidalvolume, peak pressure, mean pressure, and positive end-expiratory pressure. An adult SmartLung™ (IMT Medical)was connected to the analyzer output. The SmartLungsettings were as follows: total lung volume � 1000 mL;airway resistance � 5 mbar/L/s; and lung compliance � 30mbar/mL (Fig. 1).

From the *Department of Anesthesiology and Intensive Care, CHU de Dijon,General Hospital, Dijon; and †Department of Anesthesiology, CHU deNantes, Hotel Dieu, Nantes, France.


Supported by departmental sources except for the PF-300™ flow analyzer,which was provided free of charge by SEBAC (Gennevilliers, France).


Address correspondence and reprint requests to Dr. Francois Lenfant,Department of Anesthesiology, CHU de Dijon, General Hospital, 3 rue duFaubourg Raines, 21033 Dijon Cedex, France. Address e-mail [email protected].



Two investigators manually delivered 1-second insuffla-tions for 3 minutes at the rate of 4 breaths/min and 12breaths/min with the plastic cork in place, simulating totalupper airway obstruction, and after making a 2-mm hole inthe plastic cork, simulating partial upper airway obstruc-tion. A metronome was used to keep the rate and theduration of the insufflations constant. Data were recordedfor 3 minutes with an insufflation rate of 4 breaths/minwith the Manujet driving pressure set at 0.5, 1.0, 1.5, 2.0, 2.5,and 3 bar. The SmartLung was removed and the gas flowdelivery rates were recorded without insufflations (0breaths/min).

All of the variables were sampled and recorded every 10milliseconds using Flowlab™ software and transferred to apersonal computer as Excel™ files. Statistical analysis wasperformed using StatPlus� software (AnalystSoft, StatPlusMac, Version 2008; http://www.analystsoft.com/fr). Con-tinuous data were expressed as the mean � SD. Data wereanalyzed using analysis of variance followed by a Fisherexact test. P � 0.05 was considered significant.

RESULTSThe mean duration of the manually delivered insufflationswas 1.01 � 0.06 second at 4 breaths/min and 0.95 � 0.05second at 12 breaths/min for the ENK, and 0.97 � 0.08second at 4 breaths/min and 0.95 � 0.08 second at 12breaths/min for the Manujet. During insufflations, peakflow was 36.0 � 6.3 L � min�1 with the Manujet and 13.7 �2.42 L � min�1 with the ENK. When holes were not occluded,the ENK delivered oxygen at the rate of 0.7 � 0.1 L � min�1.

Table 1 shows the airway pressure delivered by the ENKduring total airway occlusion. With the Manujet, airwaypressure was 34 cm H2O after the first insufflation and 136cm H2O after the second, therefore the experiment wasstopped. During total airway occlusion, airway pressureswere significantly lower with the ENK.

Table 2 shows tidal volumes, minute volumes, andmean and peak airway pressures during partial upperairway obstruction. Values were significantly higherwith the Manujet. Table 3 shows how the Manujet gasflows and tidal volumes increased with an increase indriving pressure.

DISCUSSIONThe Manujet delivers oxygen at a higher peak flow ratethan the ENK. As a result, 1-second insufflations with theManujet result in larger tidal and minute volumes andhigher airway pressures than with the ENK. Our resultsconfirm the results of Flint et al.3 The tidal volumes wemeasured for the Manujet (650 mL) are close to thosecalculated from the minute volumes reported by Flint et al.3

(627 mL). For the ENK, we measured a 240-mL tidalvolume, whereas Flint et al. measured a 156-mL tidalvolume. This discrepancy may be explained by the differ-ences in both the model and the methodologies used in the2 studies.

Yildiz et al.4 found that the oxygen that flows from theENK between manual insufflations enhances oxygenation,especially at low respiratory rates. Preussler et al.2 foundthat the constant flow improves CO2 elimination. Becausethe Manujet generates higher airway pressure, it may resultin better alveolar recruitment and better lung ventilation.Both investigators found that decreasing the Manujet driv-ing pressure allows it to deliver higher respiratory rateswithout excessive airway pressure.2,4

Because the main risk of transtracheal oxygenation isbarotrauma, the Manujet should not be used in case oftotal upper airway obstruction. The ENK seems to be lessproblematic because it delivers gas at a lower pressureand has a pressure release vent that allows gas to escapebetween insufflations.5 However, even with the ENK, thepeak and mean airway pressures may reach dangerouslyhigh levels at high respiratory rates (12 breaths/min)during total airway obstruction. Increasing the respira-tory rate results in short expiratory times and, as aconsequence, an increase in the volume of the gastrapped in the lung. Our results confirm the absolutenecessity of ensuring that the insufflated gas is exhaledduring the expiratory period.6 Sufficient exhalation caneasily be checked by watching the chest fall beforedelivering a second tidal volume.

When using the Manujet during partial upper airwayobstruction, great attention should be given to deliveringinsufflations that last �1 second and using low respiratory

Figure 1. Photograph showing the ENK Oxygen Flow Modulator™device connected to the simulated trachea, the PF-300™ flowanalyzer, and the artificial lung. The trachea is occluded at one endby a plastic cork.

Table 1. Airway Pressure Delivered by the ENKOxygen Flow Modulator™ During TotalAirway Obstruction

Respiratoryrate (breaths/

min)

Mean pressureduring

inspiration(cm H2O)

Peak pressure(cm H2O)

Mean pressureduring

expiration(cm H2O)

0 13.2 � 0.04 21.9 � 4.4 27.7 � 0.7 18.8 � 3.812 62.7 � 24.3* 95.9 � 21.2* 51.4 � 21.4*

* P � 0.05, 4 vs 12 breaths/min.


rates, to avoid excessive airway pressure. The most impor-tant goal is to deliver oxygen to the lung, as recommendedby Cook and Alexander.7

REFERENCES1. Combes X, Pean D, Lenfant F, Francon D, Marciniak B, Legras

A. Difficult airway management devices: establishment andmaintenance—question 4. Societe Francaise d’Anesthesie et deReanimation. Ann Fr Anesth Reanim 2008;27:33–40

2. Preussler NP, Schreiber T, Huter L, Gottschall R, Schubert H,Rek H, Karzai W, Schwarzkopf K. Percutaneous transtrachealventilation: effects of a new oxygen flow modulator on oxygen-ation and ventilation in pigs compared with a hand triggeredemergency jet injector. Resuscitation 2003;56:329–33

3. Flint NJ, Russell WC, Thompson JP. Comparison of differentmethods of ventilation via cannula cricothyroidotomy in atrachea-lung model. Br J Anaesth 2009;103:891–5

4. Yildiz Y, Preussler NP, Schreiber T, Hueter L, Gaser E, SchubertH, Gottschalll R, Schwarzkopf K. Percutaneous transtrachealemergency ventilation during respiratory arrest: comparison ofthe oxygen flow modulator with a hand-triggered emergency jetinjector in an animal model. Am J Emerg Med 2006;24:455–9

5. Hamaekers A, Borg P, Enk D. The importance of flow andpressure release in emergency jet ventilation devices. PaediatrAnaesth 2009;19:452–7

6. Bourgain JL, Desruennes E, Fischler M, Ravussin P. Transtra-cheal high frequency jet ventilation for endoscopic airwaysurgery: a multicentre study. Br J Anaesth 2001;87:870–7

7. Cook TM, Alexander R. Major complications during anaesthesiafor elective laryngeal surgery in the UK: a national survey of theuse of high-pressure source ventilation. Br J Anaesth2008;101:266–72

Table 2. Gas Flow Rate During Inspiration, Tidal Volume, Minute Volume, and Mean and Peak AirwayPressures Measured with the Manujet and the ENK Oxygen Flow Modulator™ (ENK) During PartialAirway Obstruction

Respiratory rate(breaths/min)

Flow(L � min�1)

Tidal volume(mL)

Minute volume(L � min�1)

Mean pressure(cm H2O) Peak pressure (cm H2O)

ENK 4 12.4 � 2.0 263.0 � 26.6 0.91 � 0.01 11.3 � 1.8 14.1 � 0.3Manujet 4 39.9 � 4.9 676.6 � 62.3 2.67 � 2.04 16.7 � 6.8 27.7 � 5.0ENK 12 12.0 � 1.6 223.9 � 27.3 2.69 � 0.07 12.1 � 2.0 18.2 � 1.2Manujet 12 38.8 � 5.8 753.4 � 88.9 9.06 � 0.02 16.7 � 6.8 58.3 � 25.0

All differences between Manujet and ENK were significant at P � 0.05.

Table 3. Gas Flow Rate During Inspiration andTidal Volume Versus Driving Pressure for theManujet at a Respiratory Rate of 4 breaths/minDriving pressure

(bar)Gas flow

(L � min�1)Tidal volume

(mL)0.5 12.6 � 1.6 246.0 � 15.91.0 17.4 � 2.5 304.0 � 34.91.5 23.7 � 2.9 356.0 � 25.22.0 28.2 � 4.2 485.3 � 18.02.5 33.9 � 3.5 626.3 � 5.53.0 37.1 � 5.8 653.3 � 50.3

Comparison of Two Transtracheal Oxygenation Manual Devices


Two Serial Check Valves Can PreventCross-Contamination Through Intravenous TubingDuring Total Intravenous AnesthesiaOliver C. Radke, MD, PhD,*† Katrin Werth, MD,‡ Margarete Borg-von-Zepelin, MD, PhD,§Petra Saur, MD, PhD,� and Christian C. Apfel, MD, PhD†

BACKGROUND: Nonsterile handling of propofol for anesthesia has been linked with severesepsis and death. Placing a single check valve in the IV tubing does not prevent retrogradeascension of pathogens into propofol-filled syringes, so we designed an IV tubing set withmultiple check valves. To estimate the efficacy of this design, we measured the concentration ofpathogens detected upstream in the IV tubing in relation to the pathogen concentration in amodel of a contaminated patient.METHODS: A glass container with a rubber sealed port was filled with a suspension of eitherbacteria or phagocytes and kept at 37°C (“contaminated patient” model). A bag of normal saline wasconnected to an IV cannula, punctured through the rubber sealed port of the patient model. Twoadditional sidestream infusion lines were connected to syringes in 2 standard infusion pumps. Oneof the syringes contained propofol and the other contained normal saline as a substitute for an opioidpreparation. After 5 hours of infusion, we obtained samples from different parts of the infusion linesand syringes. The samples were streaked out on blood agar plates and incubated at 37°C for 24hours. We repeated this experiment with 6 different pathogens.RESULTS: We incubated 825 agar plates. Whereas the concentration of bacteria and phago-cytes in the “patient” had significantly increased during the 5-hour experiments (positive control),no bacterial growth could be detected in any of the incubated plates.CONCLUSION: The data from this experimental setting suggest that the design with multiplecheck valves in paired configuration prevents retrograde contamination. Of note, this doesnot permit the reuse of propofol syringes because reusing is against the manufacturer’srecommendations. (Anesth Analg 2010;111:925–8)

Propofol infusions have been identified as a mediumsupporting the growth of microorganisms.1 Contami-nation can occur during the preparation of the drug2

and during its administration.3–9 Clinicians often put a single1-way valve into the IV line to prevent backflow but, unfor-tunately, these valves may not prevent ascension of bacteriafrom the patient into the tubing and the syringes.10

To overcome the shortcomings of single check valves,we propose placing 2 check valves in series.11 We hypoth-esized that the serial placement of check valves working inpairs would improve their function as a barrier to blood-borne pathogens. Because fluids are not compressible, the 2

valves in each pathway would work in conjunction, ensur-ing that the column of fluid between them did not movebackward even if one of the valves failed.

To estimate the efficacy of this design, we conducted aseries of experiments with a model of a contaminatedpatient. Our main objective was to measure the concentra-tion of pathogens detected upstream in the IV tubing inrelation to the pathogen concentration in the patient model.

METHODSTests for Bacterial ContaminationAs a model of a contaminated patient, a 3000-mL glasscontainer with a rubber-sealed port was filled with 2500 mLof a liquid medium (Fig. 1). The medium consisted of 25 gpeptone, 5 g NaCl, 10 g dry meat extract, and 10 g glucosein distilled water. For each experimental run, we infectedthe medium with a bacteria suspension to achieve a con-centration of at least 1 � 106 mL�1 colony-forming units(CFUs). The bacteria were strains isolated from patients ofour university hospital (Staphylococcus aureus, Staphylococ-cus epidermidis, Escherichia coli, Proteus mirabilis, and Pseudo-monas aeruginosa).

We assembled the IV tubing set (“TIVA-Set”; SmithMedical Deutschland GmbH, Grasbrunn, Germany) shownin Figure 2 with 4 strategically placed check valves.11 A500-mL bag of 0.9% NaCl was connected via the TIVA-Setto an 18-gauge IV cannula. This cannula was puncturedthrough the rubber-sealed port of the patient model (Fig. 2).The 2 infusion lines of the TIVA-Set were connected to50-mL syringes (B.Braun, Melsungen, Germany) in Perfu-sor� fm and perfusor compact infusion pumps (B.Braun).

From the *Department of Anesthesiology and Critical Care Medicine,University Hospital Carl Gustav Carus Dresden at the TU Dresden,Dresden, Germany; †Perioperative Clinical Research Core, Department ofAnesthesia and Perioperative Care, University of California at San Francisco,UCSF Medical Center at Mount Zion, San Francisco, California; ‡Depart-ment of Pediatrics, Klinikum Aschaffenburg, Aschaffenburg, Germany;§Institute of Infectiology & Pathobiology, Hufeland Klinikum GmbH, Stan-dort Muhlhausen, Muhlhausen, Germany; and �Department of Anesthesia,Intensive Care Medicine and Pain Medicine, Sana Kliniken Lubeck GmbH,Lubeck, Germany.


Funded by departmental resources. Medex Medical, now Smith MedicalInternational, Watford, UK, donated the IV tubings. B.Braun, Melsungen,Germany, donated the infusion pumps for the experiments. AstraZenecaGmbH, Wedel, Germany, donated the propofol. None of these companiesprovided financial support.

Address correspondence and reprint requests to Oliver C. Radke, MD,PhD, DEAA, Klinik und Poliklinik fur Anasthesiologie und Intensiv-therapie, Fetscherstr. 74, 01307 Dresden, Germany. Address e-mail [email protected].

Copyright © 2010 International Anesthesia Research SocietyDOI: 10.1213/ANE.0b013e3181eb7194


One of the syringes contained propofol 1% (Disoprivan�;AstraZeneca, Wedel, Germany) and the other containednormal saline as a substitute for an opioid preparation. Theinfusion pumps were approximately 30 cm above the levelof the port in the patient model.

The flow of the normal saline (Table 1) was regulatedby a drip pump (Infusomat fmS; B.Braun). All pumpswere operated by a microprocessor-controlled fluid man-ager (fm controller; B.Braun) that was programmed torun a 5-hour infusion program (Table 1). This programwas designed to simulate an actual anesthetic. For eachpathogen, the experimental run was repeated at least 8times (Table 2).

Before and after the infusion program, we took a sampleof 100 �L from the bacteria suspension in the patient modeland prepared a dilution series to estimate the bacteriaconcentrations. Additional undiluted samples of 500 �Lwere obtained under sterile conditions from the followinglocations (Fig. 2): IV tubing close to the cannula; bothmicrotubings leading from the drug pumps to the main IVline; and leftover drugs in both 50-mL syringes.

Each sample was spread out on 3 blood agar plates (100 �Lper plate), and the plates were incubated at 37°C for 24 hours.At the end of the incubation, we manually counted thenumber of CFUs on the plates. Bacteria concentrations in thepatient model were compared using the Student t test forrepeated measurements (SPSS 16.0.2; SPSS Inc., Chicago, IL).

Test for Nonbacterial ContaminationBecause nonbacterial pathogens such as viruses or prions aremuch smaller than bacteria, we used a bacteriophage (Bacte-riophage T3) as a sample of such a small infectious unit. TheBacteriophage T3 infects E coli bacteria and can be detected bytheir ability to lyse the bacteria. The experimental runs withthe phage were conducted almost identically to the runs withbacteria, but because a phage cannot replicate on its own, weadded E coli B14 to the patient model. The blood agar platesfor detection of pathogens in the samples were also floodedwith E coli B14 to provide the hosts for the phages’ replication.After incubation, the concentration of phages in a sample wascalculated from the number of lytic holes in the bacterialspread on the blood agar plate.

Figure 1. Photograph of the “contaminated patient” model. Therubber disc that seals the connector at the bottom has beenpunctured with a standard 16-gauge IV cannula. The thermometer atthe top is connected to the heated stirrer to ensure a constanttemperature of 37°C.

Figure 2. Placement of check valves andsample sites. Schematic of the IV tubing.Four check valves are placed in pairedconfigurations to prevent retrograde flow.The 5 sample sites are marked.

Table 1. Infusion Program

Duration(min)

Carrierdrip rate(mL/h)

Propofolinfusion

rate (mL/h)

Salineinfusion

rate (mL/h)Before induction 30 100 0 0Maintenance I 120 100 10 10Fluids run dry 30 0 10 10Maintenance II 120 100 10 10

Standardized infusion program, controlled by a microprocessor. The programconsisted of 4 parts mimicking a total IV anesthesia in clinical practice withlow drug infusion rates. Total duration of each experimental run was 5 hours.

Can Serial Check Valves Prevent Cross-Contamination?


RESULTSThe test runs were repeated 8 to 12 times with each of the5 bacterial isolates and the phage. We performed a total of55 runs. For each bacterial strain, we found a statisticallysignificant increase in concentration in the patient model(Table 2). Proteus mirabilis had the highest increase inconcentration (67-fold). S aureus had the second highestincrease (10-fold), whereas the others increased their start-ing concentration 3- to 6-fold. The phages’ replication factorwas even higher, ranging from 10 to 333 depending on thestarting concentration.

In 55 experimental runs, we obtained 275 samples fromdifferent parts of the IV tubing and the drug syringes andincubated 825 plates. Even with the significant increase ofthe bacteria and phage concentrations in the patient model,no plates grew colonies or had lytic holes.

DISCUSSIONSoon after propofol was introduced into clinical practice,there were several cases of severe sepsis resulting from theuse of propofol that was contaminated with bacteria (e.g.,see Veber et al.12). Contamination by the anesthesia pro-vider can be found in 8% to 11% of syringes wheneverpropofol is drawn up for clinical use,13 even when themanufacturer’s recommendations are closely adhered to.3

Because bacterial growth in propofol is slow at roomtemperature and begins after a latency period of 5 to 6hours,14,15 the manufacturer recommends discarding anypropofol within 6 hours of opening the vial. Serious infec-tious outbreaks of sepsis were always attributed to propo-fol that had been opened several hours ahead of use(usually the day before); therefore, bacterial contaminationfrom opening the vials could grow to achieve bacterialcounts that were high enough to cause severe sepsis.

The risk of bacterial contamination by the anesthesiaprovider can be addressed by hand washing, aseptic tech-niques, and using propofol within 6 hours after opening.Transfer of pathogens from one patient to another bymeans of propofol syringes has not yet been documented,but in vitro studies have shown that retrograde contami-nation occurs easily.16

With our experimental setup, we wanted to measure theincidence and the magnitude, if any, of retrograde bacterialcontamination through a special IV tubing set for total IVanesthesia (TIVA).

Validity of the Patient ModelThe pathogens tested in this study are representative ofinfections found in the patients of our hospital. S aureus wasused because of its ubiquitous presence, accounting for 12%of all nosocomial infections.17,18 S aureus strains with resis-tance to a wide spectrum of common antibiotics (not onlymethicillin-resistant S aureus) are an increasing problem inmodern health care.19,20 S epidermidis is an important causeof infection in patients with a compromised immune sys-tem, or who have indwelling catheters. This bacterium hasa special ability to produce a matrix (“biofilm”) that allowsthem to adhere to artificial surfaces, such as those ofmedical prostheses.21,22 E coli is frequently found in noso-comial infections, and strains that are highly resistant toantibiotics are increasingly isolated.20 Proteus mirabilis hasan exceptionally high intrinsic motility and could poten-tially swarm actively into IV lines. Pseudomonas aeruginosahas a high intrinsic resistance to antibiotics and disinfec-tants,20 and some strains can also generate a biofilm.23

The Bacteriophage T3 is a well-known phage without anypathogenic potency for humans. It has approximately thesame size as a virus and therefore the same potential forcontamination. Its lytic effect on E coli bacteria makes iteasy to detect.

The bacterial load in our contaminated patient model wasat least 1 � 106 CFUs mL�1. Studies in patients with endocar-ditis measured a maximum of 173 CFUs per mL of blood,24

and children with manifested infections (meningitis, peritoni-tis) had bacterial loads of up to 104 CFUs per mL of blood.25

Even blood samples from infected indwelling catheters mea-sured only up to 300 CFUs per mL.26 Therefore, our contami-nated patient model was �300 times more infectious than aworst case patient in real-life conditions.

Validity of the IV SetupWith our experimental setup, we tried to model real-lifeconditions during TIVA. We expected the risk for contami-nation to increase with duration of infusion. Also, highinfusion rates would flush the IV tubing and make it lesslikely for pathogens to ascend. To facilitate the migration ofbacteria from our contaminated patient model into thetubing, we ran an infusion pattern with low infusion ratesover a period of 5 hours.

It is obvious that an IV line without any check valveswould not protect against contamination. In a survey of IVtubings without check valves in the operating room, the

Table 2. Bacteria Concentrations in the Patient Model

Before After P valueGrowth rate

(95% CI) Runs Plates106 · mL�1 CFUs 106 · mL�1 CFUs

Staphylococcus aureus 12.3 � 6.3 110.6 � 4.0 �0.001 10.4 (8.1–12.6) 12 180Staphylococcus epidermidis 9.4 � 1.2 34.0 � 25.2 0.028 3.6 (1.5–5.8) 8 120Escherichia coli 50.7 � 38.8 249.4 � 135.4 �0.001 5.9 (3.8–8.0) 9 135Proteus mirabilis 1.0 � 0.0 67.3 � 14.2 �0.001 67.3 (56.5–78.2) 8 120Pseudomonas aeruginosa 20.7 � 4.0 66.6 � 12.0 �0.001 3.4 (2.4–4.5) 9 135

106 · mL�1 PFUs 106 · mL�1 PFUsBacteriophage T3 1.4 � 1.7 37.8 � 43.7 0.015 97.2 (21.4–173.0) 9 135

CI � confidence interval; CFU � colony-forming unit; PFU � plaque-forming unit.Concentrations are given as mean � standard deviation in 106 CFUs or 106 PFUs per mL. Growth rate is the CFUafter divided by CFUbefore. Total number of plates: 825.


authors found traces of blood from the patient in 3.3% of thecases.4 In an experiment with a contaminated patient modelsimilar to our setup but where the IV tubing was filled withliquid culture medium instead of saline and propofol, theresearchers found a bacterial contamination of the line close tothe patient in 87% of the cases,27 even though the line was“protected” against backflow with a 1-way valve. The re-searchers even increased the “venous” pressure of their modelto 150 mm Hg to simulate occlusion caused by noninvasivearterial blood pressure measurements.16 This maneuver didnot cause increased ascension of the pathogens into theinfusion tubing farther away from the patient. In our model,we did not increase the venous pressure to such high read-ings. Because the authors of the aforementioned article did notfind an increase in contamination rates even with only a singlecheck valve, we do not expect to see an increase in contami-nation rates with our paired valves configuration.

The lipid preparation containing propofol supports bac-terial growth, but only after a latency period of 5 to 6hours.14,15 Additionally, the propofol preparation we usedcontained EDTA, an additive that suppresses bacterialgrowth.28 The bacteriostatic effect of the EDTA could haveinhibited the growth of small amounts of bacteria that infact entered the IV tubing and thus prevented their detec-tion. However, running a propofol/saline mixture in the IVtubing is closer to real-life conditions than running liquidmedium. Any bacteria that could not grow on our plateswith optimal conditions for the growth of human patho-gens probably would not have been capable of growing ina human organism either.

CONCLUSIONIt is highly unlikely that during 5 hours of propofol anesthesiathe syringe filled with propofol and EDTA will becomecontaminated by pathogens from the patient’s blood if 2 checkvalves are placed in series in the IV line. We did not find asingle trace of contamination in any of our 825 samples.

AUTHOR CONTRIBUTIONSOCR: Inventor of the TIVA-Set, initiator of the study, and themain author of the manuscript. KW: Performed all the micro-biological testing and did additional literature research.MB-v-Z: Designed and supervised the microbiological testing.PS: Coauthor of the manuscript and the advisor of studydesign. CCA: Senior author of the manuscript, statisticalanalysis and interpretation of data. All authors contributed tothe drafting of the article and gave final approval of the versionto be published.

DISCLOSUREThe University of Gottingen, Germany, holds a patent on theIV tubing set and is required by German law to share a fractionof all royalties with the inventor, Dr. Oliver C. Radke. To avoida conflict of interest, Dr. Radke was not involved in themicrobiological experiments.

REFERENCES1. Thomas DV. Propofol supports bacterial growth. Br J Anaesth

1991;66:2742. Lorenz IH, Kolbitsch C, Lass-Florl C, Gritznig I, Vollert B, Lingnau

W, Moser PL, Benzer A. Routine handling of propofol preventscontamination as effectively as does strict adherence to the manu-facturer’s recommendations. Can J Anaesth 2002;49:347–52

3. Lessard MR, Trepanier CA, Gourdeau M, Denault PH. Amicrobiological study of the contamination of the syringesused in anaesthesia practice. Can J Anaesth 1988;35:567–9

4. Trepanier CA, Lessard MR, Brochu JG, Denault PH. Risk ofcross-infection related to the multiple use of disposable sy-ringes. Can J Anaesth 1990;37:156–9

5. Zacher AN, Zornow MH, Evans G. Drug contamination fromopening glass ampules. Anesthesiology 1991;75:893–5

6. Bach A. Syringe— or lead change for TCI? [in German].Anaesthesist 1998;47:434 – 6

7. el Mikatti N, Dillon P, Healy TE. Hygienic practices of consul-tant anaesthetists: a survey in the north-west region of the UK.Anaesthesia 1999;54:13–8

8. Halkes MJ, Snow D. Re-use of equipment between patientsreceiving total intravenous anaesthesia: a postal survey ofcurrent practice. Anaesthesia 2003;58:582–7

9. Vonberg RP, Gastmeier P. Infection control measures in anaes-thesia [in German]. Anasthesiol Intensivmed NotfallmedSchmerzther 2005;40:453–8

10. Gretzinger DT, Cafazzo JA, Ratner J, Conly JM, Easty AC.Validating the integrity of one-way check valves for thedelivery of contrast solution to multiple patients. J Clin Engl1996;21:375–82

11. Radke O. Das TIVA-Set. Anasthesiol Intensivmed 2003;44:787–812. Veber B, Gachot B, Bedos JP, Wolff M. Severe sepsis after

intravenous injection of contaminated propofol. Anesthesiol-ogy 1994;80:712–3

13. Magee L, Godsiff L, Matthews I, Farrington M, Park GR.Anaesthetic drugs and bacterial contamination. Eur J Anaes-thesiol Suppl 1995;12:41–3

14. Langevin PB, Gravenstein N, Doyle TJ, Roberts SA, Skinner S,Langevin SO, Gulig PA. Growth of Staphylococcus aureus inDiprivan and intralipid: implications on the pathogenesis ofinfections. Anesthesiology 1999;91:1394–400

15. Sosis MB, Braverman B. Growth of Staphylococcus aureus infour intravenous anesthetics. Anesth Analg 1993;77:766–8

16. Sim J, Choi Y, Yoon M, Lee D, Leem J. The peripheral venouspressure changes during non-invasive blood pressure mea-surement. Can J Anaesth 1999;46:711–2

17. Bouza E. Intravascular catheter-related infections: a growingproblem, the search for better solutions. Clin Microbiol Infect2002;8:255

18. Emori TG, Gaynes RP. An overview of nosocomial infections,including the role of the microbiology laboratory. Clin Micro-biol Rev 1993;6:428–42

19. Gastmeier P, Geffers C. Nosocomial infections in Germany:what are the numbers, based on the estimates for 2006? [inGerman] Dtsch Med Wochenschr 2008;133:1111–5

20. Toniolo A, Endimiani A, Luzzaro F. Microbiology of postop-erative infections. Surg Infect (Larchmt) 2006;7:S13–6

21. Ziebuhr W, Hennig S, Eckart M, Kranzler H, Batzilla C,Kozitskaya S. Nosocomial infections by Staphylococcus epider-midis: how a commensal bacterium turns into a pathogen. IntJ Antimicrob Agents 2006;28:S14–20

22. Pascual A. Pathogenesis of catheter-related infections: lessonsfor new designs. Clin Microbiol Infect 2002;8:256–64

23. O’Toole GA, Kolter R. Flagellar and twitching motility arenecessary for Pseudomonas aeruginosa biofilm development.Mol Microbiol 1998;30:295–304

24. Werner AS, Cobbs CG, Kaye D, Hook EW. Studies on thebacteremia of bacterial endocarditis. JAMA 1967;202:199–203

25. Santosham M, Moxon ER. Detection and quantitation of bac-teremia in childhood. J Pediatr 1977;91:719–21

26. Flynn PM, Shenep JL, Barrett FF. Differential quantitation witha commercial blood culture tube for diagnosis of catheter-related infection. J Clin Microbiol 1988;26:1045–6

27. Eichler W, Schumacher J, Ohgke H, Klotz KF. Reuse of a set fortotal intravenous anaesthesia: safe against bacterial contami-nation? Eur J Anaesthesiol 2004;21:501–3

28. Hart B. ‘Diprivan’: a change of formulation. Eur J Anaesthesiol2000;17:71–3

Can Serial Check Valves Prevent Cross-Contamination?


TECHNICAL COMMUNICATION

Robot-Assisted Airway Support: A Simulated CasePatrick J. Tighe, MD, S. J. Badiyan, MD, I. Luria, BS, MS, S. Lampotang, PhD, and S. Parekattil, MD

Recent advances in telemedicine and robotically assisted telesurgery may offer advancedsurgical care for the geographically remote patient. Similar advances in tele-anesthesia will benecessary to optimize perioperative care for these patients. Although many preliminaryinvestigations into tele-anesthesia are underway, none involves remote performance ofanesthesia-related procedures. Here we describe simulated robotically assisted fiberopticintubations using an airway simulation mannequin. Both oral and nasal approaches tofiberoptic intubation were successful, but presented unique opportunities and challengesinherent to the robot’s design. Robotically assisted airway management is feasible usingmultipurpose surgical robotic systems. (Anesth Analg 2010;111:929–31)

Telemedicine has improved access to consultant-levelmedical care by minimizing geographic obstacles.Similar advances in telesurgery could offer expert

surgical care for the geographically remote patient. Suchsurgical advances have been historically preceded by simi-lar advances in anesthesia. True to this paradigm, earlyinvestigations into tele-anesthesia are already underway.1

However, none of these efforts has tackled one of thecentral tenets of anesthesiology: airway management.

In this report we present what we believe to be the firstdescription of a simulated robotically assisted fiberopticintubation. Instead of a procedure-specific device, we usedthe multipurpose DaVinci Surgical System type S (DVS)(Intuitive Surgical, Sunnyvale, California). This systemincorporates 4 separate robotic arms, 1 of which is mated toa high-definition stereoscopic camera. The workstationallows the person performing the procedure to view therobot’s camera output, control the limbs, and receive simul-taneous video input from third-party sources.2,3 The DVS isalready in widespread clinical use for a variety of urologic,gynecologic, and cardiothoracic surgical procedures.4 Inthis study involving an airway mannequin, we successfullyused the DVS for both oral and nasal fiberoptic intubation.

METHODSAn adult airway mannequin was placed at the head of astandard operating room bed. A stereoscopic video camera(DaVinci Surgical System, Intuitive Surgical) was mated tothe first robotic arm. This arm was situated above themannequin in a sagittal plane, angled to view the manne-quin from a caudal to rostral fashion. The second and thirdarms were equipped with large and small graspers (Figs. 1and 2). The fourth arm was equipped with a standard

fiberoptic bronchoscope (Karl Storz Endoscopy, El Seg-undo, California). The trocars used for mounting standardrobotic instruments were removed from this arm, allowingthe bronchoscope handle to slide into the mountingbracket. The bronchoscope handle was oriented to keep thetip actuator lever facing away from the arm, with thesuction port positioned to the side (Fig. 3). An externalbrace was required to keep the bronchoscope firmly seatedwithin the arm during manipulation of the actuator.

We manually loaded the endotracheal tube onto thebronchoscope. A camera was attached to the bronchoscopeand connected to the DVS Tilepro multivideo input system(Intuitive Surgical, Sunnyvale, California), allowing simul-taneous viewing of both the robot camera and broncho-scope camera in a single three-dimensional view (Fig. 4).

Before attempting intubation, the urologist spent ap-proximately 2 hours performing robotic dexterity exerciseswith the robotic surgical instruments and operator console.After training was complete, an anesthesiologist manuallyplaced the bronchoscope tip within the oropharnyx, and aurologist (S. Parekattil) used robot arms 2 and 3 to adjustthe bronchoscope actuator and steer the bronchosc-ope tip into the hypopharynx and through the vocalcords (Video 1; see Supplemental Digital Content 1,http://links.lww.com/AA/A171; see the Appendix forvideo legend). Next the bronchoscope tip was placedexternal to the nare, an anesthesiologist manually ad-vanced and rotated the bronchoscope, and an urologistused the robot to steer the tip into the hypopharynx andthrough the vocal cords.

RESULTSTwo intubation attempts were completed for this dem-onstration. During oral intubation, it took 75 seconds toadvance the bronchoscope tip from the oropharynx tocarinal visualization. For nasal intubation, it took 67seconds to advance the tip from nasal entry to carinalvisualization.

DISCUSSIONFiberoptic intubation is feasible with robotic equipment.We did not encounter significant differences between thenasal and oral intubation routes. Even if optimized foranesthetic practice, robotic-assisted anesthetic proceduresare not likely to become a part of routine anesthetic

From the University of Florida College of Medicine, Gainesville, Florida.


Funding was provided by the University of Florida College of Medicine,Department of Anesthesiology, as well as by National Institutes of Healthgrant UL1 RR029890 Clinical and Translational Science Award, NIH(NCRR).


Address correspondence to Patrick J. Tighe, MD, University of FloridaCollege of Medicine, Department of Anesthesiology, PO Box 100254, Gaines-ville, FL 32610-0254. Address e-mail to [email protected].

Copyright © 2010 International Anesthesia Research SocietyDOI: 10.1213/ANE.0b013e3181ef73ec


practice. Their optimal role may ultimately be for environ-ments hazardous to routine anesthetic practice, such as thebattlefield or space-based environments.5

Our initial attempts at robotic-assisted intubation fo-cused on direct laryngoscopy. However, we had consider-able difficulty using the DVS robotic graspers to lift andmanipulate both Macintosh and Miller laryngoscope lighthandles. Direct endotracheal tube manipulation with theDVS was quite challenging. On the other hand, the focus ofthe DVS on small-scale manipulation, coupled with itsmultiaxis flexibility and unifying video input, suggestedthat fiberoptic approaches to airway management wouldcapitalize on the capabilities of the DVS.

The robot operator controls the DVS through a dedi-cated workstation in a corner of the operating room. Theworkstation could easily be placed in another room, build-ing, or continent.6,7 Traditionally, such distances have beenlimited by latencies between user input, robot action, androbot-user feedback.8 Progress in adapting data packettransmission along existing telecommunication systems hasminimized such limitations,9 allowing long-distance ro-botic telesurgery with acceptable latencies. 10

Ideally, this exercise would have required zero nonro-botic interventions. However, because of the cost of thebronchoscope, we elected to accept a loss in simulationfidelity to minimize potential damage to the costly equip-ment. Further work will be necessary to explore how wellthe DVS can rotate and advance a bronchoscope tip.

Our experience indicates that because of the support andpreparation required for robotically assisted intubation, cur-rent systems are not ready for such deployment into complexoperating environments. The bedside presence of the anesthe-siologist was necessary even in this focused simulation toassist with important maneuvers necessary during roboticallyassisted fiberoptic intubation. Furthermore, this exercise did

Figure 1. Schematic depicting the arrangement of the DaVinciSurgical System type S (DVS), bronchoscope, and airwaymannequin.

Figure 2. Schematic depicting the arrangement of the DaVinciSurgical System type S (DVS), bronchoscope, and airwaymannequin.

Figure 3. The handle of the bronchoscope was positioned into arobotic arm. Two other robotic arms were fitted with graspers andused to control the flexion and extension of the bronchoscope tip.

Figure 4. Image from the DaVinci workstation during the roboti-cally assisted fiberoptic intubation. The robot operator can simul-taneously visualize video output from both the bronchoscope andthe robot’s main camera without turning attention away from theworkstation.

TECHNICAL COMMUNICATION


not include important factors such as patient preparation,positioning, topical and systemic medication administration,and patient monitoring.

This study demonstrated that a multipurpose surgicalrobot could be adapted for use in airway management.Although limited in its approach to direct laryngoscopy,the DVS was able to assist with both oral and nasalfiberoptic intubation. Future studies will be necessary tooptimize robotic interfaces with other airway managementtechniques.

APPENDIX: VIDEO CAPTIONSVideo 1. This video demonstrates how the robotic manipu-lation arms adjusted the flexion and extension of thebronchoscope tip, and how the bronchoscope itself wassecured to a third arm. Both oral and nasal intubations aredemonstrated. The bronchoscope advancement was per-formed manually to avoid damage to the bronchoscope andrequired very minimal human input. Advancement wasperformed at the direction of the robot operator.

REFERENCES1. Hemmerling TM. Automated anesthesia. Curr Opinion Anaes-

thesiol 2009;103:811–62. Bhayani S, Snow D. Novel dynamic information integration

during da Vinci robotic partial nephrectomy and radical ne-phrectomy. J Robotic Surg 2008;2:67–9

3. Rogers CG, Laungani R, Bhandari A, Krane LS, Eun D, PatelMN, Boris R, Shrivastava A, Menon M. Maximizing consolesurgeon independence during robot-assisted renal surgery byusing the fourth arm and TilePro. J Endourol 2009;23:115–22

4. Palep JH. Robotic assisted minimally invasive surgery. J MinimAccess Surg 2009;5:1–7

5. Harnett BM, Doarn CR, Rosen J, Hannaford B, Broderick TJ.Evaluation of unmanned airborne vehicles and mobile robotictelesurgery in an extreme environment. Telemed e-Health2008;14:539–44

6. Sterbis JR, Hanly EJ, Herman BC, Marohn MR, Broderick TJ,Shih SP, Harnett B, Doarn C, Schenkman NS. Transcontinentaltelesurgical nephrectomy using the da Vinci robot in a porcinemodel. Urology 2008;71:971–3

7. Marescaux J, Leroy J, Gagner M, Rubino F, Mutter D, Vix M,Butner SE, Smith MK. Transatlantic robot-assisted telesurgery.Nature 2001;413:379–80

8. Anvari M, Broderick T, Stein H, Chapman T, Ghodoussi M,Birch DW, Mckinley C, Trudeau P, Dutta S, Goldsmith CH.The impact of latency on surgical precision and task comple-tion during robotic-assisted remote telepresence surgery.Comp Aided Surg 2005;10:93–9

9. Rayman R, Primak S, Patel R, Moallem M, Morady R, TavakoliM, Subotic V, Galbraith N, van Wynsberghe A, Croome K.Effects of latency on telesurgery: an experimental study. Medi-cal image computing and computer-assisted intervention.MICCAI 2005;8:57–64

10. Rayman R, Croome K, Galbraith N, McClure R, Morady R,Peterson S, Smith S, Subotic V, Wynsberghe AV, Primak S.Long-distance robotic telesurgery: a feasibility study for care inremote environments. Intl Med Robotics Comp Assist Surg2006;2:216–24

Robot-Assisted Airway Support


Anesthesia Patient Safety Foundation

Section Editor: Sorin J. Brull

SPECIAL ARTICLE

Personal Protective Equipment for Care of PandemicInfluenza Patients: A Training Workshop for thePowered Air Purifying RespiratorBonnie M. Tompkins, MD,* and John P. Kerchberger, MD†

Virulent respiratory infectious diseases may present a life-threatening risk for health careprofessionals during aerosol-generating procedures, including endotracheal intubation. The2009 Pandemic Influenza A (H1N1) brings this concern to the immediate forefront. TheCenters for Disease Control and Prevention have stated that, when performing or participatingin aerosol-generating procedures on patients with virulent contagious respiratory diseases,health care professionals must wear a minimum of the N95 respirator, and they may wish toconsider using the powered air purifying respirator (PAPR). For influenza and other diseasestransmitted by both respiratory and contact modes, protective respirators must be combinedwith contact precautions.

The PAPR provides 2.5 to 100 times greater protection than the N95, when used within thecontext of an Occupational Safety and Health Administration–compliant respiratory protec-tion program. The relative protective capability of a respirator is quantified using the assignedprotection factor. The level of protection designated by the APF can only be achieved withappropriate training and correct use of the respirator.

Face seal leakage limits the protective capability of the N95 respirator, and fit testing does notassure the ability to maintain a tight face seal. The protective capability of the PAPR will be defeatedby improper handling of contaminated equipment, incorrect assembly and maintenance, and improperdon (put on) and doff (take off) procedures. Stress, discomfort, and physical encumbrance may impairperformance. Acclimatization through training will mitigate these effects.

Training in the use of PAPRs in advance of their need is strongly advised. “Just in time”training is unlikely to provide adequate preparation for groups of practitioners requiringspecialized personal protective equipment during a pandemic. Employee health departmentsin hospitals may not presently have a PAPR training program in place. Anesthesia and criticalcare providers would be well advised to take the lead in working with their hospitals’employee health departments to establish a PAPR training program where none exists.

User instructions state that the PAPR should not be used during surgery because itgenerates positive outward airflow, and may increase the risk of wound infection. Clarifica-tion of this prohibition and acceptable solutions are currently lacking and need to beaddressed. The surgical hood system is not an acceptable alternative.

We provide on line a PAPR training workshop. Supporting information is presented here.Anesthesia and critical care providers may use this workshop to supplement, but notsubstitute for, the manufacturers’ detailed use and maintenance instructions. (Anesth Analg2010;111:933–45)

Exposure to virulent respiratory pathogens during in-vasive airway procedures may present a life-threatening risk for health care providers. The Institute

of Medicine (IOM) of the National Academies Committee on

PPE (personal protective equipment) for Healthcare Workersstated that, “There is an urgent need to address the lack ofpreparedness regarding effective PPE for use in an influenzapandemic.” (Table 1, Ref. 1). In failing to anticipate andaddress PPE issues, health care workers may threaten notonly personal, colleague, and family safety, but patient safetyas well.1 Safe and effective use of PPE can only be optimizedby practicing correct procedures and by being aware of theoperational factors that defeat its protective capacity.

VIRAL PANDEMICS AND OUTBREAKS—PASTAND PRESENTInfluenza pandemics typically occur several times eachcentury. The 1957 and 1968 pandemics (Pandemic SeverityIndex [PSI] 2, case fatality ratio [CFR] 0.1��0.5%) did not

From the *Department of Anesthesiology, University of Wisconsin School ofMedicine and Public Health, Madison, Wisconsin; and †Department ofAnesthesiology, Rush University Medical Center, Chicago, Illinois.


Disclaimer: Neither author has a financial relationship with any devicemanufacturer.


Address correspondence and reprint requests to Bonnie M. Tompkins, MD,Department of Anesthesiology, University of Wisconsin Hospitals, 600 High-land Ave., Madison, WI 53792-0001. Address e-mail to [email protected].

Copyright © 2010 International Anesthesia Research SocietyDOI: 10.1213/ANE.0b013e3181e780f8


Table 1. Website ReferencesReference no. Website

1 Institute of Medicine (IOM) 2008. Preparing for an Influenza Pandemic: Personal Protective Equipment for Healthcare Workers.Washington, DC: The National Academies Press. Available at: http://www.nap.edu/catalog.php?record_id�11980.Accessed February 21, 2010

2 Interim Pre-pandemic Planning Guidance: Community Strategy for Pandemic Influenza Mitigation in the United States. Availableat: http://pandemicflu.gov/professional/community/commitigation.html#IV. Accessed February 21, 2010

3 Department of Heath and Human Services (U.S. DHHS) Pandemic Influenza Plan 2005. Available at: http://hhs.gov/pandemicflu/plan/appendixb.html. Accessed February 22, 2010

4 Occupational Safety and Health Administration (OSHA). Pandemic Influenza Preparedness and Response Guide 2007.Available at: http://www.osha.gov/Publications/3328-05-2007-English.html. Accessed February 22, 2010

5 Centers for Disease Control and Prevention (CDC). Estimates of 2009 H1N1 Influenza Cases, Hospitalizations, and Deaths in theUnited States. Available at: http://www.cdc.gov/h1n1flu/estimates/results_2009_h1n1.htm. Accessed January 16, 2010

6 Cumulative Number of Confirmed Human Cases of Avian Influenza A/(H5N1) Reported to World Health Organization (WHO).Available at: http://www.who.int/csr/disease/avian_influenza/country/cases_table_2010_02_17/en/index.html. AccessedFebruary 18, 2010

7 WHO/Pandemic Influenza Preparedness and Response. Available at: http://www.who.int/csr/disease/influenza/PIPGuidance09.pdf. Accessed February 21, 2010

8 Center for Infectious Disease Research and Policy (CIDRAP). Avian Influenza (Bird Flu): Implications for Human Disease. Lastupdated April 2, 2010. Available at: http://www.cidrap.umn.edu/cidrap/content/influenza/avianflu/biofacts/avflu_human.html. Accessed February 21, 2010

9 WHO Summary of Probable SARS Cases with Onset of Illness from 1 November 2002 to 31 July 2003. Available at: http://www.who.int/csr/sars/country/table2004_04_21/en/index.html. Accessed February 22, 2010

10 The SARS Commission Executive Summary: Spring of Fear. Available at: http://www.health.gov.on.ca/english/public/pub/ministry_reports/campbell06/online_rep/V1.html. Accessed September 28, 2009

11 Adalja AA. Will the D225G Mutation Herald More Severe Illness in Patients with 2009 H1N1 Influenza? Clinician’s BiosecurityNetwork Report. Available at: [email protected]. Accessed December 11, 2009

12 1918 Polymorphism in Ukraine H1N1? Available at: http://www.recombinomics.com/News/11090902/Ukraine_1918.html.Accessed December 11, 2009

13 Siegel JD, Rhinehart E, Jackson M, Chiarello L, and the Hospital Infection Control Practices Advisory Committee (HICPAC).2007 Guideline for Isolation Precautions: Preventing Transmission of Infectious Agents in Healthcare Settings. Available at:http://www.cdc.gov/ncidod/dhqp/pdf/guidelines/Isolation2007.pdf. Accessed February 20, 2010

14 Interim Domestic Guidance on the Use of Respirators to Prevent Transmission of SARS May 3, 2005. Available at: http://www.cdc.gov/ncidod/sars/respirators.htm. Accessed February 19, 2010

15 IOM 2009. Respiratory Protection for Healthcare Workers in the Workplace Against Novel H1N1 Influenza A: A Letter Report.Washington, DC: The National Academies Press. Available at: http://www.nap.edu/catalog/12748.html. Accessed February22, 2010

16 IOM 2006. Reusability of Facemasks During an Influenza Pandemic. Board of Health Sciences Policy. Washington, DC: TheNational Academies Press. Available at: http://books.nap.edu/openbook.php?record_id�11637&page�R1. AccessedFebruary 22, 2010

17 Interim Guidance on Planning for the Use of Surgical Masks and Respirators in Health Care Settings during an InfluenzaPandemic. Oct. 2006 Available at: http://www.flu.gov/professional/hospital/maskguidancehc.html#airborne. Accessed Feb20, 2010

18 Proposed Guidance on Workplace Stockpiling of Respirators and Facemasks for Pandemic Influenza. Available at: http://www.osha.gov/dsg/guidance/proposedGuidanceStockpilingRespirator.pdf. Accessed February 21, 2010

19 Infection Prevention and Control in Health Care for Confirmed or Suspected Cases of Pandemic (H1N1) 2009 and Influenza-Like Illnesses 25 June 2009. Available at: http://www.who.int/csr/resources/publications/SwineInfluenza_infectioncontrol.pdf.Accessed February 21, 2010

20 Interim Guidance on Infection Control Measures of 2009 H1N1 Influenza in Healthcare Settings, Including Protection ofHealthcare Personnel. October 14, 2009. Available at: http://www.cdc.gov/h1n1flu/guidelines_infection_control.htm.Accessed February 21, 2010

21 OSHA Instruction Directive #: CPL-02-02-075 Effective Nov. 11, 2009. Enforcement Procedures for High to Very HighOccupational Exposure Risk to 2009 H1N1 Influenza. Available at: http://www.osha.gov/OshDoc/Directive_pdf/CPL_02_02–075.pdf. Accessed February 18, 2010

22 OSHA, Appendix C, Pandemic Influenza Preparedness and Response Guide 2007 Implementation and Planning for RespiratoryProtection Programs in Healthcare Settings. Available at: http://www.osha.gov/Publications/OSHA_pandemic_health.pdf.Accessed February 22, 2010

23 OSHA State Occupational Safety and Health Plans. Available at: http://www.osha.gov/dcsp/osp/index.html. AccessedFebruary 22, 2010

24 Assigned Protection Factors for the Revised Respiratory Protection Standard OSHA 3352–02 2009. Available at: http://www.osha.gov/Publications/3352-APF-respirators.html. Accessed February 22, 2010

25 King JH. Assessing Powered Air Purifying and Supplied Air Respirator Performance: An Employer’s Guide to OSHA’s Final Ruleon Assigned Protection Factors for Respirators. An E.D. Bullard Company White Paper. September 21, 2006. Available at:http://www.bullard.com/V3/resources/downloads/respiratory_dwnlds.php#Technical_Publications. Accessed February 22,2010

26 3M™ Technical Data Bulletin #160; Reusable Respirator Facepieces and Powered Air Purifying Respirator Systems (PAPRs) inthe HealthCare Environment: Considerations for Use; and TDB #175 Assigned Protection Factors. 3M Occupational Health& Environmental Safety Respiratory Protection Media Library. Available at: http://www.3M.com/occsafety. AccessedFebruary 21, 2010

(Continued)

SPECIAL ARTICLE


approach the catastrophic nature of the 1918 Great Influ-enza, “Spanish Flu” (PSI 5, CFR �2%), wherein one-third(approximately 500 million) of the world’s populationbecame ill and 50 to 100 million died (Table 1, Refs. 2–4).2,3

The Centers for Disease Control (CDC) midrange esti-mates for the 2009 Pandemic Influenza A (H1N1) (hereafter2009 H1N1) in the United States from April 2009 to January2010 are 57 million total cases, 257,000 (0.45%) hospitaliza-tions, and 11,700 deaths (overall CFR 0.02%, hospitalizedCFR 4.5%) (Table 1, Ref. 5). In seasonal influenza, thegreatest mortality is among the very young and the veryold; in contrast, the 1918 and 2009 H1N1 pandemicspredominantly affected children and younger adults (Table1, Refs. 3 and 5).2

The World Health Organization (WHO) reported that 478human confirmed cases and 286 deaths (CFR 60%) occurredfrom 2003 through February 2010 from the H5N1 highlypathogenic avian influenza (HPAI) (“bird flu”). Many caseswere in children and young adults (Table 1, Ref. 6). This virushas not currently developed a high affinity for human respi-ratory tract receptors; therefore, human-to-human transmis-sion is nonsustained and cases have occurred only in smallclusters (WHO pandemic phase 3). Should H5N1 HPAI and2009 H1N1 occur simultaneously in the same individual, ahighly pathogenic reassortant pandemic strain could emerge(Table 1, Refs. 7 and 8).

The severe acute respiratory syndrome coronavirus(SARS-CoV) epidemics of 2002 to 2003 incurred �8000cases and 700 deaths in 29 countries (CFR 9.6%) (Table 1,Ref. 9). Twenty percent of the cases were in health careworkers.4 In Ontario, Canada, �50% of the 438 SARS casesand 3 of the 43 deaths were in health care workers (Table 1,Ref. 10).5–7 The SARS experience is of particular import toanesthesiology and critical care specialists.1

VIRULENCE FACTORS IN SEVERE VIRALRESPIRATORY ILLNESSThe history and pathology of influenza have been studiedand presented effectively by Taubenberger and Morens.8

Both pandemic and seasonal influenza viruses may repli-cate throughout the respiratory tract. Disease is limited tothe upper respiratory tract and trachea in nonfatal cases,but fatal cases show lung involvement.8,9 The 1918 and2009 H1N1 pandemic and H5N1 HPAI viruses exhibit agreater propensity than seasonal influenza to bind to viralreceptors in the lower respiratory tract.3,8–11 The histopa-thology from autopsy of influenza victims is that of aprimary viral hemorrhagic bronchitis and pneumonia withdiffuse alveolar damage and destruction.8,9,12 Secondarybacterial pneumonia contributes prominently to the mor-tality in seasonal and pandemic influenza.8,9,12

Table 1. (Continued)Reference no. Website

27 Understanding Respiratory Protection Against SARS. Available at: http://www.cdc.gov/niosh/npptl/topics/respirators/factsheets/respsars.html. Accessed February 26, 2010

28 Adalja AA. Update on Personal Protective Equipment. Available at: http://www.upmc-cbn.org/report_archive/2008/12_December_2008/cbnreport_12192008.html. Accessed February 21, 2010

29 General Procedures for Putting On and Taking Off a Disposable Respirator. Available at: http://www.cdc.gov/flu/freeresources/2009-10/pdf/n95respirator_instructions.pdf. Accessed February 19, 2010

30 3M Air-Mate™. Available at: http://www.3m.com/occsafety. Accessed February 21, 201031 3M Resource CD. Available on request from 3M Occupational Health and Environmental Safety Division Technical Service at

800-243-4630 and from [email protected]. Available at: http://www.3M.com/occsafety. Accessed October 11,2009

32 Powered Air Purifying System PA20™. Available at: http://www.bullard.com/Respiratory/papr/pa20page.shtml. AccessedOctober 11, 2009

33 Bullard EVA™ Competitive Comparison. Available at: http://www.bullard.com/V3/resources/downloads/respiratory_dwnlds.php#EVA. Accessed February 5, 2010

34 PAPR Training Workshop. Available at: http://www.med.wisc.edu/papr-workshop. Accessed October 12, 200935 OSHA Best Practices for Hospital-Based First Receivers of Victims from Mass Casualty Incidents Involving the Release of

Hazardous Substances. Available at: http://www.osha.gov/dts/osta/bestpractices/html/hospital_firstreceivers.html.Accessed October 11, 2009

36 Full Barrier Personal Protective Equipment (PPE) with Powered Air Purifying Respirator (PAPR). Available at: http://www.health.state.mn.us/divs/idepc/dtopics/infectioncontrol/ppe/ppepapr.html. Accessed September 28, 2009

37 3M Technical Data Bulletin #178: Maintenance and Care of 3M Powered Air Purifying Respirator (PAPR) Batteries. PublishedMarch 2007, Revised November 2008. Available at: http://www.3M.com/occsafety. Accessed February 21, 2010

38 Maintenance of Battery Packs for Bullard Powered Air-Purifying Respirators (PAPRs). Available at: http://www.bullard.com.Accessed February 21, 2010

39 Bullard Technical Advisory: Release of Filtered Particles. Available at: http://www.bullard.com. Accessed February 21, 201040 Stryker T4 Surgical Helmet System Filtration Testing Summary Report. Available at: www.sars.medtau.org/strykerreport.doc.

Accessed August 31, 200941 Using the Stryker T4 Personal Protection System for High-Risk Procedures During SARS Outbreaks. Available at:

sars.medtau.org/strykertraining.htm. Accessed April 16, 201042 Questions and Answers About CDC’s Interim Guidance on Infection Control Measures for 2009 H1N1 Influenza in Healthcare

Settings, Including Protection of Healthcare Personnel. December 1, 2009. Available at: http://www.cdc.gov/h1n1flu/guidance/control_measures_qa.htm. Accessed February 21, 2010

43 Personal Protective Equipment in Pandemic/Avian Influenza/SARS: N95 or PAPR for Intubation? ASA Newsletter. Available at:http://www.asahq.org/Newsletters/2008/01-08/tompkins01-08.html. Accessed September 28, 2009

Personal Protective Equipment for Pandemic Influenza


The genomes of the 1918 and present-day influenzaviruses have been reconstructed, studied, and compared,presenting opportunity for identifying preventive andtherapeutic approaches. This work has clarified how thevirus binds to lower as well as upper respiratory tractreceptors and has identified gene components associatedwith increased virulence of the 1918 pandemic, H5N1avian, and 2009 H1N1 pandemic influenza viruses (Table 1,Refs. 11 and 12).3,8,11,13 Identified genetic amino acid se-quences associated with host specificity and high virulencemay provide a predictive monitoring tool.14

In addition to diffuse lung damage, altered immunemechanisms and a viral-associated extreme proinflammatorycytokine response, sometimes with hemophagocytosis, havebeen observed in young, previously healthy influenza pa-tients who died or were critically ill.8,15 These pathologicfeatures also have been described in animal models inocu-lated with the reconstructed 1918 virus,11 in animals andhumans with H5N1 avian influenza,13,15,16 in SARS-CoV,17–19

and in 2009 H1N1 victims (Table 1, Ref. 12).9

TRANSMISSION MODESThe predominant transmission modes for influenza andSARS are respiratory droplet and direct and indirect con-tact (fomite). (Table 1, Refs. 13 and 14). Epidemiologic andinvestigational evidence is strongly suggestive of near-range airborne transmission (Table 1, Refs. 1, 9, 10, and13–15).20–22 Remote airborne transmission is supported bylaboratory research and theoretical modeling of aerosolbehavior and is suspected in specific outbreaks, but be-lieved to be unusual (Table 1, Refs. 1 and 15).21–25

Contact ModeThe sources of contact transmission are often overlooked(contaminated surfaces, clothing, equipment, personal pro-tective equipment, exposed skin) (Table 1, Refs. 1 and 13).The infectivity of virus on surfaces, skin, and hands decays,and the time that the fomite remains infective may vary(Table 1, Ref. 13).22

Droplet and Airborne ModesNewer investigations on aerosol transmission of influenzahave challenged the traditional belief that airborne trans-mission does not occur. Droplet and airborne modes arenow viewed as a continuum, with particle sizes rangingfrom large to fine droplet or aerosol (Table 1, Refs. 1, 14,and 15).20–22 The term “droplet” is consistent with trans-mission by larger disease-bearing particles that settle out ofthe air onto surfaces within shorter distances from thesource or are inhaled into the upper airway and trachea.Larger droplets may evaporate to small “droplet nuclei”that behave as aerosol (Table 1, Refs. 1, 13, and 16).

“Near-range airborne” transmission is through smallerdroplet particles (inspirable) that remain suspended in airfor relatively short but variable distances (1 m with breath-ing, 2 m with coughing, 6 m with sneezing), and wheninhaled, reach only the trachea and bronchi (Table 1, Refs.1, and 13–15).20–22 Near-range airborne transmissionthrough a spectrum of disease-bearing particle sizes is nowacknowledged by the CDC and others to be operational ininfluenza and SARS (Table 1, Refs. 1, 8, 13, and 14).22

The term “airborne transmission” traditionally refers tothe remote transmission and inhalation of yet smallerdroplets and aerosol-sized (respirable) particles that mayaccess the alveoli as well as the upper airway and mayremain suspended in the air for an indeterminate time anddistance (Table 1, Ref. 13).20 Airborne transmission of anyrespiratory infectious disease is difficult to prove or todefinitively exclude.23 Isolated specific outbreaks of SARS(aerosolized remote fecal source) and influenza (airplaneoutbreaks) are believed to be examples of unusual, oropportunistic, airborne transmission; aerosol research sup-ports the possibility of remote airborne transmission ofinfluenza and SARS (Table 1, Refs. 1, 14, 15, and 17).20,22,23

Transmission and subsequent infection is influenced byadditional factors that include viability of the agent, expi-ratory force, distance from and duration of exposure to thesource, and environmental conditions (humidity, tempera-ture, and wind) (Table 1, Ref. 1).26–28

AEROSOL-GENERATING PROCEDURES INANESTHESIA AND CRITICAL CAREAerosol particles are generated during all invasive airwayprocedures, noninvasive and positive pressure ventilatorysupport modes, suction, sputum induction, high-flow oxy-gen delivery, aerosolized or nebulized medication delivery,interventions that stimulate coughing, and autopsies (Table1, Refs. 1, 4, 13, 15, and 17).21,22 Infection is established witha smaller quantity of aerosol than nasal instillation (Table 1,Ref. 1).22,29 Health care professionals who perform andassist with aerosol-generating procedures and therapies areincluded in the Occupational Safety and Health Adminis-tration (OSHA) “very high occupational exposure risk”category (Table 1, Refs. 3, 4, and 18), and are at a higher riskof infection than others.

TRANSMISSION-BASED PRECAUTIONSClose Patient CareConfusion persists about whether a facemask or respiratoris indicated for close care of influenza patients and thosewith a flu-like illness. For 2009 H1N1 influenza, because ofthe limited access to N95s in many countries, the WHO hasstated that the N95 (or European equivalent, EU FFP2) isonly indicated during aerosol-generating procedures, andthat a surgical mask in addition to face shield and eyeprotection, with other contact precautions, is satisfactoryfor all other patient care activity (Table 1, Ref. 19). Incontrast to the WHO, the CDC, OSHA, and the IOM of theNational Academies all state that the N95 respirator, orhigher ([N100], or powered air purifying respirator[PAPR]), should be used during close contact with influ-enza or SARS patients. Contact precautions (gown, gloves,hat, close-fitting eye protection, shoe covers) are also rec-ommended for influenza and other virulent diseases thatare transmitted by both the respiratory and contact modes.The CDC describes close contact as within 6 to 10 ft. of thepatient or in the patient’s room (Table 1, Ref. 20).

Aerosol-Generating ProceduresThe CDC, OSHA, and the IOM further suggest that healthcare personnel consider using the higher level of protectionprovided by the PAPR when performing or assisting with

SPECIAL ARTICLE


aerosol-generating procedures (Table 1, Refs. 1, 3, 4, 13, 15,and 20). Some state and hospital pandemic influenza,SARS, and tuberculosis protocols assert definitively thatthe PAPR is to be used for endotracheal intubation andbronchoscopy. The CDC states that these proceduresshould be done in an airborne infection isolation room(Table 1, Ref. 20).

OSHA STANDARDS APPLICABLE TO HEALTHCARE FACILITIESThe General Duty Clause of the Occupational Safety andHealth Act requires that employers abate or address recog-nized workplace hazards. OSHA standards for employeesafety were developed initially for the industrial work-place; subsequently, standards for the health care venuewere developed as well (Table 1, Refs. 4 and 21). The OSHABlood-borne Pathogens, Personal Protective Equipment,and Respiratory Protection standards, as listed in the Codeof Federal Regulations, are of particular importance inpandemic influenza preparedness and response (Table 1,Ref. 4). Airborne pathogens are included in the hazardousairborne contaminants covered by the standards. OSHAhas posted inspection, investigation, and enforcement pro-cedures that address 2009 H1N1 Influenza in health careworkers in the “high” and “very high occupational expo-sure risk” categories (Table 1, Ref. 21).

Under the Respiratory Protection standard (29 CFR1910.134), all respirators must be deployed within thecontext of a written, worksite-specific respiratory protec-tion program. The requisite respiratory protection programdirector oversees and documents respirator selection,maintenance and cleaning procedures, employee medicalclearance to wear a respirator, fit testing of tight-fittingrespirators, instruction, and periodic program review(Table 1, Ref. 22).

States may elect to develop their own standards foremployee safety, which the employers must follow, butthese plans must be approved by OSHA (Table 1, Ref. 23).When applied, the OSHA standards maximize safety bysupporting the correct and consistent use of respirators.

QUANTIFICATION OF RESPIRATORY PROTECTION:THE ASSIGNED PROTECTION FACTORThe assigned protection factor (APF), applied to industrialrespirators initially, is a value given or assigned to eachrespirator by OSHA and the National Institute of Occupa-tional Safety and Health (NIOSH) denoting the factor bywhich a respirator reduces the contaminating substance in theambient air (Table 1, Ref. 24). The APF is derived fromlaboratory simulated workplace (SWPF) and actual work-place protection factors (WPF) computed from photometricdeterminations of the test contaminant outside (Co) the respi-rator relative to inside (Ci) the respirator facemask.30 For eachrespirator tested, the lowest fifth percentile of test Co/Ci

values are averaged and further divided by a safety factor of25. This final value is assigned to the respirator as the APF.APFs range from 10 to 10,000. This number represents theminimum factor by which exposure to contaminants is re-duced when wearing the respirator. Hence, a higher numberindicates greater protection (Table 1, Refs. 25 and 26).

The APF of the N95 respirator is 10, meaning that thewearer will expect to inhale no more than one-tenth of thehazardous airborne particles present (Table 1, Ref. 24).OSHA has set a minimum APF of 1000 that a PAPR mustachieve to obtain NIOSH certification, but charges theprospective buyer with obtaining confirmatory testing evi-dence of the APF from each PAPR manufacturer (Table 1,Refs. 24 and 25).30

TERMINOLOGY AND CATEGORIZATION OFFACEMASKS AND RESPIRATORSMedical MasksMedical masks of several different types are worn, asintended, for protecting the patient and the environmentfrom the respiratory secretions of health care workers(Table 3A). Masks also provide the wearer with droplet andcontact protection for the skin area covered by the mask.The degree of protection corresponds to the masks’ fluidresistance; surgical masks are fluid resistant, proceduremasks are less so, and isolation masks generally are not.

The filtering fabric of some masks is tested to 0.1-�mparticle size but this is not required for Food and DrugAdministration certification. Medical masks do not permita tight face seal, do not provide a barrier to aerosolparticles, and therefore do not meet the criteria for arespirator. Gauze masks are an effective droplet barrieronly with 6 or more layers (Table 1, Ref. 16).

The infection rate was reduced in Hong Kong during theSARS outbreak with consistent use of medical masks pluscontact barrier garb that consisted of gowns, gloves, andeye protection.6,26 Public use of masks, with hand hygiene,also reduced the infection rate (Table 1, Refs. 1 and 16). Acomparison of the infection rate between nurses wearingsurgical masks or N95s found the influenza rate to be nodifferent, but still considerable in both groups (approxi-mately 23%).31

Discussion and study continue on evidence-based choiceof respiratory personal protective equipment (PPE) for healthcare workers exposed to influenza.32,33 In the absence of arespirator or when supplies are limited, the surgical mask,with contact transmission precautions, should be used forroutine bedside care of SARS and influenza patients (Table 1,Ref. 17). Masks must be removed carefully and discarded,followed by hand washing (Table 1, Refs. 13 and 20).

RespiratorsIn the context of PPE, a “respirator” reduces exposure toinhaled environmental contaminants (Table 1, Ref. 16).Respirator classifications consider the air supply, power,filter type and efficiency, facepiece description, and pres-ence of an exhalation valve. Additional descriptive fea-tures, such as air flow direction, clarify the respirator’sprotective mechanism (Tables 2 and 3B) (Table 1, Refs. 1, 4,15, and 16).34 The particulate respirator filter removesdroplet and aerosol-sized airborne particles and an absor-bent respirator filter removes chemical vapors and gases.

The nonpowered category of air purifying respiratorsincludes the elastomeric and the filtering facepiece respira-tors. The “elastomeric” respirator has a full or half facepiecethat is nondisposable, tight fitting, and made of silicone orrubber with either a replaceable particulate or vapor/gas



absorbing filter, or a combination filter. Examples are the“gas mask,” “MOPP” respirator (Mission Oriented Protec-tive Posture) (dual filter), the painter’s respirator (vapor),and the particulate elastomeric respirator for use in healthcare if N95 stores are exhausted.

The filtering facepiece respirators (N95 and higher) areclassified by the resistance to degradation by oil, and by thepercent filtration efficiency of 0.3-�m saline test particles,the most penetrating size and substance (Table 1, Ref. 4).Both the N95 and the elastomeric are negative pressurerespirators; that is, they are dependent on the maintenanceof a tight face seal to prevent contaminated air frombypassing the filter (Table 1, Refs. 4, 14, and 27).

The notable advantages of the N95 are simplicity andaccessibility.35 The predominant disadvantages are thewell-known increased resistance to breathing36 and thepropensity for a gap to occur in the face seal (face sealleakage) (Table 4). Prior fit testing does not assure successin attaining and maintaining a tight face seal. Only thosewearing the N95 in their daily work were found to achievea face seal with some consistency, irrespective of whenprior fit testing had occurred (Table 1, Ref. 28).37 The N95don (put on) and doff (take off) procedure posted by theCDC includes a positive and negative pressure face sealcheck, which should be performed when using any tight-fitting respirator (Table 1, Refs. 4, 14, and 29).

Hospitals are urged to stockpile supplies. OSHA haspredicted that every nurse, using 4 N95s per shift, woulduse 480 over a 12-week pandemic wave, at a cost of $240per worker (Table 1, Ref. 18). All disposables will present asubstantial expense and will be depleted rapidly.38 TheCDC recommends prioritization of N95 use based on risk ofexposure, to be initiated when N95 supplies are limited(Table 1, Ref. 20).

Reuse of the N95 is strongly discouraged but permissible ifsupplies are limited, provided it is not soiled, creased, dam-aged, moist, or wet (Table 1, Refs. 14 and 16). Reaerosolizationof disease particles from the N95 does not occur (Table 1, Ref.16).34 However, the used respirator is a fomite and presents acontamination risk. Covering the N95 with a face shield willprotect it from droplets and splashes. A fluid-resistant N95 isavailable and should be used in surgery (Table 1, Refs. 14 and16). The used N95 should be stored in a paper, not plastic, bagto avoid condensation (Table 1, Ref. 26). When donning a usedmask, extreme care should be taken not to contaminateoneself, and hand washing should follow. The N95 filteringfacepiece respirator provides a 10-fold factor of protectionrelative to ambient air (APF 10) (Table 1, Refs. 24 and 26).

THE PAPR FOR BIOLOGICAL HAZARDSThe PAPR provides a higher level of protection than theN95 respirator because it supplies maximally filtered air,eliminates face seal leak, and provides contact protectionfor the head (Table 3B). It is composed of a belt-worn casethat houses a battery, fan, and filter. The fan, referred to asthe “blower,” draws ambient air through a high-efficiencyparticulate air (HEPA) filter (99.97% efficient) and blows it at�170 L/min (6 cu. ft./min) through a flexible tube and into aTyvek or Tychem hood with a plastic face shield (Table 1, Refs.4 and 30–33). The high flow of air exiting the hood prevents thewearer from entraining contaminated ambient air.

The PAPR hoods are available in 2 styles: the double-shrouded hood and the loose-fitting face covering. The formercompletely covers the head and shoulders and does notpermit use of a conventional stethoscope. The inner shroudtucks beneath the neck of a gown, and the filtered air exitsfrom under the inner shroud and gown. The PAPR used witha double-shrouded hood provides more than a 1000-foldprotection relative to ambient air (APF 1000) and 100 times the

Table 3A. Medical MasksTypes of

medical masks aIsolation;Procedure

Surgical; Laser;Dental

Water resistance b Variable YesTransmission mode

protection c

For wearer:Contact Yes YesDroplet Very little YesAerosol (airborne) No No

For wearer’s contacts:Contact Yes YesDroplet Very little YesAerosol (airborne) No No

Chemical protection d None Nonea Medical masks include various types of face masks used in the healthcarevenue; they provide no protection from airborne biologic aerosol and are notclassified as respirators. Masks are intended to protect patients and otherclose contacts from the wearer’s secretions.b Only water resistant masks and respirators provide the wearer (patient orhealthcare provider) with droplet protection from others. Water resistance isnot a requirement for certification of procedure and isolation masks and mayvary among different brands. Surgical masks are water resistant and protectthe wearer and their contacts from droplets and splashes.c Contact and droplet protection for the wearer is only over the area coveredby the mask or respirator.d Medical masks and respirators are particulate filters only and do not provideprotection from chemical agents.

Table 2. Categorizing RespiratorsType of respirators

Air purifyingNonpowered: Depend on the wearer drawing air in through filters

or cartridgesPowered air purifying respirators: A blower draws air through the

filter and delivers it to the wearerAir supplying

Self-contained breathing apparatusType of filters

Particulate filtersP (oil proof; can survive oil exposure for �1 work shift)R (oil resistant; can be used for oil exposure in 1 shift)N (not oil resistant; used for oil-free environments)

Gas-vapor respiratorCombination particulate and gas vapor

Filtering efficiencyCertified for a range of efficiency classes (e.g., 95%, 99%, 100%)

Type of facepieceFiltering facepiecesReplaceable filter components: half-mask and full-mask

elastomeric respiratorsLoose-fitting facepieces

Use or nonuse of an exhalation valve

Adapted from: Institute of Medicine. Preparing for an Influenza Pandemic:Personal Protective Equipment for Health Care Workers. Washington, DC: TheNational Academies Press, 2008. Reprinted with permission from the Na-tional Academies Press.

SPECIAL ARTICLE


protection of the N95. The double-shrouded hood offers thebest protection for aerosol-generating procedures.

The loose-fitting face covering protects the face, chin,and top of the head but leaves the neck and ears exposed,allowing use of a stethoscope. Illustrations are availableonline (Table 1, Refs. 30 –34). The high-flowing air exitsthrough perforations beneath the chin. This hood isappropriate for continuous bedside care in the absence ofaerosol-generating conditions. The protection factor ofthis hood is 25-fold greater than no protection (APF 25)

and 2.5 times the protection of the N95 respirator (APF10) (Table 1, Refs. 30 –33).

Because the PAPR is a loose-fitting respirator, fit testingis not required. The full hood, but not the loose-fitting facecover, may be used by those with a beard. Facial hair canpermit unfiltered air to be entrained under the elastic bandborder of the loose-fitting face cover.

The PAPR does not increase the work of breathing andis more comfortable for extended wear than an N95 (Table1, Refs. 1 and 4) (Table 5). The PAPR and the nondisposable

Table 3B. Features of Masks and Respirators

a Self-contained breathing apparatus.b Remote supplied air respirator.c Available with particulate filter, gas/vapor absorbent canister, or combination.d Powered air purifying respirator.e Water resistant N95 respirators are available and should be used during surgery.f Mission Oriented Protective Posture: Uniformed Services elastomeric respirator and chem warfare agent resistant clothing.g Chemical absorbent filters are agent specific; painter’s respirator is not protective against chemical and nerve agents.h The loose fitting face covering helmet covers the top of the head, the face and the chin, leaving the ears and neck exposed.i The double shrouded hood covers the entire head and shoulders.j Assigned Protection Factor (factor by which the test contaminant is reduced by the respirator).k Contact protection is only for the skin area covered by the mask or respirator.l Only with particulate filter.m Only with absorbent filter.n Storage sites and containers for biologic PAPRs should be labeled prominently, “Not for chemical protection”.o High Efficiency Particulate Air filter.p APF 50 for full face elastomeric.



elastomeric air purifying respirator are the only realisticalternatives for respirators in the hospital during a pan-demic if N95 stocks are exhausted.

THE PAPR FOR CHEMICAL ORHAZARDOUS MATERIALThe particulate HEPA filter in PAPRs used for protectionagainst biological hazard, e.g., 3M™ Air-Mate™, BullardPA20™, or Bullard EVA™, do not protect against hazard-ous chemicals, gases, or vapors. PAPRs for biologicalprotection should be prominently labeled “not for chemical

or vapor protection.” The hazardous-material PAPRs, e.g.,3M Breathe-Easy™, Bullard PA30™, and Bullard EVA™,have vapor-absorbing and dual (combination) cartridgesand chemical-resistant hoods; these must be worn withadditional full-body chemical protective suits (Table 1,Refs. 31, 33, and 35) (Appendix 1).

To be certified to wear chemical protective equipment, theOSHA standard 29 CFR 1910.120 on Hazardous Waste Op-erations and Emergency Response (HAZWOPER) requires acombination of 8 hours of OSHA-compliant awareness-level(informational) and operations-level (practical) training (Table1, Ref. 35). Hospital security services and emergency depart-ments strictly adhere to, and enforce, this mandate.

PAPR ISSUES AND LIMITATIONSAttention to many details of PAPR use and maintenance isessential for the assurance of effective protection (Table 6).

The Donning and Doffing Sequence Is ImportantExposed contaminated skin and PPE constitute indirectsources of infection (fomite) for health care workers andpotentially their contacts. Using fluorescein dye as a surrogatemarker of contamination, a comparison of the N95 and PAPR

Table 4. N95 Respirator Advantagesand DisadvantagesAdvantages

● Filters 95% of aerosol test particles● 10-fold protection (APF 10)● Easily accessible● Disposable● No setup required● No interference to using a stethoscope● Not powered● Noiseless

Disadvantages● Face seal leak common● Increases resistance to breathing (exhalation valve may reduce

discomfort but must cover with surgical mask during surgery)● Requires fit testing (costly, labor intensive, does not ensure a

tight face seal)● Facial features may preclude a satisfactory fit with any model● Headaches, dizziness, shortness of breath, possible CO2

retention● Supplies rapidly depleted when demand is high● Reuse risks self-contamination (use face shield if reuse is likely)

(store in paper bag)● Facial hair inhibits a face seal, defeats protection● Ineffective when moist, wet, creased, or damaged

Table 5. Powered Air Purifying Respirator (PAPR) Advantages and DisadvantagesPAPR advantages PAPR disadvantages

● HEPA filtered air ● Initial cost high (but may be cost effective compared with N95 stockpiling/use)

● Positive inside to outside air flow (170 L/min, 6cfm �cu. ft./min�)

● Larger requirement for storage space

● Highest level of protection for aerosol-generatingprocedures

● Self-contamination from used equipment possible

● Contact protection of head and neck: double-shrouded hood better than the loose-fitting facecovering hood; both more than N95

● Protection factor, relative to no protection:1000 � for PAPR with double-shrouded hood25 � for PAPR with loose-fitting face covering hood10 � for N95 respirator

● No fit testing required● No entrainment of contaminated air● Comfortable for extended wear (bedside care,

reprocessing worker, invasive airway procedures)● Hoods are disposable (reusable, by single user

only, after reprocessing)● Facial hair is acceptable, with double-shrouded

hood only, not with loose-fitting face covering hood

● Reprocessing of hoods required between work breaks● Longer training time than N95 (required by Occupational Safety and Health

Administration, essential for safety: awareness of equipment; limitations;assembly; preuse check; maintenance/reprocessing; don/doff sequence;acclimatization and procedure practice)

● Battery recharging schedule is crucial● Biannual discharge of unused units needed so battery maintains its capacity

to accept a charge● Some models do not have real-time air flow indicator● Care needed in seating of HEPA filter gasket after changing battery to avoid

misplacement● Equipment may impede performance● Hearing reduced because of fan noise; speech muffled because of hood● PAPR use contraindicated during surgery (issue has not been addressed,

remains unresolved)● The surgical hood system is not certified as a respiratory protection device

HEPA � high-efficiency particulate air.

Table 6. Reasons for Failure of Powered AirPurifying Respirator (PAPR) Protection● Neglect of battery maintenance procedures● Failure to check equipment and flow rate before use● Decreased air flow during use (likely incomplete battery charge)● Absence or incorrect placement of filter and gasket● Incorrect don/doff sequence● Neglect of hood care/inspection and unit cleaning/reprocessing● Removal of PAPR during a procedure (failure to practice

procedures while wearing the equipment)

Caution: A biological PAPR does not provide chemical or vapor protection.

SPECIAL ARTICLE


PPE ensembles showed that skin contamination was greaterwith the N95 ensemble. Errors in don/doff steps were morefrequent with the PAPR ensemble, thereby increasing thechance for self-contamination or exposure.39 However, infollowing the CDC proscribed sequence for removal of theN95 and contact garb, viral contamination of hands andclothing occurred. In that study, a fluorescein dye surrogatemarker was not a reliable indicator of viral contamination.Double gloving and the customary sequence for removal ofgowns and mask after surgery were recommended.40 The Min-nesota Health Department has posted an illustrated don/doffsequence for the PAPR and contact garb (Table 1, Ref. 36).

Initial Cost Is High and Storage Space Is ChallengedBudget constraints and scarce storage space limit supplies ofPAPRs and hoods in most hospitals. Hospitals may purchasePPE using federal funds for disaster preparedness.

Battery Maintenance Is CrucialThe 3M Air-Mate has an interchangeable rechargeablenickel-cadmium battery. A fully charged and maintainedbattery will run for longer than 8 hours when properlycharged. Close attention to battery-charging procedureswill avoid failure of the PAPR during use (Table 1, Refs. 37and 38).

The Bullard PA20 does not have an interchangeablebattery so it must be removed from service for recharging.A rapid charger is available (Table 1, Refs. 32 and 33). Thebattery runs for 9 to 10 hours.

A Flow Indicator Is DesirableThe 3M Air-Mate has no real-time flow or battery chargeindicator to signal when airflow or charge is reduced (Table1, Refs. 31 and 33). The air flow must be checked with allPAPRs with the test float each time the unit is turned on.The 3M Breathe-Easy PAPR for chemical protection doeshave an external air-flow gauge. The Bullard PA20 PAPRshave a low battery charge and flow sound indicator.

Reprocessing Is EssentialThe PAPR hoods are disposable and intended for singleuse. Reuse of a hood is acceptable, but only by the sameindividual. The hood must be cleaned every time it isremoved, because reuse of a soiled hood presents a contacttransmission risk. Reprocessing procedures must followmanufacturers’ instructions, should be documented inwriting, and should be approved by the hospital infectioncontrol practitioner. Reprocessing personnel and usersmust understand the equipment and be thoroughly versedin procedural details.

The HEPA Filter and Gasket PlacementRequire CareThe 3M Air-Mate filter and gasket should be checked forintegrity and correct placement. This should be done in thereprocessing room while wearing gloves and before theunit is returned to service. The PAPR reprocessing stepsrequire removal of the filter to change the battery and toconfirm that the filter gasket is present and aligned cor-rectly in the groove under the filter. The PAPR unit mustnot be used without an intact and correctly placed filtergasket. The used filter, as does a used N95, presents a risk

of contact but not respiratory transmission, becausedisease-bearing particles do not reaerosolize from the filtermaterial (Table 1, Refs. 13, 16, and 39).

The filter seldom needs to be changed unless it becomeswet or moist, which destroys its filtering capability, orunless the flow decreases despite a fully charged battery.Large industrial particles, such as wood or asbestos, mayclog a filter quickly, but accumulated biological particlesrarely cause the flow to decrease (Table 1, Ref. 31). Adecreased flow is likely to be the result of an incompletelycharged battery.

PAPR Training Is Essential for Safe UseThe requisite PAPR training is more involved and less avail-able than N95 fit testing and training. As with the N95, thePAPR training must be in cooperation with a hospital OSHA-compliant respiratory protection program director, usuallythe director of employee health. This individual may beoverextended with required N95 fit testing and may not havethe time or expertise to undertake PAPR training.

Although instructions may be sought on a “just in time”basis, OSHA strongly encourages users to achieve compe-tency with a respirator in advance. The PAPR competencycan be included in the Department of Anesthesiology initialand periodic required demonstration of competencies withvarious other equipment.

PPE Impedes PerformancePPE impedes performance of manual clinical procedures,impairs hearing and communication, and may triggerclaustrophobia in some users. Although this is a recognizedissue with military and industrial protective gear, the samecould apply with the PAPR used in health care (Table 1,Ref. 1).41 Practicing airway procedures on a mannequinwhile dressed in the PAPR is strongly advised so thatpatient care is not compromised in the clinical setting.Training is effective in improving compliance with PPE useand in reducing distressing symptoms while wearing pro-tective equipment (Table 1, Ref. 1).35,41

PAPR Use Is Contraindicated During SurgeryThe 3M Air-Mate user instructions state that the PAPRshould not be used during surgery (Table 1, Ref. 31). Thepresumed rationale is that positive (outward) air flowcould increase the risk of wound infection. The user in-structions do not specify whether the prohibition applies toboth the double-shrouded full hood and the loose-fittingface cover hood styles, or only to the latter style. The air exitholes of the loose-fitting face cover are located under thechin; thus, if the wearer were standing at the operatingtable, air would exit directly onto the surgical field. Incontrast, the air exits from the double-shrouded hoodunder the accompanying surgical gown and close to thefloor.

OSHA has acknowledged the stated prohibition of PAPRuse during surgery but has not addressed the potential impacton the safety of operating room personnel, most notablyanesthesia providers (Table 1, Ref. 4); neither has the CDCexpressed an opinion on PAPR use in surgery. Therefore, inconsort with their hospitals’ infection control practitioner and



respiratory protection program and employee health direc-tors, anesthesiology departments should address PAPR use inthe operating suite and develop procedures that maximizesafety for both patients and health care workers. At oneauthor’s hospital (Rush), a Department of Anesthesiologypolicy for using the PAPR with the double-shrouded hood inthe surgical suite was developed, approved and documentedin cooperation with the hospital’s infectious disease andinfection control sections.

Outcome DataA comparison of infection rates in health care workers whoused either a PAPR or an N95 for aerosol-generatingprocedures would be pertinent, as would data on PAPR useissues and problems. However, no such data are availableat this time.

The Surgical Hood System Is Not an AcceptableAlternative to the PAPRDuring SARS outbreaks, some health care workers used thesurgical hood system, and added goggles and an N95 uponrecommendation from Stryker (Table 1, Refs. 40 and 41).39

Although the surgical hood system used in orthopedics andspine surgery would seem to be a satisfactory substitute forthe PAPR, it is neither classified nor certified as a respirator,nor has NIOSH evaluated it for that purpose. The Strykersurgical hood system draws ambient air through the materialof the hood and gown, not through a HEPA filter. Anevaluation of 2 surgical hood systems found both to have alower filtration efficiency than either the N95 or the PAPR.42

There is further concern that the incoming surgical hoodsystem airflow blows directly over the wearer’s eyes.

CRITICAL POINTS IN PAPR MAINTENANCE AND USEThe respiratory and contact protection provided by thePAPR will fail if critical factors are neglected (Table 6).

EQUAL PROTECTION FOR HEALTH CAREWORKERS AND PATIENTSThe health and safety of health care workers and patientsalike are supported by a “hierarchy of controls” that reduceexposure, use engineering solutions, enact administrativepolicies, and lastly, facilitate the consistent and effective useof personal protective equipment and transmission-basedprecautions (Table 1, Refs. 1, 20, 42, and 43).4 A culture ofsafety is created when all parties participate with mutualrespect and support in a program that uses effectivemeasures known to minimize nosocomial and workplace-acquired infections (Table 1, Ref. 1).

PROTOCOL FOR TRACHEAL INTUBATIONOF PATIENTS WITH TRANSMISSIBLERESPIRATORY DISEASEA practical guide provides a template for anesthesia depart-ment preparedness (Appendix 2). A written protocol forendotracheal intubation of patients with a virulent transmis-sible respiratory disease should be in place.1,43,44 The protocolshould emphasize the use of appropriate PPE within thecontext of an OSHA-compliant respiratory protection pro-gram. Rehearsing the procedure, avoiding emergent intuba-tion through early intervention, and minimizing a mask leak

or avoiding manual ventilation will reduce exposure risk. Atechnique that prevents coughing and avoids nebulized medi-cation delivery will minimize aerosolization.

Noninvasive ventilation, an aerosol-generating proce-dure, is generally avoided in patients with a virulenttransmissible respiratory disease; however, noninvasiveventilation, in conjunction with infection control air han-dling systems, was used successfully during the SARSepidemic without an increase in health care worker infec-tions.17 Early noninvasive ventilation may reduce the riskof health care worker exposure by avoiding tracheal intu-bation and ventilatory support.17

OPERATING ROOM POLICY AND PROCEDURESFOR INFLUENZAPreviously published recommendations for the manage-ment of SARS and tuberculosis patients in the surgical suitemay be adapted to avian/pandemic/H1N1 influenzaplans. These include recommendations on scheduling ofoperations, air handling controls, isolation procedures,protection of the anesthesia machine, disinfecting andcleaning procedures, and communication with the directorof plant engineering (Table 1, Refs. 14 and 34).17,45 Duringthe course of a pandemic, operating room schedules couldbe profoundly disrupted by staff and equipment shortagesand policies that defer elective operations. Anesthesiamachines might be enlisted as surge capacity ventilatorsand anesthesia professionals as critical care providers.Consideration of these possibilities is encouraged.

A TRAINING WORKSHOPA PAPR workshop in PowerPoint format is available fortraining purposes (Table 1, Refs. 34 and 43). This article andPowerPoint presentation support the workshop’s practicalcomponent that incorporates PPE don/doff sequence andprocedure practice. The workshop was developed tosupplement, not replace, the manufacturers’ training mate-rials, and must be a component of an OSHA-compliantrespiratory protection program.

SUMMARYPandemic influenza is highly contagious and potentiallymany times more lethal than seasonal influenza. Thosewho participate in aerosol-generating procedures, in-cluding endotracheal intubation, are in the OSHA “veryhigh occupational exposure risk” category. OSHA stan-dards require that employers provide safe working con-ditions in a hazardous environment. We must work inconcert with our hospitals and our fellow health careprofessionals to create a culture of safety.

The N95 respirator is limited by face seal leakage.Wearing an N95 in daily work improved the success ofmaintaining a face seal. Using an N95 without having beenfit tested voids the assigned protective factor (APF 10).

When used within the context of an OSHA-compliantrespiratory protection program, the PAPR offers a higher levelof respiratory and contact protection than the N95. The PAPRhas an APF of �1000 when used with the double-shroudedhead- and shoulders-covering hood, and an APF of 25 whenused with the loose-fitting face covering hood. That is, the

SPECIAL ARTICLE


wearer may expect a 1000-fold (or more) or a 25-fold reduc-tion in contaminant within the respirator, depending on thehood choice. Anesthesia providers must be able to performprocedures comfortably while wearing a PAPR. The double-shrouded hood is most appropriate for aerosol-generatingprocedures and the loose-fitting face covering hood for usualbedside care.

The PAPR carries substantial use and maintenanceissues that must be addressed by health care workers andinstitutions alike. The package user instructions and OSHAstate that the PAPR is not to be used during surgerybecause it could increase the risk of wound infection. Thisdilemma is as yet unresolved and should be addressedwith the hospital infection control practitioner.

APPENDIX 2

PRACTICAL GUIDE: AIRWAY MANAGEMENTOF PATIENTS WITH CONTAGIOUSRESPIRATORY DISEASEBackground

• Anesthesiology and Critical Care providers are inthe Occupational Safety and Health Administration(OSHA) “very high exposure risk” category.

• Some influenza patients progress to fatal destructiveviral pneumonia and extreme inflammatory response,“cytokine storm.”

• Airborne (droplet and aerosol) and contact transmis-sion precautions are indicated.

• The limiting features of the N95 respirator are anincreased work of breathing and leakage around theface seal.

• The powered air purifying respirator (PAPR) has ahigher protection factor than the N95; however, inat-tention to details of maintenance and use will void

protection. The surgical hood system is not a respira-tor and is not a satisfactory alternative.

RecommendationsPrepare for Safety

• Get vaccinated and practice sound respiratory,hand, personal, and workplace hygiene.

• Communicate with:• Hospital director of OSHA-compliant respiratory

protection program regarding respirator choiceand training

• Infection control practitioner regarding proce-dures for personal protective equipment (PPE)don/doff sequence, and discuss/receive clear-ance for use of PAPR with full hood in theoperating room

• Plant engineer regarding air flow/filtration controls• Identify an interested anesthesia department coor-

dinator to lead pandemic (and other) emergencypreparedness/response.

• Select a respirator for aerosol-generating procedures.• If N95, be fit tested; remember your brand and

size; wear an N95 daily to improve (but notassure) your ability to maintain a tight face seal;address reuse protocol with hospital infectioncontrol practitioner; use a face shield to protectthe N95 if reuse is intended.

• If PAPR, become familiar with chosen hospitalbrand, or recommend a brand; know manufac-turer’s use and maintenance procedures; practicedon/doff sequence of PAPR and associated con-tact PPE; practice procedures while dressed inPPE; secure same protection for assisting staff.

Prepare for Invasive Airway Procedure• Anticipate/avoid the need for emergent intuba-

tion: use noninvasive ventilatory support; intu-bate preemptively.

• Use air infection isolation room where available orhigh-efficiency particulate air (HEPA) filtered roomexhaust.

• Experienced individual should intubate. Excuse:pregnant; nonessential personnel.

• Don PPE (PAPR plus contact, or N95 plus contact)per protocol, training, and hospital respiratory pro-tection plan before entering the room.

• Confirm that equipment and medications are imme-diately available.

• Use closed suction when possible.• Use disposable or dedicated monitoring equipment.

Remove unnecessary equipment.• Use a HEPA filter between mask and resuscitation bag.• Plan a procedure that will avoid all of the following:

patient coughing, nebulized/topical/transtracheallidocaine, forceful bag-mask ventilation.

• Rehearse the procedure just before intubation.

Conduct Urgent/Emergent Intubation

• Patients in respiratory failure/arrest: Above, andintubate using the practitioner’s technique of great-est success. A laryngeal mask airway may be usedas a bridge or a conduit.

APPENDIX 1. Powered Air Purifying Respirator(PAPR) Brand Name and Type of Protective Filter

Name Protection Filter type“Medical” PAPRsa

1. 3M™ Air-Mate™ Biologic Particulate (HEPA)2. Bullard PA20TM Biologic Particulate3. Bullard EVA™ Biologic Particulate

“Chemical” PAPRs1. 3M Breathe-Easy™ Chemical (gas/vapor) Absorbent

Bbiologic andchemical

Combination

2. Bullard PA30™ Chemical AbsorbentBiologic and

chemicalCombination

3. Bullard EVA Biologic andchemical

Combination

4. MSA OptimAir�b TL Chemical AbsorbentBiologic and

chemicalCombination

4. NorthCompactAir®b

Chemical AbsorbentBiologic and

chemicalCombination

HEPA � high-efficiency particulate air.See Table 1, Refs. 30, 32, and 33.a Industrial use also.b Industrial PAPR.



• Patients in impending respiratory failure: Above,and perform focused history and physical examina-tion; assess the airway; consider glycopyrrolate,preoxygenate.• For normal airway: Administer a muscle relax-

ant, ventilate gently if at all.• For difficult airway: Consider deep sedation with

titrated midazolam, fentanyl, or ketamine. Uselidocaine 1.5 mg/kg IV 1 minute before intuba-tion. Administer muscle relaxant after intubationis confirmed.

• For all cases: Remove PAPR per don/doff proto-col outside the room, with a gloved assistant toplace reusable equipment, including laryngo-scope, in a marked biohazard container for repro-cessing. Reprocess PAPR equipment according tomanufacturer’s recommendation and hospitalprotocol.

Adapt Above for Perioperative Management• Defer elective surgery.• Follow hospital infection control guidelines for pa-

tient transport.• Use HEPA filters on both limbs of anesthesia circuit;

consider disposable circuit.• Remove gloves and wash hands after each patient

contact.• Limit contamination of the anesthesia equipment.• Recover the patient in isolation.• Change PPE before transporting the patient; used

PPE and exposed skin are fomites.

References:(Table 1, Refs. 4, 13, 20, 22, 26, 29, 30–34, and 36–38)1,43,44,45

ACKNOWLEDGMENTSThe authors gratefully thank Chelsea Wanta and TraciNathans-Kelly for editing assistance.

REFERENCES1. Kamming D, Gardam M, Chung FI. Anesthesia and SARS. Br J

Anaesth 2003;90:715–82. Barry JM. The Great Influenza: The Epic Story of the Deadliest

Plague in History. New York: Penguin, 20043. Taubenberger JK, Morens DM. 1918 influenza: the mother of

all pandemics. Emerg Infect Dis 2006;12:15–224. Lau AC, Yip IK, Li MC, Wan M, Sit AW, Lee RA, Yung RW,

Yam LY. Response to SARS as a prototype for bioterrorism:lessons in a regional hospital in Hong Kong. In: McIsaac JH, ed.Hospital Preparation for Bioterror: A Medical and BiomedicalSystems Approach. Boston: Elsevier Academic Press, 2006:281–94

5. Caputo KM, Byrick R, Chapman MG, Orser BJ, Orser BA.Intubation of SARS patients: infection and perspectives ofhealthcare workers. Can J Anaesth 2006;53:122–9

6. Nicolle L. SARS safety and science. Can J Anaesth2003;50:983–8

7. Shaw K. The 2003 SARS outbreak and its impact on infectioncontrol practices. Public Health 2006;120:8–14

8. Taubenberger JK, Morens DM. The pathology of influenzavirus infections. Annu Rev Pathol 2008;3:499–522

9. Gill JR, Sheng ZM, Ely SF, Guinee DG, Beasley MB, Suh J,Deshpande C, Mollura DJ, Morens DM, Bray M, Travis WD,Taubenberger JK. Pulmonary pathologic findings of fatal 2009pandemic influenza A/H1N1 viral infections. Arch Pathol LabMed 2010;134:235–43

10. Yeh E, Luo RF, Dyner L, Hong DK, Banaei N, Baron EJ, PinskyBA. Preferential lower respiratory tract infection in swine-origin 2009 A(H1N1) influenza. Clin Infect Dis 2010;50:391–4

11. Watanabe T, Watanabe S, Shinya K, Kim JH, Hatta M,Kawaoka Y. Viral RNA polymerase complex promotes optimalgrowth of 1918 virus in the lower respiratory tract of ferrets.Proc Natl Acad Sci USA 2009;106:588–92

12. Gupta RK, George R, Nguyen-Van-Tam JS. Bacterial pneumo-nia and pandemic influenza planning. Emerg Infect Dis2008;14:1187–92

13. Stevens J, Blixt O, Tumpey TM, Taubenberger JK, Paulson JC,Wilson IA. Structure and receptor specificity of the hemagglu-tinin from an H5N1 influenza virus. Science 2006;312:404–10

14. Allen JE, Gardner SN, Vitalis EA, Slezak TR. Conserved aminoacid markers from past influenza pandemic strains. BMCMicrobiol 2009;9:77–87

15. de Jong MD, Simmons CP, Thanh TT, Hien VM, Smith GJ,Chau TN, Hoang DM, Chau NV, Khanh TH, Dong VC, Qui PT,Cam BV, Ha do Q, Guan Y, Peiris JS, Chinh NT, Hien TT,Farrar J. Fatal outcome of human influenza A (H5N1) isassociated with high viral load and hypercytokinemia. NatMed 2006;12:1203–7

16. Baskin CR, Bielefeldt-Ohmann H, Tumpey TM, Sabourin PJ,Long JP, Garcia-Sastre A, Tolnay AE, Albrecht R, Pyles JA,Olson PH, Aicher LD, Rosenzweig ER, Murali-Krishna K,Clark EA, Kotur MS, Fornek JL, Proll S, Palermo RE, SabourinCL, Katze MG. Early and sustained innate immune responsedefines pathology and death in nonhuman primates infectedby highly pathogenic influenza virus. Proc Natl Acad Sci USA2009;106:3455–60

17. Lau AC, Yam LY, So LK. Management of critically ill patientswith severe acute respiratory syndrome (SARS). Int J Med Sci2004;1:1–10

18. Ng PC, Lam CW, Li AM, Wong CK, Cheng FW, Leung TF, HonEK, Chan IH, Li CK, Fung KS, Fok TF. Inflammatory cytokineprofile in children with severe acute respiratory syndrome.Pediatrics 2004;113:e7–14

19. Pang BS, Wang Z, Zhang LM, Tong ZH, Xu LL, Huang XX,Gao WJ, Zhu M, Wang C. Dynamic changes in blood cytokinelevels as clinical indicators in severe acute respiratory syn-drome. Chin Med J (Engl) 2003;116:1283–7

20. Nicas M, Jones RM. Relative contributions of four exposurepathways to influenza infection risk. Risk Anal 2009;29:1292–303

21. Tellier R. Review of aerosol transmission of influenza A virus.Emerg Infect Dis 2006;12:1657–62

22. Tellier R. Aerosol transmission of influenza A virus: a reviewof new studies. J R Soc Interface 2009;6(Suppl 6):S783–90

23. Roy CJ, Milton DK. Airborne transmission of communicableinfection: the elusive pathway. N Engl J Med 2004;350:1710–2

24. Tong TR, Liang C, Nicastri E, Petrosillo N, Puro V, Fowler RA,Scales DC, Ilan R, Yu ITS, Li Y. Evidence of airborne transmis-sion of SARS [correspondence]. N Engl J Med 2004;351:609–11

25. Yu IT, Li Y, Wong TW, Tam W, Chan AT, Lee JH, Leung DY,Ho T. Evidence of airborne transmission of the severe acuterespiratory syndrome virus. N Engl J Med 2004;350:1731–9

26. Gamage B, Moore D, Copes R, Yassi A, Bryce E. Protectinghealth care workers from SARS and other respiratory patho-gens: a review of the infection control literature. Am J InfectControl 2005;33:114–21

27. Scales DC, Green K, Chan AK, Poutanen SM, Foster D, NowakK, Raboud JM, Saskin R, Lapinsky SE, Stewart TE. Illness inintensive care staff after brief exposure to severe acute respi-ratory syndrome. Emerg Infect Dis 2003;9:1205–10

28. Varia M, Wilson S, Sarwal S, McGeer A, Gournis E, Galanis E,Henry B. Investigation of a nosocomial outbreak of severeacute respiratory syndrome (SARS) in Toronto, Canada. CMAJ2003;169:285–92

29. Alford RH, Kasel JA, Gerone PJ, Knight V. Human influenzaresulting from aerosol inhalation. Proc Soc Exp Biol Med1966;122:800–4

30. Cohen HJ, Hecker LH, Mattheis DK, Johnson JS, Biermann AH,Foote KL. Simulated workplace protection factor study ofpowered air-purifying and supplied air respirators. AIHAJ2001;62:595–604

SPECIAL ARTICLE


31. Loeb M, Dafoe N, Mahony J, John M, Sarabia A, Glavin V,Webby R, Smieja M, Earn DJ, Chong S, Webb A, Walter SD.Surgical mask vs N95 respirator for preventing influenzaamong health care workers: a randomized trial. JAMA2009;302:1865–71

32. Radonovich LJ Jr, Perl TM, Davey V, Cohen H. Preventing thesoldiers of health care from becoming victims on the pandemicbattlefield: respirators or surgical masks as the armor of choice.Disaster Med Public Health Prep 2009;3(Suppl 2):S203–10

33. Srinivasan A, Perl TM. Respiratory protection against influ-enza. JAMA 2009;302:1903–4

34. Fennelly KP. The role of masks in preventing nosocomialtransmission of tuberculosis. Int J Tuberc Lung Dis1998;2:S103–9

35. Yassi A, Lockhart K, Copes R, Kerr M, Corbiere M, Bryce E,Danyluk Q, Keen D, Yu S, Kidd C, Fitzgerald M, Thiessen R,Gamage B, Patrick D, Bigelow P, Saunders S. Determinants ofhealthcare workers’ compliance with infection control proce-dures. Healthc Q 2007;10:44–52

36. Li Y, Tokura H, Guo YP, Wong AS, Wong T, Chung J, NewtonE. Effects of wearing N95 and surgical facemasks on heart rate,thermal stress and subjective sensations. Int Arch OccupEnviron Health 2005;78:501–9

37. Lee MC, Takaya S, Long R, Joffe AM. Respirator-fit testing:does it ensure the protection of healthcare workers againstrespirable particles carrying pathogens? Infect Control HospEpidemiol 2008;29:1149–56

38. Phin NF, Rylands AJ, Allan J, Edwards C, Enstone JE, Nguyen-Van-Tam JS. Personal protective equipment in an influenzapandemic: a UK simulation exercise. J Hosp Infect 2009;71:15–21

39. Zamora JE, Murdoch J, Simchison B, Day AG. Contamination:a comparison of two personal protective systems. CMAJ2006;175:249–54

40. Casanova L, Alfano-Sobsey E, Rutala WA, Weber DJ, SobseyM. Virus transfer from personal protective equipment tohealthcare employees’ skin and clothing. Emerg Infect Dis2008;14:1291–3

41. Barach P, Rivkind A, Israeli A, Berdugo M, Richter ED.Emergency preparedness and response in Israel during theGulf War. Ann Emerg Med 1998;32:224–33

42. Derrick JL, Gomersall CD. Surgical helmets and SARS infec-tion. Emerg Infect Dis 2004;10:277–9

43. Cooper A, Joglekar A, Adhikari N. A practical approach toairway management in patients with SARS. CMAJ 2003;169:785–7

44. Peng PW, Wong DT, Bevan D, Gardam M. Infection controland anesthesia: lessons learned from the Toronto SARS out-break. Can J Anaesth 2003;50:989–97

45. Stackhouse RA. Severe acute respiratory syndrome and tuber-culosis. Anesthesiol Clin North America 2004;22:437–55



CME

Early Postoperative Subcutaneous Tissue OxygenPredicts Surgical Site InfectionRaghavendra Govinda, MD,*† Yusuke Kasuya, MD,†‡ Endrit Bala, MD,§ Ramatia Mahboobi, MD,§Jagan Devarajan, MD,§ Daniel I. Sessler, MD,§ and Ozan Akca, MD†�

BACKGROUND: Subcutaneous oxygen partial pressure is one of several determinants of surgicalsite infections (SSIs). However, tissue partial pressure is difficult to measure and requiresinvasive techniques. We tested the hypothesis that early postoperative tissue oxygen saturation(StO2) measured with near-infrared spectroscopy predicts SSI.METHODS: We evaluated StO2 in 116 patients undergoing elective colon resection. Saturationwas measured near the surgical incision, at the upper arm, and at the thenar muscle with anInSpectra™ tissue spectrometer model 650 (Hutchinson Technology Inc., Hutchinson, MN) 75minutes after the end of surgery and on the first postoperative day. An investigator blinded toStO2 assessed patients daily for wound infection. Receiver operating characteristic curves wereused to analyze the performance of StO2 measurements as a predictor of SSI.RESULTS: In 23 patients (�20%), SSI was diagnosed 9 � 5 days (mean � SD) after surgery.Patients who did and did not develop an SSI had similar age (48 � 14 vs 48 � 15 years,respectively; P � 0.97) and gender (female:male, 15:8 vs 46:47, respectively), but patients whodeveloped SSI weighed more (body mass index 32 � 7 vs 27 � 6 kg/m2; P � 0.01). StO2 at theupper arm was lower in patients who developed SSI than in those who did not develop SSI (52 �22 vs 66 � 21; P � 0.033), and these measurements had a sensitivity of 71% and specificityof 60% for predicting SSI, using StO2 of 66% as the cutoff point.CONCLUSION: StO2 measured at the upper arm only 75 minutes after colorectal surgerypredicted development of postoperative SSI, although the infections were typically diagnosedmore than a week later. Although further testing is required, StO2 measurements may be able topredict SSI and thus allow earlier preventive measures to be implemented. (Anesth Analg 2010;111:946–52)

Surgical site infections (SSIs) are perhaps the mostcommon serious complication of anesthesia and sur-gery. They cause considerable morbidity and are

expensive to treat.1,2 The transition from contamination toestablished infection occurs during a decisive period thatprobably lasts only a few hours, even though infections aretypically detected a week or longer after surgery.3,4 Ifantibiotics are administered during this decisive period,they are more effective in reducing infection risk than whengiven later.5 Tissue oxygen tension levels are one of the bestfactors established for predicting SSI.6,7 Oxidative burstfunction of neutrophils is one of the primary defensesagainst SSI.8 Oxidative burst depends on the Po2 over the

entire physiological range of tissue values.9 Interventions toimprove tissue oxygenation during or immediately aftersurgery are thus most likely to reduce the morbidity andmortality associated with SSI.10,11

Tissue oxygenation has traditionally been measuredwith Clark-type electrodes or similar systems. However,these methods are invasive, expensive, and require exper-tise to use.12,13 Near-infrared spectroscopy (NIRS) is analternative noninvasive technique.14–17 We tested the hy-pothesis that tissue oxygen saturation (Sto2) measured soonafter surgery with NIRS predicts ultimate development ofSSI. Confirming this hypothesis would allow early clinicalinterventions, which might reduce the risk of infection.

METHODSWith approval by the Human Studies Committees at theUniversity of Louisville and the Cleveland Clinic, andwritten informed consent by the participants, we included116 colorectal surgery patients. Forty-three of these patientshad laparoscopic-assisted colorectal procedures; the re-mainder had laparotomies (Table 1).

In a preliminary study, we found that patients whodeveloped SSI had lower Sto2 measurements at the thenareminence �15% �20% (mean � SD) than those whoremained uninfected postoperatively; the analogous abso-lute difference was �20% �30% at the upper arm.18 Usingthese estimates, a sample size of 80 measurements at thethenar eminence and 132 at the upper arm was calculatedto achieve 90% power with an � of 0.05. We therefore

From the *Department of Anesthesiology, Tufts Medical Center, Boston,Massachusetts; †Department of Anesthesiology and Perioperative Medicine,and �Neuroscience ICU, University of Louisville, Louisville, Kentucky;‡Tokyo Women’s Medical University, Tokyo, Japan; and §Department ofOutcomes Research, Cleveland Clinic, Cleveland, Ohio.


Supported in part by Hutchinson Technology Inc. (Hutchinson, MN) and theJoseph Drown Foundation (Los Angeles, CA).

Data were presented in part at the American Society of AnesthesiologistsAnnual Meeting in Orlando, FL (October 2008).

All authors are also affiliated with the Outcomes Research Consortium.


Address correspondence and reprint requests to Ozan Akca, MD, Depart-ment of Anesthesiology and Perioperative Medicine, University of Louis-ville Hospital, 530 S. Jackson St., Louisville, KY 40202. Address e-mail [email protected].

Copyright © 2010 International Anesthesia Research SocietyDOI: 10.1213/ANE.0b013e3181e80a94


enrolled 116 patients to complete the study with sufficientpower for both outcomes.

The study enrollment period was from July 2007 to May2008. Adults between 18 and 80 years of age with an ASAphysical status I to III were included in the study. Patientswith intestinal obstruction, those in whom the surgeon didnot anticipate a primary wound closure, and those with adiagnosed or suspected intraabdominal abscess were ex-cluded from the study. Patients were also excluded if theyhad severe chronic obstructive pulmonary disease, recentmyocardial infarction, unstable angina, or required oxygenpreoperatively.

ProtocolAll patients received antibiotics before surgery according tothe protocols of the 2 participating institutions, each ofwhich required administration of appropriate prophylacticantibiotic within an hour before surgical incision. Patientswere given midazolam for premedication, propofol oretomidate for induction of anesthesia, and succinylcholineor rocuronium for initiation of muscle relaxation. Musclerelaxation was maintained with rocuronium or vecuronium.

Anesthesia was maintained with sevoflurane, desflu-rane, or isoflurane in 30% to 80% oxygen, supplementedwith fentanyl or morphine. Intraoperative IV fluid manage-ment was at the discretion of the attending anesthesiologist,and consisted of 8 to 10 mL/kg. Core temperature wasmaintained near 36°C by using forced-air warming blan-kets and fluid warmers. Hair was clipped from the surgicalsite immediately preoperatively, and the skin was preparedwith a chlorhexidine-based antiseptic kit. Postoperative

analgesia was provided with IV intermittent boluses or viapatient-controlled analgesia with morphine or hydromor-phone. If an epidural catheter was placed preoperatively,epidural analgesia with fentanyl (�25 to 50 �m/h) wasused intraoperatively and approximately for the first 2hours of the recovery if it was the preference of theattending anesthesiologist. A combination of local anes-thetic and opioid via the epidural was only administeredafter the study was completed.

Spontaneously breathing patients were given oxygenvia nasal cannula or Venturi mask. The inspired oxygenconcentration was adjusted to maintain pulse oximetersaturation (Spo2) �95%. Sto2 was measured 15 minutesafter weaning from oxygen and thereafter patients receivedsupplemental oxygen as required to maintain Spo2 �92%.

MeasurementsThe duration of surgery, blood loss, intraoperative crystal-loid and colloid administration, hemodynamic variables,blood transfusion requirements, and urine output wererecorded.

Two systems were used to evaluate the SSI risk: theStudy on the Efficacy of Nosocomial Infection Control(SENIC) and National Nosocomial Infection SurveillanceSystem (NNISS). The SENIC scoring system assigns 1 pointfor each of the following factors: �3 underlying diagnoses,surgery that lasts �2 hours, an abdominal site of surgery,and the presence of a contaminated or infected wound.19

The NNISS predicts risk on the basis of the contamination

Table 1. Demographic and Morphometric Characteristics, and Potential Confounding FactorsPatients withSSI (n � 23)

Patients withoutSSI (n � 93) P value

Demographic and morphometric characteristicsAge (y) 48 � 14 48 � 15 0.971Sex (female/male) 15/8 46/47 0.175Weight (kg) 92 � 21 78 � 20 0.006BMI (kg/m2) 32 � 6.8 27 � 5.5 �0.001ASA physical status I/II/III, n 2/15/6 10/57/26 0.426Smokers, n (%) 4 (17%) 18 (19%) 0.830Cardiac disease, n (%) 7 (30%) 25 (27%) 0.732Diabetes mellitus, n (%) 3 (13%) 4 (4%) 0.115Previous chemotherapy, n (%) 2 (9%) 4 (4%) 0.394Preoperative Hct (%) 40 � 4 40 � 5 0.612SENIC I/II/III (%) 17/78/5 9/80/11 0.373NNISS I/II/III (%) 43.5/43.5/13 39/48/13 0.824

Intraoperative confoundersCrystalloids (mL) 3273 � 1414 3274 � 1500 0.996Crystalloids (mL)a 3150 (2550–3875) 3000 (2400–4000)Colloids (mL) 620 � 532 439 � 452 0.101Colloids (mL)a 500 (0–1000) 500 (0–813)Blood loss (mL) 549 � 511 373 � 378 0.078Blood loss (mL)a 350 (200–800) 250 (150–463)PRBC transfusion (U) 0.1 � 0.4 0.2 � 0.5 0.541Duration of surgery (h) 3.4 � 1.6 3.5 � 1.5 0.784Intraoperative core temperature (°C) 36.1 � 0.5 36.1 � 0.6 0.911Intraoperative mean FIO2 62 � 18 61 � 19 0.816Mean arterial blood pressure (mm Hg) 85 � 8 86 � 9 0.470Laparoscopic-assisted procedures, n (%) 6 (26%) 37 (40%) 0.335Epidural analgesia, n (%) 0 (0%) 5 (5%) 0.581

SSI � surgical site infection; BMI � body mass index; SENIC � Study on the Effect of Nosocomial Infection Control; NNISS � National Nosocomial InfectionSurveillance System; Hct � hematocrit; PRBC � packed red blood cell; FIO2 � fraction of inspired oxygen.Data are presented as mean � standard deviation, count (percentage), or a Median (25th–75th quartile).


of surgery, the rating of physical status on a scale devel-oped by the American Society of Anesthesiologists, and theduration of surgery.20

Sto2 and tissue hemoglobin index (THI) were measuredby an InSpectra™ tissue spectrometer model 650 (HutchinsonTechnology Inc., Hutchinson, MN) 75 minutes after surgeryand on the first postoperative day (Fig. 1). A time point of75 minutes postoperatively was chosen because patients areusually weaned from supplemental oxygen at the end of thefirst postoperative hour, and waiting an additional 15 minutesgenerally allows full washout of supplemental oxygen.

The InSpectra tissue spectrometer measures Sto2 usingwide-gap second-derivative NIRS.21,22 The InSpectra spec-trometer makes use of the characteristic absorption prop-erties of hemoglobin in the near-infrared wavelength rangebetween 680 and 800 nm. The absorption spectrum of lightremitted from the tissue sample varies mainly with oxyhe-moglobin and deoxyhemoglobin concentration; otherchromophores have minimal effect. Sto2 is a measure ofhemoglobin oxygen saturation of the blood contained in thevolume of tissue illuminated by the near-infrared light. Themaximum depth of the tissue sampled is the distancebetween the probe’s send and receive fibers. Mean mea-surement depth is half the probe’s spacing. We used a15-mm probe that measures Sto2 of 5- to 8-mm tissue depth.For the forearm and wound sites, this corresponds tosubcutaneous tissue above the skeletal muscle. For thethenar muscle site, this depth corresponds to muscle.

Sto2 near the wound was measured 2.5 cm lateral to theincision at the upper, middle, and lower third of the incision.The upper lateral arm site was chosen for measurementbecause this site reflects the Sto2 of operative wounds in thechest and abdomen even though it is approximately 10 mmHg higher than in the wound area.6,12 The thenar muscle sitewas chosen because it is one of the best established sites fortissue oximetry measurement.23,24 At each of these sites, theprobe was placed for 15 to 30 seconds until a stable oxygensaturation value was obtained.

THI was measured simultaneously throughout the studyfrom the same probe as the Sto2 was monitored. THI is not yeta well-established value. As with several others, this NIRS-based device can provide a hemoglobin value obtained from

the probe’s detection site, but this hemoglobin value does notnecessarily represent the body’s total hemoglobin concentra-tion. THI monitors the total amount of hemoglobin in thetissue site where the NIRS probe is applied. Therefore, it maybe considered more of a perfusion variable.

During each set of measurements, mean arterial bloodpressure, heart rate, and Spo2 were recorded simultaneously.We asked the patients to rate their pain using a 10-cm visualanalog scale.

An independent investigator not aware of the Sto2

measurements evaluated patients’ wounds daily through-out their hospitalization. SSI was diagnosed according tothe surgical wound infection definitions from the Centersfor Disease Control and Prevention (CDC).19,25

Data AnalysisPrimary outcomes were the Sto2 measured around thesurgical incision, at the upper arm, and at the thenarmuscle. Surgical site Sto2 measurements were summarizedas a mean value and presented as a post hoc analysis.Specifically, subcutaneous Sto2 values at the upper one-third, middle one-third, and lower one-third of the surgicalincision were averaged into a single “incision” value foreach patient.

Normally distributed data are presented as means � SD.Skewed data are presented as medians and interquartileranges. After descriptive analysis of all parameters, univar-iate analysis was performed using the �2 test for categoricalvariables and unpaired, 2-tailed t and Kruskal-Wallis testsfor continuous variables. P values �0.05 were consideredstatistically significant. Additionally, a multivariate analy-sis was performed to assess the independent contributionof each potential variable (SPSS Inc., Chicago, IL).

Receiver operating characteristic (ROC) curves weredeveloped for Sto2 in the upper arm and Spo2 on the firstpostoperative day to predict surgical wound infections. Wecompared our predictions based only on early postopera-tive upper arm Sto2, using an Sto2 of 66% as the cutoffpoint, with SENIC predictions. The CDC in the SENICdeveloped a predictive model for the risks of SSI that hasbecome the standard.19

RESULTSOf the 116 patients enrolled, 23 developed SSI (20%). If thediverticulitis cases were excluded, the SSI rate woulddecrease to 14%. All of the 23 patients who developed SSIhad superficial incisional site infections. Three of thesepatients also developed deep incisional site infections, and4 developed peritoneal infections as defined by CDC crite-ria. Infections were diagnosed an average of 9 � 5 daysafter surgery.

Age, ASA physical status, and SENIC and NNISS riskscores were similar in the patients who developed SSI andthose who remained uninfected (Table 1). Patients with SSIhad a greater body mass index. There were no statisticallysignificant or clinically significant differences in the dura-tion of surgery, IV fluid administration, intraoperativetemperature, or blood transfusion requirements. Surgicaltechnique and use of epidural analgesia were also similar inthe patients who developed SSI and those who did not(Table 1).

Figure 1. InSpectra™ StO2 tissue oxygenation monitor.

Tissue Oxygen and Surgical Site Infection


A significant proportion of the patients (35%–38%)could not be weaned from supplemental oxygen at the75th-minute measurement period (Table 2). Additionally,from this group of patients, there were 14 patients (3 withSSI and 11 with no SSI) who required oxygen rates of morethan 2 to 3 L/min. During the first postoperative daymeasurements, 70% to 75% of the patients were weanedfrom supplemental oxygen. Therefore, some of the study

measurements were done under supplemental oxygen be-cause of the oxygenation needs of the patients as specifiedabove.

In the early postoperative period (Table 2), Sto2 at thesurgical incision did not differ significantly between pa-tients who eventually developed SSI and those who re-mained uninfected. The THI near the surgical incision was25% lower in patients who developed SSI, but this differ-ence did not reach statistical significance.

In contrast, Sto2 at the upper lateral arm was signifi-cantly lower in the patients who developed SSI (52 � 22mm Hg) than in those who did not (66 � 21 mm Hg; P �0.033). ROC curve for Sto2 at the upper lateral arm had asensitivity of 71% and specificity of 60% using an Sto2 of66% as the cutoff point for predicting SSI (Fig. 2). Thepositive predictive value of this cutoff value was 29%, witha negative predictive value of 90%. The THI measured inthe early postoperative period at the upper arm was alsostatistically lower in the patients who developed SSI (3.8 �1.3 vs 5.5 � 3.2 g/dL). ROC curves for THI at the upperlateral arm in the early postoperative period had a sensi-tivity of 88% and specificity of 55% for predicting SSI at aTHI of 4.3. Figure 3 shows the difference between observedand expected (based on SENIC scores) infection risk as afunction of early postoperative Sto2 at the upper arm.Because intraoperative oxygen concentrations were notcontrolled per protocol, we calculated whether there wasany correlation between inspired intraoperative oxygenconcentrations and postoperative Sto2 at the upper lateralarm. There was no correlation (r2 � 0.007).

Figure 2. A, Upper arm tissue oxygen saturation(StO2) measured 75 minutes after surgery in pa-tients who did and did not develop surgical siteinfections (SSIs). B, The associated receiver oper-ating characteristic curve.

Table 2. Major Study Outcomes: Tissue OxygenSaturation Given as Percentage

Measurements

Patientswith SSI(n � 23)

Patientswithout SSI

(n � 93)P

value75 min after surgery (early

postoperative period)StO2 at surgical incision (%) 44 � 16 51 � 20 0.184THI at surgical incision 3.8 � 1.1 4.3 � 2.0 0.229StO2 at upper arm (%) 52 � 22 66 � 21 0.033THI at upper arm 3.8 � 1.3 5.3 � 3.2 0.041StO2 at thenar eminence (%) 69 � 17 74 � 16 0.210THI at thenar eminence 8.2 � 3.6 8.9 � 3.1 0.403SpO2 (%) 97 � 3 98 � 2 0.193MAP (mm Hg) 94 � 15 93 � 14 0.831Pain (cm VAS) 7.6 � 1.5 6.3 � 2.6 0.048Pain (cm VAS)a 8.0 (7.0–8.3) 6.0 (4.0–8.0) 0.043No. of patients on room air 13 (57%) 48 (52%) 0.403No. of patients receiving

nasal O2 at 2–3 L/min8 (35%) 35 (38%) 0.746

No. of patients with FIO2

�0.602 (8%) 10 (10%) 0.467

First postoperative dayStO2 at surgical incision (%) 44 � 25 50 � 24 0.300THI at surgical incision 4.5 � 1.3 5.0 � 2.1 0.410StO2 at upper arm (%) 61 � 13 60 � 22 0.825THI at upper arm 4.1 � 0.9 4.9 � 2.6 0.193StO2 at thenar eminence (%) 77 � 12 79 � 13 0.538THI at thenar eminence 11.0 � 2.7 10.9 � 2.8 0.824SpO2 (%) 95 � 3 97 � 2 0.001MAP (mm Hg) 89 � 11 84 � 13 0.068Pain (cm VAS) 5.8 � 2.2 4.9 � 2.5 0.152Pain (cm VAS)a 5.0 (4.6–7.8) 5.0 (3.0–6.5)Hct (%) 35 � 5 32 � 5 0.436No. of patients on room air 16 (70%) 69 (74%)No. of patients receiving

nasal O2 at 2–3 L/min6 (26%) 23 (25%)

No. of patients with FIO2

�0.601 (4%) 1 (1%)

SSI � surgical site infection; MAP � mean arterial blood pressure; Hct �hematocrit; VAS � 10-cm-long visual analog scale for pain; StO2 � tissueoxygen saturation; THI � tissue hemoglobin index; SpO2 � pulse oximeteroxygen saturation; FIO2 � fraction of inspired oxygen.Data are presented as mean � SD, count (percentage), or a Median(25th–75th quartile).

Figure 3. Surgical wound infection rates (%) stratified by earlypostoperative tissue oxygen saturation (StO2) (%) values measuredat the upper arm. The bars represent the difference between theobserved infection rates and those expected based on the Study onthe Effect of Nosocomial Infection Control (SENIC) multivariate riskindex.


Thenar muscle oxygenation and THI values were bothhigher than those of the subcutaneous tissue values in bothgroups of patients. Neither the Sto2 nor the THI measuredat the thenar eminence 75 minutes after surgery differedsignificantly in patients who did or did not develop SSI(Table 2). Early postoperative visual analog scale painscores were slightly, but statistically significantly, higher inpatients who developed SSI (Table 2).

Patients who developed SSI had lower Spo2 values onthe first postoperative day (95% � 3% vs 97% � 2%; P �0.001). The ROC curve for the first postoperative day Spo2

had a sensitivity of 75% and specificity of 73% using anSpo2 of 95% as the cutoff point for predicting SSI.

Multivariate logistic regression analyses indicated thatthe first postoperative day Spo2 value was the only statis-tically significant independent factor contributing to SSI.Although body mass index, postoperative pain, intraopera-tive blood loss, and upper arm Sto2 data provided a goodclinical and statistical difference in univariate analysis, theydid not reach independent significance levels with multi-variate statistics (Table 3).

DISCUSSIONThe link between tissue oxygenation and surgical woundinfection is well established. This concept, developed byThomas Hunt in the 1960s and 1970s26,27 led to a landmarkarticle by Hopf et al.6 more than a decade ago. Hopf et al.used a subcutaneous needle-guided tonometric siliconcatheter system into which they inserted a polarographicClark-type electrode to monitor tissue oxygen partial pres-sure. They obtained measurements at 3 different times:within 6 hours of surgery, on the first postoperative day,and on the second postoperative day. They observed thattissue oxygenation was a strong predictor of SSI.

A subcutaneous tissue oxygen monitoring system (LICOX;Medical Systems Corp., Greenvale, NY) has been used byvarious investigators, including us, for decades and is consid-ered the “gold standard” for tissue oxygen monitoring. How-ever, the LICOX system is invasive, expensive, and requiresapproximately 45 minutes of calibration and equilibrationwith tissue as well as considerable operator experience to beaccurate. Therefore, a simple and noninvasive tissue oxygenmonitoring system might be easier and more feasible forperioperative use. The InSpectra Sto2 system is noninvasiveand relatively easy to use. Our main study question waswhether NIRS Sto2, measured immediately postoperatively,would be able to predict SSI. Our principal finding was that

patients who eventually developed SSI according to the CDCcriteria had lower Sto2 at the upper arm only 75 minutes aftersurgery, which was approximately 9 days before the diagno-sis of SSI was made clinically.

Upper arm Sto2 provided a “fair” area under the ROCcurve and was a better predictor of infection than theestablished SENIC or NNISS risk scores. This point isimportant because the high-risk patients we identifiedwould generally not have been detected using the SENICrisk score or other routinely available clinical systems.

As we have shown previously, subcutaneous tissue inthe upper lateral arm is relatively well oxygenated andperfused under general anesthesia and in the awakestate.1,28,29 An average of 48 � 19 perforator arterioles from15 vascular territories supply the integument of the upperextremity. Septocutaneous arteries predominate in theshoulder, elbow, distal forearm, and hand regions. Muscu-locutaneous perforators are more numerous in the upperarm and proximal forearm. The average perforator size inthe upper extremity is approximately 0.7 � 0.2 mm indiameter and supplies an average area of 35 cm2.30,31

Therefore, we can extrapolate that our upper arm Sto2

measurement likely included a site perfused with at least 1perforator artery.

Leukocyte-mediated oxidative burst and collagen for-mation require oxygen partial pressures of at least 40 mmHg.32 Upper arm subcutaneous tissue oxygen tension istypically �50 mm Hg under a fraction of inspired oxygen�0.4 even during sympathetic vasoconstriction. Our tissueoxygen tension values were lower than those reportedpreviously (27 to 35 mm Hg). This difference might resultfrom differences in oxygen monitoring techniques. Sto2

measures hemoglobin’s oxygen saturation in a focal area.Sto2 systems calculate the hemoglobin’s oxygen saturationin the volume of tissue illuminated by near-infrared light.Because in the majority of the peripheral tissues tested(surgical wound area and upper arm subcutaneous tissue)the THI values were low, we can extrapolate that there wasless perfusion in the site of interest. Therefore, Sto2 pro-vides a different oxygenation value than tissue oxygentension, which monitors the partial pressure of free oxygenin the tissue. Free oxygen was reported to be the mainsource of oxygen available for the tissues.33 This interestingconcept of tissue oxygenation will continue to be an in-tensely debated topic until applicability of monitoring,reproducibility of oxygenation data, and until it is provenwith clinical outcomes.

We are not the first to use Sto2 to predict SSI. Ives etal.16,34 reported that Sto2 of 53% at the surgical incision site,measured 12 hours postoperatively, provided 71% sensitiv-ity and 76% specificity as a test to predict SSI. Our studyextends previous work by showing that infection can bepredicted shortly after surgery, during the decisive period.But interestingly, Ives et al. could not find any difference inthe upper arm Sto2 measurements between the patientswho did and did not develop SSI. A possible explanation isthat Ives et al. measured tissue oxygenation 10 cm belowthe shoulder tip over the bulk of the biceps muscle. Incontrast, we used the lateral upper arm, which was ap-proximately 15 cm below the shoulder tip, the area betweenthe biceps and triceps brachii’s lateral head. This is the site

Table 3. Independent Contributors to SurgicalSite Infections (Multivariate Analysis)

Oddsratio

95% Confidenceinterval P value

Body mass index (kg/m2) 1.081 0.984–1.188 0.106Blood loss 1.001 0.998–1.002 0.157Upper arm StO2

(postoperative 75 min)0.975 0.945–1.006 0.108

Pain VAS (postoperative75 min)

1.292 0.944–1.767 0.109

SpO2 (postoperative day 1) 0.765 0.603–0.971 0.044

StO2 � tissue oxygen saturation; VAS � visual analog scale; SpO2 � pulseoximeter oxygen saturation.



used in most previous studies.6,28,35 The reason behind ourfailure to find a statistically significant difference at thesurgical site can be explained (1) by the lack of power andsmall sample size, and (2) because of the stretch of theperi-incisional tissues during surgery, there might havebeen tissue edema to blunt the extent of oxygenationdifference. If the difference between the groups continuedconsistently, 170 to 180 patients would have provided astatistically significant difference.

Another interesting finding of the current trial is thatSpo2 values of the first postoperative day apparentlypredicted SSI. Before taken into consideration, some majorlimitations of these data need to be addressed: (1) Spo2 wasnot planned as one of the a priori outcomes of this study;(2) the data were gathered from only a few values from asnapshot period; and (3) although the difference wasstatistically significant, clinical meaning may be of limitedvalue.

Early detection of patients at special risk of woundinfection is important because there are well-establishedinterventions that improve tissue oxygenation and maythus reduce infection rate and possibly improve surgicaloutcomes. For example, sympathetic nervous system acti-vation triggers vasoconstriction and reduces tissue oxygen-ation.36 A major mediator of sympathetic activity, and onethat can often be treated, is surgical pain. In fact, it is wellestablished that adequate analgesia improves tissue oxy-genation,37,38 which is supported by the fact that patients inour study who eventually developed SSI had higher painscores in the early postoperative period.

Thermoregulatory vasoconstriction is another factorthat decreases tissue oxygen tension and perfusion.39,40

As might thus be expected, maintaining perioperativenormothermia41 and local warming42 decreases SSI rate.Supplemental fluid administration increases tissue oxy-genation,43,44 but it does not necessarily improve the SSIrate.45 However, supplemental fluids may have short-term tissue oxygenation benefits that should be consid-ered despite the ongoing perioperative fluid debates.46

Finally, supplemental oxygen (i.e., 80% inspired oxygen)approximately doubles tissue oxygen partial pressure1

without causing atelectasis.37,47

There are thus at least 4 established interventions thatimprove tissue oxygenation and could be implementedrelatively easily in patients found to have low Sto2 duringthe decisive period. We caution, however, that although thelink between tissue oxygenation and SSI is clear, whethertitrating interventions perioperatively to improve Sto2 ac-tually reduces infection risk is yet to be proven.

Despite the promising data provided in this trial sup-porting the link between oxygenation and SSI, there aresome important limitations that need to be addressed. Thevalues of the upper arm Sto2 were relatively weak (lowROC area), which were also apparent with the nonsignifi-cant odds ratio obtained from the multivariate analysis. Alarger trial aiming to reconfirm the current results in alarger and wider surgical population will be needed. Moreimportantly, no study in human subjects has validated theaccuracy of Sto2 by comparing it with the “gold standard”tissue oxygen tension monitoring. Additionally, it is impor-tant to recognize that the reflectance spectroscopy-based

oximeters calculate the mean value of oxyhemoglobin acrossall the vessels of the microcirculation of the skin. Therefore,the derived oximeter saturation is a mean blood oxygensaturation across arterioles, capillaries, and venules.48

Another important limitation is in the claims made bydiagnosing potential infections in a decisive period. Thisdecisive period was established for antibiotic prophy-laxis as the intervention. Therefore, the decisive periodresponding to increased oxygenation needs to be testedbefore considering the validity of the proposed interven-tions. Using an animal model, Knighton et al.7 showedthat oxygen might have a decisive period of at least 6hours.

In summary, Sto2 measured at the upper arm only 75minutes postoperatively predicts development of surgicalwound infection after colorectal surgery, even thoughinfections were typically diagnosed more than a week later.Because high-risk patients can potentially be identifiednoninvasively, it may also be possible to intervene toimprove tissue oxygenation.

ACKNOWLEDGMENTSSamual Chen, MD, Luke Reynolds, MSc, Adrian Alvarez, MD,and Gena Harrison, BA (Department of Outcomes Research,Cleveland Clinic) are acknowledged for their assistance withdata acquisition. We appreciate the efforts of Nancy Alsip,PhD, and Gilbert Haugh, MS (OCRSS, University of Louis-ville), in medical editing and statistical assistance; and JosephOrtner and Hutchinson Technology Inc. (Hutchinson, MN) fortechnical support and for providing tissue oximeters and theirprobes.

REFERENCES1. Greif R, Akca O, Horn EP, Kurz A, Sessler DI, Outcomes

Research™ Group. Supplemental perioperative oxygen to re-duce the incidence of surgical wound infection. N Engl J Med2000;342:161–7

2. Olsen MA, Chu-Ongsakul S, Brandt KE, Dietz JR, Mayfield J,Fraser VJ. Hospital-associated costs due to surgical site infec-tion after breast surgery. Arch Surg 2008;143:53–60

3. Miles AA, Miles EM, Burke J. The value and duration ofdefence reactions of the skin to the primary lodgement ofbacteria. Br J Exp Pathol 1957;38:79–96

4. Polk HC Jr. The prophylaxis of infection following operativeprocedures. J Ky Med Assoc 1974;72:139–43

5. Classen DC, Evans RS, Pestotnik SL, Horn SD, Menlove RL,Burke JP. The timing of prophylactic administration of antibi-otics and the risk of surgical-wound infection. N Engl J Med1992;326:281–6

6. Hopf HW, Hunt TK, West JM, Blomquist P, Goodson WH III,Jensen JA, Jonsson K, Paty PB, Rabkin JM, Upton RA, vonSmitten K, Whitney JD. Wound tissue oxygen tension predictsthe risk of wound infection in surgical patients. Arch Surg1997;132:997–1004

7. Knighton DR, Halliday B, Hunt TK. Oxygen as an antibiotic:the effect of inspired oxygen on infection. Arch Surg1984;119:199–204

8. Allen DB, Maguire JJ, Mahdavian M, Wicke C, Marcocci L,Scheuenstuhl H, Chang M, Le AX, Hopf HW, Hunt TK.Wound hypoxia and acidosis limit neutrophil bacterial killingmechanisms. Arch Surg 1997;132:991–6

9. Babior BM. Oxygen-dependent microbial killing by phagocytes(first of two parts). N Engl J Med 1978;298:659–68

10. Leaper D, Burman-Roy S, Palanca A, Cullen K, Worster D,Gautam-Aitken E, Whittle M. Prevention and treatment ofsurgical site infection: summary of NICE guidance. BMJ2008;337:a1924


11. Lee JT. A new surgical site infection (SSI) prevention guideline.Surg Infect (Larchmt) 2000;1:127–31

12. Gottrup F, Firmin R, Chang N, Goodson WH III, Hunt TK.Continuous direct tissue oxygen tension measurement by anew method using an implantable silastic tonometer andoxygen polarography. Am J Surg 1983;146:399–403

13. Hopf HW, Hunt TK. Comparison of Clark electrode andoptode for measurement of tissue oxygen tension. Adv ExpMed Biol 1994;345:841–7

14. Soller BR, Idwasi PO, Balaguer J, Levin S, Simsir SA, VanderSalm TJ, Collette H, Heard SO. Noninvasive, near infraredspectroscopic-measured muscle pH and PO2 indicate tissueperfusion for cardiac surgical patients undergoing cardiopul-monary bypass. Crit Care Med 2003;31:2324–31

15. Soller BR, Ryan KL, Rickards CA, Cooke WH, Yang Y, SoyemiOO, Crookes BA, Heard SO, Convertino VA. Oxygen satura-tion determined from deep muscle, not thenar tissue, is anearly indicator of central hypovolemia in humans. Crit CareMed 2008;36:176–82

16. Ives CL, Harrison DK, Stansby GS. Tissue oxygen saturation,measured by near-infrared spectroscopy, and its relationshipto surgical-site infections. Br J Surg 2007;94:87–91

17. Ives CL, Harrison DK, Stansby G. Prediction of surgical siteinfections using spectrophotometry: preliminary results. AdvExp Med Biol 2006;578:149–54

18. Govinda R, Kasuya Y, Mahboobi R, Devarajan J, Akca O.Oxygen saturation at thenar eminence predicts surgical woundinfections. American Society of Anesthesiologists AnnualMeeting, Orlando, FL, 2008

19. Haley RW, Culver DH, Morgan WM, White JW, Emori TG,Hooton TM. Identifying patients at high risk of surgical woundinfection: a simple multivariate index of patient susceptibilityand wound contamination. Am J Epidemiol 1985;121:206–15

20. Culver DH, Horan TC, Gaynes RP, Martone WJ, Jarvis WR,Emori TG, Banerjee SN, Edwards JR, Tolson JS, Henderson TS,Hughes JM; National Nosocomial Infections Surveillance Sys-tem. Surgical wound infection rates by wound class, operativeprocedure, and patient risk index. Am J Med 1991;91:152S–7S

21. Myers DE, Anderson LD, Seifert RP, Ortner JP, Cooper CE,Beilman GJ, Mowlem JD. Noninvasive method for measuringlocal hemoglobin oxygen saturation in tissue using wide gapsecond derivative near-infrared spectroscopy. J Biomed Opt2005;10:034017

22. Myers DE, Cooper CE, Beilman GJ, Mowlem JD, Anderson LD,Seifert RP, Ortner JP. A wide gap second derivative NIRspectroscopic method for measuring tissue hemoglobin oxy-gen saturation. Adv Exp Med Biol 2006;578:217–22

23. Creteur J. Muscle StO2 in critically ill patients. Curr Opin CritCare 2008;14:361–6

24. Cohn SM, Nathens AB, Moore FA, Rhee P, Puyana JC, MooreEE, Beilman GJ. Tissue oxygen saturation predicts the devel-opment of organ dysfunction during traumatic shock resusci-tation. J Trauma 2007;62:44–54

25. Horan TC, Gaynes RP, Martone WJ, Jarvis WR, Emori TG.CDC definitions of nosocomial surgical site infections, 1992: amodification of CDC definitions of surgical wound infections.Infect Control Hosp Epidemiol 1992;13:606–8

26. Hunt TK, Dunphy JE. Effects of increasing oxygen supply tohealing wounds. Br J Surg 1969;56:705

27. Hunt TK, Linsey M, Sonne M, Jawetz E. Oxygen tension andwound infection. Surg Forum 1972;23:47–9

28. Akca O, Sessler DI, Delong D, Keijner R, Ganzel B, Doufas AG.Tissue oxygenation response to mild hypercapnia during car-diopulmonary bypass with constant pump output. Br J An-aesth 2006;96:708–14

29. Akca O, Doufas A, Morioka N, Iscoe S, Fisher J, Sessler D.Hypercapnia improves tissue oxygenation. Anesthesiology2002;97:801–6

30. Morris S, Tang M, Geddes C. Vascular anatomical basis ofperforator skin flaps. Cir Plast Iberlatinamer 2006;32:1–5

31. Offman SL, Geddes CR, Tang M, Morris SF. The vascular basisof perforator flaps based on the source arteries of the laterallumbar region. Plast Reconstr Surg 2005;115:1651–9

32. Hopf HW, Holm J. Hyperoxia and infection. Best Pract ResClin Anaesthesiol 2008;22:553–69

33. Bitterman H. Bench-to-bedside review: oxygen as a drug. CritCare 2009;13:1–8

34. Ives CL, Harrison DK, Stansby GS. Prediction of surgical siteinfections after major surgery using visible and near-infraredspectroscopy. Adv Exp Med Biol 2007;599:37–44

35. Hopf HW, Viele M, Watson JJ, Feiner J, Weiskopf R, Hunt TK,Noorani M, Yeap H, Ho R, Toy P. Subcutaneous perfusion andoxygen during acute severe isovolemic hemodilution inhealthy volunteers. Arch Surg 2000;135:1443–9

36. Jensen JA, Jonsson K, Goodson WH III, Hunt TK, Roizen MF.Epinephrine lowers subcutaneous wound oxygen tension.Curr Surg 1985;42:472–4

37. Akca O, Melischek M, Scheck T, Hellwagner K, Arkilic C, KurzA, Kapral S, Heinz T, Lackner FX, Sessler DI. Postoperativepain and subcutaneous oxygen tension. Lancet 1999;354:41–2

38. Buggy DJ, Kerin MJ. Paravertebral analgesia with levobupiva-caine increases postoperative flap tissue oxygen tension afterimmediate latissimus dorsi breast reconstruction comparedwith intravenous opioid analgesia. Anesthesiology 2004;100:375–80

39. Rabkin JM, Hunt TK. Local heat increases blood flow andoxygen tension in wounds. Arch Surg 1987;122:221–5

40. Sheffield CW, Sessler DI, Hopf HW, Schroeder M, Moayeri A,Hunt TK, West JM. Centrally and locally mediated thermo-regulatory responses alter subcutaneous oxygen tension.Wound Repair Regen 1996;4:339–45

41. Kurz A, Sessler DI, Lenhardt R. Perioperative normothermia toreduce the incidence of surgical-wound infection and shortenhospitalization. Study of Wound Infection and TemperatureGroup. N Engl J Med 1996;334:1209–15

42. Melling AC, Ali B, Scott EM, Leaper DJ. Effects of preoperativewarming on the incidence of wound infection after cleansurgery: a randomised controlled trial. Lancet 2001;358:876–80

43. Arkilic CF, Taguchi A, Sharma N, Ratnaraj J, Sessler DI, ReadTE, Fleshman JW, Kurz A. Supplemental perioperative fluidadministration increases tissue oxygen pressure. Surgery2003;133:49 –55

44. Jonsson K, Jensen JA, Goodson WH III, West JM, Hunt TK.Assessment of perfusion in postoperative patients using tissueoxygen measurements. Br J Surg 1987;74:263–7

45. Kabon B, Akca O, Taguchi A, Nagele A, Jebadurai R, ArkilicCF, Sharma N, Ahluwalia A, Galandiuk S, Fleshman J, SesslerDI, Kurz A. Supplemental intravenous crystalloid administra-tion does not reduce the risk of surgical wound infection.Anesth Analg 2005;101:1546–53

46. Liu B, Finfer S. Intravenous fluids in adults undergoingsurgery. BMJ 2009;338:b2418

47. Edmark L, Kostova-Aherdan K, Enlund M, Hedenstierna G.Optimal oxygen concentration during induction of generalanesthesia. Anesthesiology 2003;98:28–33

48. Thorn CE, Matcher SJ, Meglinski IV, Shore AC. Is mean bloodsaturation a useful marker of tissue oxygenation? Am J PhysiolHeart Circ Physiol 2009;296:H1289–95



American Society of Critical Care Anesthesiologists

Section Editor: Michael J. Murray

High-Fidelity Simulation Demonstrates the Influenceof Anesthesiologists’ Age and Years from Residencyon Emergency Cricothyroidotomy SkillsLyndon W. Siu, MBBS, FANZCA,* Sylvain Boet, MD,* Bruno C. R. Borges, MD,*Heinz R. Bruppacher, MD, FMH,* Vicki LeBlanc, PhD,† Viren N. Naik, MD, MEd, FRCPC,*Nicole Riem, MD,* Deven B. Chandra, MD, MEd, FRCPC,* and Hwan S. Joo, MD, FRCPC*

BACKGROUND: Age-related deterioration in both cognitive function and the capacity to control finemotor movements has been demonstrated in numerous studies. However, this decline has not beendescribed with respect to complex clinical anesthesia skills. Cricothyroidotomy is an example of acomplex, lifesaving procedure that requires competency in the domains of both cognitive processingand fine motor control. Proficiency in this skill is vital to minimize time to reestablish oxygenationduring a “cannot intubate, cannot ventilate” scenario. In this prospective, controlled, single-blinded study,we tested the hypothesis that age affects the learning and performance of emergency percutaneouscricothyroidotomy in a high-fidelity simulated cannot intubate/cannot ventilate scenario.METHODS: Thirty-six staff anesthesiologists (19 aged younger than 45 years and 17 older than45 years) managed a high-fidelity cannot intubate/cannot ventilate scenario in a high-fidelitysimulator before and after a 1-hour standardized training session. The group division cutoff ageof 45 years was based on the median age of our sample subject population before enrollment.The scenarios required the insertion of an emergency percutaneous cricothyroidotomy. Wecompared cricothyroidotomy skills in the older group with those in the younger group usingprocedural time, 5-point task-specific checklist score, and global rating scale score. Correlationbased on age, years from residency, weekly clinical hours worked, previous continuing medicaleducation in airway management, and previous simulation experience was also performed.RESULTS: In both prestandardization and poststandardization, age and years from residencycorrelated with procedural time, checklist scores, and global rating scores. Baseline, prestandard-ization variables were all better for the younger group, with a mean age of 37 years, compared withthe older group, with a mean age of 58 years. Procedural time was 100 (72–128) seconds versus152 (120–261) seconds. Checklist scores were 7.0 (6.1–8.0) versus 6.0 (4.8–8.0). Global ratingscale scores were 22.0 (17.8–29.8) versus 17.5 (10.4–20.6). After the 1-hour standardized trainingsession, the younger group continued to perform better than the older group with procedural time of75 (66–91) seconds versus 87 (78–123) seconds, checklist scores of 10.0 (9.1–10.0) versus 9.0(8.0–10.0), and global rating scale scores of 35.0 (32.1–35.0) versus 32.0 (29.0–33.8). Regressionanalysis was performed on the poststandardization data. Both age and years from residency indepen-dently affected procedural time, checklist scores, and global rating scale scores (all P � 0.05).CONCLUSIONS: Baseline proficiency with simulated emergency cricothyroidotomy is associatedwith age and years from residency. Despite standardized training, operator age and years fromresidency were associated with decreased proficiency. Further research should explore thepotential of using age and years from residency as factors for implementing periodic continuingmedical education. (Anesth Analg 2010;111:955–60)

Age-related deterioration in cognitive functioning andfine motor skills has been demonstrated in numerousstudies and reviews.1,2 The theoretical clinical signifi-

cance of aging has been extensively addressed in the litera-ture.3–5 However, the specific impact of age on a particularanesthetic procedure has never been objectively assessed.

Cricothyroidotomy is a complex lifesaving procedurethat requires competency in the domains of both cognitiveprocessing and fine motor control. Proficiency in this skill isvital because minimizing time to achieve oxygenation isessential in a cannot intubate/cannot ventilate situation.Age-related impact on this skill, and potentially otherimportant skills, should be recognized to facilitate thedevelopment of appropriate educational strategies.

Using a low-fidelity static model, Wong et al.6 observedthat anesthesiologists aged younger than 45 years per-formed cricothyroidotomies faster than those older than45 years. In another study, John et al.7 demonstrated thatunder stressful conditions in simulated cannot intubate/cannot ventilate scenarios, procedural times were longercompared with times achieved on static mannequinmodels. However, neither of the 2 studies controlled for

From the *Department of Anesthesia, St. Michael’s Hospital, University ofToronto; and †Wilson Centre, University Health Network, Department ofMedicine, University of Toronto, Toronto, Ontario, Canada.


Funded by departmental funds.


Address correspondence and reprint requests to Hwan S. Joo, MD, FRCPC,St. Michael’s Hospital, 30 Bond St., Toronto, ON, M5B1W8, Canada. Addresse-mail to [email protected].

Copyright © 2010 International Anesthesia Research SocietyDOI: 10.1213/ANE.0b013e3181ee7f4f


confounders such as previous procedural and/or simula-tion experience.

The primary purpose of this prospective, controlled,single-blinded study was to investigate whether age affectslearning and performance of simulated cannot intubate/cannotventilate emergency percutaneous cricothyroidotomies.

We hypothesize that age may affect the ability to learnand perform emergency cricothyroidotomy in a high-fidelity simulated setting.

METHODSAfter institutional ethics approval, 36 attending anesthesi-ologists in a tertiary care teaching hospital (19 agedyounger than 45 years and 17 older than 45 years) manageda high-fidelity cannot intubate/cannot ventilate scenarioimmediately before and after a standardization process thatincluded cricothyroidotomy training. The age 45 years waschosen to divide the anesthesiologists into equal groups,based on the median age of 39 possible participants beforestudy enrollment. To clarify for this article, we have namedthe “older than 45 years” group the “older” group and the“younger than 45 years” group the “younger” group. All 39attending anesthesiologists in the Department of Anesthe-sia of our institution were approached for recruitment; 3declined to participate. Informed consent and confidential-ity agreements were obtained from all participants.

Standardization ProcessAll participants received a 1-hour introduction of thesimulation center consisting of an orientation session to thehuman patient simulator and participation in an introduc-tory high-fidelity airway management scenario, which wasnot part of the study.

The prestandardization scenario was a cannot intubate/cannot ventilate scenario. In this videotaped session, anactor playing a second-year resident calls for help after 2unsuccessful intubation attempts and difficulty with face-mask ventilation. The simulated patient’s oxygen satura-tion was 89% when the subject arrived and decreased by10% every minute. All alternative methods of intubationwere intended to be unsuccessful. This was accomplishedby changing the mannequin’s airway anatomy. The cervicalspine was immobilized, the tongue was made macroglos-sic, and the vocal cords were adducted. When requested bythe participant, an ear-nose-throat surgeon and secondanesthesiologist were called but would not be available.The scenario was designed to necessitate an emergencypercutaneous cricothyroidotomy using a 4.0-mm Melkeremergency cricothyroidotomy catheter set (C-TCCS-400;Cook Inc., Bloomington, IN). The scenario only ended withsuccessful cricothyroidotomy, defined by positive capnog-raphy on the monitor.

Immediately after the first cannot intubate/cannot ven-tilate scenario, a teaching session including practical in-structions on percutaneous cricothyroidotomy insertionand video-assisted debriefing was provided. The sameindividual conducted all debriefing sessions (LWS).

Immediately after the standardization session, all par-ticipants managed the poststandardization scenario, anidentical cannot intubate/cannot ventilate scenario to thatpresented during the prestandardization session. Subjects

did not have prior knowledge of the content of anyscenario.

All scenarios were completed in a simulated operatingroom environment containing a high-fidelity mannequin(Sim Man; Laerdal, Kent, UK) equipped with an anatomi-cally accurate larynx, properly designed for performance ofcricothyroidotomies, and standard monitors (electrocardio-gram, noninvasive arterial blood pressure, oxygen satura-tion as measured by pulse oximetry, and end-tidal CO2).Appropriate equipment and airway devices in the roomincluded multiple-sized laryngoscopes, endotracheal tubes,laryngeal mask airways, gum elastic bougie, and anesthesiadrug cart. Airway adjuncts kept outside the room includeda fiberoptic bronchoscope, a videolaryngoscope (Glide-Scope®; Verathon, Inc., Bothell, WA), intubating laryngealmask airways, and a cricothyroidotomy kit. Two simula-tion assistants played the scripted roles of a nurse and ajunior resident.

Sample Size CalculationWe hypothesized that there would be a difference in theprocedural time in favor of the younger group. One stan-dard deviation for procedural times between the 2 groupswas defined to be a clinically important difference. Wecalculated that 17 participants per group would be requiredto achieve a difference of 1 SD between groups in thepoststandardization procedural times based on a 2-tailed �of 0.05 and a power of 0.8.

Data CollectionDemographic data including age, the number of years aftergraduation from anesthesia residency, the number of hoursof clinical practice per week, previous simulation and/orairway simulation experience, and previous cricothyroid-otomy experience on both patients and mannequins werecollected. All cricothyroidotomy performances (2 per sub-ject) were video-recorded and later evaluated by 2 blindedevaluators. The evaluators assigned were blinded to thestudy outcome, each other’s scores, and whether thevideo was from the pre- or posttest cannot intubate/cannot ventilate session. Blinding to age was attemptedbut may not have been possible as the approximate agemay have been guessed because participants may havebeen recognized.

Outcome MeasurementThe primary outcome was the comparison of cricothyroid-otomy performances between the younger and the oldergroups. Performance was assessed with 3 variables: proce-dural time and 2 previously validated tools for proceduralskill evaluation (a 3-point task-specific checklist [Appendix1] and global rating scale [GRS] [Appendix 2]).8 Proceduraltime was measured during video review and was definedas the time between the first instances when the subjectgrasps any equipment from the cricothyroidotomy kit tothe time of successful cricothyroidotomy. The cricothyroid-otomy checklist was based on observation of common butimportant mistakes made by novice operators based on thestudy by Friedman et al.8 A score of 0, 1, or 2 was givenwhen a stage was not performed, poorly performed, orperformed well, respectively. The GRS uses more general

Age and Percutaneous Cricothyroidotomy Skills


descriptors and focuses on the overall performance of thesubject, not the specifics of the manual task.

Statistical AnalysisAnalysis was performed using SPSS 11.0 software (SPSS,Inc., Chicago, IL). To determine the reliability of assess-ments provided by the 2 evaluators, intraclass correlationcoefficients (ICCs) were calculated for checklist and GRSscores.9 We compared cricothyroidotomy performancesbetween younger and older groups both before and afterstandardization. Procedural times, checklist scores, andGRS scores were compared using the Mann-Whitney test.Correlations between demographic variables and proce-dural time, checklist scores, and GRS scores were per-formed using Spearman’s correlation. All P values were 2sided except for correlation, which was 1 sided. Signifi-cance for correlation was set at a value of P � 0.05.Significance for comparison between younger and older,and between pre- and poststandardization variables wasset at a value of P � 0.025 to account for Bonferronicorrection.

We used multiple regression analysis to account fordifferences between the 2 groups regarding variables thatcould influence our primary outcomes. Initially, we enteredour primary outcomes measurements (procedural times,GRS score, and checklist score) as dependent variables andthe other factors such as age, the number of hours of clinicalpractice per week, previous simulation and/or airwaysimulation experience, and previous cricothyroidotomyexperience on both patients and mannequins as predictorvariables. Any of these factors that correlated with the

dependent variables at P � 0.1 would be taken inconsideration in the final analysis of age as a predictorvariable.

RESULTSThirty-six attending anesthesiologists were recruited over 9months. General demographics of the younger (aged 37 �3 years) and older (aged 58 � 8 years) groups are shown inTable 1.

Interrater reliability was strong for both checklist andGRS scores (checklist: ICC � 0.911; GRS: ICC � 0.837) (bothP � 0.05).

The performance of both age groups significantly im-proved after teaching for all 3 variables (Table 2). Prestan-dardization procedural time was longer and checklist andGRS scores were lower in the older group (Table 2).Poststandardization procedural time was longer for theolder group compared with the younger group. Bothchecklist and GRS scores were lower in the older groupcompared with the younger group (Table 2).

Age and years from residency correlated with proce-dural time, checklist scores, and GRS scores, both beforeand after standardization (Table 3). Previous simulationexperience also correlated with GRS scores, and withprocedural time but only before standardization. Weeklyclinical hours correlated with GRS scores, but only afterstandardization (Table 3).

Multiple regression analysis was performed on post-standardization data. Both age and years from residencyindependently affected procedural time, checklist scores,and GRS scores (all P � 0.05).

Table 1. DemographicsCharacteristics Age <45 y (n � 19) Age >45 y (n � 17) P value

Age 36.7 (�3.4) 58.0 (�8.3) �0.001Gender (male/female) 13 (68.4%)/6 (32.6%) 15 (88.2%)/2 (11.8%) NSYears from residency graduation 4 (0.5–7.0) 28 (14.8–33.3) �0.001Hours of clinical work per week 43.4 (�3.8) 38.5 (�8.6) 0.031Previous simulation experience 16 (84.2%) 5 (29.4%) 0.002Previous simulation experience in airway

management4 (21.1%) 2 (11.8%) NS

Attended airway lecture or continued medicaleducation within 10 y

12 (63.1%) 11 (64.7%) NS

Previous percutaneous cricothyroidotomy experience 15 (78.9%) 7 (41.2%) 0.039Patient 3 (15.8%) 3 (17.7%) NSMannequin or pig 12 (63.2%) 4 (23.5%) 0.023

NS � not significant.Values are mean (�standard deviation), median (interquartile range), or number (percentage).

Table 2. Cricothyroidotomy Performance Before and After StandardizationPrestandardization Poststandardization

Age <45 y Age >45 y P value Age <45 y Age >45 y P valueProcedural time (s) 100 (72–128) 152 (120–261) 0.003 75 (66–91)* 87 (78–123)* 0.018Checklist scorea 7.0 (6.1–8.0) 6.0 (4.8–8.0) 0.024 10.0 (9.1–10.0)* 9.0 (8.0–10.0)* 0.005Global rating scale scoreb 22.0 (17.8–29.8) 17.5 (10.4–20.6) 0.004 35.0 (32.1–35.0)* 32.0 (29.0–33.8)* 0.012

Data are median (interquartile range).a Checklist score ranges from 0 to 10.b Global rating scale ranges from 0 to 35.* Significant difference with corresponding age group prestandardization data (P � 0.01).


DISCUSSIONBefore standardization, the younger group significantlyoutperformed the older group as assessed by all 3 variables.This difference in baseline (prestandardization) perfor-mances between the 2 groups may reflect differences inprevious simulation exposure and/or cricothyroidotomyexperience on mannequins (Table 1). Of note, however,those who had previously performed cricothyroidotomieson mannequins had only done so once or twice and none inthe 6 months before the study. In addition, a previous studysuggests that cricothyroidotomy performance in low-fidelity models declines to near baseline after 3 months.10

Both groups improved significantly in time and checklistand GRS scores after standardization. This trend suggeststhat training is effective in improving both the cognitiveand psychomotor aspects of emergency cricothyroidotomyskills.

To investigate whether the difference in performancewas attributable to age, we focused our comparative anal-ysis on the poststandardization data. By priming partici-pants with a 1-hour standardization session consisting of 2airway management simulation scenarios, debriefing, anda practical instructional session on cricothyroidotomy im-mediately before the study scenario, we aimed to minimizebias caused by the variable simulation and cricothyroid-otomy experience of participants. The results of this studydemonstrated that after standardized exposure to high-fidelity simulation and after standardized teaching, an-esthesiologists who are older and further away fromresidency training require more time to perform emer-gency percutaneous cricothyroidotomies and have lowerchecklist and GRS scores.

We failed to find studies in the literature that addressedthe impact of age on the performance of a specific anes-thetic procedure, except for the study by Wong et al.,6

which also used cricothyroidotomies but that study wasnot primarily powered for this specific objective. However,the literature is rich with articles that address patient safetyconcerns and aging physicians, including surgeons andanesthesiologists.3,4 Psychomotor and perceptual processesthat are required in aviation that deteriorate with age

include the ability to perform complex tasks rapidly, toprocess incoming information and make complex deci-sions, and to perform effectively in a stressful environ-ment.11 All of them are required for the practice ofanesthesiology as well.

Proficient performance of emergency cricothyroidotomyrequires both the cognitive ability to recall essential stepsfor the procedure and psychomotor skills to executeplanned actions efficiently. Both aspects may be adverselyaffected by increasing age as a result of the general slowingof central cognitive processes.12–14 One probable neuro-physiological mechanism is the loss of neural connectivityor decreased levels of neurotransmitters in the agingbrain.13,14 The difference in poststandardization proceduraltimes, however, supports an age-related decline in psy-chomotor skills required for this procedure. Another pos-sibility is that years from residency or training is just asimportant. Time from formal residency or training mayaffect technical skills, even after standardized training.Nonetheless, the clinical significance of the difference inprocedural times, measured within a simulation setting, isunknown and may be difficult to determine.

It should be clear that our study was not focused onfinding a cutoff age beyond which more training on crico-thyrotomies is necessary, rather, simply on showing thatage may well be a factor that affects learning and perfor-mance of emergency cricothyroidotomies. The cutoff age of45 years was used mainly as a median-age dividing point.

Further research should explore the potential of usingage and years from residency as factors for implementingperiodic continuing medical education. As shown in ourstudy, the older anesthesiologists benefited more fromstandardized cricothyroidotomy training than the youngeranesthesiologists. However, this may also be attributable toa ceiling effect because the younger anesthesiologists mayhave been closer to their maximal potential in their pres-tandardization scenario.

This study has several limitations. Even with our stan-dardized training, we might not have equalized previousknowledge base in performing cricothyroidotomies. Previ-ous knowledge and recent airway training in many

Table 3. Predictor Variables: Correlation and P ValuesAirway continuedmedical education

Previous simulationexperience

Weekly clinicalhours

Years fromresidency Age

Statistical analysis ofprestandardization data

Procedural time �0.257 �0.437 �0.103 0.370 0.362P � 0.065 P � 0.004* P � 0.275 P � 0.013* P � 0.015*

Checklist scores �0.031 0.186 0.130 �0.411 �0.503P � 0.43 P � 0.139 P � 0.225 P � 0.006* P � 0.001*

Global rating scale scores 0.257 0.437 0.007 �0.454 �0.497P � 0.065 P � 0.004* P � 0.483 P � 0.003* P � 0.001*

Statistical analysis ofpoststandardization data

Procedural time �0.007 �0.041 0.070 0.341 0.310P � 0.484 P � 0.407 P � 0.344 P � 0.021* P � 0.033*

Checklist scores 0.096 0.177 0.187 �0.333 �0.333P � 0.289 P � 0.151 P � 0.138 P � 0.024* P � 0.024*

Global rating scale scores 0.076 0.302 0.344 �0.564 �0.521P � 0.330 P � 0.037* P � 0.02* P � 0.001* P � 0.001*

Values represent Spearman correlation coefficient and P values.* Statistically significant (P � 0.05).



younger participants may have contributed to better timesand scores. All participants were attending anesthesiolo-gists recruited from a single tertiary academic center. Amulticenter study would render the results more general-izable. Because our study was conducted using 4.0-mmMelker cricothyroidotomy kits (C-TCCS-400; Cook Inc.),the results might not reflect performance using othercommercial kits, especially those that do not useSeldinger’s technique. In addition, as for all simulationresearch, there are conflicting views on whether results canbe extrapolated to real-life clinical settings.

In conclusion, there may well be an age/number ofyears from residency–related decline in proficiency foremergency cricothyroidotomy skills in a high-fidelity simu-lated setting. Because emergency cricothyroidotomies are

performed rarely, if at all during routine clinical practice,lack of experience further contributes to failure in reallife.15–17 Further research should explore the potential ofusing age and years from residency as factors for imple-menting periodic specific continuing medical education.

AUTHOR CONTRIBUTIONSLWS helped with study design, conduct of study, and manuscriptpreparation; SB helped with conduct of study, data collection andanalysis, and manuscript preparation; BCRB helped with conductof study and manuscript preparation; HRB helped with datacollection; VL helped with data analysis; VNN helped with studydesign; NR and DBC helped with manuscript preparation; andHSJ helped with study design, conduct of study, and manuscriptpreparation, and is responsible for archival.

Appendix 1. Task-Specific Checklist for CricothyroidotomyScore

Task 0 1 2Aspiration to identify trachea Does not aspirate Performed inadequately Aspirates with air-filled

or fluid-filled syringeVentilation during

cricothyroidotomyDoes not ventilate during the

cricothyroidotomyVentilates for part of the

duration of thecricothyroidotomy

Ventilates for the entireduration of thecricothyroidotomy

Correct caudal angling duringguidewire insertion (notduring needle insertion)

Cephalad angle 90° to trachea 45° caudad

Adequate skin and membraneincision

Does not use the scalpel forincision

Performed inadequately Cuts skin andcricothyroidmembrane with thescalpel

Correct use of dilator andcricothyroidotomy

Attempts to insert cricothyroidotomycannula without dilator in place

Dilates separately andthen railroad entireassembly

Railroad entireassembly (dilator andcricothyroidotomycannula)

Appendix 2. Global Rating Scale for CricothyroidotomyScore

1 2 3 4 5Preparation for procedure Did not organize equipment well.

Had to stop procedurefrequently to prepareequipment

Equipment generally organized.Occasionally had to stop andprepare items

All equipment neatly organized,prepared, and ready for use

Respect for tissue Frequently used unnecessaryforce on tissue or causeddamage

Careful handling of tissue butoccasionally causedinadvertent damage

Consistently handled tissuesappropriately with minimaldamage

Time and motion Many unnecessary moves Efficient time/motion but someunnecessary moves

Clear economy of movementand maximum efficiency

Instrument handling Repeatedly made tentative orawkward moves withinstruments

Competent use of instrumentsbut occasionally appearedstiff or awkward

Fluid moves with instrumentsand no awkwardness

Flow of procedure Frequently stopped procedureand seemed unsure of nextmove

Demonstrated some forwardplanning with reasonableprogression of procedure

Obviously planned course ofprocedure with effortlessflow from 1 move to the next

Knowledge of procedure Deficient knowledge Knew all important steps ofprocedure

Demonstrated familiarity withall aspects of procedure

Overall performance Very poor Competent Clearly superior


REFERENCES1. Krampe RT. Aging, expertise and fine motor movement. Neu-

rosci Biobehav Rev 2002;26:769–762. Birren J, Schaie K. Handbook of the Psychology of Aging. 5th

ed. San Diego: Academic Press, 2001:313–483. Katz JD. Issues of concern for the aging anesthesiologist. Anesth

Analg 2001;92:1487–924. Greenfield LJ, Proctor MC. When should a surgeon retire? Adv

Surg 1999;32:385–935. Boom-Saad Z, Langenecker SA, Bieliauskas LA, Graver CJ,

O’Neill JR, Caveney AF, Greenfield LJ, Minter RM. Surgeonsoutperform normative controls on neuropsychologic tests, butage-related decay of skills persists. Am J Surg 2008;195:205–9

6. Wong DT, Prabhu AJ, Coloma M, Imasogie N, Chung FF. Whatis the minimum training required for successful cricothyroid-otomy. Anesthesiology 2003;98:349–53

7. John B, Suri I, Hillermann C, Mendonca C. Comparison ofcricothyroidotomy on manikin vs. simulator: a randomisedcross-over study. Anaesthesia 2007;62:1029

8. Friedman Z, You-Ten KE, Bould MD, Naik V. Teaching lifesav-ing procedures: the impact of model fidelity on acquisition andtransfer of cricothyrotomy skills to performance on cadavers.Anesth Analg 2008;107:1663–9

9. Landis JR, Koch GG. An application of hierarchical kappa-typestatistics in the assessment of majority agreement among mul-tiple observers. Biometrics 1977;33:363–74

10. Amrhein PC, Stelmach GE, Goggin NL. Age differences in themaintenance and restructuring of movement preparation. Psy-chol Aging 1991;6:451–66

11. Eyraud MY, Borowsky MS. Age and pilot performance. AviatSpace Environ Med 1985;56:553–8

12. Birren J, Schaie K. Handbook of the Psychology of Aging. 5thed. San Diego: Academic Press, 2001:288–312

13. Salthouse TA. The processing-speed theory of adult age differ-ences in cognition. Psychol Rev 1996;103:403–28

14. Birren J, Schaie K. Handbook of the Psychology of Aging. 5thed. San Diego: Academic Press, 2001:135–60

15. Schaumann N, Lorenz V, Schellongowski P, Staudinger T,Locker GJ, Burgmann H, Pikula B, Hofbauer R, Schuster E,Frass M. Evaluation of Seldinger technique emergency crico-thyroidotomy versus standard surgical cricothyroidotomy in200 cadavers. Anesthesiology 2005;102:7–11

16. Ravlo O, Bach V, Lybecker H, Moller JT, Werner M, NielsenHK. A comparison between two emergency cricothyroidotomyinstruments. Acta Anaesthesiol Scand 1987;31:317–9

17. Chang RS, Hamilton RJ, Carter WA. Declining rate of crico-thyrotomy in trauma patients with an emergency medicineresidency: implications for skills training. Acad Emerg Med1998;5:247–51



Adaptive Support Ventilation with ProtocolizedDe-Escalation and Escalation Does Not AccelerateTracheal Extubation of Patients After Nonfast-TrackCardiothoracic SurgeryDave A. Dongelmans, MD, MSc,* Denise P. Veelo, MD, PhD,*†‡ Jan M. Binnekade, PhD,*Bas A.J.M. de Mol, MD, PhD,§ Anna Kudoga, MS,* Frederique Paulus, MSc,*and Marcus J. Schultz, MD, PhD*‡�

BACKGROUND: It is uncertain whether adaptive support ventilation (ASV) accelerates weaningof nonfast-track cardiothoracic surgery patients. A lower operator set %-minute ventilation withASV may allow for an earlier definite switch from controlled to assisted ventilation, potentiallyhastening tracheal extubation. We hypothesized that ASV using protocolized de-escalation andescalation of operator set %-minute ventilation (ASV-DE) reduces time until tracheal extubationcompared with ASV using a fixed operator set %-minute ventilation (standard ASV) in uncompli-cated patients after nonfast-track coronary artery bypass graft.METHODS: We performed a randomized controlled trial comparing ASV-DE with standard ASV.With ASV-DE, as soon as body temperature was �35.0°C with pH �7.25, operator set %-minuteventilation was decreased stepwise to a minimum of 70%.RESULTS: Sixty-three patients were randomized to ASV-DE, and 63 patients to standard ASV.The duration of mechanical ventilation was not different between groups (10.8 [6.5–16.1] vs10.7 [6.6–13.9] hours, ASV-DE versus standard ASV; P � 0.32). Time until the first assistedbreathing period was shorter (3.1 [2.0–6.7] vs 3.9 [2.1–7.5] hours) and the number of assistedventilation episodes was higher (78 [34–176] vs 57 [32–116] episodes), but differences did notreach statistical significance. The duration of assisted ventilation episodes that ended withtracheal extubation was different between groups (2.5 [0.9–4.6] vs 1.4 [0.3–3.5] hours, ASV-DEversus standard ASV; P � 0.05).CONCLUSION: Compared with standard ASV, weaning of patients after nonfast-track coronaryartery bypass graft using ASV with protocolized de-escalation and escalation does not shortentime to tracheal extubation. (Anesth Analg 2010;111:961–7)

Adaptive support ventilation (ASV) is an advancedclosed-loop mode of mechanical ventilation (MV)that maintains an operator preset minute ventila-

tion. ASV adjusts respiratory rates and pressure levelsaccording to measured lung mechanics at each breath.1 Inaddition, ASV automatically switches between controlledand assisted ventilation according to the patient’s status.2

Previous randomized controlled trials have tested the effi-cacy of ASV in patients after cardiothoracic surgery.3–5 In astudy of fast-track cardiothoracic surgery patients, ASVcompared with synchronized intermittent mandatory ven-tilation or traditional pressure support ventilation short-ened the time to tracheal extubation.5 This was confirmed

in another study of fast-track cardiothoracic surgery pa-tients in which ASV was compared with pressure-regulatedvolume controlled ventilation.4 However, in a recent studyof nonfast-track cardiothoracic surgery patients, ASV com-pared with traditional pressure support ventilation did notshorten the time to tracheal extubation.3

A lower operator set %-minute ventilation with ASVmay allow for earlier and more frequent switches fromcontrolled to assisted ventilation. Indeed, patients whoselungs are ventilated with lower minute volumes could beforced to breathe spontaneously sooner because arterialPco2 thresholds for breathing are reached faster.6 Second,in the above-mentioned study of nonfast-track cardiotho-racic surgery patients,3 patients were able to trigger theventilator early in the weaning process, at least suggestingthat a lower operator set %-minute could push patientstoward longer periods of assisted ventilation and therebyearlier tracheal extubation.

In a randomized controlled trial of patients afterplanned and uncomplicated nonfast-track coronary arterybypass graft (CABG), we compared ASV using protocol-ized de-escalation and escalation of operator set %-minuteventilation (ASV-DE) with ASV using a fixed operator set%-minute ventilation (standard ASV). We hypothesizedthat ASV-DE reduces time to tracheal extubation comparedwith standard ASV.

From the Departments of *Intensive Care Medicine, †Anesthesiology, and§Cardiothoracic Surgery; and ‡Laboratory of Experimental IntensiveCare and Anesthesiology (L.E.I.C.A.), Academic Medical Center, Univer-sity of Amsterdam; and �HERMES Critical Care Group, Amsterdam, TheNetherlands.


Supported by the Department of Intensive Care Medicine, Academic Medi-cal Center.

Presented, in part, at the ATS conference, San Diego, CA, May 18, 2009.


Address correspondence and reprint requests to Dave A. Dongelmans, MD,MSc, Department of Intensive Care Medicine, G3-212, Academic MedicalCenter, Meibergdreef 9, 1105 AZ Amsterdam, The Netherlands. Addresse-mail to [email protected].

Copyright © 2010 International Anesthesia Research SocietyDOI: 10.1213/ANE.0b013e3181efb316


METHODSPatients and SettingConsecutive patients after elective and uncomplicatedCABG admitted to the 28-bed intensive care unit (ICU) ofthe Academic Medical Center, Amsterdam, The Nether-lands, were eligible for inclusion. The study protocol wasapproved by the local IRB, and preoperative written andsigned informed consent was obtained from eligible pa-tients programmed for surgery.

Study DesignAccording to an open-label randomized controlled design,patients were assigned to receive MV with either ASV-DEor standard ASV.

Inclusion CriteriaPatients were included and randomized using sealed num-bered envelopes on ICU arrival after surgery. We created ahomogeneous group of patients after elective and uncom-plicated CABG, i.e., without a history of chronic obstructivepulmonary disease or hemodynamic instability. Patientswith a history of chronic obstructive pulmonary disease ora history of pulmonary surgery, and patients with anintraaortic balloon pump or inotropics and/or vasopres-sors at a more than usual rate (in milligrams per hour:dopamine 20, norepinephrine 0.5, dobutamine 25, or epi-nephrine [any rate]) on ICU arrival were excluded.

Cardiothoracic Surgery/Anesthesia ProceduresAll patients in both groups were anesthetized according toour standard institutional protocol, starting with 1 or 2 mglorazepam as premedication, followed by etomidate, sufen-tanil, and rocuronium for induction of anesthesia andfacilitation of intubation. During the surgical procedure,sufentanil was used as analgesic, and sevoflurane pluspropofol were used to maintain anesthesia. Muscle relax-ants were not given during the surgical procedure. Mor-phine and midazolam could be administered at the end ofthe procedure.

Cardiopulmonary bypass was performed under moder-ate hypothermia (28°C–35°C), using a membrane oxygen-ator and nonpulsatile blood flow. At the end of anesthesia,all patients were transferred to the ICU with trachealintubation. Anesthesiologists and surgeons at the operatingroom were blinded for inclusion or randomization ofpatients.

ICU ManagementUnit policy comprised that a patient was cared for by adedicated ICU nurse, responsible for 1 or 2 patients.Attending ICU nurses were constantly at the bedside, andchanges in treatment according to the postoperative ICUprotocol, based on observations by ICU nurses, were ex-ecuted immediately.

The postoperative ICU protocol was similar for bothstudy groups and involved fluid resuscitation with normalsaline and starch solutions, blood transfusion to maintainhemoglobin concentration (�5.0 mmol/L), norepinephrinein continuous infusion to achieve mean arterial bloodpressure �70 mm Hg, and dobutamine and/or enoximone

to achieve a cardiac index �2.5 L/min/m2 or a mixedvenous oxygenation �60%.

Sedation and analgesics were given according to localprotocol for postoperative cardiothoracic surgery patients.Propofol was given for sedation via continuous infusion.Infusion of propofol was stopped instantly when coretemperature reached 35°C. The confusion assessmentmethod for the ICU was used to screen for delirium, and ifpresent, haloperidol was started. Acetaminophen (4 g/d)was started in all patients. The requirement for additionalanalgesia was assessed by attending ICU nurses. Morphinewas given in boluses of 1 to 2 mg IV until patients were freeof pain. The boluses were repeated as needed. Postopera-tive shivering, if present, was treated with meperidine (25mg IV). Muscle relaxants were not given in the ICU.

MV ProtocolsAll patients’ lungs were ventilated by a Hamilton Galileoventilator (software version GMP03.41f, GCP03.40a,GTP01.00; Hamilton Medical AG, Rhazuns, Switzerland).Passive humidification of the ventilatory circuit was ap-plied by means of an HME filter (Medisize Hygrovent S;Medisize, Hillegom, The Netherlands).

In both groups, the initial levels of fraction of inspiredoxygen (Fio2) (50%) positive end-expiratory pressure(PEEP) (5 cm H2O), peak airway pressure (35 cm H2O), and%-minute ventilation (a theoretical value based on idealbody weight, 100%) were set by the attending ICU physi-cian. Flow trigger sensitivity was set at 2 L/s; activepatients could trigger the ventilator (i.e., actual minuteventilation could exceed set %-minute ventilation). Anarterial blood gas analysis was performed 30 minutes afterconnection to the ICU ventilator, and 30 minutes after eachmodification of ventilator settings (except for Fio2), it wasadvised to perform an additional arterial blood gas analy-sis. Fio2 could be adjusted to maintain arterial oxygensaturation of �95%.

In both groups, patients were tracheally extubated afterachieving general tracheal extubation criteria (i.e., respon-sive and cooperative, urine output �0.5 mL/kg/h, chesttube drainage �100 mL last hour, no uncontrolled arrhyth-mia, and having a core temperature �36.0°C and a respi-ratory frequency of �10 breaths per minute withoutmachine-controlled breaths for at least 30 minutes). T-pieceweaning was not used; patients were tracheally extubatedwhen they reached the above-described extubation criteria.

ASV-DE Versus Standard ASVWith ASV-DE, as soon as body temperature reached 35.0°Cwith pH �7.25, irrespective of arterial Pco2, %-minuteventilation was decreased stepwise by 10% (de-escalation)until 70% of the theoretical value based on ideal bodyweight, only if pH declined �7.25%-minute ventilation wasincreased again (escalation) (Fig. 1). Indeed, by neglectingthe Pco2 safety limits as defined for standard ASV, wecreated a span for de-escalation with ASV-DE.

With standard ASV, %-minute ventilation was onlychanged if Pco2 was �3.5 or �5.5 kPa. Indeed, settingswith ASV were based on previously used safety limits toguarantee sufficient minute ventilation at all times.3

Adaptive Support Ventilation Protocolized with De-Escalation After Cardiothoracic Surgery


Data CollectionCollected data included the patient characteristics of gender,age, weight, and height, and the operation characteristics ofnumber of bypasses, cardiopulmonary bypass time (pumptime), and aortic clamp time. Intraoperative and ICU sedativeand analgesic prescriptions; time from admission to ICU untilreaching a central body temperature of 36.0°C; ventilationcharacteristics, tidal volume size, PEEP, inspiratory pressure(defined as the maximum airway pressure minus the level ofPEEP) and respiratory rate, %-minute ventilation, and arterialblood gas data in time were collected. Respiratory data werecollected by a data logger connected to the ventilator (Ham-ilton data logger, version 3.27.1; Hamilton Medical AG) andfrom our patient Data Management System (IMDsoft, Sassen-heim, The Netherlands).

Outcome variables (see Definitions) were calculated foreach patient. Primary end point was total duration oftracheal intubation; secondary end points were summedsingle assisted ventilation episodes, time to first singleassisted ventilation episode, time to the assisted ventilationepisode that was followed by tracheal extubation, andlength of stay in the ICU.

DefinitionsTotal duration of tracheal intubation was defined as theperiod from ICU admission until tracheal extubation, and asingle assisted ventilation episode was defined as an epi-sode of �20 minutes during which the patient was breath-ing at least 5 breaths per minute.

Opiate doses were all recalculated as morphine equipo-tent doses with the following formula: 10 mg morphine �0.1 mg fentanyl � 0.01 mg sufentanil.7 Doses of benzodi-azepines were similarly converted to equipotent doses ofdiazepam using the following formula: 5 mg midazolam �10 mg diazepam � 50 mg oxazepam.8

Sample SizeThe study was powered on total duration of trachealintubation. Sample size assumptions were based on resultsof our previous study, i.e., a mean duration of ventilation of

16.4 hours in the ASV group.3 A reduction of approxi-mately 10% was expected for the total duration of trachealintubation. A sample size of 61 patients in each group wasdeemed to have 80% power to detect a difference in theduration of MV of 10%, assuming a common standarddeviation of 3.3 hours, using a 2-sided t test with a 0.052-sided significance level.

Statistical AnalysisDescriptive statistics were used to summarize patient char-acteristics. Categorical variables were compared betweengroups by �2 tests. If normally distributed, continuousvalues were expressed as means � SD; otherwise, mediansand interquartile ranges were used. All analyses wereperformed in SPSS version 16.0 (SPSS, Inc., Chicago, IL).

RESULTSPatientsWe included 126 consecutive patients after elective and un-complicated CABG: 63 patients were randomized to ASV-DEand 63 to standard ASV (Fig. 2). Of patients enrolled in thestudy, 2 patients were lost for analysis of the secondary endpoints because of data logger failures: 1 patient randomized toASV-DE and 1 randomized to standard ASV.

Baseline, Perioperative, and ICU CharacteristicsGroups were well balanced (Table 1). Arterial blood gasanalyses on ICU admission were not different (data notshown). Core temperature on ICU admission was notdifferent (35.5°C � 1.1°C vs 35.7°C � 0.6°C, ASV-DE versusstandard ASV; P � 0.32). The number of patients with atemperature �36°C on ICU admission was also not differ-ent (27 [44%] vs 32 [51%], ASV-DE versus standard ASV;P � 0.24). There were no differences in time to rewarmingto 36°C (2.1 � 3.0 vs 1.7 � 2.1 hours, ASV-DE versusstandard ASV; P � 0.64). ICU survival was 100% for the 2randomization groups.

Figure 1. Flow sheet as used by intensive care unit (ICU) nurses andICU physicians [translated from Dutch]. ASV � adaptive supportventilation; %-MV � percentage minute ventilation; F-spont � fre-quency of spontaneous breath; ABG � arterial blood gas.

Figure 2. Flow diagram showing the flow of the patients through eachstage of the clinical trial. CABG � coronary artery bypass graft;COPD � chronic obstructive pulmonary disease; ASV-DE � adaptivesupport ventilation de-escalation escalation.


Protocol AdherenceIn the ASV-DE group, mean %-minute ventilation at tra-cheal extubation was 92% � 13% (14% were tracheallyextubated at %-minute ventilation level of 70%, 77% at alevel between 70% and 100%, and 9% at a level �100%). Inthe standard ASV group, mean %-minute ventilation at thetime of tracheal extubation was 103% � 10% (2% at a level�100%, 78% at a level of 100%, and 20% at a level �100%)(P � 0.05 versus ASV-DE).

Sedation and analgesic use was not different betweenrandomization groups (Table 1). There were no patientswho fulfilled the criteria for delirium, and haloperidol wasnever started.

Duration of MV and AssistedVentilation EpisodesDuration of tracheal intubation was not different betweengroups (10.8 [6.5–16.1] vs 10.7 [6.6–13.9] hours, ASV-DEversus standard ASV; P � 0.32) (Fig. 3; Table 2). Neithertime from admission to the first assisted breathing period(3.1 [2.0–6.7] vs 3.9 [2.1–7.5] hours; P � 0.49) nor thenumber of assisted ventilation episodes (78 [34–176] vs 57[32–116] episodes; P � 0.20) was different. However, dura-tion of assisted ventilation episodes that ended with tra-cheal extubation was longer with ASV-DE (2.5 [0.9–4.6] vs1.4 [0.3–3.5] hours; P � 0.05).

In a per-protocol analysis in which only patients whoreached �80%-minute ventilation before tracheal extuba-tion were compared with patients in the standard ASVgroup, there were also no significant differences in timeuntil extubation: 10.5 (8.0–16.0) vs 10.0 (6.0–13.0) hours(n � 16 vs n � 63). The choice for 80%-minute ventilation wasarbitrary; however, we believed that 90%-minute ventilation

was not clinically significant and with 70%-minute ventilationthe number of patients would result in too few patients.

Ventilator and Ventilation VariablesVentilator and ventilation variables are presented in Figure4. There were no differences between groups regardingtidal volume, respiratory rate, arterial pH, Pco2, and Po2.The highest levels of arterial Pco2 were similar in the 2randomization groups (5.9 [range, 5.2–6.4] vs 5.8 [range,5.0–6.4] kPa, ASV-DE versus standard ASV; P � 0.82).

DISCUSSIONIn this study of postoperative weaning of patients afterplanned and uncomplicated nonfast-track CABG, we foundthat ASV with protocolized de-escalation and escalation

Figure 3. Tracheally intubated patients (%) expressed as Kaplan-Meier curve in the adaptive support ventilation (ASV) and in the ASVde-escalation escalation (ASV-DE) groups.

Table 1. Patient Characteristics, and Intraoperative and Intensive Care CharacteristicsASV-DE Standard ASV

Patient characteristicsPatients, n 63 63Male gender, n (%) 56 (89) 55 (87)Age (y), mean � SD 68 � 10 65 � 9Actual body weight (kg), mean � SD 82 � 14 83 � 10Ideal body weight (kg), mean � SD 69 � 8 70 � 8Set body weight (kg), mean � SD 69 � 8 70 � 8Height (cm), mean � SD 172 � 15 175 � 8

Intraoperative characteristicsNo. of bypasses, mean � SD 3 � 1 3 � 1ECC time (h), mean � SD 1.7 � 0.7 1.7 � 0.6AOX time (h), mean � SD 1.1 � 0.5 1.1 � 0.5Lowest core temperature (°C), mean � SD 34 � 1.2 34 � 1.3Sufenta dose (�g), n; median (IQR)a 63; 200 (135–250) 63; 200 (150–250)Midazolam dose (mg), n; median (IQR)a 50; 15 (5–25) 49; 15 (5–20)Morphine dose (mg), n; median (IQR)a 28; 20 (20–20) 26; 20 (20–20)Clonidine dose (�g), n; median (IQR)a 6; 150 (150–225) 3; 150 (150–300)

Intensive care unit characteristicsAPACHE II score 17 � 5 16 � 5Morphine dose (mg/kg) n; median (IQR)a 0.05 (0.03–0.11) 0.05 (0.02–0.07)Propofol dose (mg) ICU, n; median (IQR)a 63; 1022 (621–1963) 63; 1125 (500–2122)Sedation duration (h), median (IQR)a 3.6 (2.2–6.8) 4.6 (2.6–7.0)Length of stay ICU (h), median (IQR) 27 (21–49) 27 (22–40)

ASV � adaptive support ventilation; ASV-DE � ASV de-escalation escalation; ECC time � duration of extracorporeal circulation; AOX time � duration of aorticcross-clamping; IQR � interquartile range; APACHE � Acute Physiology and Chronic Health Evaluation; ICU � intensive care unit.a Only dose of patients who actually received medication is displayed.There were no significant differences between the randomization groups.



compared with standard ASV did not shorten duration oftracheal intubation. The time to the first assisted breathingperiod was also not different between groups. There was,however, a difference in the duration of the assistedbreathing period ending with extubation.

MV could harm all patients, including those whose lungsare ventilated for only hours.9 Therefore, it is imperative tostrive for shorter duration of MV and tracheal intubation at alltimes, including the weaning phase after surgery. In addition,controlled forms of MV could rapidly cause muscle atrophy ofthe diaphragm.10 To counteract this phenomenon, it is impor-tant to allow patients to use their diaphragm as soon aspossible while still being mechanically ventilated. ASV allowsfor automatic switches between controlled ventilation andassisted ventilation, depending on the patient’s activity. Assuch, ASV with protocolized de-escalation and escalation mayimprove outcome because there was a trend to shorter timeuntil the first assisted breathing period, and more assistedventilation episodes. Our study, however, was underpoweredto show statistical difference regarding these secondary endpoints.

Although we hypothesized that protocolized de-escalationwould prevent longer intubation times, our study showed

otherwise. Of note, improved patient-ventilator synchronywith ASV could also lengthen duration of tracheal intubation.Indeed, this could be attributable to increased comfort, feweralarms, and a gradual transmission from controlled to spon-taneous ventilation thus not leading to apnea. Our study wasnot designed to test this hypothesis.

In contrast to our study, a significant reduction of timeuntil tracheal extubation with ASV as compared withsynchronized intermittent mandatory ventilation/pressuresupport was found in patients after fast-track cardiotho-racic surgery.5 In this trial by Sulzer et al., initially ASV wasset at 100%-minute ventilation (phase 1). When spontane-ous breathing occurred, %-minute ventilation was reducedby 50% (phase 2), and if necessary again by 50% (phase 3).This weaning approach can be described as much moreaggressive with respect to de-escalation, compared withour study. We chose to de-escalate stepwise until %-minuteventilation was 70%, as suggested on the Web site of themanufacturer.¶ Our stepwise approach and the minimumlevel of %-minute ventilation may have been a flaw.

¶Hamilton Medical. Available at: http://www.hamilton-medical.com/. Ac-cessed May 14, 2009.

Figure 4. Graphical display of ventilatorand ventilation variables over time.Minute ventilation in %, positive end-expiratory pressure, tidal volume (VT) inmL/kg ideal body weight (IBW), L/min,mean pH, and PCO2 and PO2 in kPa; blackcircles represent adaptive support venti-lation (ASV) group; open circles representASV de-escalation escalation (ASV-DE)group. RR � respiratory rate in breathsper minute; I � after stabilization; II �after 4 hours; III � before extubation.

Table 2. Respiratory CharacteristicsASV-DE Standard ASV P value

Duration of tracheal intubation (h), median (IQR) 10.8 (6.5–16.1) 10.7 (6.6–13.9) 0.32Time (h) from admission to first spontaneous breathing period, median (IQR) 3.1 (2.0–6.7) 3.9 (2.1–7.5) 0.49No. of spontaneous breathing periods, median (IQR) 78 (34–176) 57 (32–116) 0.20Spontaneous breathing/total time (%), median (IQR) 36 (21–61) 31 (21–50) 0.39Controlled breathing/total time (%), median (IQR) 64 (39–79) 69 (50–79) 0.39Last spontaneous breathing period before extubation (h), median (IQR) 2.5 (0.9–4.6) 1.5 (0.3–3.5) 0.045Tidal volume (mL), mean � SD 573 � 105 580 � 105 0.62Tidal volume (mL/kg ideal body weight), mean � SD 8.1 � 1.5 8.4 � 1.5 0.27Respiratory rate during spontaneous breathing, mean � SD 12.4 � 1.4 12.7 � 1.3 0.22PEEP level (cm H2O), mean � SD 5.3 � 1.2 5.3 � 1.0 0.60

ASV � adaptive support ventilation; ASV-DE � ASV de-escalation escalation; IQR � interquartile range; PEEP � positive end-expiratory pressure.


In a more recent study in patients after fast-track cardio-thoracic surgery, time to extubation was also significantlyshorter with ASV as compared with pressure-regulated vol-ume controlled with automode ventilation.4 In this trial byGruber et al., weaning consisted of 3 phases: controlledventilation (phase 1), assisted ventilation (phase 2), followedby a T-piece trial (phase 3) that ended with extubation. Thistrial did not use de-escalation. An important similarity be-tween the trials by Sulzer et al. and Gruber et al., and incontrast to our trial, is that the former trials were bothperformed in fast-track cardiothoracic surgery patients,whereas we explicitly included nonfast-track patients.

There were no differences in arterial blood gas variables,including arterial pH and Pco2, which could be explainedby the fact that patients in the ASV-DE group could adjusttheir minute ventilation when operator set %-minute ven-tilation was decreased (resulting in higher actual minuteventilation than the operator set %-minute ventilation).Indeed, time until the first assisted breathing period wasshorter, and more assisted ventilation episodes were foundin the ASV-DE group. It is also important to note that thehighest levels of arterial Pco2 were not different betweenthe 2 randomization groups, indicating that ASV-DE is atleast as safe as standard ASV.

There are multiple reasons why we were unable to showa difference between standard ASV and ASV-DE, includingstandard of care in our setting and type of patients studied.In a study comparing SmartCare® (a knowledge-basedweaning tool including an automatic gradual reduction ofpressure support, automatic performance of spontaneousbreathing trials, and generation of an incentive messagewhen a breathing trial was successfully passed) with con-ventional weaning, Rose et al.11 found no differencesregarding duration of MV. This finding was in sharpcontrast to the results from a study by Lellouche et al.12

showing SmartCare to significantly reduce duration of MV.One important difference, however, between the controlgroups of the 2 studies was that duration of MV wasshorter in the study by Rose et al. The shorter duration ofMV could have masked any beneficial effect of the interven-tion in the first study, whereas it allowed for an importanteffect of the intervention in the second study. Differences induration of MV could have resulted from differences inpatient case mix and differences in standard care surroundingthe studied patient populations in these 2 studies. We mayhave encountered a similar problem: in our study, duration oftracheal intubation was rather long compared with other trialsof ASV. This may very well relate to the fact that we includednonfast-track patients instead of most other studies of wean-ing of cardiothoracic surgery patients. The rather long dura-tion of MV may have precluded any effect of ASV-DE overstandard ASV in our study.

Because this was an open-label, i.e., not blinded, ran-domized clinical trial, we could not exclude the possibilitythat patients randomized to the standard ASV group alsobenefited from early de-escalation. This could have mini-mized contrast between the study groups. To promoteprotocol adherence, nurses and physicians were able toassess only 1 of the 2 flowcharts (as presented in Fig. 1), forASV-DE or standard ASV, depending on randomizationgroup. We could not always prevent both flowcharts being

available in the unit. Notably, however, we noticed a signifi-cant difference between the study groups regarding %-minuteventilation at the time of extubation and the period of assistedventilation leading to tracheal extubation.

Apart from the fact that this was a single center study,which limits the generalizability of our conclusions, thereare other limitations of the study. Sedation and analgesiarequirements were not reported using specific scales. How-ever, we did not gradually wean patients off of sedation,but stopped infusion of sedation completely when the coretemperature was �35°C. Of note, sedation and analgesicrequirements were the same in the 2 study groups. Anotherlimitation is that we excluded patients with chronic ob-structive pulmonary disease, also limiting generalizability.

Compared with a previous study of ASV by our group,overall weaning time in the present study was considerablyshorter. Indeed, median time to tracheal extubation was16.4 (12.5–20.8) hours in the previous study.3 The questionmust be raised whether tracheal extubation of cardiothoracicsurgery patients is dependent on the ventilatory strategyalone, or (also) on factors independent of the ventilationstrategy. The above-mentioned studies by Gruber et al.4 andSulzer et al.5 certainly show that the ventilator strategyinfluences weaning time in these patients. One ventilatorystrategy factor that could have influenced weaning time in ourstudies was the use of different levels of PEEP. Specifically, inour previous study of ASV, patients received 10 cm H2OPEEP in the first 4 hours after arrival in the ICU, andthereafter 5 cm H2O PEEP until tracheal extubation. In thepresent study, patients received 5 cm H2O PEEP throughoutthe complete weaning phase. This extra step in the weaningprocess could have accelerated weaning in the present study.In addition, this change in practice could have resulted in achange in use of sedatives because we continued sedatives forat least the first 4 hours, or as long as the higher level of PEEPwas used, in the first study. This usually took longer than thetime needed to reach a core temperature �35°C, which wasthe time to stop sedatives in the present study. Factorsindependent of ventilation strategy could also have a role.Both the surgical/anesthesiological team and the ICU teamgained experience over time, whereas the local guidelines(apart from the advice on PEEP) of these teams as well as theircomposition did not change. Better understanding of theneeds of patients after cardiothoracic surgery could have ledto the use of less sedatives both intra- and postoperativelydespite the fact that no formal protocol changes wereimplemented.13 Also, more experience with ASV in thisparticular patient group could have led to more confidencein earlier tracheal extubation in our department. Finally,better awareness of long weaning times in our institutioncould have led to a more proactive behavior with regard totracheal extubation.14 Although we performed a random-ized controlled trial, and as such all these factors should nothave affected the primary outcome differently in the 2study arms, one could certainly suggest that other factorsthan the ventilatory strategy have key roles in time untiltracheal extubation in these patients.

In conclusion, compared with standard ASV, weaning ofpatients after nonfast-track CABG using ASV with proto-colized de-escalation and escalation does not shorten timeto tracheal extubation.



AUTHOR CONTRIBUTIONSDAD helped design and conduct the study, analyze the data,and write the manuscript. This author has seen the originalstudy data, reviewed the analysis of the data, approved thefinal manuscript, and is the author responsible for archivingthe study files. DPV helped conduct the study, analyze thedata, and write the manuscript. This author has seen theoriginal study data, reviewed the analysis of the data, andapproved the final manuscript. JMB helped design the study,analyze the data, and write the manuscript. This author hasseen the original study data, reviewed the analysis of the data,and approved the final manuscript. BAJMdM helped designthe study. This author has seen the original study data andapproved the final manuscript. AK helped conduct the studyand analyze the data. This author has seen the original studydata, reviewed the analysis of the data, and approved the finalmanuscript. FP helped conduct the study. This author has seenthe original study data and approved the final manuscript. MJShelped design the study, analyze the data, and write themanuscript. This author has seen the original study data,reviewed the analysis of the data, and approved the finalmanuscript.

REFERENCES1. Brunner JX, Iotti GA. Adaptive support ventilation (ASV).

Minerva Anestesiol 2002;68:365–82. Arnal JM, Wysocki M, Nafati C, Donati S, Graniet I, Corno G,

Durrand-Gasselin J. Automatic selection of breathing patternusing adaptive support ventilation. Intensive Care Med2008;34:75– 81

3. Dongelmans DA, Veelo DP, Paulus F, de Mol BA, Korevaar JC,Kudoga A, Middelhoek P, Binnekade JM, Schultz MJ. Weaningautomation with adaptive support ventilation: a randomizedcontrolled trial in cardiothoracic surgery patients. AnesthAnalg 2009;108:565–71

4. Gruber PC, Gomersall CD, Leung P, Joynt GM, Ng SK, Ho KM,Underwood MJ. Randomized controlled trial comparingadaptive-support ventilation with pressure-regulated volume-controlled ventilation with automode in weaning patients aftercardiac surgery. Anesthesiology 2008;109:81–7

5. Sulzer CF, Chiolero R, Chassot PG, Mueller XM, Revelly JP.Adaptive support ventilation for fast tracheal extubation aftercardiac surgery: a randomized controlled study. Anesthesiol-ogy 2001;95:1339–45

6. Santiago TV, Edelman NH. Opioids and breathing. J ApplPhysiol 1985;59:1675–85

7. Reisine T, Pasternak GW. Opioid analgesics and antagonists. InHardman JG, Limbird LE, eds. Goodman & Gilman’s ThePharmacological Basis of Therapeutics. 9th ed. New York:McGraw-Hill, 1996:521–55

8. Wilson WC, Smedira NG, Fink C, McDowell JA, Luce JM.Ordering and administration of sedatives and analgesics dur-ing the withholding and withdrawal of life support fromcritically ill patients. JAMA 1992;267:949–53

9. Schultz MJ, Haitsma JJ, Slutsky AS, Gajic O. What tidalvolumes should be used in patients without acute lung injury?Anesthesiology 2007;106:1226–3

10. Levine S, Nguyen T, Taylor N, Friscia ME, Budak MT, Rothen-berg P, Zhu J, Sachdeva R, Sonnad S, Kaiser LR, RubinsteinNA, Powers SK, Shrager JB. Rapid disuse atrophy of dia-phragm fibers in mechanically ventilated humans. N EnglJ Med 2008;358:1327–35

11. Rose L, Presneill JJ, Johnston L, Cade JF. A randomised,controlled trial of conventional versus automated weaningfrom mechanical ventilation using SmartCare/PS. IntensiveCare Med 2008;34:1788–95

12. Lellouche F, Mancebo J, Jolliet P, Roeseler J, Shortgen F, DojatM, Cabello B, Bouadma L, Rodriguez P, Maggiore S, ReynaertM, Mersmann S, Brochard L. A multicenter randomized trial ofcomputer-driven protocolized weaning from mechanical ven-tilation. Am J Respir Crit Care Med 2006;174:894–900

13. De Hert SG, Van der Linden PJ, Cromheecke S, Meeus R, tenBroecke PW, De Blier IG, Stockman BA, Rodrigus IE. Anesthe-siology 2004;101:9–20

14. Hawkes CA, Dhileepan S, Foxcroft D. Early extubation foradult cardiac surgical patients. Cochrane Database Syst Rev2003;4:CD003587


Lung Recruitment and Positive End-ExpiratoryPressure Have Different Effects on CO2 Elimination inHealthy and Sick LungsGerardo Tusman, MD,* Stephan H. Bohm, MD,† Fernando Suarez-Sipmann, PhD,‡Adriana Scandurra, Eng,§ and Goran Hedenstierna, PhD�

BACKGROUND: We studied the effects that the lung recruitment maneuver (RM) and positiveend-expiratory pressure (PEEP) have on the elimination of CO2 per breath (VTCO2,br).METHODS: In 7 healthy and 7 lung-lavaged pigs at constant ventilation, PEEP was increasedfrom 0 to 18 cm H2O and then decreased to 0 in steps of 6 cm H2O every 10 minutes. CyclingRMs with plateau pressure/PEEP of 40/20 (healthy) and 50/25 (lavaged) cm H2O were appliedfor 2 minutes between 18-PEEP steps. Volumetric capnography, respiratory mechanics, bloodgas, and hemodynamic data were recorded.RESULTS: In healthy lungs before the RM, VTCO2,br was inversely proportional to PEEP decreasingfrom 4.0 (3.6–4.4) mL (median and interquartile range) at 0-PEEP to 3.1 (2.8–3.4) mL at18-PEEP (P � 0.05). After the RM, VTCO2,br increased from 3.3 (3–3.6) mL at 18-PEEP to 4.0(3.5–4.5) mL at 0-PEEP (P � 0.05). In lavaged lungs before the RM, VTCO2,br increased initiallyfrom 2.0 (1.7–2.3) mL at 0-PEEP to 2.6 (2.2–3) mL at 12-PEEP (P � 0.05) but then decreasedto 2.4 (2–2.8) mL when PEEP was increased further to 18 cm H2O (P � 0.05). After the RM, thehighest VTCO2,br of 2.9 (2.1–3.7) mL was observed at 12-PEEP and then decreased to 2.5(1.9–3.1) mL at 0-PEEP (P � 0.05). VTCO2,br was directly related to changes in lung perfusion, thearea of gas exchange, and alveolar ventilation but inversely related to changes in dead space.CONCLUSIONS: CO2 elimination by the lungs was dependent on PEEP and recruitment and showedmajor differences between healthy and lavaged lungs. (Anesth Analg 2010;111:968–77)

The effect of positive end-expiratory pressure (PEEP)on CO2 kinetics has been described. PEEP, at con-stant ventilation and body metabolism, is related to a

decrement in the elimination of CO2 by the lungs becauseof several factors: (1) a decrement in CO2 transport to thelungs by a decrease in venous return and thus cardiacoutput (CO),1–3 (2) a transient decrease in expired tidalvolume (Vt) caused by the sequential accumulation of airwithin the lungs right after the increase in PEEP,3,4 (3) again in functional residual capacity (FRC) leading to afilling of the lungs with inspiratory gases free of CO2

thereby inducing a transient dilution of alveolar CO2,3–5

and (4) an increase in airway and alveolar dead spacecausing a decrease in alveolar ventilation (Va).6,7

The collapse of healthy anesthetized and acutely injuredlungs of patients is well described and the main ventilatorytreatment of such collapse conditions is built uponPEEP.8–11 PEEP and low Vt ventilation are parts of alung-protective ventilatory strategy that aims to minimizethe injury caused by tidal recruitment and overdisten-sion.12,13 However, because PEEP has been related to theretention of CO2 within the body,14–17 this protectiveventilatory strategy could lead to hypercapnia, especially inthe context of low Vt ventilation.12

We have observed in patients that after a lung recruitmentmaneuver (RM), a ventilatory intervention aimed at restoringpulmonary aeration, CO2 elimination increased despite theuse of high PEEP levels and low Vt values.18–20 These resultscontradict the classical understanding of the effects that PEEPseems to have on CO2 kinetics.3–5 To our knowledge, asystematic analysis of the elimination of CO2 during changesin PEEP combined with a lung RM has not been performed.These ventilatory interventions are likely to show differentresponses in healthy and sick lungs because of the patho-physiology of acute lung injury resulting from massive lungcollapse, lung edema, and tissue inflammation.

Therefore, the aim of this study was to describe suchchanges in the elimination of CO2 in animals with healthy andsurfactant-depleted lungs and to determine whether lungrecruitment and PEEP induced retention of CO2 within theblood.

METHODSAfter approval by the Animal Research Committee of Upp-sala University in Sweden, 14 Swedish mixed country breedpigs (body weight � 24.5 � 3 kg) were anesthetized with IVketamine 25 to 50 mg/kg/h, midazolam 90 to 180 �g/kg/h,fentanyl 3 to 6 �g/kg/h, and pancuronium 0.25 to 0.50mg/kg/h. The trachea was intubated with a 7-mm innerdiameter cuffed endotracheal tube and air leaks were identi-fied by incomplete flow-volume loops or changes in thecapnograms. The lungs were ventilated with a SERVO-i(Maquet Critical Care, Wayne, NJ) using a volume controlmode with a Vt of 6 mL/kg, respiratory rate of 30breaths/min, I:E ratio of 1:2, a fraction of inspired oxygen of 1,and initially without PEEP. Intravenous saline solution wasmaintained at a fixed rate of 4 mL/kg during the study. Body

Author affiliations are provided at the end of the article.



Reprints will not be available from the author.

Address correspondence to Gerardo Tusman, MD, Department of Anesthe-siology, Hospital Privado de Comunidad, Mar del Plata, Argentina. Addresse-mail to [email protected].

Copyright © 2010 International Anesthesia Research SocietyDOI: 10.1213/ANE.0b013e3181f0c2da


temperature was maintained by a warm blanket to keep rectaltemperature within a range of 37.5°C � 0.5°C.

CO2 DataVolumetric capnography was recorded on-line using theNICO capnograph (Respironics, Wallingford, CT). The air-way flow and CO2 mainstream sensor were placed at the “Y”piece of the ventilator circuit and delivered data into acustom-made MatLab program (MathWorks, Natick, MA),which constructed volumetric capnograms by a Levenberg-Marquardt fitting method, and all capnography-derived pa-rameters were calculated from this mathematical function.21

The Vtco2,br is the amount of CO2 eliminated in 1 breathobtained by integrating expired airway flow and Pco2 signals.

Petco2 is the partial pressure of CO2 at the end ofexpiration and Peco2 is the mixed partial pressure of CO2

in 1 breath. Airway dead space was calculated as theinflection point of the capnogram, which defines theairway-alveolar interface or the limit between Vdaw andthe alveolar Vt (Vtalv).21

Dead space to Vt ratio (Vd/Vt) is an index that deter-mines the global inefficiency of ventilation. It was calcu-lated using the Bohr-Enghoff formula22:

Vd/Vt � (Paco2 � Peco2)/Paco2

The mean value of 10 slopes of phase III (SIII) wascalculated as previously described.21 SIII is a qualitative andnoninvasive marker of the global ventilation-perfusion(V/Q) ratio, where low values are related to normal V/Qratios whereas high values are indicators of an increaseddispersion of V/Q.23–26

Respiratory Mechanics DataRespiratory mechanics data were recorded using the flowand pressure sensors of the same NICO capnograph. Va isthe effective portion of ventilation and is defined as theproduct of Vtalv and respiratory rate.

Respiratory system dynamic compliance (Crs) was cal-culated as �Vt/�Paw. Changes in FRC (�FRC) induced byPEEP were calculated by the following formula3:

�FRC � �i�1

i�n

Vt(0) � Vt(i)

where Vt(0) was the reference exhaled Vt and n was thenumber of breaths with decreased Vt [Vt(i)] after a PEEPincrease or increased Vt [Vt(i)] after a PEEP decrease,respectively.

Gas Exchange DataArterial blood gases were monitored on-line using themultiparameter intraarterial sensor TrendCare (DiametricsMedical Ltd., High Newcombe, UK) inserted into the rightcarotid artery. Independent arterial and mixed venousblood gas samples were extracted at each protocol step andanalyzed within 5 minutes using an ABL 300 (Radiometer,Copenhagen, Denmark).

Pao2, Paco2, and Pvco2 are the Po2 and CO2 in thearterial and venous blood, respectively. The Pa-etco2 is thedifference between arterial and end-tidal partial pressure ofCO2 representing the area for eliminating CO2 through thelungs where a low difference corresponds to a large area for

exchange with a better diffusion of CO2 and vice versa.20

Pv-aco2 is the gradient between venous and arterial CO2.Shunt was calculated using a standard formula.27

Hemodynamic DataElectrocardiogram and pulse oxymetry were recorded, and acatheter for mean arterial blood pressure measurement wasplaced in the femoral artery. A 7.5F pulmonary artery catheterCCOmbo (Edwards Lifesciences, Irvine, CA) placed into theright jugular vein provided continuous CO and pulmonarypressures. CO was then subdivided into (1) an ineffectiveshunting part (COSHUNT), calculated as the product betweenshunt and CO to get the absolute value in L/min, and (2) aneffective pulmonary perfusion part (COEPP) or the portion ofthe CO that participates in CO2 exchange calculated bysubtracting the COSHUNT value from CO. This last parameterwas used to represent the effect of pulmonary blood flow onVtco2,br during the protocol.

ProtocolAnimals were randomly assigned to 2 groups: healthy (n �7) or surfactant-depleted lungs (n � 7). Lung lavages with35 mL/kg warm isotonic saline solution were performed inthe lavaged lung group.28 Lavages were repeated every 5minutes until Pao2 stabilized between 100 to 150 mm Hg atpure oxygen and PEEP of 8 cm H2O.

The protocol consisted of 3 sequential parts:

1. An increasing PEEP limb, where PEEP was increasedin steps of 6 cm H2O from 0 to 18 cm H2O using avolume control mode of ventilation. These data rep-resent the isolated effect that PEEP would have onthe elimination of CO2.

2. An RM, which consisted of a 2-minute cycling RM inpressure control ventilation using 40/20 cm H2O inhealthy lungs29 or 50/25 cm H2O in sick lungs30 forplateau pressure and PEEP, respectively.

3. A decreasing PEEP limb, as a mere mirror image of part1 of the protocol where PEEP was decreased from 18 to0 PEEP in steps of 6 cm H2O. These data represent thecumulative effect that PEEP in combination with a priorRM would have on CO2 elimination.

Each level of PEEP in parts 1 and 3 was maintained for 10minutes because previous publications3–5 and our ownresults from pilot studies showed that �90% of all changesin Vtco2,br caused by PEEP occurred within this timeframe.

Data AnalysisBefore starting the protocol, in vitro and in vivo calibrations ofdevices were performed following the manufacturer’s guides.Hemodynamic and on-line blood gas data were stored in alaptop by an acquisition system programmed in LabView(National Instruments, Austin, TX) while CO2 and respiratorydata were recorded in another laptop using the dedicatedsoftware Aplus (Respironics, Wallingford, CT). Both laptopswere synchronized in time. CO2, respiratory mechanics, andhemodynamic data belonging to the last minute of each PEEPstep (30 breaths � 30 data points), including the blood gassamples taken at this time, were analyzed.

A descriptive statistical analysis was performed using theMatLab program. Lilliefors test determined a non-Gaussian


distribution of the data. Friedman nonparametric test wasused to compare the results of the same level of PEEP beforewith those after RM in a 2-way direction. The same test wasused to compare differences between results of consecutivelevels of PEEP. Values are expressed as median (interquartilerange) and P values �0.05 were considered significant.

RESULTSAll pigs completed the protocol successfully. Absolute valuesof the main variables belonging to the last minute for eachPEEP step are presented in Tables 1 to 3. In general, PEEPapplied after lung recruitment improved lung function whencompared with PEEP alone in both healthy and lavagedlungs. The recruitment effect was characterized by a gain in�FRC, an increase in Crs, and decreases in Vd/Vt and shunt,paralleled by improvements in gas exchange.

In healthy pigs before recruitment, Vtco2,br decreasedfrom 4.0 (3.6–4.4) mL (0-PEEP) to 3.1 (2.8–3.4) mL (18-PEEP, P � 0.05) as PEEP increased (Fig. 1). Figure 2 shows

that the relative changes in the elimination of CO2 withincreasing PEEP levels were mainly related to a decrease inthe efficacy of ventilation (�Va) and the decrease in COEPP.The area for eliminating CO2 (Pa-etco2) showed a smallincrease with increasing PEEP levels but with little effect onCO2 elimination. Dead space (Vd/Vt) and SIII increasedproportionally to PEEP (Table 1).

After recruitment, Vtco2,br increased from 3.3 (3–3.6)mL (18-PEEP) to 4.0 (3.5–4.5) mL (0-PEEP, P � 0.05) asPEEP was reduced. This increased CO2 elimination wasassociated with an increase in Va and COEPP. Initially,Pa-etco2 decreased when going from 18 to 12 cm H2O ofPEEP, but progressively increased again when going fur-ther down to 0-PEEP (Table 1).

Vtco2,br presented a different behavior in lavaged ani-mals (Figs. 1 and 2). Before recruitment, Vtco2,br initiallyincreased from 2.0 (1.7–2.3) mL (0-PEEP) to 2.6 (2.2–3) mL(12-PEEP) (P � 0.05). This increment in CO2 eliminationwent along with an increased COEPP and a decreased

Table 1. Ventilation-Related Data During the Protocol in Healthy AnimalsPEEP (cm H2O)

Increasing limb of PEEP

0 6 12 18VT (mL) 196 (184–208) 197 (187–207) 192 (180–204) 188 (165–211)VA (L) 3.5 (3.1–3.9) 3.3 (2.8–3.8) 3.0 (2.6–3.4) 2.7 (2.3–3.1)Pplat (cm H2O) 12 (10–14) 18 (17–19) 26 (25–27) 33 (31–35)�FRC (mL) 96 (55–137) 279 (213–342) 570 (497–643)Crs (mL/cm H2O) 17 (14–20) 16 (14–18) 15 (14–16) 15 (14–16)VD/VT 0.43 (0.41–0.45) 0.49 (0.45–54) 0.48 (0.44–0.52) 0.52 (0.48–56)SIII (mm Hg/mL) 0.022 (0.012–0.032) 0.018 (0.008–0.028) 0.023 (0.001–0.044) 0.036 (0.010–0.062)pH 7.44 (0.43–0.45) 7.43 (0.38–0.48) 7.43 (0.37–49) 7.40 (0.33–0.47)PaO2 (mm Hg) 496 (424–568) 483 (421–544) 514 (483–545) 552 (536–568)Pa–ETCO2 (mm Hg) 3 (2–4) 2 (1–3) 1 (0–2) 1 (0–2)Pv�–aCO2 (mm Hg) 9 (5–13) 9 (6–12) 12 (9–15) 13 (5–21)

Values are presented in median and interquartile range. Comparison of the same level of positive end-expiratory pressure (PEEP) before (increasing limb) and after(decreasing limb) a lung recruitment maneuver.VT � expired tidal volume; VA � alveolar ventilation; Pplat � plateau pressure; �FRC � change in functional residual capacity; Crs � dynamic respiratory compliance;VD/VT � ratio between physiological dead space and tidal volume; SIII � slope of phase III; PaO2 � arterial partial pressure of oxygen; Pa-ETCO2 � difference betweenarterial and end-tidal partial pressure of carbon dioxide; Pv�-aCO2 � difference between mixed venous and arterial partial pressure of carbon dioxide.* P � 0.05.

Table 2. Ventilation-Related Data During the Protocol in Surfactant-Depleted AnimalsPEEP (cm H2O)


0 6 12 18VT (mL) 176 (152–200) 177 (161–183) 175 (150–190) 172 (158–186)VA (L) 3.0 (2.4–3.6) 2.6 (2.1–3.1) 2.5 (2–3) 2.3 (1.8–2.8)Pplat (cm H2O) 27 (25–29) 26 (23–29) 27 (26–28) 32 (30–34)�FRC (mL) 92 (76–108) 246 (206–286) 457 (406–508)Crs (mL/cm H2O) 7 (6–8) 11 (9–11) 16 (14–18) 15 (12–18)VD/VT 0.73 (0.62–0.84) 0.64 (0.54–0.74) 0.58 (0.44–0.72) 0.58 (0.53–0.63)SIII (mm Hg/mL) 0.090 (0.070–0.110) 0.072 (0.047–0.097) 0.039 (0.013–0.065) 0.067 (0.027–0.107)pH 7.29 (0.26–0.32) 7.28 (0.24–0.32) 7.34 (0.30–0.38) 7.36 (0.31–0.41)PaO2 (mm Hg) 40 (18–62) 93 (32–154) 240 (155–325) 494 (421–565)Pa-ETCO2 (mm Hg) 33 (25–41) 13 (7–19) 6 (2–10) 4 (2–6)Pv�-aCO2 (mm Hg) 7 (4–10) 8 (6–10) 10 (7–13) 12 (10–14)

Values are presented in median and interquartile range. Comparison of the same level of positive end-expiratory pressure (PEEP) before (increasing limb) and after(decreasing limb) a lung recruitment maneuver.VT � expired tidal volume; VA � alveolar ventilation; Pplat � plateau pressure; �FRC � change in functional residual capacity; Crs � dynamic respiratory compliance;VD/VT � ratio between physiological dead space and tidal volume; SIII � slope of phase III; PaO2 � arterial partial pressure of oxygen; Pa-ETCO2 � difference betweenarterial and end-tidal partial pressure of carbon dioxide; Pv�-aCO2 � difference between mixed venous and arterial partial pressure of carbon dioxide.* P � 0.05.

CO2 Elimination by Lung Recruitment and PEEP


Pa-etco2. Vtco2,br then decreased from 2.6 (2.2–3) mL(12-PEEP) to 2.4 (2–2.8) mL (18-PEEP) (P � 0.05). This time,the impairment of CO2 elimination was associated with areduced Va and COEPP despite Pa-etco2 showing thelowest value of the increasing limb of PEEP.

After recruitment, the highest Vtco2,br was observed at12-PEEP (2.9 [2.1–3.7] mL, P � 0.05) which decreased to 2.5(1.9–3.1) mL at 0-PEEP (P � 0.05). The progressive decre-ments in the area for CO2 exchange and in COEPP wereassociated with a lower elimination of CO2 at 0-PEEP. Vd/Vtand SIII increased with reductions in PEEP (Table 2).

Figure 3A represents the elimination of CO2 over time fora single step change of PEEP from 6 to 12 cm H2O beforerecruitment. Compared with 6-PEEP, the change in medianVtco2,br was �6% in healthy lungs and �8% in lavaged lungsat 12-PEEP (both P � 0.05). In both healthy and surfactant-depleted animals, Vtco2,br decreased in the first breaths afterthe PEEP increase with �90% of the effect occurring within 5minutes at the higher PEEP.

Figure 3B shows the effect of a PEEP change from 6- to12-PEEP before recruitment on the main variables for 18consecutive respiratory cycles. The effects were qualita-tively similar, but quantitatively different in healthy andin lavaged lungs. Vtco2,br decreased almost to zero inthe first breath but recovered within 5 to 6 consecutivebreaths in both healthy and lavaged lungs. This changein Vtco2,br was related to a decrease in expired Vt rightafter the PEEP change resulting in an expansion of �FRCby approximately 183 mL in healthy and 154 mL inlavaged lungs. As opposed to Vtco2,br, CO was littleaffected by the change in PEEP.

Pvco2, Paco2, and Petco2 are presented in Figure 1 andTables 1 and 2. In healthy lungs, no clinically significantchanges in those variables were observed. In surfactant-depleted lungs, however, the difference between Paco2 andPetco2 narrowed significantly but in an inverse relation tothe applied PEEP before and after lung recruitment with

Table 1. (Continued)PEEP (cm H2O)

Decreasing limb of PEEP

18 12 6 0191 (169–213) 195 (175–215) 200 (178–222) 201 (179–223)2.9 (2.5–3.3) 3.0 (2.5–3.5) 3.6 (3.2–4.0)* 3.9 (3.4–4.4)*29 (27–31) 19 (18–20)* 13 (12–14)* 9.6 (7–12)

�628 (�597 to 659)* �401 (�370 to 432)* �406 (�371 to 441)* �181 (�151 to 211)18 (16–20) 30 (27–33)* 31 (28–34)* 21 (18–24)*

0.51 (0.48–0.54) 0.44 (0.39–49)* 0.41 (0.38–0.44)* 0.42 (0.39–45)0.031 (0.006–0.056)* 0.033 (0.023–0.043)* 0.014 (0.009–0.019)* 0.023 (0.010–0.036)7.39 (0.32–0.46) 7.42 (0.37–0.47) 7.43 (0.39–47) 7.42 (0.38–0.47)591 (570–612)* 576 (553–599)* 556 (540–572)* 476 (424–528)

1 (0–2) 0.6 (0.4–1)* 1 (0.7–2.1)* 5 (3–7)*13 (8–17) 15 (7–22)* 12 (8–16) 11 (8–14)*



18 12 6 0171 (158–184) 172 (149–185) 174 (150–198) 177 (153–201)2.4 (1.4–3.4) 2.8 (2.2–3.4)* 2.8 (2.3–3.3)* 2.9 (2.3–3.5)29 (27–31)* 22 (19–25)* 20 (18–22)* 22 (19–25)*

�607 (�541 to 673)* �377 (�354 to 400)* �384 (�353 to 415)* �185 (�165 to 205)19 (15–23)* 23 (17–29)* 15 (12–18)* 9 (8–10)

0.56 (0.50–0.62) 0.56 (0.48–0.64) 0.59 (0.56–0.62)* 0.63 (0.57–0.69)*0.035 (0.010–0.069)* 0.024 (0.004–0.048)* 0.031 (0.011–0.051)* 0.050 (0.028–0.072)*7.40 (0.32–0.48)* 7.40 (0.33–0.47)* 7.34 (0.23–45)* 7.25 (0.20–0.30)527 (501–553)* 399 (354–444)* 179 (90–268)* 80 (50–110)*

2 (1–3)* 6 (3–9) 9 (3–15)* 17 (13–21)*14 (12–16) 10 (9–11) 9 (8–10) 7 (3–11)


corresponding changes in pH. Both Pvco2 and Paco2

decreased with lung recruitment and PEEP.

DISCUSSIONThe main findings of this study can be summarized asfollows:

1. Lung recruitment and PEEP have different effects onCO2 elimination in healthy and surfactant-depleted lungs.

2. At constant metabolism and ventilatory settings, anychange in the elimination of CO2 can be explained bya combination of changes in (a) the effectiveness oflung perfusion, (b) the area for CO2 exchange, and (c)the amount of Va and, thus, in the global V/Qrelationship of the lungs.

3. In healthy and lavaged lungs, changes in PEEPaltered the elimination of CO2 because of immediateeffects on both expired Vt values and �FRC for 5 to6 consecutive breaths. Stable new values for Vtco2,br,lung perfusion, area of CO2 exchange, and Va werereached within 5 minutes.

4. In lavaged lungs, the efficacy of CO2 elimination wasdirectly related to the recruitment/derecruitment ef-fect. Lung recruitment and PEEP did not retain CO2

in the blood during the study periods.

Effect of PEEP and Lung Recruitment onCO2 EliminationBreen and Mazumdar3 and Johnson and Breen5 were thefirst to analyze the effect of PEEP on the non–steady-stateCO2 kinetics in healthy lungs. Our main goal was to furthertheir work and to describe the same effect in the context ofa lung RM. Thus, to our knowledge, the data presented inthe current study are the first to describe the effect of PEEPand lung recruitments on the elimination of CO2 innon–steady-state conditions of healthy and surfactant-depleted lungs.

Figure 1. The elimination of CO2 per breath and tension-based CO2values during the protocol. PVCO2 (dots), PaCO2 (squares) and PETCO2(triangles). White columns in the lower panel are the elimination ofCO2 per breath (VTCO2,br) values. *Comparisons between values of thesame level of positive end-expiratory pressure (PEEP) before and after alung recruitment maneuver (RM); P � 0.05.

Table 3. Hemodynamic Data in Healthy and Surfactant-Depleted AnimalsPEEP (cm H2O)


0 6 12 18Healthy lungs

HR (bpm) 95 (68–122) 86 (48–124) 94 (66–122) 90 (68–112)MAP (mm Hg) 74 (47–101) 82 (61–103) 69 (49–89) 65 (44–86)MPAP (mm Hg) 21 (15–27) 25 (17–33) 25 (19–31) 28 (21–35)CVP (mm Hg) 7 (5–9) 8 (6–10) 9 (5–13) 11 (7–15)CO (L/min) 2.60 (2–3.2) 2.60 (2.1–3.1) 2.10 (1.6–2.6) 2.10 (1.4–2.8)COEPP (L/min) 2.03 (1.1–3) 2.42 (2–2.8) 1.93 (1.3–2.5) 1.93 (1.2–2.6)COSHUNT (L/min) 0.31 (0.1–0.5) 0.29 (0.1–0.4) 0.17 (0.1–0.3) 0.09 (0.05–0.15)

Lavaged lungsHR (bpm) 106 (75–137) 104 (71–135) 101 (63–139) 114 (85–143)MAP (mm Hg) 84 (62–106) 95 (70–120) 85 (65–105) 80 (55–105)MPAP (mm Hg) 45 (24–66) 38 (29–47) 34 (26–42) 35 (28–42)CVP (mm Hg) 9 (6–12) 9 (6–12) 9 (5–13) 11 (8–14)CO (L/min) 5.50 (4.3–6.7) 5.40 (4.2–6.6) 4.31 (3.5–5.1) 4.30 (3.4–5.1)COEPP (L/min) 3.03 (2–4) 3.08 (2.3–3.8) 3.77 (3.3–4.3) 3.50 (2.9–4.1)COSHUNT (L/min) 2.48 (2–3) 1.70 (1.2–2.2) 0.84 (0.5–1.1) 0.30 (0.1–0.5)

Values are presented in median and interquartile range. Comparison of the same level of positive end-expiratory pressure (PEEP) before (increasing limb) and after(decreasing limb) a lung recruitment maneuver.HR � heart rate; MAP � mean systemic arterial pressure; MPAP � mean pulmonary artery pressure; CVP � central venous pressure; CO � cardiac output;COEPP � effective part of CO; COSHUNT � ineffective part of CO.* P � 0.05.



The elimination of CO2 by the lungs, at constant venti-lation and body metabolism, depends on its transport bythe blood into the lungs, its diffusion through the alveolar-capillary membrane, and its elimination by the Va.3,31

Therefore, the final Vtco2,br value measured at the airwayopening in response to a PEEP challenge with or without alung RM is the result of a complex interaction of severalfactors. Figure 2 and Tables 1 to 3 show that these interactionsdiffer between healthy and lavaged lungs and, sometimes, the

influence on CO2 elimination of 1 factor differed from otherfactors.

Figure 3 shows the change in the elimination of CO2

during an increase in PEEP from 6 to 12 cm H2O without RM.The layout of this figure is similar to the ones presentedin the articles by Breen and Mazumdar3 and Johnson andBreen5 to facilitate the reader’s comparison of the data.According to the results shown in this figure, changes inVtco2,br, lung perfusion, CO2 exchange, and Va

Figure 2. Relative changes in the CO2 elimination per breath (VTCO2,br) during sequential changes of positive end-expiratory pressure(PEEP). The determinants of VTCO2,br are represented by the effective portion of cardiac output (COEPP), the area for CO2 exchange(Pa-ETCO2), and the effective portion of ventilation (VA). The relative changes in each of the variables as they are induced by the PEEPchange are expressed in percentage, from zero to a maximum cutoff of 50%. An improvement is defined as a change toward more normalvalues. Such improvements in CO2 elimination are characterized by increases in VA and COEPP and decreases in Pa-ETCO2 and arerepresented as white bars above the zero line. Impairment is defined as a change toward more abnormal values. Such impairments inCO2 elimination are characterized by decreases in VA and COEPP and increases in Pa-ETCO2 and are represented by black bars below thezero line. *P � 0.05.



18 12 6 0

92 (67–117) 91 (67–115) 89 (68–102) 92 (64–120)67 (45–89) 66 (38–94) 86 (69–103) 83 (65–101)*20 (14–26)* 23 (15–31) 20 (12–28)* 26 (18–34)*12 (6–18) 10 (7–13) 9 (6–12) 6 (4–8)

1.80 (1.3–2.3)* 2.00 (1–3) 3.10 (2.1–4.1)* 3.70 (2.5–5.2)*1.63 (1.1–2.1)* 1.93 (0.9–2.9) 2.95 (2–4)* 3.29 (1.3–5.3)*0.07 (0.05–0.1) 0.11 (0.1–0.2)* 0.19 (0.1–0.3)* 0.47 (0.1–1)*

113 (73–153) 119 (91–149)* 117 (92–142)* 120 (85–155)*77 (53–104) 80 (59–99)* 87 (65–109)* 85 (65–105)36 (25–47) 36 (26–46) 40 (32–48) 46 (38–54)12 (8–16) 11 (9–13) 9 (7–11) 9 (7–11)

3.29 (2–4.4)* 4.00 (2.8–5.2)* 5.00 (3.8–6.2)* 5.80 (4.5–7.2)2.99 (2.3–3.6)* 3.52 (2.6–4.4) 3.95 (3.1–4.8)* 2.96 (2.3–3.7)0.22 (0.2–0.3)* 0.48 (0.2–0.8)* 1.07 (0.5–1.5)* 2.56 (2–3.2)


after a PEEP challenge vary between healthy and lavagedlungs.

In healthy lungs, 12-PEEP decreased Vtco2,br because ofdecreased Va and CO despite a slight improvement in gasexchange (Fig. 3A). More than 90% of the effects on Vtco2,br

induced by 12-PEEP occurred within 5 minutes. These resultsare similar to the findings of Breen and Mazumdar3 observedin healthy dogs and those of Johnson and Breen5 in healthyanesthetized humans. They also found that a PEEP challengedecreased Vtco2,br because of a decrement in Va and CO,with most of the recovery of Vtco2,br within 10 minutes.Differences in the recovery of Vtco2,br observed betweenthese protocols could be explained by differences in thespecies studied, differences in the levels of PEEP applied (11cm H2O in the study by Breen and Mazumdar and 10 cm H2Oin the study by Johnson and Breen), or by differences inhemodynamic status and treatments.

In lavaged lungs, Vtco2,br increased until 12-PEEP causedby an increment in the area for CO2 exchange despite

decreasing Va and CO (Fig. 3A). The effects of a partialalveolar recruitment induced by the increasing PEEP levelsare supported by concomitant increases in Pao2 and compli-ance at reduced dead spaces (Table 2). Although CO wasreduced by 21%, its COEPP increased by 22% because of therecruitment of shunt areas (Table 3). The final result was abetter elimination of CO2 because of an improved V/Q ratioas indicated by a decrement in SIII. The different findings inlavaged and healthy lungs point toward differences in thephysiological effects that PEEP and recruitment have onnormally aerated and on collapsed lungs.

Figure 3B provides a zoomed view of the first breaths aftera change in a PEEP step. The decrement in Vtco2,br after anincrease in PEEP was caused by the trapped gas within lungsat the onset of the higher PEEP, which diluted alveolar CO2.The opposite results were observed in cases of PEEP reduc-tion. After a PEEP change, a fast recovery in Vtco2,br wasfound within 5 to 6 consecutive breaths. This finding is similarto the one described by Johnson and Breen5 in anesthetized

Figure 3. Time course of CO2 elimination during 2 consecutive positive end-expiratory pressure (PEEP) steps. A, Data of each of the animals isdisplayed for two 10-minute periods at 6 and 12 cm H2O of PEEP, respectively. VTCO2,br is the elimination of CO2 per breath, VA the alveolar ventilation,Pa-ETCO2 the arterial to end-tidal difference in PCO2 (calculated by the on-line PaCO2 value from the arterial catheter minus PETCO2 from capnograms),CO the continuous cardiac output, and �FRC the change in functional residual capacity. B, Baseline is the mean value of the last 10 breaths on 6cm H2O of PEEP (here summarized in breath 0) followed by 18 consecutive breaths at 12-PEEP.



patients in which the recovery of Vtco2,br after a change ofPEEP from 0 to 10 cm H2O took 8 breaths.

Does PEEP and Lung Recruitment Retain CO2 inthe Blood?A protective ventilatory management with low Vt andplateau pressures is currently mandatory when treatingacutely injured lungs.12,13 Arterial hypercapnia is thus aclinically tolerated negative consequence of an intentionaldecrease in Va. This hypercapnic state could become evenworse if PEEP, as many authors have postulated, causedCO2 retention in the body.14–17

In healthy lungs, Vtco2,br decreased in proportion to PEEPbefore the recruitment but increased after it. At constantventilation and metabolism, this decrease in CO2 eliminationshould lead to retention of CO2 within the body when PEEPis applied without a prior recruitment. However, Pvco2,Paco2, and thus P�v-aco2 were not affected much during theprotocol (Table 1 and Figs. 1 and 2), which must be inter-preted that CO2 was not retained within the blood, at leastduring the study period. After recruitment, the elimination ofCO2 and global lung physiology improved at PEEP levelsdown to 12 cm H2O (Table 1 and Figs. 1 and 2) with chancesfor CO2 retention even lower than before the recruitment.

Figure 3. Continued.


Our results differ from those of Breen and Mazumdar3 andJohnson and Breen5 who found increased Pvco2 and Paco2 inhealthy lungs at 10 and 11 cm H2O of PEEP, respectively.Differences in the hemodynamic status, protocol time, andexperimental models might explain these differences.

In surfactant-depleted lungs, however, results weretotally different. Pv-aco2 increased whereas Pa-etco2

decreased with positive-pressure ventilation as long aslung collapse was low and overall lung function pre-served (Table 2 and Fig. 1). The increased area of gasexchange leading to an augmented diffusion of CO2

across the alveolar-capillary membrane after a lungrecruitment can explain this effect and was confirmed byparallel increments in Pao2 and Crs, 2 well-known mark-ers of lung recruitment.32,33 COEPP increased with in-creasing airway pressures because of the recruitment ofpreviously collapsed capillaries in the atelectatic areas(Table 3). Because Vtco2,br is directly proportional tolung perfusion, this increment in COEPP facilitates thetransport of CO2 from the body stores toward the lungswhere it is eliminated. Note that the absolute values ofPvco2 and Paco2 decreased with both lung recruitmentand PEEP, thereby confirming that CO2 was not retainedwithin the blood but rather eliminated more efficiently.

In lavaged lungs, the highest recruitment pressuresapplied in pressure control ventilation resulted in a meanVt of 7.4 mL/kg (221 [202–251] mL). This small andtransient increase in minute ventilation could have influ-enced the elimination of CO2 beyond the actual lungrecruitment effect during the descending PEEP steps.However, we believe that the decreased values for Pvco2

and Paco2 on the descending limb of the PEEP titrationwere mainly attributable to the recruitment of alveolibecause: (1) changes in the area of gas exchange, and notin Va, had a dominating influence on Vtco2,br (Fig. 2),(2) variables representing the recruitment effect andwhich are independent of Va (Crs, SIII, �FRC, Pao2, orshunt) improved after lung recruitment, and (3) a tran-sient and marginal increase in minute ventilation for 2minutes only would not have had a lasting impact onVtco2,br after a lapse of 10 minutes, the point in timewhen the blood samples were taken.

Clinical ImplicationsIn contrast to continuous positive airway pressure ma-neuvers, cycling RMs have the following advantages: (a)they are hemodynamically better tolerated,34 (b) thestep-wise and sequential increments in PEEP allow thegained volumes of air to spread progressively instead ofabruptly throughout the lung parenchyma,35,36 and (c)they allow real-time monitoring of respiratory variableson a breath-by-breath basis. Today, the way to conductand optimize RM is a topic of much debate, where theadvent of volume-based capnographic monitoring mayadd important new arguments in favor of such anapproach because of the noninvasive and real-time na-ture of this methodology.

Because CO2 kinetics are context sensitive and highlydependent on the sequence of steps during a cycling RM, inour protocol, the levels of PEEP were not assigned inrandom order. Had we randomized the protocol steps, we

would not have been able to show reproducibly the sequen-tial nature and the time dependence of CO2 elimination.For the same reason, we chose to study the CO2 kineticsduring a short lapse of time because we were interested inthe CO2 kinetics during the non–steady-state conditionsinduced by RMs and a PEEP titration process. This iscrucial new information, which should contribute to abetter understanding of CO2 kinetics, and we hope that itwill finally have clinical implications for the monitoring ofpatients during mechanical ventilation.

LimitationsSurfactant-depleted lungs as used in our experimentalstudy do not adequately represent the complex nature ofacute lung injury in real patients, and thus our resultsshould be interpreted with caution.

It is well documented that healthy and sick human lungshave different opening pressures29,30and therefore wechose our recruitment pressures in health and diseaseassuming that this difference would be true for animalsalso. To achieve a complete recruitment effect in oursurfactant-depleted animals, we decided to use an arbitraryand fixed opening pressure of 50 cm H2O for all animalswith sick lungs based on our own previous experience33

because we did not have access to lung imaging bycomputed tomographic scan to determine an optimal open-ing pressure individually for each animal. Had we used thesame pressure as we did for healthy animals, the risk ofhaving incompletely recruited diseased lungs and thusinconclusive study results would have been high.

CONCLUSIONSThe results of this study show that lung recruitment andPEEP have different effects on the elimination of CO2 inhealthy and lavaged lungs. These differences can be ex-plained by a complex interaction between the key factors oflung perfusion, diffusion through the alveolar-capillarymembrane, and Va. Our results suggest that sufficientlyhigh levels of PEEP applied after lung recruitment mayhelp decrease hypercapnia in patients treated with lung-protective ventilation and low Vt strategies.

AUTHOR AFFILIATIONSFrom the *Department of Anesthesiology, Hospital Privado deComunidad, Mar del Plata, Argentina; †CSEM Centre Suissed’Electronique et de Microtechnique SA, Research Centre forNanomedicine, Landquart, Switzerland; ‡Department of Criti-cal Care Medicine, Fundacion Jimenez Díaz-UTE, Madrid,Spain; §Bioengineering Laboratory, Electronic Department,University of Mar del Plata, Mar del Plata, Argentina; and�Department of Medical Sciences, Clinical Physiology, Univer-sity Hospital, Uppsala Sweden.

REFERENCES1. Qvist J, Pontoppidan H, Wilson RS, Lowenstein E, Laver MB.

Hemodynamic response to mechanical ventilation with PEEP:the effect of hypervolemia. Anesthesiology 1975;42:45–55

2. Berglund JE, Halden E, Jakonson S, Landelius J. Echocardio-graphic analysis of cardiac function during high PEEP venti-lation. Intensive Care Med 1994;20:174–80

3. Breen PH, Mazumdar B. How does positive end-expiratorypressure decrease CO2 elimination from the lung? RespirPhysiol 1996;103:233–42



4. Elliott WR, Harris AE, Philip JH. Positive end-expiratorypressure: implications for tidal volume changes in anesthesiamachine ventilation. J Clin Monit 1989;5:100–4

5. Johnson JL, Breen PH. How does positive end-expiratorypressure decrease pulmonary CO2 elimination in anesthetizedpatients? Respir Physiol 1999;118:227–36

6. Hedenstierna G, Lundberg S. Airway compliance during arti-ficial ventilation. Br J Anaesth 1975;47:1277–81

7. Coffey RL, Albert RK, Robertson HT. Mechanisms of physio-logical dead space response to PEEP after acute oleic acid lunginjury. J Appl Physiol 1983;55:1550–7

8. Lundquist H, Hedenstierna G, Strandberg A, Tokics L, BrismarB. CT-assessment of dependent lung densities in man duringgeneral anaesthesia. Acta Radiol 1995;36:626–32

9. Brismar B, Hedenstierna G, Lundquist H. Pulmonary densitiesduring anesthesia with muscular relaxation: a proposal ofatelectasis. Anesthesiology 1985;62:422–8

10. Gattinoni L, Caironi P, Valenza F, Carlesso E. The role ofCT-scan studies for the diagnosis and therapy of acute respi-ratory distress syndrome. Clin Chest Med 2006;27:559–70

11. Ware LB, Matthay MA. The acute respiratory distress syn-drome. N Engl J Med 2000;342:1334–49

12. The National Heart, Lung, and Blood Institute ARDS ClinicalTrials Network. Ventilation with lower tidal volumes as com-pared with traditional tidal volumes for acute lung injury andthe acute respiratory distress syndrome. The Acute Respira-tory Distress Syndrome Network. N Engl J Med 2000;342:1301–8

13. Amato MBP, Barbas CSV, Medeiros DM, Magaldi RB, Schet-tino GP, Lorenzi-Filho G, Kairalla RA, Deheinzelin D, MunozC, Oliveira R, Takagaki TY, Carvalho CRR. Effect of a protec-tive ventilation strategy on mortality in the acute respiratorydistress syndrome. N Engl J Med 1998;338:347–54

14. Dueck R, Wagner PD, West JB. Effects of positive end-expiratory pressure on gas exchange in dogs with normal andedematous lungs. Anesthesiology 1977;47:359–66

15. Pesenti A, Marcolin R, Prato P, Borelli M, Riboni A, GattinoniL. Mean airway pressure vs positive end-expiratory pressureduring mechanical ventilation. Crit Care Med 1985;13:34–7

16. Isserles SA, Breen PH. Can changes in end-tidal PCO2 measurechanges in cardiac output? Anesth Analg 1991;73:808–14

17. Hedenstierna G, White FC, Mazzone R, Wagner PD. Redistri-bution of pulmonary blood flow in the dog with PEEP venti-lation. J Appl Physiol 1979;46:278–87

18. Tusman G, Bohm SH, Suarez-Sipmann F, Turchetto E. Alveolarrecruitment improves ventilatory efficiency of the lungs dur-ing anesthesia. Can J Anaesth 2004;51:723–7

19. Tusman G, Bohm SH, Suarez-Sipmann F, Maisch S. Lungrecruitment improves the efficiency of ventilation and gasexchange during one-lung ventilation anesthesia. AnesthAnalg 2004;98:1604–9

20. Tusman G, Suarez-Sipmann F, Bohm SH, Pech T, ReissmannH, Meschino G, Scandurra A, Hedenstierna G. Monitoringdead space during recruitment and PEEP titration in anexperimental model. Intensive Care Med 2006;32:1863–71

21. Tusman G, Scandurra A, Bohm SH, Suarez-Sipmann F, Clara F.Model fitting of volumetric capnograms improves calculationsof airway dead space and slope of phase III. J Clin MonitComput 2009;23:197–206

22. Englhoff H. Volumen inefficax. Bemerkungen zur Frage desschadlichen Raumes. Uppsala Lakareforen Forhandl 1938;44:191–218

23. Harris B, Bailey DL, Chicco P, Bailey EA, Roach PJ, King GG.Objective analysis of whole lung and lobar ventilation/perfusion relationships in pulmonary embolism. Clin PhysiolFunct Imaging 2008;28:14–26

24. Blanch LL, Fernandez R, Saura P, Baigorri F, Artigas A.Relationship between expired capnogram and respiratory sys-tem resistance in critically ill patients during total ventilatorysupport. Eur Respir J 1999;13:1048–54

25. Hofbrand BI. The expiratory capnogram: a measure ofventilation-perfusion inequalities. Thorax 1966;21:518–24

26. Stromberg NO, Gustafsson PM. Ventilatory inhomogeneityassessed by nitrogen washout and ventilation-perfusion mis-match by capnography in stable and induced airway obstruc-tion. Pediatr Pulmonol 2000;29:94–102

27. Berggren SM. The oxygen deficit of arterial blood caused bynon-ventilated parts of the lungs. Acta Physiol Scand1942;Suppl 4:4–9

28. Lachmann B, Jonson B, Lindroth M, Robertson B. Modes ofartificial ventilation in severe respiratory distress syndrome:lung function and morphology in rabbits after wash-out ofalveolar surfactant. Crit Care Med 1982;10:724–32

29. Rothen HU, Sporre B, Englberg G, Wegenius G, HedenstiernaG. Reexpansion of atelectasis during general anaesthesia: acomputed tomography study. Br J Anaesth 1993;71:788–95

30. Borges JB, Okamoto VN, Matos GF, Caramez MPR, ArantesPR, Barros F, Souza CE, Victorino JA, Kacmarek RM, BarbasCSV, Carvalho CRR, Amato MBP. Reversibility of lung col-lapse and hypoxemia in early acute respiratory distress syn-drome. Am J Respir Crit Care Med 2006;174:268–78

31. Cherniack NS, Longobardo GS. Oxygen and carbon dioxidegas stores of the body. Physiol Rev 1970;50:196–243

32. Lachmann B. Open up the lung and keep the lung open.Intensive Care Med 1992;18:319–21

33. Suarez-Sipmann F, Bohm SH, Tusman G, Pesch T, Thamm O,Reissmann H, Reske A, Magnusson A, Hedenstierna G. Use ofdynamic compliance for open lung positive end-expiratorypressure titration in an experimental study. Crit Care Med2007;35:214–21

34. Celebi S, Koner O, Menda F, Korkut K, Suzer K, Cakar N. Thepulmonary and hemodynamic effects of two different recruit-ment maneuvers after cardiac surgery. Anesth Analg2007;104:384–90

35. Hickling KG. Best compliance during a decremental, but notincremental, positive end-expiratory pressure trial is related toopen-lung PEEP: a mathematical model of acute respiratorydistress syndrome lungs. Am J Respir Crit Care Med2001;163:69–78

36. Albaiceta GM, Taboada F, Parra D, Luyando LH, Calvo J,Menendez R, Otero J. Tomographic study of the inflectionpoints of the pressure-volume curve in acute lung injury. Am JRespir Crit Care Med 2004;170:1066–72


Society for Obstetric Anesthesia and Perinatology

Section Editor: Cynthia A. Wong

Pitfalls in Chronobiology: A Suggested Analysis UsingIntrathecal Bupivacaine Analgesia as an ExampleSteven L. Shafer, MD,* Bjoern Lemmer, MD, PhD,† Emmanuel Boselli, MD, PhD,‡ Fabienne Boiste, MD,‡Lionel Bouvet, MD,‡ Bernard Allaouchiche, MD, PhD,§ and Dominique Chassard, MD, PhD§

BACKGROUND: The duration of analgesia from epidural administration of local anesthetics toparturients has been shown to follow a rhythmic pattern according to the time of drug administration.We studied whether there was a similar pattern after intrathecal administration of bupivacaine inparturients. In the course of the analysis, we came to believe that some data points coincident withprovider shift changes were influenced by nonbiological, health care system factors, thus incorrectlysuggesting a periodic signal in duration of labor analgesia. We developed graphical and analyticaltools to help assess the influence of individual points on the chronobiological analysis.METHODS: Women with singleton term pregnancies in vertex presentation, cervical dilation 3 to5 cm, pain score �50 mm (of 100 mm), and requesting labor analgesia were enrolled in thisstudy. Patients received 2.5 mg of intrathecal bupivacaine in 2 mL using a combinedspinal-epidural technique. Analgesia duration was the time from intrathecal injection until thefirst request for additional analgesia. The duration of analgesia was analyzed by visual inspectionof the data, application of smoothing functions (Supersmoother; LOWESS and LOESS [locallyweighted scatterplot smoothing functions]), analysis of variance, Cosinor (Chronos-Fit), Excel,and NONMEM (nonlinear mixed effect modeling). Confidence intervals (CIs) were determined bybootstrap analysis (1000 replications with replacement) using PLT Tools.RESULTS: Eighty-two women were included in the study. Examination of the raw data using 3smoothing functions revealed a bimodal pattern, with a peak at approximately 0630 and asubsequent peak in the afternoon or evening, depending on the smoother. Analysis of variance didnot identify any statistically significant difference between the duration of analgesia when intrathecalinjection was given from midnight to 0600 compared with the duration of analgesia after intrathecalinjection at other times. Chronos-Fit, Excel, and NONMEM produced identical results, with a meanduration of analgesia of 38.4 minutes (95% CI: 35.4–41.6 minutes), an 8-hour periodic waveformwith an amplitude of 5.8 minutes (95% CI: 2.1–10.7 minutes), and a phase offset of 6.5 hours (95%CI: 5.4–8.0 hours) relative to midnight. The 8-hour periodic model did not reach statisticalsignificance in 40% of bootstrap analyses, implying that statistical significance of the 8-hour periodicmodel was dependent on a subset of the data. Two data points before the change of shift at 0700contributed most strongly to the statistical significance of the periodic waveform. Without these datapoints, there was no evidence of an 8-hour periodic waveform for intrathecal bupivacaine analgesia.CONCLUSION: Chronobiology includes the influence of external daily rhythms in the environment(e.g., nursing shifts) as well as human biological rhythms. We were able to distinguish the influenceof an external rhythm by combining several novel analyses: (1) graphical presentation superimposingthe raw data, external rhythms (e.g., nursing and anesthesia provider shifts), and smoothingfunctions; (2) graphical display of the contribution of each data point to the statistical significance;and (3) bootstrap analysis to identify whether the statistical significance was highly dependent on adata subset. These approaches suggested that 2 data points were likely artifacts of the change in nursingand anesthesia shifts. When these points were removed, there was no suggestion of biological rhythm inthe duration of intrathecal bupivacaine analgesia. (Anesth Analg 2010;111:980–5)

The duration of epidural analgesia, and the hypnoticeffect of many drugs used for anesthesia or sedation,have been shown to vary in a periodic pattern

through the day.1 The transcutaneous passage of lidocaine

has been investigated in the morning or in the evening;significantly higher plasma lidocaine levels were obtainedafter evening dosing in children.2 It has also been shownunder conditions of daily dental practice that the durationof local anesthesia after subcutaneous mepivacaine or arti-caine is not constant throughout the day.3 These studiessuggest a longer duration of anesthesia in the afternoon atapproximately 1500 for a variety of local anesthetic drugs.4

We have demonstrated a significant effect of the time ofadministration on the duration of epidural ropivacaineanalgesia in parturients, with a 30% longer duration ofanalgesia when the block is placed in the afternoon.5

Similarly, spinal opioids such as fentanyl or sufentanil donot have a constant duration of analgesia throughout a

Authors’ affiliations are listed at the end of the article.

Accepted for publication March 10, 2010.

Supported solely by institutional sources.


Address correspondence and reprint requests to Dominique Chassard, MD,PhD, Department of Anesthesiology and Intensive Care, Hopital Mere Enfant,Bron 69500, France. Address e-mail to dominique.chassard @ chu-lyon.fr.

Copyright © 2010 International Anesthesia Research SocietyDOI: 10.1213/ANE.0b013e3181dd22d4


24-hour period.6,7 These studies indicate that epidurally orspinally injected anesthetic drugs exhibit a circadian pat-tern. Because there are no such data on intrathecal localanesthetics, we tested the hypothesis that the duration ofaction of intrathecal local anesthetics has a circadianrhythm in parturients during the first stage of labor. Inexploring this hypothesis, we developed several novelapproaches to assess the influence of external rhythms onchronobiological models.

METHODSAfter obtaining agreement from the Ethical Committee ofthe University of Lyon (France) and written informedconsent, we enrolled pregnant women, ASA physical statusI, parity 0 and 1, with a singleton vertex pregnancy of atleast 36 weeks of gestation. Each woman was in the firststage of spontaneous labor with a cervical dilation 2 to 4 cmwhen requesting labor analgesia. Only patients with un-complicated pregnancy and normal fetal heart rate tracingsrequesting analgesia were enrolled. We excluded patientsreceiving antihypertensive drugs and patients who hadalready received opioids during labor.

Each patient received an intrathecal injection of 2.5 mg ofplain bupivacaine in 2 mL (a mixture of 1 mL 0.25% bupiva-caine and 1 mL normal saline) using a combined spinal-epidural technique with the Braun Escopan� combinationneedle (Braun Medical SA, Boulogne, France). After insertionof a 27-gauge pencil-point spinal needle and intrathecalbupivacaine injection, a 20-gauge multiorifice catheter wasinserted in the epidural space. The epidural catheter was leftin position, but no drug was delivered epidurally until theend of the intrathecal bupivacaine effect.

Pain was assessed using a visual analog scale (VAS) ofpain intensity at 15, 30, and 45 minutes after the block, andwhen secondary analgesia was requested. Each patient waspresented with a 100-mm line, labeled as no pain (left end)and worst pain imaginable (right end). Patients were askedto mark the line to indicate the intensity of their pain at thepeak of a contraction. Patients could request additionalanalgesia if pain relief was unsatisfactory by 15 minutesafter injection of the intrathecal bupivacaine. If a patientrequested relief within 15 minutes after injection, this wasconsidered a technical failure; these patients were excludedfrom further data collection. When additional analgesiawas requested, the study protocol and data collection wereterminated, and analgesia duration was recorded. Analge-sia duration was the time from intrathecal bupivacaineadministration to the first request for additional analgesia.

Data are presented as mean � SD. Circadian rhythmswere explored graphically using the 3 smoothing functions,Supersmoother,* LOWESS,8 and LOESS,9 as implementedin the R statistical programming language (R Project forStatistical Computing: http://www.r-project.org). Super-smoother is a variable span scatterplot smoothing function,whereas LOWESS and LOESS are locally weighted scatterplotsmoothing functions. To prevent edge effects, the smoothedvalue for each time point was calculated by rescaling the Xdata to place the point in the center of the data. This was done

by subtracting 24 hours from all data points �12 hours afterthe point, and adding 24 hours for each data point �12 hoursbefore the data point. This required calculating a newsmoother function for every data point, rather than just oncefor across all data. However, for this data set, the compu-tational time was negligible. The R code for Figure 2appears in Appendix 1 (see Supplemental Digital Con-tent1, http://links.lww.com/AA/A123).

The influence of parity and time of day (day [0800–2000]versus night [2000–0800]) on the VAS score before intra-thecal injection was analyzed using Mann-Whitney tests.Analgesia durations were divided into 4 periods based onthe time of intrathecal injection: midnight to 0600, 0600 to1200, 1200 to 1800, and 1800 to midnight. Times exactly onthe border of 2 periods (e.g., 0600) were assigned to thelater period. Differences in analgesia duration among peri-ods were analyzed using analysis of variance.

Parametric analysis estimated the duration of analgesiaas a cosine function of time, based on the equation:

analgesia duration � mesor � �1

i �amplitudei

� cos� 2�

periodi

(t � acrophasei)�� (1)

where t is the time of initiation of intrathecal analgesia,mesor is the mean of the fitted curve (i.e., average durationof anesthesia), amplitude is half of the magnitude of thecircadian variation, and acrophase is the phase offsetrelative to midnight (time 0). These parameters were calcu-lated for periods of 6, 8, 12, and 24 hours.

Three fitting programs were used: Chronos-Fit,†10 Excel2007 (Microsoft Corp., Redmond, WA) (using the Solverfunction to minimize �2 log likelihood), and nonlinear mixedeffect modeling (NONMEM) (version 6.2, Icon DevelopmentSolutions, Ellicott City, MD) (which also minimized �2 loglikelihood). Results were considered significant if P valueswere �0.05. No adjustment was made for multiple tests (i.e.,we considered 4 different periodic signals but did not adjustour P value to reflect the multiple tests).

The contribution of each subject to the final model wasestimated using NONMEM, comparing the individual �2log likelihood of a null model (analgesia duration � mesor[e.g., mean]) with the final model (Eq. 1, period � 8 hours).Ninety-five percent confidence intervals for the estimatedparameters were calculated using a NONMEM bootstrapanalysis of 1000 random bootstrap samples with replace-ment. The dependency of the statistical significance on asubset of the data was estimated by NONMEM usingbootstrap analysis to calculate the difference in �2 loglikelihood between the null (no period) and final (8-hourperiod) models in 1000 random samples with replacement.NONMEM was run using the PLT Tools platform (PLTsoft,San Francisco, CA; www.pltsoft.com).

No formal power analysis was done. The sample size

*Friedman JH. A variable span scatterplot smoother. Laboratory for Com-putational Statistics, Stanford University Technical Report No. 5, 1984.

†Zuther P, Lemmer B. Chronos-Fit, version 1.05, 2004. Available at:http://www.ma.uni-heidelberg.de/inst/phar/forschungLemmer.html. Ac-cessed June 6, 2008.


was chosen based simply on the experience of the investi-gators. Power analysis for chronobiological studies is re-viewed in more detail in the Discussion section.

The data files for this analysis, the R programming code,the Excel spreadsheet for the nonlinear regression, andNONMEM control and output files for chronobiologyanalysis using these data as an example are available as aWeb supplement to this article.

RESULTSNinety-one parturients were enrolled in this study fromJune 2005 through October 2005. Nine were excluded forspinal analgesia failure, protocol violation, or emergencycesarean delivery before a request for analgesia. Patientcharacteristics (n � 82) are provided in Table 1.

The mean VAS score before intrathecal injection was70 � 17 mm (median, 70 mm; range, 40–100 mm). VAS scoresbefore intrathecal injection were not significantly affected bythe period of day (74 � 13 mm from 0800 to 2000 vs 69 � 12mm from 2000 to 0800; P � 0.08) or parity (72 � 16 mm forparity 0 vs 68 � 10 mm for parity 1; P � 0.17). Visualinspection of the baseline VAS data over the 24-hour timecourse did not suggest any periodic signal. Fifteen minutesafter intrathecal bupivacaine, the mean VAS score decreasedto a median of 20 mm (range, 0–70 mm).

Figure 1 shows analgesia duration versus the time ofintrathecal injection. The black horizontal line is the aver-age analgesic duration. The smoothers (red � Super-smoother, blue � LOWESS, and green � LOESS) all reveala bimodal pattern with peaks (colored arrows) at 0624 and

2349 (Supersmoother), 0609 and 2318 (LOWESS smoother),and 0722 and 1812 (LOESS smoother). The orange horizon-tal bar shows the change in the certified registered nurseanesthetist (CRNA) and midwife shift, whereas the purplehorizontal bar shows the change in the shift for the anes-thesia fellow and anesthesia attending.

The duration of analgesia was modestly decreased dur-ing the period from midnight to 0600 compared with theother periods (Table 2). This trend did not reach statisticalsignificance by analysis of variance (P � 0.42).

Chronos-Fit, Excel, and NONMEM produced identicalresults. Excel and NONMEM returned identical �2 loglikelihood estimates. There was just 1 statistically signifi-cant (P � 0.04) periodic waveform, with an 8-hour period,amplitude of 5.8 minutes, acrophase (phase offset) of 6.5hours, and a mean duration of analgesia (mesor) of 38.4minutes (Table 3 and Fig. 2). The periodic waveformcaptured the peak seen in the original data at approxi-mately 0630, and approximately captured the peaks at 18hours (LOESS) and 24 hours (Supersmoother, LOWESS).

In Figure 2, the contribution of each data point can beinferred from the size and color of the dot. Black datapoints represent observations better fit by the model withthe 8-hour period than by the null model, which onlymodeled the mesor (average analgesic duration). Red datapoints represent observations worse fit by the model withthe 8-hour period than by the null model. The size of thedata point represents the increase (black) or decrease (red)in �2 log likelihood of the 8-hour periodic model over thenull model.

As shown in Figure 2, the 2 points that contributed most tothe improvement in �2 log likelihood were from the 2intrathecal injections performed shortly before the change innursing shift at 0700. One was placed at 0624 and lasted 81minutes, and the other was placed at 0653 and lasted 90minutes.

Figure 3 shows the result of the bootstrap analysis basedon 1000 samples of the data set with replacement. Thex-axis shows the difference in �2 log likelihood, and they-axis shows the cumulative fraction. The vertical linesmark differences in �2 log likelihood associated with P �0.05, 0.01, 0.001, and 0.0001. If a model was highly statisti-cally significant, regardless of which subset of the data was

Figure 1. Analgesia duration versus the time of intrathecal injection.Black horizontal line � median analgesic duration; red line �Supersmoother; blue line � LOWESS smoother; green line � LOESSsmoother. Arrows denote the peaks for each smoother of the samecolor. Orange horizontal bar � daytime certified registered nurseanesthetist and midwife shift; purple horizontal bar � daytimeanesthesia fellow and anesthesia attending shift.

Table 1. Demographic Data (Mean � SD)Age (y) 29 � 4Height (cm) 164 � 5Weight (kg) 72 � 9Dilation (cm) 3 (2–4)Parity 0/1 (n) 38/44

Table 2. Analgesia Duration by PeriodPeriod Analgesia duration (min � SD)

1 (0:01 to 6:00) 33.1 � 10.42 (6:01 to 12:00) 41.2 � 17.83 (12:01 to 18:00) 40.0 � 14.54 (18:01 to midnight) 40.0 � 14.2

Table 3. Periodic ModelParameter Typical value 95% CI

Mean (min) 38.4 35.4–41.6Period (h) 8 naAmplitude (min) 5.8 2.1–10.7Offset (h) 6.5 5.4–8.0

CI � confidence interval; na � not applicable.

Pitfalls in Chronobiological Studies


considered, then the vast majority of replicates in a boot-strap analysis would show a P value �0.05. That is not thecase. Statistical significance was not achieved in 40% of thereplicates, suggesting that statistical significance is highlydependent on the subset of data included. Given that the 2data points that contribute most to the model occur shortlybefore the change in nurse and anesthesia provider, weinvestigated the performance of the model with thoseoutliers excluded.

Figure 4 shows the data with the 2 suspected outlying datapoints removed. When those 2 data points were excluded,there was no longer a statistically significant periodic signalaccording to our reanalysis with Chronos-Fit, Excel, andNONMEM. Only the LOWESS smoother (blue line) offeredany evidence of periodicity. The Supersmoother and LOESSsmoother functions were essentially flat.

DISCUSSIONThe mean duration of action of intrathecal bupivacainereported in our study, 39 � 15 minutes, is similar to that

obtained by Wong et al.,11 39 � 25 minutes. Stocks et al.,12

using a technique of up-down sequential allocation, dem-onstrated that the minimum local analgesic dose of intra-thecal bupivacaine was 1.99 mg, and that the duration ofanalgesia with this dose was 43 � 19 minutes. However,using the same dosage, Lim et al.13 reported a longerduration of action, 76.3 � 5.9 minutes. The results couldhave been influenced by several factors other than the timeof the day: for example, the method to calculate analgesiaduration, VAS scores, cervical dilation at the time ofinclusion, parity, and the use of oxytocin. These factorsmust be incorporated in future studies of chronobiology inparturients, as well as the observation that the perception oflabor pain varies with time of day.14

This article was initially submitted to Anesthesia &Analgesia as a demonstration of the chronobiology of anal-gesia after intrathecal bupivacaine. Over the course of peerreview and additional analysis, this article has metamor-phosed into a cautionary tale of the influence of externalrhythms on chronobiological analysis, and an explorationof data analysis methods to facilitate detection of artifacts.Our presentation of the methods and results in this articlereflects the way we think this type of analysis shouldproceed. The actual analysis was much more convoluted.

The most critical element in data analysis is visualizingthe data. Figure 1 is our proposal for visualization of theraw data in chronobiological analyses. The figure showsthe raw data and adds 3 smoothing functions to draw theeye toward the underlying signal. Because smoothers areprone to edge effects, the smoothing functions have beencalculated by centering each observation in a 24-hour timewindow, permitting the smoothing function to properly“wrap around” midnight. The smoothers show that therecan be, at most, 2 peaks. Thus, the data themselves suggestthat there is little reason to explore a periodic signal with aperiod shorter than 12 hours.

In addition, Figure 1 shows the primary externalrhythms likely to influence the data: the start and end of thedaytime nursing and anesthesia shifts. It is immediatelyobvious that the spike in analgesia duration for intrathecal

Figure 2. Same result as Figure 1, with the 8-hour periodic waveformshown in gold. The size of the points reflects the magnitude of thecontribution of each point toward (black) or against (red) statisticalsignificance of the final model.

Figure 3. The cumulative distribution of the improvement in �2 loglikelihood comparing the null model with the final model in 1000bootstraps. In 40% of the bootstrap analyses, the difference was notstatistically significant, suggesting that the statistical significancedepended on the inclusion of a subset of the data points in theanalysis.

Figure 4. The same analysis as in Figure 1, but with the 2 outlyingdata points at 0624 and 0653 removed. Only the LOWESS smoother(blue line) continues to show evidence of periodicity. The Super-smoother and LOESS smoother are essentially flat.


injections before the change in CRNA shift might beexplained because the CRNAs, and subsequently the anes-thesiologists, are starting their daytime shift, delayingassessment of patients’ pain. We suggest that future chro-nobiological analyses include a figure similar to Figure 1 toidentify the basic patterns in the data, assess the likelyperiod of any chronobiological signal by using smoothingfunctions, and assess the relationship, if any, to externalrhythms. Of course, it is entirely possible that rhythmsunknown to the investigators (e.g., lunch breaks, the timingof the change in bed sheets, and the regular 4 am test of thebackup generators) might be present. If that was the case,isolated spikes would appear at given times, but an under-lying sine rhythm would not be evident.

Our second step was to divide the data into 4 time periodsthat approximately divided the 24-hour day into “night”(midnight to 0600), “morning” (0600 to 1200), “afternoon”(1200 to 1800), and “evening” (1800 to midnight). We did notsee any statistically significant differences.

Our third step was modeling the data. The “model” is asimple cosine function. We chose to model periods of 24,12, 8, and 6 hours. However, the only period that makes apriori sense biologically is the 24-hour period. Of course,there could be 2 peaks in some biological function, but whyshould they necessarily be 12 hours apart? If the day isdivided into a rhythmic pattern with multiple peaks, whyshould it be sinusoidal rather than (for example) 16 hoursalternating with 8 hours, as in the awake/asleep cycle? It iseven more mysterious why the day should be divided intosinusoidal 6- or 8-hour periods. Interestingly, even thoughwe only saw 2 peaks with the smoothing functions, the12-hour rhythm was not statistically significant. The onlyrhythm that was statistically significant was a cosine signalwith a period of 8 hours, shown in Figure 2. Figure 2 alsointroduces another new graphic element: showing theextent to which each observation contributes for, or against,the statistical significance of the model.

Inspection of Figure 2 reveals a problem with the model.It captures the observed peak at 0630, but the other 2 peaksin the sinusoidal model do not correspond with peaks inthe data identified by the smoothing functions. Examiningthe individual points, the 2 largest black dots are theobservations that contribute the most to the statisticalsignificance of the 8-hour periodic waveform. Looking atthe external rhythms (horizontal orange and purple bars),these are the very observations that are most likely toreflect the influence of the change in nursing and anesthesiashifts.

Before discarding the fit entirely, we turned to a boot-strap analysis to address whether the statistical significanceof the model was highly dependent on the specific datasampled. If the data broadly supported the statisticalsignificance of the more complex model, then statisticalsignificance should be evident from any subset of the dataanalyzed. However, if statistical significance depends onjust a few points, then a significant fraction of the bootstrapanalyses will fail to be statistically significant (the boot-straps that do not include the critical observations),whereas some bootstraps will have greatly increased sta-tistical significance (bootstraps that have �1 copy of the

critical observations). Because 40% of the bootstrap analy-ses did not reach statistical significance (Fig. 3), a smallnumber of data points are driving the statistical result.

Supported by the evidence in Figures 1, 2, and 3 that thestatistical significance was driven by an artifact, the obser-vations at 0624 (81 minutes of analgesia) and 0653 (90minutes of analgesia) were removed from the data. Figure4 shows that there is still a peak in duration at the time ofthe nursing shift change, but that the peak is significantlyattenuated by the loss of these 2 points. Indeed, Super-smoother and the LOESS smoother were essentially flat.Similarly, no periodic signal was statistically significantonce these points were removed.

Our lives are governed by powerful external rhythms.Most obviously, we are synchronized to the rotation of theearth, and the resulting periods of light and darkness.Other rhythms infuse our lives. These can affect studies ofchronobiology in 2 ways. First, they may induce realphysiological changes. For example, arterial blood pressuremight increase twice daily with the commute to and fromwork. However, external rhythms might produce artifactsthat have nothing to do with biology, such as delayedpatient assessment during the changing of provider shifts.

We suggest 6 steps to assess the influence of externalartifacts on chronobiological analyses:

1. Graph the raw data over 24 hours, along with 1 ormore smoothing functions. To avoid edge effects, thefunction needs to wrap around midnight. We elimi-nated edge effects by calculating the value of thesmoothing function for each observation with thetime centered in a 24-hour window. If the smoothingfunctions do not suggest a periodic waveform, thenno further analysis should be undertaken.

2. If a periodic pattern is evident in the smoothingfunctions, compare the smoothed functions with ex-ternal rhythms, such as shift changes, that mightaffect the observations.

3. If a periodic pattern is evident in the smoothingfunctions, fit the data with a regression program,such as Chronos-Fit, Excel, or NONMEM. If there isonly 1 point per subject, then it makes no differencewhat program is used. In this analysis, all 3 programsreturned results that were identical in all reporteddigits. If there are multiple observations per subject,only NONMEM can separate interindividual vari-ability from intraindividual variability.

4. Represent the data points to show the contributionfor, or against, the final model. This will identifywhether the model is being driven by a subset of thedata.

5. Undertake a bootstrap analysis to confirm whetherstatistical significance is driven by a small subset ofthe data points.

6. If data artifacts are driving the result, remove themand reassess the model.

Removing data that appear to be artifacts may bias studyresults in favor of the investigator’s expectation. For example,we do not know for certain that the observations at 0624 and0653 are artifacts. These data points could be completely validobservations. However, this analysis demonstrates that even

Pitfalls in Chronobiological Studies


if these points are valid (which we doubt), the lack ofcorrelation between the data and the other 2 peaks in the“statistically significant” 8-hour periodic signal preclude ourhaving any confidence in the model.

We have only explored rhythms that are synchronizedamong all patients. If each patient had a strong individualrhythm, but a random phase relationship to every otherpatient, then the underlying chronobiology would bemissed in this type of analysis. Individual rhythms withvariable phase offset could only be detected by analyzingmultiple observations in each individual, and using amixed effects model with intersubject variability in thephase offset (“acrophase”).

Conventional power analyses do not readily handle adata analysis plan that involves comparing 1 model to asecond model, as in this case in which a straight line isbeing compared with a sine wave. In our view, the best wayto approach power analysis is to perform a Monte Carlosimulation, in which multiple synthetic data sets are gen-erated for a range of sample sizes, and the percent ofsuccessful trials is calculated. The study size is based on thesample size in the Monte Carlo simulation that producesthe desired percent of models correctly identified.

In conclusion, circadian rhythms are an important as-pect of human biology. On the basis of these data, there isno important chronobiological rhythm in spinal bupiva-caine analgesia during labor. External daily rhythms maycontaminate the data when assessing biological rhythms.We suggest that the approach used in this study may helpto separate true biological signals from external artifact infuture studies.

AUTHORS’ AFFILIATIONSFrom the *Department of Anesthesiology, Columbia University,New York, New York; †Institute of Pharmacology & Toxicology,University of Heidelberg, Heidelberg, Germany; ‡Department ofAnesthesiology and Intensive Care, Hotel-Dieu Hospital; and§University Claude Bernard-Lyon 1, Lyon, France.

RECUSE NOTESteven L. Shafer is Editor-in-Chief of the Journal. The manu-script was handled by James G. Bovill, Guest Editor-in-Chief,and Dr. Shafer was not involved in any way with the editorialprocess or decision.

DISCLOSUREDr. Shafer is codeveloper of PLT Tools and has a financialinterest in the software.

ACKNOWLEDGMENTSThe authors thank Peter Tucker (American Translator, Univer-sity of Aix-Marseille, France) for assistance in the preparationof the manuscript. The authors also acknowledge the manyinsightful suggestions of Dr. Dennis Fisher to this work.

REFERENCES1. Chassard D, Bruguerolle B. Chronobiology and anesthesia.

Anesthesiology 2004;100:413–272. Bruguerolle B, Giaufre E, Prat M. Temporal variations in

transcutaneous passage of drugs: the example of lidocaine inchildren and in rats. Chronobiol Int 1991;8:277–82

3. Lemmer B, Wiemers R. Circadian changes in stimulus thresh-old and in the effect of a local anaesthetic drug in human teeth:studies with an electronic pulptester. Chronobiol Int 1989;6:157–62

4. Reinberg A, Reinberg MA. Circadian changes of the durationof action of local anaesthetic agents. Naunyn SchmiedebergsArch Pharmacol 1977;297:149–52

5. Debon R, Chassard D, Duflo F, Boselli E, Bryssine B, Allaouch-iche B. Chronobiology of epidural ropivacaine. Anesthesiology2002;96:542–5

6. Debon R, Boselli E, Guyot R, Allaouchiche B, Lemmer B,Chassard D. Chronopharmacology of intrathecal sufentanil forlabor analgesia: daily variations in duration of action. Anes-thesiology 2004;101:978–82

7. Pan PH, Lee S, Harris L. Chronobiology of subarachnoidfentanyl for labor analgesia. Anesthesiology 2005;103:595–9

8. Cleveland WS. Robust locally weighted regression andsmoothing scatterplots. J Am Stat Assoc 1979;74:829–36

9. Cleveland WS, Grosse E, Shyu WM. Local regression models.In: Chambers JM, Hastie TJ, eds. Statistical Models in S. PacificGrove, CA: Wadsworth & Brooks/Cole, 1992:309–76

10. Zuther P, Witte K, Lemmer B. ABPM-FIT and CV-SORT: aneasy-to-use software package for detailed analysis of data fromambulatory blood pressure monitoring. Blood Press Monit1996;1:347–54

11. Wong CA, Scavone BM, Loffredi M, Wang WY, Peaceman AM,Ganchiff JN. The dose-response of intrathecal sufentanil addedto bupivacaine for labor analgesia. Anesthesiology 2000;92:1553–8

12. Stocks GM, Hallworth SP, Fernando R, Engl AJ, Columb MO,Lyons G. Minimum local analgesic dose of intrathecal bupiv-acaine in labor and the effect of intrathecal fentanyl. Anesthe-siology 2001;94:593–8

13. Lim Y, Ocampo CE, Sia AT. A comparison of duration ofanalgesia of intrathecal 2.5 mg of bupivacaine, ropivacaine,and levobupivacaine in combined spinal epidural analgesia forpatients in labor. Anesth Analg 2004;98:235–9

14. Aya AG, Vialles N, Mangin R, Robert C, Ferrer JM, Ripart J, deLa Coussaye JE. Chronobiology of labour pain perception: anobservational study. Br J Anaesth 2004;93:451–3


The Influence of Time of Day of Administration onDuration of Opioid Labor Analgesia

Barbara M. Scavone, MD, Robert J. McCarthy, PharmD, Cynthia A. Wong, MD, and John T. Sullivan, MD

BACKGROUND: Medications administered into the epidural or intrathecal space for laboranalgesia may demonstrate variable effects dependent on time of day, and this may affectclinical research trials investigating the pharmacology of specific drugs. In this retrospectivestudy, we evaluated the effect of time of day of administration of intrathecal fentanyl andsystemic hydromorphone labor analgesia from data collected as part of a randomized clinical trialexamining the influence of analgesia method on labor outcome.METHODS: Six hundred ninety-two healthy parturients were randomized early in labor to receivecombined spinal-epidural (intrathecal fentanyl 25 �g followed by a lidocaine and epinephrinecontaining epidural test dose) versus systemic (hydromorphone 1 mg IV and 1 mg IM) laboranalgesia at first analgesia request. No further analgesics were administered until the patientrequested additional analgesia (second analgesia request). Subjects were assigned to thedaytime group (DAY) if initial analgesia (neuraxial or systemic) was administered between thehours of 07:01 and 23:00 and to the nighttime group (NIGHT) if it was administered between23:01 and 07:00. Within each mode of analgesia study arm (neuraxial or systemic), the DAY andNIGHT groups were compared. The primary outcome variable was analgesia duration, defined asthe time interval from administration of labor analgesia until the second analgesia request.Cervical dilation at first and second analgesia requests, pain score at first analgesia request, andaverage amount of pain between analgesia administration and second analgesia request werealso compared between DAY and NIGHT groups. Rhythm analyses for duration of analgesia,cervical dilation, and pain scores were performed.RESULTS: There was no difference in the median duration of either neuraxial or systemicanalgesia in DAY versus NIGHT subjects, and no harmonic variation was observed for analgesiaduration. Rhythm analysis demonstrated a 24-h harmonic cycle for cervical dilation at firstanalgesia request with maximum values occurring near 17:00 and minimum values near 05:00,but the amplitude of the difference was very small. Rhythm analysis demonstrated a 24-hharmonic cycle with maximum values occurring near 22:00 and minimum values near 10:00 forthe average amount of pain between analgesia administration and second analgesia request inneuraxial group patients, but amplitude was small.CONCLUSIONS: Time of day of administration did not seem to influence combined spinal-epidural or systemic labor analgesia duration under these study conditions. (Anesth Analg2010;111:986–91)

Chronobiology is the study of the effect of time,especially biologic rhythms, on living organisms;chronopharmacology is the study of the relation

between time of administration and the effects of drugs.Chronopharmacologic effects can be either chronopharmaco-dynamic (time-dependent differences in organism re-sponse to drugs) or chronopharmacokinetic (time-dependent differences in pharmacokinetic parameters,such as absorption, distribution, and clearance).1

Several authors have noted time-dependent effects ofmedications administered into the epidural or intrathecal

space for labor analgesia; specifically, differences in analgesiaduration after administration of epidural ropivacaine andintrathecal sufentanil and fentanyl depending on time of dayof administration.2–4 This has prompted concern regardingthe impact of chronopharmacology on clinical research trialsinvestigating the pharmacology of specific drugs.1 However,these trials involved small numbers of subjects, especiallyduring the night hours, and did not always control forexternal factors that may indirectly influence duration ofanalgesia (e.g., differences in labor unit staffing, visitinghours, and sleep deprivation).

The purpose of this retrospective study was to evaluatepossible effects time of administration may have on intra-thecal fentanyl and systemic hydromorphone labor analge-sia from data collected as part of a randomized clinical trialexamining the influence of analgesia method on laboroutcome.5 In the aforementioned study, duration of laboranalgesia was a secondary outcome, but the influence oftime of day was not examined. This study is a secondaryanalysis of the previously obtained data, examining theinfluence of time of administration of analgesia on durationof labor analgesia. The null hypothesis for this analysis wasthat duration of labor analgesia would not vary with timeof day.

From the Department of Anesthesiology, Northwestern University FeinbergSchool of Medicine, Chicago, Illinois.

Accepted for publication September 7, 2009.

Supported by Departmental funds.

Cynthia A. Wong is Section Editor of Obstetric Anesthesiology for theJournal. This manuscript was handled by Steve Shafer, Editor-in-Chief, andDr. Wong was not involved in any way with the editorial process ordecision.


Address correspondence to Barbara M. Scavone, MD, 251 E. Huron St.,F5-704, Chicago, IL 60611. Address e-mail to [email protected].

Copyright © 2010 International Anesthesia Research SocietyDOI: 10.1213/ANE.0b013e3181c29390


METHODSThe study was approved by the Northwestern UniversityOffice for the Protection of Research Subjects IRB. Details ofthe study design and protocol have been published.5

Briefly, written informed consent was obtained from 884healthy, opioid naïve, nulliparous women. Additional in-clusion criteria included uncomplicated, term, singletonpregnancy with vertex position, and spontaneous labor orspontaneous rupture of membranes. Parturients were re-cruited shortly after admission to the Labor and DeliveryUnit by anesthesia personnel 24 h per day, 7 days per wkover a 38-mo period commencing October 2000. Subjectswere enrolled (n � 750) in the study if the cervix wasdilated �4 cm at the first request for analgesia and wererandomized to receive neuraxial versus systemic opioidanalgesia. At the first analgesia request, subjects in theneuraxial group received a combined spinal-epidural via aneedle through needle technique. They received intrathecalfentanyl 25 �g and an epidural test dose of lidocaine 45 mgand epinephrine 15 �g in 3 mL total volume. Subjects in thesystemic analgesia group received 1 mg IV and 1 mg IMhydromorphone. No further analgesics were administereduntil the patient requested additional analgesia (secondanalgesia request). At the second analgesia request, acervical examination was performed and the subject re-ceived additional medications per study protocol.

Subjects rated peak contraction pain using a verbal ratingscale for pain: a 0–10 scale in which 0 � no pain and 10 � worstimaginable pain. Pain scores were obtained at the first analgesiarequest (pain score1st request). In addition, at the second anal-gesia request, when the effects of analgesia were diminish-ing, the patient was asked to rate the “average amount ofpain” experienced between receiving intrathecal fentanylor systemic hydromorphone and the second analgesiarequest (pain scoreaverage).

In this secondary analysis, subjects were assigned to thedaytime group (DAY) if initial analgesia (neuraxial orsystemic) was administered between 07:01 and 23:00 h, andto the nighttime group (NIGHT) if it was administeredbetween 23:01 and 07:00 h.2 Within each mode of analgesiastudy arm (neuraxial or systemic), the DAY and NIGHTgroups were compared. Duration of analgesia was theprimary outcome variable and was defined as the timeinterval between the administration of intrathecal fentanylor systemic hydromorphone and the second analgesiarequest. Age, height, weight, gestational age, oxytocinadministration, time interval from analgesia administrationto delivery, infant weight, mode of delivery, cervical dila-tion, and pain scores were also compared in DAY versusNIGHT group subjects.

Previous studies examining the chronopharmacologiceffect of intrathecal opioids have demonstrated a 30-mindifference between day and night administration in theduration of analgesia after administration of intrathecalfentanyl 20 �g in 77 subjects,4 and a 23-min difference afteradministration of intrathecal sufentanil 10 �g in 91 sub-jects.3 A sample of 345 subjects would achieve 96% powerto detect a difference of 7 min between day and nightduration of analgesia assuming an average duration of 90min with an estimated standard deviation of 25 min at an �equal to 0.05 using a 2-sided Mann-Whitney test.

DAY and NIGHT groups within each mode of analgesiaarm (neuraxial or systemic) were compared using the Mann-Whitney U-test (age, height, weight, gestational age, time todelivery, analgesia duration, cervical dilation, and pain scores) or�2 test (oxytocin administration and mode of delivery). The timedata were plotted against the data for duration of analgesia, aswell as cervical dilation and pain scores at first and secondrequest and average pain score between analgesia interven-tions and evaluated using weighted least squares regression(lowess), weighted quadratic least squares regression (loess),and nonparametric smoothing (the “super smoother” ofFriedman).*6 Curve fitting trend lines were determined usingthe R functions “lowess,” “loess,” and “supsmu”). To im-prove smoothness at the beginning and end of the 24-h cycle,data were concatenated from 25 h before and after the 24-hcycle. Lowess and loess curves were estimated at a smootherspan of 0.15. Data were analyzed using R version 2.9.1(http://www.r-project.org).

Rhythm analyses for duration of analgesia, cervical dila-tion, and pain scores were performed using a nonlinearWIN-ABPM-FIT program.† This analysis calculates the mesor(rhythm-adjusted mean), amplitude (half of the peak totrough of the rhythm-adjusted harmonic), and acrophase(time of the occurrence of the rhythm-adjusted harmonic).Data were evaluated for 24-, 12-, 8-, and 6-h harmonics. Thesolver add in for Microsoft Excel 2007 was used to calculatethe r2, �2 log likelihood, and P value for the harmonicsidentified. P � 0.05 was required to reject the null hypothesis.

RESULTSThere were 366 patients assigned to the neuraxial analgesiaarm and 362 to the systemic analgesia arm of the originalstudy. Data from 343 of the neuraxial group subjects and349 of the systemic group subjects are included in thepresent analysis (the remainder of subjects had protocolviolations or missing data). Comparison of demographicdata between the DAY versus NIGHT groups is presentedin Table 1. A significantly larger proportion of DAY sub-jects than NIGHT subjects were receiving oxytocin at thetime of analgesia in both the neuraxial and systemic arms ofthe study. DAY subjects in both analgesic arms of the studyhad a shorter analgesia to delivery time interval thanNIGHT subjects.

Outcome data are presented in Table 2, and scatter plotsand smoother curves for duration of analgesia, cervicaldilation, and pain scores are shown in Figures 1–4. Therewas no difference in the median duration of eitherneuraxial or systemic analgesia in DAY versus NIGHTsubjects (Table 2). No harmonic variation was apparent foreither neuraxial or systemic analgesia duration (Fig. 1).Despite a similar median value, there was a difference indistribution of cervical dilation at first analgesia requestbetween the DAY and NIGHT subjects who receivedneuraxial analgesia. Subjects who received systemic anal-gesia did not demonstrate the same variation. Rhythmanalysis of all subjects demonstrated a 24-h harmonic cyclefor cervical dilation at first analgesia request (Figs. 2 and 5).

*Friedman JH. A variable span smoother: technical report LC55. Palo Alto,CA: Department of Statistics, Stanford University, 1984.†Available at: http://www.ma.uni-heidelberg.de/inst/phar/forschungLemmer.html. 2004. Accessed February 8, 2008.


There was no difference observed between DAY andNIGHT subjects regarding cervical dilation at second anal-gesia request, and no harmonic variation was observed forthis variable (Fig. 3).

The distribution of pain score1st request was differentduring the night period compared with the day in subjectswho received neuraxial analgesia despite a similar medianvalue. Subjects who received systemic analgesia did notdemonstrate a difference. Pain score1st request did not exhibitany harmonic variation (Fig. 2). Median pain scoreaverage didnot differ between DAY and NIGHT subjects in either theneuraxial or systemic arms of the study; however, a 24-hharmonic variation was observed for pain scoreaverage in theneuraxial group (Figs. 4 and 5); no harmonic variation wasobserved among subjects in the systemic group (Fig. 4).

DISCUSSIONTime of day of administration of analgesia did not seem toinfluence combined spinal-epidural or systemic labor anal-gesia duration under the conditions of this study. We didobserve a harmonic cycle in the cervical dilation at whichlabor analgesia was first requested; however, the smallamplitude associated with the harmonic variation and thesmall r2 value imply that any clinical implications would be

minimal. We also observed harmonic variations in theaverage amount of pain experienced between intrathecalfentanyl administration and second analgesia request,again, with small amplitude and r2 value.

Biologic rhythms are both genetically determined andinfluenced by environmental factors known as synchroniz-ers, such as the light-dark cycle, eating, and fasting periods.In humans, the main circadian (referring to a 24-h cycle)pacemaker is located in the hypothalamus. It is modulatedby the external environment, receiving inputs from theretina, as well as nonphotic synchronizers, such as thoseinvolved with locomotor activity and eating. Certain drugsalso influence the pacemaker. The output from this pace-maker in turn influences a myriad of biologic activities,such as melatonin production, cortisol secretion, adreno-corticotropin releasing hormone secretion, and sensitivityof end organs to adrenocorticotropin releasing hormone.Organs such as the heart and liver, and even isolated cells,demonstrate their own circadian rhythms independent ofthe central circadian pacemaker.1

It can be difficult to separate true time of day effectsfrom external factors that may influence circadian humanbehaviors. Our study subjects demonstrated a minimumcervical dilation at first analgesia request in the very early

Table 1. Subject Demographic CharacteristicsNeuraxial analgesia

P

Systemic analgesia

P

Time of analgesia administration Time of analgesia administration

07:01–23:00 h 23:01–07:00 h 07:01–23:00 h 23:01–07:00 hAge (yr) 32 (29–35) 31 (28–35) 0.34 31 (28–35) 32 (30–35) 0.22Height (cm) 165 (160–170) 163 (160–169) 0.58 165 (160–170) 163 (160–170) 0.61Weight (kg) 77 (69–88) 77 (70–85) 0.78 79 (71–88) 76 (69–87) 0.09Gestational age (wk) 39 (38–40) 40 (39–40) 0.37 40 (39–40) 39.5 (39–40) 0.63Oxytocin administered at

time of firstanalgesia, n (%)

180/227 (79) 66/110 (60) �0.001 196/240 (82) 57/99 (58) �0.001

Time to delivery (min) 419 (295–562) 491 (314–651) 0.03 505 (345–653) 559 (276–575) 0.03Infant weight (g) 3490 (3168–3768) 3415 (3136–3679) 0.42 3410 (3120–3665) 3450 (3185–3732) 0.17Vaginal delivery (%) 83 82 0.89 80 76 0.40

Table 2. Outcome VariablesNeuraxial analgesia

P

Systemic analgesia

P

Time of analgesia administration Time of analgesia administration

07:01–23:00 h 23:01–07:00 h 07:01–23:00 h 23:01–07:00 hDuration of analgesia (min) 95 (73–117) 92 (75–125) 0.36 108 (80–143) 105 (80–145) 0.85Cervical dilation at 1st analgesia

request (cm)2 (1.5–3) 2 (1–3) 0.001 2 (1–3) 2 (1–3) 0.25

Distribution, n (%)�1.5 cm 65 (27) 48 (39) 97 (38) 55 (50)�1.5 to �3.0 cm 80 (33) 50 (49) 83 (33) 28 (26)�3.0 cm 97 (40) 26 (21) 73 (29) 26 (24)

Pain score1st request 8 (7–9) 8 (7–9) 0.02 8 (7–9) 8 (7–9) 0.36Distribution, n (%)

0–3 3 (2) 2 (2) 7 (3) 5 (5)4–7 93 (39) 32 (26) 92 (38) 34 (31)8–10 140 (59) 88 (72) 142 (59) 69 (64)

Cervical dilation at 2ndanalgesia request (cm)

4 (3–5) 4 (3–5) 0.05 4 (3–5) 4 (2–5) 0.36

Pain scoreaverage 2 (1–3) 2 (0–3) 0.33 6 (4–7) 5 (4–7) 0.11

Opioid Labor Analgesia and Time of Day


morning hours, perhaps indicating increased early laborpain perception during that time interval. Also, patients inthe neuraxial arm of the study had higher pain scores atfirst analgesia request between the hours of 23:01 and 07:00than between 07:01 and 23:00. These findings are consistentwith a previous study in which nulliparous women inspontaneous labor with cervical dilation between 3 and 5cm and ruptured membranes reported lower pain scoresbetween 07:00 and 13:00 h than at other times of the day.7

This circadian rhythmicity to pain perception could beeither attributable to true time of day effects or influencedby external factors. For instance, it could be explained bythe circadian rhythm associated with �-endorphin secre-tion (highest in the morning).8 However, it also could beexplained by the negative effects that factors such as sleepdeprivation may have on pain perception. Alternatively,subjects who experience labor pain during the day may bemore likely to go to the hospital when they are in lessdistress than patients who experience that pain at night.Although we did not observe any time-related differencesin analgesia duration, we did find some harmonic variationin effectiveness of intrathecal opioid analgesia as measuredby pain scoreaverage, which showed a peak in the lateevening and a nadir in the morning. Once again, our study

design does not distinguish between true time of dayexplanations for this versus external factors that influencehuman behavior.

To our knowledge, there have been no previous studiesexamining the effect of time of administration on systemicopioid labor analgesia. Several authors have demonstratedtime-related effects for neuraxial labor analgesia. Epiduralropivacaine had a longer duration between 07:00 and 13:00h compared with other times of the day.2 Interestingly, thisis the same time interval associated with lowest labor painscores.7 Intrathecal opioids given for labor analgesia alsoexhibited chronobiologic characteristics. The duration ofintrathecal fentanyl 20 �g was longer between 12:00 and18:00 h than between 20:00 and 02:00 h.4 The duration ofintrathecal sufentanil 10 �g followed a 12-h cycle, withpeaks at 12:00 and 00:00 h.3

In contrast to these previous studies, we did not observeany harmonic cycle in the duration of labor analgesia. Onepossible explanation for the difference in our results fromprevious studies is that our patients were recruited in earlylabor, with a median cervical dilation of 2 cm, whereasthose in other studies were in more advanced labor. It ispossible that chronopharmacologic analgesic effects aremore likely to be observed as labor progresses and painworsens. Another explanation may lie in the fact that theanalgesic doses differed. Not only did subjects in this study

Figure 1. Analgesia duration versus the time of administration ofanalgesia (top panel � neuraxial group; bottom panel � systemicgroup). Black horizontal line � mean analgesic duration; blue line �lowess smoother; green line � loess smoother; and red line �supersmoother. Diamonds denote the maximum and minimumvalues for each smoother of the same color.

Figure 2. Cervical dilation (top panel) and pain score at first requestfor analgesia (bottom panel) versus time of administration of anal-gesia. Black horizontal line � mean cervical dilation/pain score; blueline � lowess smoother; green line � loess smoother; and redline � supersmoother. Diamonds denote the maximum and mini-mum values for each smoother of the same color.


receive a relatively high dose of opioid but they alsoreceived a lidocaine/epinephrine epidural test dose. Noone has studied the time of day effect of a neuraxial mixtureof opioid and local anesthetic. It is possible that pharmaco-logic rhythms are less likely to be observed with higherdoses of drugs or mixtures of drugs. Our results areconsistent with a study examining chronic neuropathicpain, which demonstrated chronobiology to pain percep-tion but not to analgesic efficacy.9

The number of subjects in previous studies was small(largest n � 194) compared with this study. Additionally, inprevious studies, explaining the study objective to thesubject may have influenced the subject’s perception ofanalgesic efficacy. It is well established that a significantplacebo effect can be measured in studies involving anal-gesia administration.10 Pain scores in the previous studieswere assessed at regular intervals after the initial analgesicdose, and the presence of an investigator assessing painmay have influenced when the subject requested additionalanalgesia. Subjects in previous studies were recruited overa relatively brief interval (months) compared with ourstudy (years), and it is possible that this may also haveinfluenced results.

Finally, none of the studies, including ours, controlledfor other factors that may influence pain and analgesia,including presence or absence of medical personnel, laysupport people or family members, room lighting, sleepdeprivation before enrollment in the study, time since lastsleep, sleep during the study period, the subjects’ normalactivity and sleep cycles, and other unknown factors. Wemanage a high volume of patients (approximately 10,000deliveries per year) on our delivery unit, and therefore, thelight and noise levels at night may differ from those ofsmaller units. It would be desirable to control variablessuch as these in a study specifically investigating theinfluence of chronobiology on labor pain and analgesia.

There are other limitations to this study. Variables suchas the administration of oxytocin were not controlledresulting in a larger fraction of subjects during the dayreceiving oxytocin at the initiation of analgesia. It is pos-sible that both nurses and obstetricians are less likely toaugment labor at night because of inconvenience or otherfactors; alternatively, patients who come to the hospital atnight may be a class of patients in more active labor (inmore pain even at low cervical dilations) and therefore inless need of oxytocin augmentation. It is possible that theincreased use of oxytocin during the day masked time-related effects that would have been evident had this

Figure 3. Cervical dilation at second request for analgesia (toppanel � neuraxial group; bottom panel � systemic group). Blackhorizontal line � mean cervical dilation; blue line � lowesssmoother; green line � loess smoother; and red line � super-smoother. Diamonds denote the maximum and minimum values foreach smoother of the same color.

Figure 4. Average pain score between first and second analgesiaadministration (top panel � neuraxial group; bottom panel � sys-temic group). Black horizontal line � mean pain score; blue line �lowess smoother; green line � loess smoother; and red line �supersmoother. Diamonds denote the maximum and minimumvalues for each smoother of the same color.

Opioid Labor Analgesia and Time of Day


variable been controlled. In addition, more subjects wererecruited during the day than at night, and this may haveconfounded results. Patients at night may be different thanpatients during the day (for instance, a group in more painand therefore more “motivated” to travel to the hospital atnight).

Our aim in this analysis was not to study the chronop-harmacology of systemic hydromorphone and intrathecalfentanyl per se, but to determine whether time-relatedeffects confounded our original study results, as someauthors have suggested they might. Time-related variablesdo not seem to have played a major role in our previousstudy. There were no time-related differences in analgesiaduration, and time-related differences that we did observe,such as differences in cervical dilation and pain scores,were minimal. Furthermore, our study design does notallow us to draw conclusions regarding the cause for theseobserved differences; they may be related to true chrono-biology or to environmental factors that influence humanbehavior. Chronobiology as a possible confounding factorin labor analgesia trials requires further examination.

ACKNOWLEDGMENTSThe authors thank an anonymous reviewer for his insight intothe analysis and presentation of the data in this article.

REFERENCES1. Chassard D, Bruguerolle B. Chronobiology and anesthesia.

Anesthesiology 2004;100:413–272. Debon R, Chassard D, Duflo F, Boselli E, Bryssine B,

Allaouchiche B. Chronobiology of epidural ropivacaine:variations in the duration of action related to the hour ofadministration. Anesthesiology 2002;96:542–5

3. Debon R, Boselli E, Guyot R, Allaouchiche B, Lemmer B,Chassard D. Chronopharmacology of intrathecal sufentanilfor labor analgesia: daily variations in duration of action.Anesthesiology 2004;101:978 – 82

4. Pan PH, Lee S, Harris L. Chronobiology of subarachnoidfentanyl for labor analgesia. Anesthesiology 2005;103:595–9

5. Wong CA, Scavone BM, Peaceman AM, McCarthy RJ, SullivanJT, Diaz NT, Yaghmour E, Marcus RJ, Sherwani SS, SprovieroMT, Yilmaz M, Patel R, Robles C, Grouper S. The risk ofcesarean delivery with neuraxial analgesia given early versuslate in labor. N Engl J Med 2005;352:655–65

6. Cleveland WS, Devlin SJ. Locally weighted regression: an ap-proach to regression analysis by local fitting. JASA 1988;83:596–610

7. Aya AG, Vialles N, Mangin R, Robert C, Ferrer JM, Ripart J, deLa Coussaye JE. Chronobiology of labour pain perception: anobservational study. Br J Anaesth 2004;93:451–3

8. Lindow SW, Newham A, Hendricks MS, Thompson JW, van derSpuy ZM. The 24-hour rhythm of oxytocin and beta-endorphinsecretion in human pregnancy. Clin Endocrinol (Oxf) 1996;45:443–6

9. Odrcich M, Bailey JM, Cahill CM, Gilron I. Chronobiologicalcharacteristics of painful diabetic neuropathy and postherpeticneuralgia: diurnal pain variation and effects of analgesictherapy. Pain 2006;120:207–12

10. Zubieta JK, Yau WY, Scott DJ, Stohler CS. Belief or need?Accounting for individual variations in the neurochemistry ofthe placebo effect. Brain Behav Immun 2006;20:15–26

Figure 5. Harmonic cycles (a) identified by fitting the equation (b).Upper panel: pain scoreaverage, neuraxial group. Orange line �predicted values from function with periodi � 24 h, mesor � 2,amplitudei � 0.6, acrophasei � 21.3 h (r2 � 0.012, P � 0.005),with maximum values occurring near 22:00 h and minimum valuesnear 10:00 h. Lower panel: cervical dilation at first request foranalgesia. Orange line � predicted values from function with pe-riodi � 24 h, mesor � 2 cm, amplitudei � 0.1 cm, acrophasei � 16.4 h(r2 � 0.011, P � 0.02), with maximum values occurring near 17:00 h,and minimum values near 05:00 h.


CASE REPORT

Epidural Hematoma Nine Days After Removal of aLabor Epidural CatheterPatrick J. Guffey, MD, Warren R. McKay, MD, and Rachel Eshima McKay, MD

Timely recognition and surgical decompression are crucial to minimize risk of permanentneurologic deficit from epidural hematoma. We present the case of a patient who developedacute back pain, sensory deficit, and ascending weakness 9 days after removal of a laborepidural catheter. Magnetic resonance imaging revealed a heterogeneous fluid collectionextending from C6-7 through the lumbar region, with cord deformity at T9-11. Decompressionlaminectomy was performed within 4 hours of symptom onset. Twelve hours later, her motorfunction had fully recovered. Subsequent anatomic and hematologic workup was inconclu-sive. This presentation is atypical given the delayed presentation of symptoms after epiduralplacement. (Anesth Analg 2010;111:992–5)

Epidural hematoma related to neuraxial anesthesia isa rare but potentially devastating complication; pub-lished reports in patients undergoing epidural cath-

eter placement estimate an incidence of about 1 case in150,000 anesthetics.1–3 Most often, symptoms arise within afew hours after placement or removal of the catheter.3

Cases of spontaneous epidural hematoma occurring inpregnancy are exceedingly rare; only 6 have been re-ported in the English-language literature, only 1 of whichoccurred after delivery.3–7 We found no previous casereports of epidural hematoma presenting more than 1week after catheter removal. We report a case of anepidural hematoma in a previously healthy woman whopresented 9 days after labor and vaginal delivery withepidural analgesia. Written informed consent was ob-tained from the patient before preparation and submis-sion of the manuscript.

CASE DESCRIPTIONA 32-year-old, G1P0 woman with body mass index of 27kg/m2 underwent epidural catheter placement during la-bor. Localization of the epidural space required 2 passes inthe midline of the L2-3 interspace with an 18-gauge Tuohyneedle. After the epidural space was identified by loss ofresistance to saline at 7 cm from the skin, a flexible,wire-embedded catheter (FlexTip Plus®; Arrow Interna-tional, Reading, PA) was passed without resistance, pares-thesia, or discomfort to a distance 6 cm beyond the tip ofthe needle. The catheter was secured at 13 cm at the skin.Aspiration through the catheter with a 3-mL syringe failedto yield blood or fluid. Lidocaine 45 mg and epinephrine 15�g were administered without change in heart rate orinitiation of motor or sensory loss. Subsequent analgesia

was effective, and the remainder of labor and delivery (5hours) progressed without complication. Shortly after de-livery, the patient developed profuse vaginal bleeding andwas taken to the operating room for cervical dilation andcurettage; surgical anesthesia was successfully adminis-tered via the in situ epidural catheter. The cause of thebleeding was not clearly attributed to retained placentalfragments or cervical tear. Total estimated blood loss was750 mL, and 2.5 L lactated Ringer solution was adminis-tered. After the procedure, the epidural catheter was re-moved, and the patient was discharged home 2 days laterin excellent condition without complaints.

Nine days later, she presented to the emergency depart-ment with severe low back pain and acute, ascending lowerextremity paralysis that began 2 hours before arrival.Immediate evaluation by obstetric and anesthesia physi-cians revealed arterial blood pressure of 190/100 mm Hgand lower extremity paralysis and sensory deficit ascend-ing to the umbilicus. Laboratory values included hemato-crit 30%, platelet count 358 � 109/�L, prothrombin time(PT) 14.0 seconds (normal, 12.5–16.0 seconds; internationalnormalized ratio, 1.0), and fibrinogen 236 mg/dL (normal,202–430 mg/dL). Dexamethasone 12 mg was administeredIV and urgent neurosurgical consultation was requested.Magnetic resonance imaging revealed a large, cylindricalheterogeneous fluid collection around the spine extendingfrom C6-7 through the lumbar region, with significant masseffect, cord deformity, and T2 signal prolongation at T9-11;the majority of the blood had collected between T6 and T12(Fig. 1). Within 2 hours of her arrival at the hospital, shewas taken to the operating room to undergo decompressionsurgery. Baseline laboratory values revealed a hematocritof 28%, and platelet count, PT, and partial thromboplastintime within normal limits. Neither thromboelastographynor platelet function assay was available in our institution.Because of ongoing blood loss and oozing in the field, 1pheresis pack of platelets and 4 U fresh frozen plasma wereadministered, resulting in satisfactory hemostasis. Threeunits of packed red blood cells were administered duringsurgery, and at the conclusion of the 4-hour case, blood losswas estimated to be 1.5 L. The paralysis resolved com-pletely over the next 12 hours, and the sensory examinationreturned to normal over the next week with the exceptionof an area of paresthesia along the medial aspect of herlower right medial leg.

From the Department of Anesthesia and Perioperative Care, University ofCalifornia San Francisco, San Francisco, California.




Address correspondence to Rachel Eshima McKay, MD, Department ofAnesthesia and Perioperative Care, University of California San Francisco,521 Parnassus Ave., C450, San Francisco, CA 94143-0648. Address e-mail [email protected].

Copyright © 2010 International Anesthesia Research SocietyDOI: 10.1213/ANE.0b013e3181effd8f


Neuroangiographic evaluation for spinal arteriovascularmalformation was negative. A subsequent detailed history,performed in conjunction with the hematology service, didnot reveal personal or family history of abnormal bleeding.The patient was noted to have used ibuprofen 800 mg every8 hours for 9 days after delivery for low back and general-ized musculoskeletal pain. Further laboratory workup con-ducted 2 weeks after hematoma evacuation included aristocetin cofactor assay of 88% (reference range, 42%–200%),fibrinogen of 301 mg/dL (normal range, 202–430 mg/dL),and factor VIII activity of 162% (normal, 43%–168%). Thesevalues were all within 2 standard deviations of meanvalues at our institution for a patient who is not pregnant.At 2-month follow-up, the patient had made a completeneurologic recovery.

DISCUSSIONSpinal epidural hematoma is a rare but potentially devas-tating event, classified as spontaneous (occurring withoutapparent cause or with delayed onset after minor injury), orsecondary to identifiable cause, including spinal or epi-dural anesthesia. Patients at higher risk for hematoma afterepidural anesthesia include those with advanced age, spi-nal stenosis, coagulopathy (longstanding or acute), plateletinhibition, arteriovenous malformation, or multiple punc-tures.8 Overall, hematomas related to neuraxial anesthesiaseem to be less frequent in the obstetrical compared withthe elderly surgical population.2 There are also rare reportsof spontaneous epidural hematoma in parturients thatoccur in the absence of neuraxial block, and delay inrecognition and management has resulted in devastatingconsequences.2,9 This particular case was unusual becauseit occurred more than 1 week after epidural catheterremoval, beyond the timeframe previously described afteran inciting event.10

A systematic review of 613 reported cases of spinal andepidural hematoma demonstrates that trauma, neuraxialanesthesia, coagulopathy, arteriovenous malformation, or acombination of these factors, are frequently associated withhematoma; 30% were found to have no identifiable cause.3

In all, 63 cases were associated with neuraxial anesthesia.Of these 63 cases, coagulation status was known in 49; 34 of

49 had confirmed coagulopathy or were receiving concur-rent anticoagulation therapy; and 41 of these 49 patientsdeveloped symptoms within 72 hours of the neuraxialprocedure (Fig. 2).3 Five pregnant patients were among thisgroup of 63; 1 had an elevated PT from hepatic diseaserelated to the pregnancy, another was classified as hyper-tensive, and the remaining 3 had no identifiable riskfactors.

What may have put our patient at risk? The presence ofpostpartum hemorrhage may have been related to numer-ous factors other than coagulation disorder, but could alsohave been related to a subtle disorder of coagulation thatwas not detected during her workup after surgery.11,12

Although the values of her PT, fibrinogen, and subsequentristocetin cofactor assay and factor VIII activity were within

Figure 1. A, Sagittal T1-weighted magnetic resonance image showing presence of extensive hematoma around the spinal cord, visible herefrom C7 to T12 and below. Blood here is visible anterior to the cord, and compression is visible most prominently between T9 to T12. B, AxialT2-weighted magnetic resonance image showing hematoma surrounding the spinal cord at the level of T12. C, Sagittal T2-weighted magneticresonance image demonstrating extension of blood throughout the thoracic and lumbar region (T12 to sacrum shown).

Figure 2. Graph showing hours between initiation of neuraxialprocedure (lumbar puncture or spinal/epidural anesthesia) andsymptoms of epidural hematoma. Of the 63 cases of epiduralhematoma associated with spinal or epidural anesthesia among613 reported, 49 had documented evidence confirming or refutingpresence of antiplatelet or anticoagulant medication use, coagulopa-thy, or platelet abnormality. (The graph was constructed from datafrom Kreppel et al.3)

Hematoma After Labor Epidural Removal


the normal range for a nonpregnant patient, these mayhave been falsely normal given the typical changes incoagulation state that accompany pregnancy and the earlypostpartum period, when circulating levels of clottingfactors may be 20% to 200% above nonpregnant levels anddo not fully return to nonpregnancy baseline values until 8weeks postpartum.13

Additionally, we questioned whether the hematomacould have been related to the patient’s nonsteroidal anti-inflammatory drug (NSAID) use, or to NSAIDs in combi-nation with a subtle impairment in coagulation function.Antiplatelet drugs, including nonselective NSAIDs andaspirin, taken within 1 week of surgery, have been associ-ated with increased risk of hematoma after craniotomy andhip arthroplasty.14,15 However, there is consensus amongexperts that NSAID use does not increase risk of epiduralhematoma. This consensus is based on prospective studiesof hundreds of NSAID users undergoing neuraxial anes-thesia8 and epidural steroid injection.11 Use of throm-boelastography or other platelet function assays at the timeof presentation might have established whether NSAID usecontributed to the development of this epidural hematoma.A more thorough hematologic evaluation, performed fur-ther out from the peripartum period, would be needed toexclude an underlying bleeding disorder; to date, thepatient has declined further workup.

In either case, it is possible that inadequate hemostasiscaused gradual expansion of a small collection of blood thatwas initiated by epidural placement or removal. Over days,this small asymptomatic hematoma may have slowly con-tinued to bleed and expand, eventually reaching a criticalsize sufficient to cause dramatic symptoms. The predomi-nantly thoracic location of the hemorrhage, higher than thesite of epidural puncture, does not argue against theepidural catheter as the primary cause. First, the cathetertip may have extended a considerable distance in a cepha-lad direction, and catheter removal may have initiatedtrauma at that location. Second, blood from a hematomasecondary to needle trauma in the lumbar region may havecollected preferentially in the thoracic epidural space be-cause of anatomic factors such as the relatively narrowerdimension of the spinal cord, convexity of the thoracicspine, the relatively lower intrathoracic pressure, andgreater capacity of the thoracic epidural space.16,17

Given the prolonged period of time between cathetermanipulation and presentation of symptoms, it is possiblethat this patient’s condition resulted from factors other thanepidural catheter placement. However, we think it morelikely that the event was related to epidural catheterplacement/removal, with possible contributing factors in-cluding elevated venous pressure related to the pregnancy,platelet inhibition from persistent NSAID use, and hyper-tension. The relative contribution of these various factors, ifany, however, remains undetermined.

Pregnancy is characterized by a relatively hypercoagu-lable state, effectively reducing the risk of epidural hema-toma. However, the epidural space houses an extensivevenous system containing tributaries from the spinal cordand vertebral bodies that drain into the external vertebralvenous plexus and become more engorged during preg-nancy. Because this venous system does not contain valves,

and exists within a low-pressure environment, any pres-sure exerted proximally because of a Valsalva maneuver,vomiting, or mechanical compression of the vena cava candramatically increase the transmural venous pressure, leav-ing the veins vulnerable to injury. It has been postulatedthat spontaneous epidural hematoma may occur frompreexisting faults in the venous wall exacerbated by well-described mechanical and hyperdynamic conditions exist-ing in the peripartum period.4–7,18

Numerous studies have demonstrated the most favor-able outcomes when a symptomatic hematoma is decom-pressed within 36 hours, and some authors suggest evenfaster intervention, within 6 hours.19,20 In this case, inter-vention occurred within 4 hours of the onset of symptoms.Through emergent, definitive treatment, the patient made afull recovery, illustrating the importance of prompt recog-nition and treatment of epidural hematoma to maximizethe patient’s chance of a favorable outcome.

AUTHOR CONTRIBUTIONSPJG helped with patient consent, data collection, and manu-script preparation; WRM helped with manuscript preparation;and REM helped with data collection, manuscript preparation,and construction of figures.

REFERENCES1. Loo CC, Dahlgren G, Irestedt L. Neurological complications in

obstetric regional anaesthesia. Int Obstet Anesth 2000;9:99–1242. Kopp SL, Horlocker TT. Anticoagulation in pregnancy and

neuraxial blocks. Anesthesiol Clin 2008;26:1–223. Kreppel D, Antoniadis G, Seeling W. Spinal hematoma: a

literature survey with meta-analysis of 613 patients. NeurosurgRev 2003;26:1–49

4. Bidzinski J. Spontaneous spinal epidural hematoma duringpregnancy: case report. J Neurosurg 1966;24:1017

5. Yonekawa Y, Mehdorn HM, Nishikawa M. Spontaneous spinalepidural hematoma during pregnancy. Surg Neurol 1975;3:327–8

6. Carroll SG. Spontaneous spinal extradural hematoma duringpregnancy. J Matern Fetal Med 1997;6:218–9

7. Bose S, Ali Z, Rath P, Prabhakar H. Spontaneous spinalhaematoma: a rare cause of quadriplegia in the post-partumperiod. Br J Anaesth 2007;99:855–7

8. Horlocker TT, Wedel DJ, Schroeder DR, Rose SH, Elliot BA,McGregor DG, Wong GY. Preoperative anti-platelet therapydoes not increase the risk of spinal hematoma associated withregional anesthesia. Anesth Analg 1995;80:303–9

9. Doblar DD, Schumacher SD. Spontaneous acute thoracic epi-dural hematoma causing paraplegia in a patient with severepreeclampsia in early labor. Int J Obstet Anesth 2005;14:256–60

10. Pear BL. Spinal epidural hematoma. Am J Roentgenol RadiumTher Nucl Med 1972;155:155–64

11. Horlocker TT, Bajwa ZH, Ashraf Z, Khan S, Wilson JL, Sami N,Peeters-Asdourian C, Powers CA, Schroeder DR, Decker PA,Warfield CA. Risk assessment of hemorrhagic complicationsassociated with nonsteroidal anti-inflammatory medications inambulatory pain clinic patients undergoing epidural steroidinjection. Anesth Analg 2002;95:1691–7

12. Kadir RA, Kingmam CEC, Chi C, Lee CA, Economides DL. Isprimary postpartum haemorrhage a good predictor of inher-ited bleeding disorders? Haemophilia 2007;13:178–81

13. Bremme KA. Haemostatic changes in pregnancy. Best PractRes Clin Haematol 2003;16:153–68

14. Palmer JD, Sparrow OC, Ianotti F. Postoperative hematoma: a5-year survey and identification of risk factors. Neurosurgery1994;35:1061–5

15. Robinson CM, Christie J, Malcolm-Smith N. Nonsteroidalanti-inflammatory drugs, perioperative blood loss, and trans-fusion requirements in elective hip arthroplasty. J Arthroplasty1993;8:607–10

CASE REPORT


16. Visser WA, Liem TH, van Egmond J, Gielen MJ. Extension ofsensory blockade after thoracic epidural administration of atest dose of lidocaine at three different levels. Anesth Analg1998;86:332–5

17. Igarashi T, Hirabayashi Y, Shimizu R, Saitoh K, Fukuda H.Thoracic and lumbar extradural structure examined by extra-duroscope. Br J Anaesth 1998;81:121–5

18. Beatty RM, Winston KR. Spontaneous cervical epiduralhaematoma: a consideration of etiology. J Neurosurg 1984;61:143–8

19. Groen RJ, van Alphen HA. Operative treatment of sponta-neous spinal epidural hematomas: a study of the factorsdetermining postoperative outcome. Neurosurgery 1996;39:494 –508

20. Lawton MT, Porter RW, Heiserman JE, Jacobowitz R, SonntagVK, Dickman CA. Surgical management of spinal epiduralhematoma: relationship between surgical timing and neuro-logical outcome. J Neurosurg 1995;83:1–7

Hematoma After Labor Epidural Removal


Society for Pediatric Anesthesia

Section Editor: Peter J. Davis

Ipsilateral Transversus Abdominis Plane BlockProvides Effective Analgesia After Appendectomy inChildren: A Randomized Controlled TrialJohn Carney, MB,*† Olivia Finnerty, MB, FCARCSI,*† Jassim Rauf, MB,† Gerard Curley, MB, FCARCSI,*†John G. McDonnell, MB, FCARCSI,*† and John G. Laffey, MD, MA, BSc, FCARCSI*†‡

BACKGROUND: The transversus abdominis plane (TAP) block provides effective postoperativeanalgesia in adults undergoing major abdominal surgery. Its efficacy in children remains unclear,with no randomized clinical trials in this population. In this study, we evaluated its analgesicefficacy over the first 48 postoperative hours after appendectomy performed through an openabdominal incision, in a randomized, controlled, double-blind clinical trial.METHODS: Forty children undergoing appendectomy were randomized to undergo unilateral TAPblock with ropivacaine (n � 19) versus placebo (n � 21) in addition to standard postoperativeanalgesia comprising IV morphine analgesia and regular diclofenac and acetaminophen. Allpatients received a standard general anesthetic, and after induction of anesthesia, a TAP blockwas performed using the landmark technique with 2.5 mg � kg�1 ropivacaine 0.75% or an equalvolume (0.3 mL � kg�1) of saline on the ipsilateral side to the incision.RESULTS: The TAP block with ropivacaine reduced mean (�SD) morphine requirements in thefirst 48 postoperative hours (10.3 � 12.7 vs 22.3 � 14.7 mg; P � 0.01) compared with placeboblock. The TAP block also reduced postoperative visual analog scale pain scores at rest and onmovement compared with placebo. Interval morphine consumption was reduced over the first 24postoperative hours. There were no between-group differences in the incidence of sedation ornausea and vomiting. There were no complications attributable to the TAP block.CONCLUSIONS: Unilateral TAP block, as a component of a multimodal analgesic regimen,provided superior analgesia compared with placebo in the first 48 postoperative hours afterappendectomy in children. (Anesth Analg 2010;111:998–1003)

Appendectomy is one of the most frequently per-formed surgical procedures in children and is asso-ciated with significant postoperative discomfort and

pain.1 Multimodal approaches to the provision of postopera-tive analgesia often incorporate blockade of the abdominalwall, such as ilioinguinal blockade2 or wound infiltration.3 How-ever, the efficacy of these approaches is unclear.3,4 The optimalapproach to the blockade of the abdominal wall in childrenundergoing appendectomy remains to be determined.

An important component of the pain experienced bychildren after appendectomy derives from the abdominalwall incision. The nerves that supply the abdominal wallcourse through the neurofascial transversus abdominis

plane (TAP) between the internal oblique and the transver-sus abdominis muscles.5 Using anatomic studies, our groupidentified the lumbar triangle of Petit as an access point forintroducing local anesthetic drugs into the TAP.6 Thistriangle is bounded posteriorly by the latissimus dorsimuscle, anteriorly by the external oblique, and inferiorly bythe iliac crest.7 The floor of the triangle, from superficial todeep, is composed of subcutaneous tissue, and the fascialborders of the external oblique, the internal oblique, andthe transversus abdominis muscles, respectively.

The efficacy of the TAP block in providing postoperativeanalgesia has been shown in adults undergoing bowelsurgery,8 cesarean delivery,9 and total abdominal hysterec-tomy.9 Most recently, an ultrasound-guided approach tothe TAP block has been described in a series of 8 childrenundergoing inguinal hernia repair10 and has shown anal-gesic efficacy in a randomized controlled trial for appen-dicectomy in adults.11 However, there are no randomizedcontrolled clinical trials demonstrating the efficacy of theTAP block in children. We hypothesized that, comparedwith a placebo block, the TAP block, as part of a multimo-dal analgesic regimen, would result in decreased opioidconsumption and improved analgesia in the first 48 hoursfor children undergoing appendectomy.

METHODSAfter obtaining approval by the Hospital Ethics Committee,and written informed parental consent, we studied children

From the *Department of Anaesthesia, Clinical Sciences Institute, NationalUniversity of Ireland, Galway; and †Department of Anaesthesia and Inten-sive Care Medicine, and ‡Clinical Research Facility, Galway UniversityHospitals, Galway, Ireland.



Funded by departmental resources.



Address correspondence to John G. Laffey, MD, MA, BSc, FCARCSI,Department of Anaesthesia, Clinical Sciences Institute, National Universityof Ireland, Galway, Ireland. Address e-mail to [email protected].

Copyright © 2010 International Anesthesia Research SocietyDOI: 10.1213/ANE.0b013e3181ee7bba


younger than 16 years, ASA grade I to III, scheduled foremergency open appendectomy, in a randomized, double-blind, controlled clinical trial. Patients were excluded ifthere was a history of relevant drug allergy, if they werereceiving medical therapies considered to result in toler-ance to opioids, or if they were deemed to be unable toindependently assess their pain.

Patients were randomly allocated to undergo TAP blockwith 2.5 mg � kg�1 ropivacaine 0.75% (to a maximal dose of150 mg) on the right side or TAP block with an equalvolume (0.3 mL � kg�1) of 0.9% saline. The allocation se-quence was generated by a random number table, andgroup allocation was concealed in sealed, opaque enve-lopes, which were not opened until patient consent hadbeen obtained. The patients, their anesthetists, and staffproviding postoperative care were blinded to group assign-ment. All patients received a standardized rapid sequenceinduction of anesthesia. After application of standardmonitoring and oxygen administration for 3 minutes, an-esthesia was induced with propofol (2–3 mg � kg�1), cricoidpressure was applied and muscle relaxation was producedwith succinylcholine (1–1.5 mg � kg�1), and the trachea wasintubated. Anesthesia was maintained using 1 to 1.5 mini-mum alveolar concentration sevoflurane in oxygen and air.All patients also received IV morphine 0.15 mg � kg�1,rectal diclofenac 1 mg � kg�1, and rectal acetaminophen 20mg � kg�1 immediately before surgical incision. Prophylac-tic antiemetics were not administered.

The TAP block was performed after induction of anes-thesia but before surgical incision by 1 of 2 investigators(JC, OF) using the following technique.8 A 22-gauge 25-mmor 50-mm blunted regional anesthesia needle (Plexufix�; B.Braun, Melsungen AG, Germany) was attached with flex-ible tubing to a syringe filled with the study solution. Withthe patient in a supine position and the investigator stand-ing on the contralateral side, the iliac crest was palpatedfrom anterior to posterior until the latissimus dorsi muscleinsertion was appreciated. The triangle of Petit was pal-pated between the anterior border of latissimus dorsimuscle, the posterior border of the external oblique muscle,and the iliac crest. The skin over the triangle of Petit waspierced with the needle held at right angles to the coronalplane. The needle was stabilized and advanced at rightangles to the skin in a coronal plane until resistance wasencountered. This first resistance indicated that the needletip was abutting the fascial extension of the externaloblique muscle. Further gentle advancement of the needleresulted in a loss of resistance, or “pop” sensation, as theneedle entered the plane between the external and internaloblique fascial layers. Further gentle advancement resultedin the appreciation of a second increased resistance and itsloss indicated entry into the TAP. After careful aspiration toexclude vascular puncture, a test dose of 1 mL was injected.The presence of substantial resistance to this injectionindicates the needle is not between fascial planes and theneedle should be repositioned. The study solution (0.3mL � kg�1 ropivacaine 7.5 mg � mL�1 or saline 0.9%) wasinjected through the needle in increments while observingclosely for signs of toxicity.

After completion of the surgical procedure, patientswere transferred to the postanesthesia care unit. All pa-tients received oral acetaminophen 20 mg � kg�1 every 6hours and rectal diclofenac 1 mg � kg�1 every 12 hourspostoperatively. Children older than 8 years were given IVmorphine patient-controlled analgesia (PCA) (bolus dose20 �g � kg�1; lockout 6 minutes). Children younger than 8years were given nurse-administered IV morphine (20�g � kg�1 bolus) on demand. The presence and severity ofpain, nausea, and sedation were assessed systematically byan investigator blinded to group allocation. These assess-ments were performed in the postanesthesia care unit andat 2, 4, 6, 12, 24, 36, and 48 hours after TAP blockade. Allpatients were asked to give scores for their pain at rest andon movement (knee flexion) and for the degree of nausea ateach time point. Pain severity was measured using a visualanalog scale (VAS) with superimposed faces pain severityscale (10-cm line in which 0 cm � no pain and10 cm �worst pain imaginable) and a categorical pain scoringsystem (none � 0, mild � 1, moderate � 2, and severe � 3).Nausea was measured using a categorical scoring system(none � 0, mild � 1, moderate � 2, and severe � 3). Thepatient was deemed to have been nauseated if they had anausea score �0 at any postoperative time point. Sedationscores were assigned by the investigator using a sedationscale (awake and alert � 0, quietly awake � 1, asleep buteasily roused � 2, and deep sleep � 3). The patient wasdeemed to have been sedated if they had a sedation score�0 at any postoperative time point. Rescue antiemeticswere offered to any patient who complained of nausea orvomiting. The study ended 48 hours after TAP blockade.

The primary outcome measure in this study was 48-hourmorphine consumption. Secondary outcome measures in-cluded time to first request for morphine, VAS scores, andside effects associated with morphine consumption. A pilotstudy of children undergoing open appendectomy found amean 48-hour morphine requirement of 24 mg, with astandard deviation of 8 mg in the control group. Weintended to be able to detect a minimum 33% reduction inmorphine requirement in the patients receiving TAP block-ade. Based on these projections, we calculated that at least17 patients would be required per group for an experimen-tal design incorporating 2 groups, with � � 0.05 and � �0.2. We therefore planned to recruit 40 patients into thestudy.

Statistical analyses were performed using a standardstatistical program (SigmaStat 3.5; Systat Software, SanJose, CA). Demographic data were analyzed using Studentt, �2, or Fisher exact tests as appropriate. The data weretested for normality using the Kolmogorov-Smirnov nor-mality test. Repeated measurements (pain scores, nauseascores) were analyzed by repeated-measures analysis ofvariance where normally distributed, with further pairedcomparisons at each time interval performed using the ttest. For non-normally distributed data, between-groupcomparisons at each time point were made using Wilcoxonrank sum test. Categorical data were analyzed using the �2

analysis or Fisher exact test. The time to first request formorphine was analyzed using the log rank test. Normallydistributed data are presented as mean � SD, non-normallydistributed data are presented as median (interquartile


range), and categorical data are presented as raw data andfrequencies. The � level for all analyses was set at P � 0.05,and the Bonferroni correction for multiple comparisonswas used where appropriate.

RESULTSForty-two children participated in this study. No patientswere excluded on the basis of the exclusion criteria (Ap-pendix Figure; see Supplemental Digital Content 1,http://links.lww.com/AA/A161). Two patients, 1 fromeach group, were excluded after enrollment because ofpostoperative analgesic protocol violations. Of the remain-ing 40 patients, 19 were randomized to undergo TAPblockade with ropivacaine and 21 were randomized to

undergo TAP blockade with normal saline. Groups werecomparable in terms of age, which ranged from 4 to 16years in the TAP group and 5 to 16 years in the controlgroup. There were no differences between the groupsregarding weight and height, history of prior abdominalsurgery, method of postoperative morphine administra-tion, or surgical technique (Tables 1 and 2). A similarnumber of patients in both groups had appendicitis, with14 (74%) in the TAP group versus 14 (67%) in the controlgroup having a histologic diagnosis of appendicitis with orwithout perforation (Table 1). Five patients in the TAPgroup and 4 patients in the control group had histologicevidence of perforation with peritonitis. In all patients, thetriangle of Petit was located easily on palpation, the TAPwas localized after 1 or 2 attempts, and the block performedwithout complication.

Children undergoing unilateral TAP block with ropiva-caine had reduced 48-hour morphine requirements (Fig. 1)and increased time to first requirement for morphine (Fig.2). The total amount of morphine required in the first 48postoperative hours was 10.3 � 12.7 mg in the TAP groupcompared with 22.3 � 14.7 mg (P � 0.01) in the controlgroup. The median (interquartile range) time to first re-quirement for morphine was 55 (30–300) minutes in chil-dren who received a TAP block compared with 16 (7–30)minutes in the control group (P � 0.001). The TAP blockwith ropivacaine reduced cumulative postoperative mor-phine consumption compared with placebo block at alltime points (Fig. 1). Interval morphine consumption wasalso significantly lower at 6, 12, and 24 hours but not at thelater time points in the patients who had TAP blockade(Table 2). There were no differences within either the TAPor control group regarding the amounts of morphinerequired between patients in whom the morphine wasnurse administered versus those who received morphinevia PCA. In addition, morphine consumption within eachgroup did not differ significantly depending on whether ornot the patient had histologic evidence of appendicitis.Postoperative VAS pain scores at rest and on movementwere reduced after TAP block at all time points assessed(Figs. 3 and 4). Categorical pain scores were significantlylower in patients who received the TAP block at some butnot all postoperative time points (data not shown). Painseverity was low in both groups after the first 24 hours, andseveral patients were discharged in both groups between24 and 48 hours.

There was no significant difference in the incidence ofnausea or distribution of nausea scores between groups atany time point (Table 2). The limited nausea experiencedresulted in low scores in both groups; the median andinterquartile scores are shown in Table 2. Six patients (27%)in the control group and 4 patients (16%) in the TAPblockade group developed postoperative nausea. Therewas no significant difference in the incidence of sedation ordistribution of sedation scores between groups at any timepoint (Table 2).

DISCUSSIONOpen appendectomy is one of the most frequently per-formed surgical procedures in the pediatric population

Table 1. Baseline Patient CharacteristicsGroup

CharacteristicControl

(n � 21)TAP block(n � 19)

Weight (kg) 47.5 � 18.6 37.7 � 15.2Height (m) 1.5 � 0.2 1.4 � 0.2Body mass index (kg/m2) 20.2 � 4.9 18.7 � 3.2Prior abdominal surgery �n (%)� 0 (0) 0 (0)Histologic diagnosis

Appendicitis 14 14Appendicitis with perforation

of appendix5 4

TAP � transversus abdominis plane.Continuous data are presented as mean � SD. Categorical variables arepresented as number and proportion.There were no significant differences between groups.

Table 2. Postoperative Analgesia, Nausea,and Sedation

Group

Control(n � 21)

TAP block(n � 19)

Patient-controlled morphineanalgesia

17 14

Nurse-controlled morphineanalgesia

4 5

Interval morphine requirement(�g/kg)

0–6 h 117 (60–203) 55 (10–150)*6–12 h 40 (0–115) 0 (0–17)*12–24 h 60 (19–175) 0 (0–159)*24–36 h 0 (0–40) 0 (0–0)36–48 h 0 (0–0) 0 (0–0)

Postoperative nausea scores0–6 h 0 (0–0) 0 (0–0)6–12 h 0 (0–0) 0 (0–0)12–24 h 0 (0–0) 0 (0–0)24–36 h 0 (0–0) 0 (0–0)36–48 h 0 (0–0) 0 (0–0)

Postoperative sedation scores2 h 1 (1–1) 1 (0.5–1)4 h 1 (0.25–1) 1 (0–1)6 h 1 (0–1) 1 (0–1)12 h 1 (0–2) 1 (0–1)24 h 1 (0–1) 0 (0–0)36 h 0 (0–0) 0 (0–0)48 h 0 (0–0) 0 (0–0)

TAP � transversus abdominis plane.The data are presented as medians (interquartile range).*P � 0.05 (control versus TAP block).

Transversus Abdominis Plane Block in Children


worldwide and is a cause of significant pain in the postop-erative period.1 The optimal analgesic regimen shouldprovide safe, effective analgesia, with minimal side effectsfor the child. A multimodal analgesic regimen is most likelyto achieve these goals; however, the optimal componentsremain to be determined. The TAP block provides blockadeof nociception from the abdominal wall; however, there isalso nociceptive input from the abdominal organs and theonset of the block is not immediate. Therefore, the block isused as part of a multimodal approach. This trial demon-strated that an ipsilateral TAP block provides effectiveanalgesia in children undergoing open appendectomy.

A TAP block on the side of the surgical incision reducedoverall postoperative morphine requirements by approxi-mately 50% and the interval morphine requirements for up

to 24 hours after surgery. Thereafter, morphine require-ments were very low in both groups, with no morphineconsumed over the second 24 postoperative hours in anypatient studied. The TAP block delayed the time to firstrequest for supplemental opioid analgesia and reducedpain scores at rest and on movement. Because the surgerywas 1 sided, the dose of local anesthetic used was limited to2.5 mg � kg�1 on the ipsilateral side, which is within therecommended safe dose range. Alternative abdominal tech-niques, such as the ilioinguinal iliohypogastric nerve block,have been performed in children at a dose of 5 mg � kg�1

ropivacaine without central nervous system or systemictoxicity.12 The reduction in opioid use, coupled with thereduction in postoperative pain, highlights the potential ofthe TAP block in children undergoing appendectomy.

Figure 1. A box plot of postoperative cumulativemorphine consumption in each group in the first 48postoperative hours. The middle line in each boxrepresents the median value, the outer margins ofthe box represent the interquartile range, and thewhiskers represent the 10th and 90th centile foreach time point. *Significantly (P � 0.05, Wilcoxonrank sum test) higher morphine consumption com-pared with the transversus abdominis plane (TAP)block group.

Figure 2. A Kaplan-Meier graph of the proportion ofpatients in each group over time that did not requiresupplemental morphine (P � 0.004, log rank test).


These findings, when taken in context with the demon-strated efficacy of the TAP block in adults,6,8–10 attest thatthe TAP block may provide effective postoperative analge-sia for a wide variety of abdominal procedures in bothchildren and adults.

We used the landmark-based technique for the TAPblock, which was performed without difficulty in thechildren in this study. Alternative approaches to the TAPblock using ultrasound guidance have recently been de-scribed in a case series of children undergoing inguinalhernia repair,10 in adults for appendicectomy,12 and anadult study for laparoscopic cholecystectomy.11 The opti-mal approach remains to be demonstrated. Other regionalanesthesia–based approaches for analgesia after appendec-tomy include ilioinguinal/iliohypogastric nerve blockadeand local wound infiltration. However, in the case ofilioinguinal/iliohypogastric block, the landmark technique

has been shown to be relatively unreliable, producingdeposition of local anesthetic solution in close proximity tothe nerves in as few as 14% of children.4 This may beimproved by the use of ultrasound to guide needle posi-tion.4 Potentially serious complications, such as bowelwall hematoma,13 have been reported after the use ofilioinguinal/iliohypogastric blocks in children. In our ex-perience, performing the TAP block in a pediatric popula-tion has been technically easier than in adults because thedegree of obesity is usually less than that of an adultpopulation, there is a lesser degree of muscle laxity, and the1-in. Plexufix needle that is available allows easier percep-tion of the loss of resistance as described. Local infiltrationof the appendectomy wound with local anesthetic drugs isalso widely practiced. However, the efficacy of this ap-proach is also unclear, with studies reporting no demon-strable reduction in morphine consumption compared with

Figure 3. Box plots of postoperative visual analogscale (VAS) pain scores at rest in each group overthe first 48 postoperative hours. The middle line ineach box represents the median value, the outermargins of the box represent the interquartilerange, and the whiskers represent the 10th and90th centile for each time point. *Significantly (P �0.05, Wilcoxon rank sum test) higher VAS painscores compared with the transversus abdominisplane (TAP) block group.

Figure 4. Box plots of postoperative visual analogscale (VAS) pain scores on movement in each groupover the first 48 postoperative hours. The middleline in each box represents the median value, theouter margins of the box represent the interquartilerange, and the whiskers represent the 10th and90th centile for each time point. *Significantly (P �0.05, Wilcoxon rank sum test) higher VAS painscores compared with the transversus abdominisplane (TAP) block group.

Transversus Abdominis Plane Block in Children


standard care.3 Indeed, the technique has demonstrated tobe inconsistent with a lack of evidence after surgery on theabdominal wall apart from herniorrhaphy.14

There are now a variety of techniques for the TAP blockand the analgesic merit of each is being elucidated inongoing studies. Although it is possible to ultrasonicallyvisualize the 3 muscle layers of the abdominal wall, there isvariation in these muscle layers that can restrict the use ofultrasound over the triangle of Petit.15 As a result, theneedle insertion point as described in the ultrasound stud-ies, which is dependent on the adequate identification ofthe 3 muscle layers, can vary. This will alter the location ofthe injectate as will the angle of the needle insertion to skin,which contrasts to the landmark approach’s description.Anteriorly placed injectate may not spread throughout theTAP and does not according to ongoing work at thisinstitution. Although there is an ever-increasing access toultrasound, it is far from universal and there is a continuinginterest in landmark techniques.

There are a number of limitations to this study. First,there are difficulties in adequately blinding these types ofstudies because of the loss of sensation of the abdominalwall. Although patients and the investigator conducting thepostoperative assessments were technically blinded togroup allocation, true blinding may not have been possible.Second, 2 methods of postoperative administration of mor-phine were used: patients younger than 8 years receivednurse-administered morphine whereas children older than8 years received PCA morphine. However, there were nodifferences in the proportions of each method between thegroups, so these differences do not seem to have contrib-uted to the reported findings. Third, rates of appendicitis inour study (66%–74%) are somewhat lower than those ob-tained internationally.16 This likely reflects a lower thresholdfor operative intervention in this pediatric population, asevidenced by lower rates of appendiceal perforation in ourpatients compared with those reported internationally.16

Fourth, the study was not large enough to independentlyassess safety. There is a risk of inadvertent peritonealpuncture with this block as with other abdominalblocks.17,18 Although the incidence is not known, if theblock is performed as described, the risk of peritonealpuncture is likely to be low. This institution has notencountered complications relating to peritoneal puncture.Finally, a dose-response study has not yet been performedto determine whether effective postoperative analgesiacould be produced with a lower dose of ropivacaine.

In conclusion, the TAP block holds considerable promiseas part of a multimodal analgesic regimen for postappen-dectomy analgesia. With experience, the TAP block is easyto perform, has provided reliable and effective analgesia inthis study, and no complications from the TAP block weredetected.

ACKNOWLEDGMENTSThe authors thank the nursing staff on the pediatric ward fortheir assistance with this study.

REFERENCES1. Hale DA, Molloy M, Pearl RH, Schutt DC, Jaques DP. Appen-

dectomy: a contemporary appraisal. Ann Surg 1997;225:252–612. Willschke H, Marhofer P, Bosenberg A, Johnston S, Wanzel O,

Cox SG, Sitzwohl C, Kapral S. Ultrasonography forilioinguinal/iliohypogastric nerve blocks in children. Br JAnaesth 2005;95:226–30

3. Jensen SI, Andersen M, Nielsen J, Qvist N. Incisional localanaesthesia versus placebo for pain relief after appendectomyin children: a double-blinded controlled randomised trial. EurJ Pediatr Surg 2004;14:410–3

4. Weintraud M, Marhofer P, Bosenberg A, Kapral S, WillschkeH, Felfernig M, Kettner S. Ilioinguinal/iliohypogastric blocksin children: where do we administer the local anestheticwithout direct visualization? Anesth Analg 2008;106:89–93

5. Netter FH. Back and spinal cord. In: Netter FH, ed. Atlas ofHuman Anatomy. Summit, NJ: Ciba-Geigy Corporation,1989:145–55

6. McDonnell JG, O’Donnell BD, Farrell T, Gough N, Tuite D,Power C, Laffey JG. Transversus abdominis plane block: acadaveric and radiological evaluation. Reg Anesth Pain Med2007;32:399–404

7. Netter FH. Abdomen posterolateral abdominal wall. In: NetterFH, ed. Atlas of Human Anatomy. Summit, NJ: Ciba-GeigyCorporation, 1989:230–40

8. McDonnell JG, O’Donnell B, Curley G, Heffernan A, Power C,Laffey JG. The analgesic efficacy of transversus abdominisplane block after abdominal surgery: a prospective random-ized controlled trial. Anesth Analg 2007;104:193–7

9. Carney J, McDonnell JG, Ochana A, Bhinder R, Laffey JG. Thetransversus abdominis plane block provides effective postop-erative analgesia in patients undergoing total abdominal hys-terectomy. Anesth Analg 2008;107:2056–60

10. Fredrickson M, Seal P, Houghton J. Early experience with thetransversus abdominis plane block in children. Paediatr An-aesth 2008;18:891–2

11. Niraj G, Searle A, Mathews M, Misra V, Baban M, Kiani S,Wong M. Analgesic efficacy of ultrasound-guided transversusabdominis plane block in patients undergoing open appen-dicectomy. Br J Anaesth 2009;103:601–5

12. Dalens B, Ecoffey C, Joly A, Giaufre E, Gustafsson U, HuledalG, Larsson LE. Pharmacokinetics and analgesic effect of ropi-vacaine following ilioinguinal/iliohypogastric nerve block inchildren. Paediatr Anaesth 2001;11:415–20

13. Frigon C, Mai R, Valois-Gomez T, Desparmet J. Bowel hema-toma following an iliohypogastric-ilioinguinal nerve block.Paediatr Anaesth 2006;16:993–6

14. Moiniche S, Mikkelsen S, Wetterslev J, Dahl JB. A qualitativesystematic review of incisional local anaesthesia for postopera-tive pain relief after abdominal operations. Br J Anaesth1998;81:377–83

15. Loukas M, Tubbs RS, El-Sedfy A, Jester A, Polepalli S, KinselaC, Wu S. The clinical anatomy of the triangle of Petit. Hernia2007;11:441–4

16. Newman K, Ponsky T, Kittle K, Dyk L, Throop C, Gieseker K,Sills M, Gilbert J. Appendicitis 2000: variability in practice,outcomes, and resource utilization at thirty pediatric hospitals.J Pediatr Surg 2003;38:372–9

17. Farooq M, Carey M. A case of liver trauma with a bluntregional anesthesia needle while performing transversus ab-dominis plane block. Reg Anesth Pain Med 2008;33:274–5

18. Frigon C, Mai R, Valois-Gomez T, Desparmet J. Bowel hema-toma following an iliohypogastric-ilioinguinal nerve block.Paediatr Anaesth 2006;16:993–6


Dexmedetomidine Infusion for Analgesia andPrevention of Emergence Agitation in Children withObstructive Sleep Apnea Syndrome UndergoingTonsillectomy and AdenoidectomyAnuradha Patel, MD, FRCA,* Melissa Davidson, MD,* Minh C. J. Tran, MD, MPH,*Huma Quraishi, MD,† Catherine Schoenberg, BSN,* Manasee Sant, MD,* Albert Lin, MD,*and Xiuru Sun, MS*

BACKGROUND: Dexmedetomidine, a specific �2 agonist, has an analgesic-sparing effect andreduces emergence agitation. We compared an intraoperative dexmedetomidine infusion withbolus fentanyl to reduce perioperative opioid use and decrease emergence agitation in childrenwith obstructive sleep apnea syndrome undergoing adenotonsillectomy (T&A).METHODS: One hundred twenty-two patients with obstructive sleep apnea syndrome undergoingT&A, ages 2 to 10 years, completed this prospective, randomized, U.S. Food and DrugAdministration–approved study. After mask induction with sevoflurane, group D received IVdexmedetomidine 2 �g � kg–1 over 10 minutes, followed by 0.7 �g � kg–1 � h–1, and group Freceived IV fentanyl bolus 1 �g � kg–1. Anesthesia was maintained with sevoflurane, oxygen, andnitrous oxide. Fentanyl 0.5 to 1 �g � kg–1 was given to subjects in both groups for an increasein heart rate or systolic blood pressure 30% above preincision values that continued for 5minutes. Observers in the postanesthesia care unit (PACU) were blinded to treatment groups.Pain was evaluated using the objective pain score in the PACU on arrival, at 5 minutes, at 15minutes, then every 15 minutes for 120 minutes. Emergence agitation was evaluated at thesame intervals by 2 scales: the Pediatric Anesthesia Emergence Delirium scale and a 5-pointscale described by Cole. Morphine (0.05 to 0.1 mg � kg–1) was given for pain (score �4) or severeagitation (score 4 or 5) lasting more than 5 minutes.RESULTS: In group D, 9.8% patients needed intraoperative rescue fentanyl in comparison with36% in group F (P � 0.001). Mean systolic blood pressure and heart rate were significantly lowerin group D (P � 0.05). Minimum alveolar concentration values were significantly differentbetween the 2 groups (P � 0.015). The median objective pain score was 3 for group D and 5 forgroup F (P � 0.001). In group D, 10 (16.3%) patients required rescue morphine, in comparisonwith 29 (47.5%) in group F (P � 0.002). The frequency of severe emergence agitation on arrivalin the PACU was 18% in group D and 45.9% in group F (P � 0.004); at 5 minutes and at 15minutes, it was lower in group D (P � 0.028). The duration of agitation on the Cole scale wasstatistically lower in group D (P � 0.004). In group D, 18% of patients and 40.9% in group F hadan episode of SPO2 below 95% (P � 0.01).CONCLUSIONS: An intraoperative infusion of dexmedetomidine combined with inhalation anes-thetics provided satisfactory intraoperative conditions for T&A without adverse hemodynamiceffects. Postoperative opioid requirements were significantly reduced, and the incidence andduration of severe emergence agitation was lower with fewer patients having desaturationepisodes. (Anesth Analg 2010;111:1004–10)

Adenotonsillectomy (T&A) is one of the most com-mon surgical procedures performed in children.The presence of obstructive symptoms is replacing

recurrent tonsillitis as the primary indication for T&A. Theprevalence of obstructive sleep apnea syndrome (OSAS) inchildren, also referred to as sleep disordered breathing, is

estimated to be 1%–3%.1 Postoperative pain can be severeafter T&A, and providing effective and safe perioperativeanalgesia in this group of patients is challenging. Not onlyare children with OSAS undergoing T&A at significant riskof respiratory and cardiovascular complications,2 they alsohave enhanced analgesic sensitivity to opiates and reducedmorphine requirements after T&A.3 A high incidence ofemergence agitation (EA) in patients having otolarygologicprocedures adds another challenge.4 Dexmedetomidine(Dex) (Precedex, Hospira Worldwide, Lake Forest, Illinois),a specific � 2-adrenergic receptor agonist, has sedative,anxiolytic, and analgesic properties5 and is very effective inprevention of EA in children.6,7 An intraoperative infusionof Dex used as a substitute for fentanyl has been shown toreduce opiate use in the postoperative period in adultpatients undergoing bariatric surgery,8 but clinical data onthe analgesic-sparing effect of Dex in children are conflict-ing. The present study was performed to evaluate whether

From the *Department of Anesthesiology and Perioperative Medicine andthe †Department of Otolaryngology, University of Medicine and Dentistryof New Jersey, New Jersey Medical School, Newark, New Jersey.


Financial support was provided by an Institutional Cost of Drug supportgrant from Hospira Worldwide, Lake Forest, Illinois.

Address correspondence and reprint requests to Dr. Anuradha Patel,Associate Professor, Department of Anesthesiology and Perioperative Medi-cine, New Jersey Medical School, University of Medicine and Dentistry ofNew Jersey, Medical Science Building E-581, 185 South Orange Avenue,Newark, NJ 07101. Address e-mail to [email protected].

Copyright © 2010 International Anesthesia Research SocietyDOI: 10.1213/ANE.0b013e3181ee82fa


an intraoperative infusion of Dex combined with generalanesthesia would be a safe and effective substitute toopiates intraoperatively, reduce opiate requirements post-operatively, and also be effective in reducing the incidenceand severity of EA in children with OSAS undergoingT&A.

METHODSAn Investigational New Drug number (76,041) was ob-tained from the U.S. Food and Drug Administration. Thestudy was registered at www.clinicaltrials.gov (registrationnumber NCT00468052) and approved by the IRB of theUniversity of Medicine and Dentistry of New Jersey. Onehundred thirty-seven children ages 2 to 10 years, ASAphysical status II–III, undergoing elective T&A, were en-rolled in this investigator-initiated, prospective, random-ized, blinded, controlled study. Informed, written consentto participate in the study was obtained from the parent orlegal guardian and assent from children older than 7 yearsof age. All patients had OSAS on the basis of clinicalsymptoms or diagnostic polysomnography. Clinical grad-ing of OSAS was done by the surgeon on the basis ofseverity of symptoms such as restless sleep, severe snoring,apnea witnessed by the parents, nocturnal enuresis, stertor,hyperactivity, or failure to thrive. Exclusion criteria wereknown allergy to �2 agonists, developmental delay, cardiacand craniofacial abnormalities, anxiety disorder, chronicdisabilities or pain syndrome, and use of psychotherapeuticmedications, � blockers, digoxin, cimetidine, �2 agonists,anticonvulsants, or psychotropic medications. A randomnumber table was used to assign subjects into 1 of 2treatment groups: Dex infusion (group D) or IV fentanyl(group F). The anesthesiologists and data collectors in theoperating room (OR) were not blinded; the subjects, theirparents, and observers in the postanesthesia care unit(PACU) were blinded to treatment group.

No premedication was given. Monitoring includedpulse oximetry, electrocardiogram, noninvasive arterialblood pressure (NIBP), end-tidal CO2 (etco2), and a depthof anesthesia monitor, the Bispectral Index (BIS; AspectMedical Systems, Natick, Massachusetts). Anesthesia wasinduced with 8% inspired sevoflurane and 60% nitrousoxide (N2O) in oxygen by facemask. Group D received IVDex (2 �g � kg–1 over 10 minutes, followed by 0.7 �g � kg–1 � h–1

until 5 minutes before the end of the surgery), andgroup F received IV fentanyl (1 �g � kg–1) as a bolus, as soonas IV access was obtained. A balanced salt solution wasadministered according to standard fluid administrationguidelines. Rocuronium 0.6 mg � kg–1 was used to facilitatetracheal intubation. End-tidal sevoflurane concentrationwas maintained at 1 minimum alveolar concentration(MAC) with 60% N2O as long as the BIS remained below 60during surgery. If the BIS reached 60 or more, the sevoflu-rane concentration was increased to reduce the BIS below60. All patients received IV dexamethasone 0.5 mg � kg–1

(maximum 10 mg) and rectal acetaminophen 30 to 40 mg �kg–1 up to a maximum of 1000 mg before the start ofsurgery. The data collector recorded the heart rate (HR),systolic and diastolic blood pressures (NIBP), hemoglobinoxygen saturation (Spo2), etco2 tension, MAC, and BISevery 5 minutes during the anesthetic. The values in the

holding area for HR and systolic blood pressure were usedas baseline. Both groups received fentanyl 0.5 to 1 �g � kg–1

for an increase in HR or systolic NIBP 30% above the valuebefore start of surgery and sustained for 5 minutes. Lac-tated Ringer’s solution 15 mL/kg was administered as afluid bolus for a 30% decrease of systolic blood pressurefrom baseline, which continued for 2 readings and glyco-pyrrolate 0.01 mg � kg–1 for a 30% decrease in HR. Sevoflu-rane was discontinued once hemostasis was achieved andmuscle relaxation was reversed with neostigmine 0.05 mg �kg–1 and glycopyrrolate 0.01 mg � kg–1. The time to awak-ening (TA), defined as spontaneous eye opening or oncommand from end of surgery, and the time to extubation(TE), defined as time from end of surgery to trachealextubation, were recorded. All patients were observedcontinuously in the PACU for 2 hours by observers blindedto study group. Pain was evaluated using the objective painscore (OPS)9 in the PACU on arrival and at 5 minutes, at 15minutes, and then every 15 minutes for 120 minutes. EAwas evaluated at the same intervals by 2 scales: thePediatric Anesthesia Emergence Delirium (PAED) scale10

and a 5-point agitation scale described by Cole.11 Durationof severe EA was noted on the Cole scale. Morphine (0.05 to0.1 mg � kg–1) was given for pain (score �4) or severeagitation (score 4 or 5) lasting more than 5 minutes. HR,systolic and diastolic NIBP, respiratory rate (RR), and Spo2

were recorded in the PACU every 5 minutes for the first 15minutes, then at 15-minute intervals for the next 2 hours.Any desaturation episode with Spo2 below 95% was noted.

Statistical MethodsA power analysis indicated that 60 subjects were requiredper group to show that the number of patients needingintraoperative rescue fentanyl and rescue morphine in thePACU would be 50% lower in the subjects receiving Dex.Sixty subjects were also required per group to determinethat treatment with Dex would decrease the incidence ofsevere EA after surgery by 50% with 80% power (� � 0.05)in comparison with the control group.

Data were analyzed using SPSS software (version 16,Chicago, Illinois), and are presented as number (n) orpercentage (%), mean � sd, or median as appropriate.Student t test was used to compare the mean value ofquantitative data between the 2 groups. Two-wayrepeated-measures analysis of variance (ANOVA) wasused for NIBP, HR, Spo2, MAC, and BIS. Student t test wasused for the comparisons of intragroup values of intraop-erative and postoperative systolic blood pressure and HR.Nonparametric data such as pain score, PAED score, andEA score on the Cole scale were compared between groupswith Mann–Whitney U test. Fischer exact test was used forcomparision of gender; percentage of patients in eachgroup with a preoperative diagnosis of mild, moderate, orsevere OSAS; and number of patients rescued with fentanylor morphine and those with episodes of severe EA. P valueof 0.05 or less was considered statistically significant.

RESULTSResults are presented for 122 patients. One hundred thirty-seven subjects were enrolled in this study; 15 subjects wereeliminated from data analysis for the following reasons:


surgery was cancelled for 2 patients, 1 refused to participateafter enrolling, and 1 patient had an intraoperative complica-tion. Eleven subjects who completed the study had deviationfrom this strictly controlled protocol or incomplete data andwere also removed before data analysis.

The 2 groups were comparable in age, gender, baselineHR, systolic NIBP, and diagnosis of OSAS (Table 1). Theage range of patients in the study was 2 to 10 years, 90% ofpatients were 6 years or younger, and 26 patients (46.2%) ineach group were 2 to 3 years old.

Intraoperative data are presented in Table 2. In group D,6 patients (9.8%) needed rescue fentanyl in comparisonwith 22 (36%) in group F (P � 0.001). Mean HR (P � 0.001)(Fig. 1A) and mean systolic NIBP (Fig. 1B) were signifi-cantly lower in group D during the first 60 minutes (P �0.019). Mean diastolic NIBP was not statistically different inthe 2 groups (P � 0.29). During the first 60 minutes of theanesthetic, MAC values of sevoflurane were significantly

different between the 2 groups (P � 0.015); MAC was lowerin group D, ranging from 5.7% to 41.6%. There was astatistical difference in TA and TE, both lower in group Dthan in group F (P � 0.05). Duration of surgery wasstatistically lower in group D (P � 0.041). There was nodifference in the average dose of intraoperative fentanyland dexamethasone between the 2 groups. The dose ofacetaminophen was lower in group D. None of the subjectsneeded glycopyrrolate for bradycardia or fluid bolus forhypotension in the OR.

The variables measured in the PACU are shown in Table3. In group D 10 (16.3%) patients required rescue morphine,in comparison with 29 (47.5%) in group F (P � 0.002). Themedian of the maximum OPS was 3 for group D and 5 forgroup F (P � 0.001). The percentage of patients with anOPS score of 4 and above (Fig. 2A) from arrival (P � 0.001)and at 5 and 15 minutes was statistically lower in group D(P � 0.05). On the Cole scale (5-point scale), severe EA wasdefined as a score of 4 to 5. The frequency of severe EA isshown in Figure 2B. On arrival in the PACU it wasstatistically lower, 18% in group D and 45.9% in group F(P � 0.004). At 5 and 15 minutes it was statistically lower ingroup D (P � 0.028). At 30 minutes none of the patients hadsevere EA in group D, and 1.6% of patients in group F hadsevere EA. The duration of agitation on the Cole scaleshowed statistical significance; it was 6.59 � 7.4 minutes(mean � sd) for group D and 11.85 � 12.0 minutes (mean �sd) for group F (P � 0.004). There was a statistical differ-ence in the median of the highest score on the Cole scale, 3for group D and 4 for group F (P � 0.001). The percentageof patients with a score of 10 and above for the PAED (Fig.2C) was statistically lower in group D at arrival (P � 0.004)

Table 1. Demographic DataGroup F(n � 61)

Group D(n � 61) P value

Age (years) 3.8 � 1.5 4.2 � 2.1 0.162–3 years old (%) 26 (42.6) 26 (42.6) 1Gender (M/F) 35/26 35/26 1Weight (kg) 18.3 � 5.7 20.4 � 8.6 0.12Baseline HR (beats/

minute)105 � 18 104 � 15 0.77

Baseline systolicNIBP (mm Hg)

101 � 13.7 104 � 12.6 0.29

OSAS (% patients)Mild 30 26Moderate 50 60Severe 20 14 0.55

Data are expressed as n (%) and mean � SD.Group D � dexmedetomidine group; group F � fentanyl group; HR � heartrate; NIBP � noninvasive arterial blood pressure; OSAS � obstructive sleepapnea syndrome.

Table 2. Intraoperative DataGroup F(n � 61)


Rescue by fentanyl,n (%)

22 (36.1) 6 (9.8) 0.001*

Fentanyl rescuedosage (�g/kg)

1.04 � 0.67 0.73 � 0.25 0.312

Time of rescue(minutes)

10.82 � 12.5 17.6 � 6.77 0.256

Acetaminophendosage (mg/kg)

31.51 � 4.96 28.30 � 6.59 0.02*

Dexamethasonedosage (mg/kg)

0.30 � 0.12 0.30 � 0.14 0.847

Duration of surgery(minutes)

43.33 � 17.36 37.54 � 13.33 0.041*

Duration ofanesthesia(minutes)

75.08 � 24.73 69.80 � 16.82 0.175

Time to awake(minutes)

8.75 � 4.06 7.18 � 4.05 0.03*

Time to extubate(minutes)

10.44 � 4.15 8.59 � 4.51 0.02*

Data are expressed as n, mean � SD, and percentage.Group D � dexmedetomidine group; group F � fentanyl group; HR � heartrate; NIBP � noninvasive arterial blood pressure.* P � 0.05.

Figure 1. A, Heart rate and B, systolic blood pressure during the first60 minutes of the procedure. Both variables were statistically lowerin the dexmedetomidine group (P � 0.05).

Dexmedetomidine Infusion for Adenotonsillectomy


and at 5 and 15 minutes (P � 0.05). The median of thehighest score on the PAED scale did not show a statisticaldifference, 10 for group D and 14 for group F (P � 0.051).

On arrival in the PACU until 90 minutes, HR wasstatistically lower in group D (P � 0.001) than in group F.There was no statistical difference in mean systolic NIBP,respiratory rate, and Spo2 in the PACU. There was astatistically significant difference in the number of patientswith Spo2 below 95% between the 2 groups, 11 (18%) ingroup D and 25 (40.9%) in group F (P � 0.01).

DISCUSSIONIn this study of children undergoing T&A, an intraopera-tive initial loading dose of 2 �g � kg–1 Dex followed by aninfusion at 0.7 �g � kg–1 � h–1 decreased intraoperativeopiate and anesthetic requirements and decreased opiaterequirements in the PACU, in comparison with a controlgroup receiving intraoperative IV fentanyl. Additionally,there was a significantly lower incidence and duration ofsevere EA in children who received Dex.

We are reporting on the use of a high initial loading doseof Dex (2 �g � kg–1) followed by a relatively high dosecontinuous intraoperative infusion, because T&A is a pro-cedure with an intense surgical stimulus, and surgery startsimmediately with no preparation time. An analgesic-sparing effect of Dex has been shown,5 and when combinedwith N2O there is an additive interaction to enhanceanalgesia.12 We aimed to use the analgesic-sparing effect ofDex and evaluated whether a continuous infusion could beused as a substitute for bolus fentanyl. In our study, 90% ofpatients in group D did not require any other intraopera-tive analgesics. Because there are no studies using a high-dose continuous infusion of Dex combined with inhalationanesthetics, the U.S. Food and Drug Administration man-dated reporting hemodynamic changes as a safety concern.

Children in group D had statistically lower systolic bloodpressure and HR, almost the entire duration of the anes-thetic (Fig. 1), but none of the patients needed interventionfor bradycardia or hypotension on the basis of studycriteria. Hemodynamic data are consistent with reports byother investigators. Mason et al.13 used higher doses of Dex(2 to 3 �g � kg–1 loading dose and infusion of 1.5 to 2 �g �kg–1 � h–1) as the sole drug for sedation in children andobserved a decrease in HR and blood pressure, which werewithin 20% of awake normal range. In children anesthe-tized with one MAC sevoflurane or desflurane and given asingle, lower dose of Dex (0.5 �g � kg–1), Deutsch et al.14

found a significant decrease in HR, but neither the systolicnor diastolic blood pressure was statistically lower. Thebiphasic response usually seen in adults, with an initialincrease in systolic blood pressure and a reflex decrease in

Table 3. Postanesthesia Recovery Unit DataGroup F(n � 61)


OPS maximum(range)

5 (0–10) 3 (0–10) 0.001*

EA scoremaximum(range)

4 (1–5) 3 (1–5) 0.001*

Duration ofsevere EA(minutes)

11.85 � 12.02 6.59 � 7.42 0.004*

PAED scoremaximum(range)

14 (0–20) 10 (0–20) 0.051

Rescue bymorphine,n (%)

29 (48) 10 (17) 0.0003*

Morphine dosage(mg/kg)

0.073 � 0.033 0.074 � 0.033 0.928

SpO2 below95%, n (%)

25 (41) 11 (18) 0.01*

OPS, PAED, and EA (Cole scale) are expressed as median values of themaximum score.Other data are expressed as n (%) and mean � SD.Group D � dexmedetomidine group; group F � fentanyl group; OPS �objective pain score; EA � emergence agitation; PAED � pediatric anesthesiaemergence delirium.* P � 0.05.

Figure 2. A, Percentage of patients with an objective pain score(OPS) of 4 and above. Score 4 and above lasting more than 5minutes was treated. Statistically lower in group D (dexmedetomi-dine) at arrival (P � 0.001), at 5 minutes (P � 0.028), and at 15minutes (P � 0.011). B, Percentage of patients with severe emer-gence agitation (EA), defined as a score of 4 or 5 on the 5-pointscale. Lower in group D at arrival (P � 0.001), at 5 minutes (P �0.028), and at 15 minutes (P � 0.028). C, Percentage of patientswith Pediatric Anesthesia Emergence Delirium (PAED) score of 10and above. Statistically lower in group D at arrival (P � 0.001), at 5minutes (P � 0.028), and at 15 minutes (P � 0.011).


HR followed by stabilization of these variables belowbaseline, is not observed in children.15

On the basis of routine clinical practice, fentanyl wasgiven in a dose of 1 �g � kg–1 as a bolus to the control group.This lower dose is based on the enhanced analgesic sensi-tivity to opiates in children with OSAS.3 It is noteworthythat in the control group only 36% of patients neededrescue fentanyl, indicating that our technique of low-dosefentanyl is effective in almost two thirds of patients. HRand systolic blood pressure increase was used as the triggerfor rescue fentanyl in both groups in response to surgicalstimulation. The BIS monitor was used to ensure thatpatients in group D had an adequate depth of anesthesiabecause they may not display hemodynamic changes dueto the inherent sympatholytic properties of Dex. In anattempt to maintain equivalent depth of anesthesia in bothgroups, the sevoflurane concentration was titrated to main-tain a BIS value below 60. Consistent with studies in adultpatients, the concentration of sevoflurane required to main-tain the BIS below 60 was smaller in patients receiving Dex(MAC in group D was 5.7% to 41.6% lower). Tufanoguallariet al.8 found reductions in the average end-tidal desfluraneconcentration of 19%–22%, depending on the rate of Dexinfusion, which ranged from 0.2 to 0.8 �g � kg–1 � h–1. Theanesthetic-sparing effect of Dex appears to have an addedadvantage in facilitating earlier awakening and trachealextubation. In the present study, TA and TE were statisti-cally lower in group D, despite the high dose used. Mostinvestigators using Dex as a low-dose intraoperative infu-sion or as a single bolus reported no difference in TA andTE in comparison with placebo.6,16 Only 1 study reportedthat a single dose of 0.5 �g � kg–1 Dex, 5 minutes before theend of surgery significantly prolonged TA and TE incomparison with placebo in patients having T&A.7

Evaluation of postoperative pain is complicated by thedifficulty in assessing pain in younger children and by theoccurrence of EA. It is often difficult to distinguish betweenpain and EA because of the overlapping clinical picture,and pain itself can be the source of agitation.17 Mostinvestigators have used different assessment tools to tryand separate the two, but there is generally overlap in thescales, because a child who is restless or thrashing willscore high on both scales. We did find a positive correlationbetween agitation and pain; group F had higher pain andEA scores than did group D. Results on the OPS, Cole scale,and PAED showed a very similar trend in both groups;scores were highest on arrival in the PACU and decreasedover time (Fig. 2, A–C). A significantly smaller number ofpatients needed rescue morphine in group D, 18% incomparison with 44% in group F. Because it is difficult toseparate pain and EA, and the fact that the rescue drug forboth agitation and pain in our study was morphine, it is notpossible to determine whether the morphine was given forpain or for agitation. On the basis of the effectiveness ofsmaller doses of intraoperative Dex in adult patients forreducing postoperative morphine consumption for 24hours,8,18 we could assume that an analgesic effect wouldbe present in our study patients in the immediate postop-erative period. In children, Guler et al.7 found that 23%patients who received a single dose of 0.5 �g/kg Dexbefore the end of the procedure (T&A) required opoiods for

analgesia in the PACU in comparison with 53% in theplacebo group. Erdil et al.19 compared a single dose of 0.5�g � kg–1 Dex with 2.5 �g � kg–1 fentanyl in patientsundergoing adenoidectomy and concluded that Dex pro-vided residual analgesia similar to that of fentanyl.

Pain can be severe after T&A, and it is commonly treatedwith opioids, despite a known sensitivity of patients withOSAS and recurrent hypoxemia to opiates. Brown et al.3

reported enhanced analgesic morphine sensitivity in chil-dren with OSAS during T&A and reduced morphinerequirements after T&A. Therefore, several nonopioid an-algesics such as ketorolac, ketamine, and tramadol havebeen evaluated for pain management after T&A,20–22 butnone have gained widespread use or acceptance because ofconcerns with side effects or inadequate analgesia. Amorphine-sparing effect of acetaminophen has been dem-onstrated in pediatric day-case surgery,23 and dexametha-sone also reduces post-tonsillectomy pain.24 In the presentstudy, all patients were given 30 to 40 mg � kg�1 of acetamin-ophen rectally before start of surgery and intraoperative IVdexamethasone. A multimodal, opioid-sparing, analgesic ap-proach including Dex, such as the one used in our study, isworth considering in this patient population with a highpotential for adverse respiratory events. The incidence ofnausea or vomiting was extremely low in this study. Only 1patient needed an antiemetic in the PACU, probably becauseof the antiemetic effect of dexamethasone.

EA is a complex phenomenon, the etiology of which ismultifactorial. The wide variability in the incidence ofagitation in the different studies on EA may be due to thecriteria used to define this phenomenon and the time in thePACU when EA was measured.17 We did repeated mea-surements at frequent time intervals, because a singlemeasurement may not reflect the true incidence of EA.11

Group D had a statistically lower frequency of severe EAthan did group F until 30 minutes (Fig. 2B). At 30 minutesthere was no incidence of severe EA in group D, and ingroup F it was 1.6%. Severe EA lasting more than 5 minuteswas treated. The incidence of severe EA on arrival in thePACU in group D (18%) was similar to that reported byGuler et al.7 (17%), who used a single dose of Dex 5 minutesbefore the end of the procedure in children undergoingT&A. The occurrence of EA in younger patients andotolaryngologic procedures is reported to be high, althoughthe exact reason for this is not known.4 Ninety percent ofpatients in our study were 6 years old or younger, and 26patients (46.2%) in each group were 2 to 3 years old.Hyperactivity and attention deficit disorder are frequentlyseen in children with OSAS, possibly explaining or contrib-uting to a high incidence of EA in our T&A patients.Dexmedetomidine has been used successfully as an infu-sion (0.2 �g � kg–1 � h–1) continued into the postoperativeperiod for 15 minutes or single dose at the end of surgery(0.5 �g � kg–1) to prevent or reduce emergence delirium inchildren.6,7,16 It must be noted that these studies comparedDex with placebo, whereas our control group receivedfentanyl 1 �g � kg–1, which also reduces EA. However, ahigher dose is reported to be effective in patients havingpainful procedures.25 From our study and others, it re-mains difficult to discern whether the analgesic or sedativeeffects of �2 agonists are responsible for reducing EA in



children; regardless of the mechanism, Dex appears to beeffective in a wide range of doses. The half-life of Dex isreported to be 1.8 hours in children,15 but there are no dataon duration of sedative or analgesic effects after discontinu-ation of Dex infusion. The HRs were significantly slower ingroup D until 90 minutes in the PACU. The residual effectson HR of an intraoperative Dex infusion and the potentialfor an attenuated response to postoperative bleeding inT&A patients may be a concern and a disadvantage ofusing a Dex infusion.

The risk of respiratory morbidity after T&A in childrenwith OSAS is reported to be about 20%.2 Sanders et al.26

reported that although the patients with OSAS were morelikely to require supplemental oxygen, oral airway use, orassisted ventilation on emergence, severe complicationssuch as laryngospasm and bronchospasm were uncom-mon. In the present study, there were no instances oflaryngospasm or bronchospasm after extubation. One pa-tient developed intraoperative pulmonary edema and wasexcluded from the study because she remained intubatedovernight. Although not a study variable, we noted thatextubation was much smoother with less coughing andbreath-holding in patients given Dex. All patients wereobserved continuously in the PACU for 2 hours, and theobservers were asked to record the lowest Spo2 during thisperiod. There was a statistically significant difference in thenumber of patients with Spo2 below 95% in the PACUbetween the 2 groups, 11 in group D and 25 in group F. Thiscould be related to the smaller requirement for opiates inthe PACU in group D or to the lower incidence andduration of severe EA in group D. The goal of having achild who was settled, comfortable, and less restless, withapplication of monitors and administration of supplemen-tal oxygen in the PACU, was easier to achieve in patientswho received Dex.

A few methological considerations of this study need tobe mentioned. The anesthesiologist and the data recorder inthe OR were not blinded to the study group. We believethat knowledge of study group assignment did not bias theconduct of the anesthetic, because the study protocol wastightly controlled, with specific criteria regarding intraop-erative rescue fentanyl, sevoflurane concentration, the timeto discontinue sevoflurane, extubation criteria, and use ofrescue morphine in the PACU.

The PAED is the only validated rating scale for emer-gence delirium.10 The investigators who developed thePAED scale rated emergence behavior 10 minutes after thechild awakened and remained awake (did not fall back tosleep). Early in the present study, we found this to be apotential problem because children who were asleep werereceiving ratings of 4 on the first 3 items of the scalebecause they could not make eye contact, their actions werenot purposeful, and they were not aware of their surround-ings. Therefore we had to modify the scoring on the scaleand rate these items as zero. Clearly, the children were notagitated if they were sleeping. Because we used a modifiedversion of the PAED, we used a second scale (Cole) to runconcomitantly to support the findings with the modifiedversion of the PAED. The 1 to 5 scale described by Cole etal.11 has been used in several studies of EA. It is not a

validated scale, but is easy to use, and defining the catego-ries of mild or severe is clear.

The OPS is not a validated scale, but this scale or somemodification of it has been used in several studies inchildren. Although 2 other studies on EA6,19 have used theOPS, it is perhaps not the best scale to use in a study on EAbecause of considerable overlap on the items being scored.

We did not follow patients once they were dischargedfrom the PACU. A future study with overnight pulseoximetry data and use of postoperative analgesics wouldbe worthwhile to perform.

CONCLUSIONIn children undergoing T&A, the goal is to minimizerespiratory and airway compromise and have an awake,settled, comfortable child after the surgery. An opioid-sparing technique is particularly appealing in children withOSAS, when airway obstruction is known to preexist andmay persist on the night after surgery. An intraoperativeinfusion of Dex combined with sevoflurane and N2Oprovided satisfactory intraoperative conditions for T&Awithout adverse hemodynamic effects. TA and TE wereshorter than they were for the patients receiving fentanyl.Postoperative opoiod requirements were significantly re-duced, and the incidence and duration of severe EA waslower, resulting in a smooth recovery. We have described apractical, effective, and safe technique for using Dex infu-sion. A multimodal, opioid-sparing, analgesic approachincluding Dex, such as the one used in our study, can beuseful in children with OSAS undergoing other surgicalprocedures besides T&A, wherein the advantages of de-creased perioperative opioid requirements and a reducedoccurrence of EA will be beneficial.

REFERENCES1. Brown KA. What we don’t know about childhood obstructive

sleep apnea. Paediatr Anaesth 2001;11:385–92. McColley SA, April MM, Carroll JL, Naclerio RM, Loughlin

GM. Respiratory compromise after adenotonsillectomy in chil-dren with obstructive sleep apnea. Arch Otolaryngol HeadNeck Surg 1992;118(9):940–3

3. Brown KA, Laferriere A, Lakheeram I, Moss IR. Recurrenthypoxemia in children is associated with increased analgesicsensitivity to opiates. Anesthesiology 2006;105(4):645–7

4. Voepel-Lewis T, Malviya S, Tait AR. A prospective cohortstudy of emergence agitation in the pediatric postanesthesiacare unit. Anesth Analg 2003;96(6):1625–30

5. Hall JE, Uhrich TD, Barney JA, Arain SR, Ebert TJ. Sedative,amnestic, and analgesic properties of small-dose dexmedeto-midine infusions. Anesth Analg 2000;90:699–705

6. Shukry M, Clyde MC, Kalarickal PL, Ramadhyani U. Doesdexmedetomidine prevent emergence delirium in childrenafter sevoflurane-based general anesthesia? Paediatr Anaesth2005;15:1098–104

7. Guler G, Akin A, Tosun Z, Ors S, Esmaoglu A, Boyaci A.Single-dose dexmedetomidine reduces agitation and providessmooth extubation after pediatric adenotonsillectomy. PaediatrAnaesth 2005;15(9):762–6

8. Tufanoguallari B, White PF, Peixoto MP, Kianpour D, LacourT, Griffin J, Skrivanek G, Macaluso A, Shah M, Provost DA.Dexmedetomidine infusion during laparoscopic bariatric sur-gery: the effect on recovery outcome variables. Anesth Analg2008;106(6):1741–8

9. Hannallah RS, Broadman LM, Belman AB, Abramowitz MD,Epstein BS. Comparison of caudal and ilioinguinal/iliohypogastric nerve blocks for control of post-orchiopexy painin pediatric ambulatory surgery. Anesthesiology 1987;66(6):832–4


10. Sikich N, Lerman J. Development and psychometric evaluationof the pediatric anesthesia emergence delirium scale. Anesthe-siology 2004;100(5):1138–45

11. Cole JW, Murray DJ, McAllister JD, Hirshberg GE. Emergencebehaviour in children: defining the incidence of excitementand agitation following anaesthesia. Paediatr Anaesth 2002;12:442–7

12. Dawson C, Ma D, Chow A, Maze M. Dexmedetomidineenhances analgesic action of nitrous oxide: mechanisms ofaction. Anesthesiology 2004;100(4):894–904

13. Mason KP, Zurakowski D, Zgleszewski SE, Robson CD, Car-rier M, Hickey PR, Dinardo JA. High dose dexmedetomidineas the sole sedative for pediatric MRI. Paediatr Anaesth2008;18(5):403–11

14. Deutsch E, Tobias JD. Hemodynamic and respiratory changesfollowing dexmedetomidine administration during generalanesthesia: sevoflurane vs desflurane. Paediatr Anaesth2007;17:438–44

15. Petroz GC, Sikich N, James M, van Dyk H, Shafer SL, Schily M,Lerman J. A phase I, two center study of the pharmacokineticsand pharmacodynamics of dexmedtomidine in children.Anesthesiology 2006;105:1098 –110

16. Ibacache ME, Munoz HR, Brandes V, Morales AL. Single-dosedexmedetomidine reduces agitation after sevoflurane anesthe-sia in children. Anesth Analg 2004;98(1):60–3

17. Vlajkovic GP, Sindjelic RP. Emergence delirium in children:many questions, few answers. Anesth Analg 2007;104:84–91

18. Gurbet A, Basagan-Mogol E, Turker G, Ugun F, Kaya FN,Ozcan B. Intraoperative infusion of dexmedetomidine reducesperioperative analgesic requirements. Can J Anaesth 2006;53(7):646–52

19. Erdil F, Demirbilek S, Begec Z, Ozturk E, Ulger MH, Ersoy MO.The effects of dexmedetomidine and fentanyl on emergencecharacteristics after adenoidectomy in children. Anaesth Inten-sive Care 2009;37(4):571–6

20. Marret E, Flahault A, Samama CM, Bonnet F. Effects ofpostoperative, nonsteroidal, antiinflammatory drugs on bleed-ing risk after tonsillectomy: meta-analysis of randomized,controlled trials. Anesthesiology 2003;98(6):1497–502

21. Erhan OL, Goksu H, Alpay C, Beçstaçs A. Ketamine inpost-tonsillectomy pain. Int J Pediatr Otorhinolaryngol 2007;71(5):735–9

22. Hullett BJ, Chambers NA, Pascoe EM, Johnson C. Tramadol vsmorphine during adenotonsillectomy for obstructive sleepapnea in children. Paediatr Anaesth 2006;16(6):648–53

23. Korpela R, Korvenoja P, Meretoja OA. Morphine-sparing effectof acetaminophen in pediatric day-case surgery. Anesthesiol-ogy 1999;91:442–7

24. Afman CE, Welge JA, Steward DL. Steroids for posttonsillec-tomy pain reduction: meta-analysis of randomized controlledtrials. Otolaryngol Head Neck Surg 2006;134:181–6

25. Cohen IT, Finkel JC, Hannallah RS, Hummer KA, Patel KM.The effect of fentanyl on the emergence characteristics afterdesflurane or sevoflurane anaesthesia. Paediatr Anaesth2002;12:442–7

26. Sanders JC, King MA, Mitchell RB, Kelly JP. Perioperativecomplications of adenotonsillectomy in children with obstruc-tive sleep apnea syndrome. Anesth Analg 2006;103(5):1115–21



Epidemiology of Ambulatory Anesthesia for Children inthe United States: 2006 and 1996Jennifer A. Rabbitts, MB, ChB,* Cornelius B. Groenewald, MB, ChB,* James P. Moriarty, MSc,†and Randall Flick, MD, MPH*

BACKGROUND: There are few data that describe the frequency, anesthetic type, provider, ordisposition of children requiring outpatient anesthesia in the United States (US). Since the early1980s, the frequency of ambulatory surgery has increased dramatically because of advances inmedical technology and changes in payment arrangements. Our primary aim in this study was toquantify the number of ambulatory anesthetics for children that occur annually and to study thechange in utilization of pediatric anesthetic care over a decade.METHODS: The US National Center for Health Statistics performed the National Survey ofAmbulatory Surgery in 1994 through 1996 and again in 2006. The survey is based on dataabstracted from a national sample of ambulatory surgery centers and provides data on visits forsurgical and nonsurgical procedures for patients of all ages. We abstracted data for childrenwho had general anesthesia, regional anesthesia, or monitored anesthesia care during theambulatory visit. We obtained the information from the 2006 and 1996 databases and usedpopulation census data to estimate the annual utilization of ambulatory anesthesia per 1000children in the US.RESULTS: In 2006, an estimated 2.3 million ambulatory anesthesia episodes of care wereprovided in the US to children younger than 15 years (38 of 1000 children). This amountcompares with 26 per 1000 children of the same age group in 1996. In most cases, ananesthesiologist was involved in both time periods (74% in 2006 and 85% in 1996). Of thechildren, 14,200 were admitted to the hospital postoperatively, a rate of 6 per 1000 ambulatoryanesthesia episodes.CONCLUSION: The number and rate of ambulatory anesthesia episodes for US childrenincreased dramatically over a decade. This study provides an example of how databases canprovide useful information to health care policy makers and educators on the utilization ofambulatory surgical centers by children. (Anesth Analg 2010;111:1011–5)

The introduction of the first freestanding ambulatorysurgery centers (ASCs) in the 1970s resulted in arapid increase in the proportion of operations per-

formed on an outpatient basis, from �10% in 1979 to 50% in1990.1 The number of ASCs continues to increase, with a150% increase per 100,000 population reported in metro-politan areas from 1993 to 2001.2 The number of Medicare-certified ASCs increased 64% between 2000 and 2007, from3028 to 4964.3 Improvements in surgical and anesthetictechniques have increased the proportion of proceduresperformed on an outpatient basis to �70% of the totalsurgical interventions currently performed in the UnitedStates (US).1

No quantification has been made of the pediatric proce-dures occurring on an outpatient basis in the US. As thecountry enters an era of health care reform, epidemiologicdata on the utilization of medical resources may be helpfulto policy makers as health care expenditures are analyzed.For example, current Medicare payments to freestanding

ASCs are less than for corresponding services in hospital-based outpatient departments. In addition, copaymentsand charges to patients are generally less at ASCs than athospitals. Almost 90% of all US freestanding ASCs arewholly or partially owned by physicians and 96% arefor-profit institutions.4

The purpose of this study was to describe, for the firsttime, the utilization of freestanding and hospital-basedASCs in regard to their care of children. We quantified thenumber of ambulatory anesthesia episodes occurring an-nually for children in accordance with age group, anes-thetic type, and anesthesia provider and described thechange in utilization over a decade. Secondary analysesexamined the distribution of perioperative time and dispo-sition and used unplanned admission as an end point.

METHODSThe National Survey of Ambulatory Surgery (NSAS) is theonly US national study of ambulatory surgery in hospital-based and freestanding ASCs.5 We abstracted the data forambulatory anesthesia of children from this public data-base for 1996 and 2006. National census data were used toestimate utilization rates.

The NSAS DatabaseThe NSAS was performed by the National Center forHealth Statistics on a nationally representative sample ofsurgery centers that perform ambulatory procedures. Thecomplete sampling and survey methods have been de-scribed5 and select data have been published for patients of

From the *Department of Anesthesiology, and the †Division of Health CarePolicy & Research, Mayo Clinic, Rochester, Minnesota.


Supported by the Department of Anesthesiology, Mayo Clinic, Rochester, MN.


Address correspondence and reprint requests to Randall Flick, MD, MPH,Department of Anesthesiology, Mayo Clinic, 200 First St. SW, Rochester, MN55905. Address e-mail to [email protected].

Copyright © 2010 International Anesthesia Research SocietyDOI: 10.1213/ANE.0b013e3181ee8479


all ages who had both surgical and nonsurgical proce-dures.6,7 In summary, eligible hospital-based facilities wereidentified from the SMG Marketing Group, Inc., HospitalMarket Database5 and included all short-stay or general(medical, surgical, or children’s) noninstitutional, nonfed-eral hospitals in the 50 states and the District of Columbiawith 6 or more beds staffed for patient use. Eligiblefreestanding facilities were identified from the SMG Free-Standing Outpatient Surgery Center Database and theHealth Care Financing Administration Provider of ServicesPublic Use File.

Hospital-based and freestanding ASCs consisted of hos-pitals that were state regulated or certified for Medicarethat performed at least 50 ambulatory procedures in theprevious year and excluded dental, podiatry, pain, abor-tion, family planning, and birthing centers. The includedprocedures were both surgical and nonsurgical (e.g., lum-bar puncture, computed tomographic scanning) proceduresperformed on an ambulatory basis in general operatingrooms, dedicated ambulatory surgery rooms, and otherspecialized rooms, including endoscopy units and cardiaccatheterization laboratories.

A multistage probability design was used, in whichindependent samples of hospitals and freestanding ASCswere selected at the first or second stages and visits to thesefacilities were selected at the final sampling stages.5 AnNSAS medical abstract form (Appendix) was used tocollect data for each sampled visit during which �1 proce-dure may have been performed. Data were abstracted fromthe medical record by facility staff in 30% of cases and byUS Census Bureau personnel in 70% of cases. Data ab-stracted for the NSAS database included patient character-istics, payment information, surgical and nonsurgicalprocedures, surgical visit information (e.g., perioperativetimes, anesthesia provider, type of anesthesia), and patientdisposition.

In 2006, data were collected for approximately 52,000ASC visits at 437 centers (142 hospital-based and 295freestanding centers), with an overall response rate of 74%of sampled centers.6 Survey responses were received from75% of sampled hospital-based ASCs and 74% of sampledfreestanding ASCs. In 1996, data were collected for 125,000ASC visits to 488 centers, with an overall response rate of81% of sampled centers.7 Survey responses were receivedfrom 91% of sampled hospital-based centers and 70% ofsampled freestanding ASCs.

Data Abstraction from the NSAS DatabaseWe abstracted data pertaining to type of anesthetic admin-istered, anesthesia provider present, procedure time vari-ables, primary procedure, gender, source of payment, anddischarge status. We combined the data of patients youngerthan 15 years with data from the population census toestimate the rate of visits to an ASC for ambulatoryprocedures with anesthesia for US pediatric patients. Agecategories were �1 year, 1 to 4 years, and 5 to 14 yearsbased on available census data. All statistical analyses wereconducted with Stata/SE 10.1 software (StataCorp LP,College Station, TX). Where data were missing, we catego-rized the result as “not specified” (e.g., for the anesthesiaprovider category in 1996 data).

Data Abstraction from the National HospitalDischarge SurveyTo help interpret the trends observed in the utilization ofambulatory surgery facilities, we abstracted a limitedamount of information from the National Hospital Dis-charge Survey. The survey is a national database of inpa-tient medical and surgical care that is similar to the NSASdatabase.8,9 It does not include information on the admin-istration of anesthesia during procedures performed atinpatient facilities, and therefore, we could not differentiatethe procedures performed with anesthesia from the nonin-vasive procedures (including imaging studies) or proce-dures performed without anesthesia. To help interpret thechange in rate of utilization of ASCs for surgical proce-dures, we abstracted the number of inpatient visits in both1996 and 2006 for which tonsillectomy or adenoidectomy,or both, was listed as the first procedure. We combined thedata on inpatients younger than 15 years with populationcensus data to estimate the rate of these procedures.

RESULTSUtilization of ASCs for Children, 2006In 2006, 2,300,651 (standard error [SE], 315,651) ambulatoryanesthesia episodes of care were performed for patientsyounger than 15 years in the US, which is a rate of 38ambulatory anesthetic procedures per 1000 children (Fig.1). Among these cases, anesthetics were given to 1,329,976(SE, 160,647) boys and 1,071,650 (SE, 168,697) girls, or ratesof 43 (SE, 5.2) per 1000 boys and 36 (SE, 5.7) per 1000 girls.Data by age group and type of anesthesia are provided inTable 1.

The 3 most frequently performed procedures were ton-sillectomy, adenoidectomy, and myringotomy with eartube.6 Data regarding the provider of anesthesia are dis-played in Table 2.

Perioperative DataThe breakdown of perioperative times is displayed inFigure 2. Of the children who received anesthetics, 12,030were admitted postoperatively to an inpatient facility (dataon those patients readmitted after discharge were notavailable), for a rate of 6 (SE, 1.3) inpatient admissions per1000 ambulatory anesthetics. An estimated 2,193,686 (SE,311,507) of the 2,401,626 children receiving ambulatory

Figure 1. Rate of ambulatory anesthesia for children in 1996 and2006. Rate increased from 26 per 1000 children younger than 15years in 1996 to 38 per 1000 children of this age group in 2006.

Ambulatory Anesthesia in Children


anesthesia were recorded as having routine discharge (913of 1000 ambulatory anesthetics; SE, 138).

Payment Information, 2006In 2006, the cost of 1,547,744 visits to ASCs for childrenyounger than 15 years was paid by private or commercial

insurance or through self-pay. For the other visits, the costof 816,185 visits was paid through public forms of funding(e.g., Medicaid, TRICARE). Of the visits for which fundingwas known, the cost for 65% of visits was paid from aprivate or commercial source and for 35% of visits from agovernment source.

Utilization of ASCs for Children, 1996In 1996, an estimated 1,522,883 ASC visits included anes-thesia administration, which is a rate of 26 ambulatoryanesthetic procedures per 1000 children younger than 15years. Data by age group and type of anesthetic areprovided in Table 1. Data regarding the provider of anes-thesia are displayed in Table 2.

Payment Information, 1996In 1996, most (1,142,481) of the ASC visits for children werefunded through private or commercial insurance or self-pay; 494,665 (30%) were funded through public sources(including Medicaid and TRICARE). Of the visits for whichfunding was known, 70% of visits were paid from a privateor commercial source and 30% from a government source.

Rate of Inpatient and Ambulatory Tonsillectomyand Adenoidectomy, 1996 and 2006The rate of inpatient tonsillectomy or adenoidectomy, orboth, in 1996 was 0.39 (SE, 0.08) per 1000 children youngerthan 15 years. In 2006, it was 0.18 (SE, 0.04) per 1000

Figure 2. Mean perioperative times for children younger than 15years. Postoperative time accounted for the largest portion of theperioperative period during pediatric visits to ambulatory surgerycenters for surgical procedures. Room time � the difference be-tween total operating room time (from entrance into until exit out ofthe operating room, or 45 [2] minutes) and surgical time (from theoperation’s start to its finish, or 26 [1] minutes); postoperativetime � from entrance into until exit from the recovery room, or 71 (3)minutes; perioperative time � from entrance into the operating roomuntil exit from the recovery room.

Table 1. Ambulatory Anesthesia Sessions for Monitored Anesthesia Care or Regional or GeneralAnesthetics Only by Age Group, 2006 and 1996

Age, y MAC (SE) Regional (SE) General (SE) Total (SE)Overall rate

per 1000 children2006

�15 44,462 (10,149) 26,484 (7036) 2,241,985 (313,649) 2,300,651 (315,651) 38�1 —a —a

196,991 (36,173) 202,412 (36,363) 491–4 —a —a

963,733 (141,654) 974,915 (141,977) 605–14 38,215 (9823) 11,156 (3741) 1,081,261 (145,287) 1,123,295 (147,728) 28

1996b

�15 53,943 14,776 1,490,686 1,522,883 26�1 —a —a

138,661 140,639 371–4 12,283 3177 633,454 640,424 415–14 39,351 9338 718,571 741,820 19

MAC � monitored anesthesia care; SE � standard error.a Sample size was too small or SE was too large.b 1996 Data did not contain some of the survey sampling variables needed to accurately estimate the SEs and thus the SEs are not reported.

Table 2. Anesthesia Provider Involved During Admission to Ambulatory Center When Anesthesia WasProvided by an Anesthesiologist or CRNA Only, 2006 and 1996

Age, y Anesthesiologist only (SE) CRNA only (SE)Both anesthesiologist

and CRNA (SE)2006

�15 1,389,393 (209,784) 603,695 (158,713) 292,630 (52,055)�1 130,681 (23,159) 52,145a 18,375a

1–4 577,712 (89,577) 256,924a 135,772 (25,799)5–14 681,000 (104,418) 294,626 (69,382) 138,483 (26,847)

1996b

�15 936,944 219,716 314,919�1 95,883 17,738 24,9101–4 387,108 93,878 137,5965–14 453,953 108,100 152,413

CRNA � certified registered nurse anesthetist; SE � standard error.a Sample size too small or SE too large.b Data of 1996 did not contain some of the survey sampling variables needed to accurately estimate the SEs and thus the SEs are not reported.


children of that age. By comparison, the rate of ambulatorytonsillectomy or adenoidectomy, or both, in 1996 was 5.3per 1000 children younger than 15 years; in 2006, it was 9.7(SE, 2.0) per 1000 children of that age. Information by age isprovided in Table 3.

DISCUSSIONOver the 10 years between 1996 and 2006, pediatric visits toASCs during which anesthesia was administered increasedalmost 50%, from approximately 1.6 million in 1996 to 2.3million in 2006. During that period, the population ofpediatric patients increased only 5.3%, suggesting that theincrease in ASC visits requiring anesthesia was the result ofa change in overall utilization or a shift in practice frominpatient to outpatient, or both. Overall utilization in-creased from 26 to 38 ASC visits per 1000 children, repre-senting an almost 40% increase.

Whether this increase in rate of ambulatory anesthesia isattributable to an increase in surgical procedures or a shiftof procedures from inpatient to outpatient settings hasimportant implications for health care spending. No dataare available that permit a direct comparison of inpatientand outpatient utilization rates for procedures requiringanesthesia.

Therefore, we abstracted the rate of either tonsillectomyor adenoidectomy and of both procedures from the NSASdatabase and the National Hospital Discharge Survey da-tabase, because tonsillectomy and adenoidectomy are com-mon pediatric procedures that may be performed in aninpatient or an outpatient setting and always require anes-thesia. The rate of these procedures as an inpatient opera-tion decreased approximately 54% from 1996 to 2006whereas the rate for the ambulatory setting increased 82%.This change suggests that there may have been a shift ofprocedures from the inpatient, short-stay hospitals to thehospital-based and freestanding ASCs during these 10years. This shift is consistent with data from the MedicareOnline Survey Certification and Reporting System and the

American Hospital Association Annual Surveys of Hospi-tals, which showed a 28% increase in hospital-based out-patient surgery and a 4.5% decrease in inpatient surgeryfrom 1993 to 2001.2 However, these data must be inter-preted with caution because there may be a differentexplanation for this change. For example, surgeons mayschedule tonsillectomies as outpatient procedures in chil-dren who stay overnight for payment reasons.

During both 1996 and 2006, the highest rate of ASC visitswith general anesthesia administration was in the 1 to 4years age group and the lowest rate was in the 5 to 14 yearsage group. Most of the ambulatory pediatric anesthesia wasdelivered by an anesthesiologist in both time periods (74%in 2006 and 85% in 1996). However, with the increased useof ambulatory anesthesia, the proportion of anestheticsprovided by a certified registered nurse anesthetist aloneincreased whereas the proportion of anesthetics providedby a certified registered nurse anesthetist working with ananesthesiologist decreased (Fig. 3). Nongovernmentalgroups (private and commercial insurance and self-pay)were the funding source for most visits in both 1996 and2006.

Economic and Educational ImplicationsThis study is an example of how a database can be used toabstract data useful to health care policy makers, adminis-trators, and educators and to provide important informa-tion when changes have to be made in health care systems.

The increase in ambulatory anesthesia itself may beinterpreted as an increase in health care spending. How-ever, it may be associated with a decrease in inpatientanesthesia, which could decrease health care expenditures.4

If this trend continues, further savings may occur.The dramatic increase in pediatric ambulatory surgery

has direct implications for residency and fellowship train-ing, and this effect may be the most important impact ofthis trend. Currently, programs are based at inpatientmedical centers, and training at ambulatory anesthesiacenters may be limited. As pediatric anesthesia shifts tooutpatient and ambulatory centers, education for residentsand fellows may need to be adapted to adequately prepareanesthesiologists to manage the unique challenges of am-bulatory anesthesia in children.10,11

Table 3. Rates of Tonsillectomy orAdenoidectomy, or Both, per 1000 ChildrenPerformed on an Ambulatory Basisa and anInpatient Basis,b 2006 and 1996

Age, y

Rate per 1000 children (SE)

Ambulatoryc Inpatient2006

�15 9.7 (2.0) 0.18 (0.04)�1 —d —d

1–4 13.2 (2.8) —d

5–14 9.2 (2.0) —d

1996�15 5.3 0.39 (0.08)�1 —d —d

1–4 6.2 —d

5–14 5.5 —d

SE � standard error.a From National Survey of Ambulatory Surgery data.b From National Hospital Discharge Survey data.c Ambulatory data from 1996 did not contain some of the survey samplingvariables needed to accurately estimate SEs and thus SEs are not reported.d Sample size was too small or SE too large.

Figure 3. Provider of ambulatory anesthesia for children in 2006 and1996. An anesthesiologist was involved in most anesthesia epi-sodes for ambulatory surgery in both time periods (61% in 2006 and64% in 1996). CRNA � certified registered nurse anesthetist.

Ambulatory Anesthesia in Children


LimitationsThe main limitations of this study are those inherent to theNSAS database and the medical charts that were reviewedfor it, because our study was reliant on data collected by theNational Center for Health Statistics for the NSAS database.There was an average response rate of 74% by sampledhospitals in 2006 and 81% in 1996. Data were extracted fromthe medical records of sampled patients by nonmedicalpersonnel after training,5 and it is possible that the medicalabstract form (Appendix) was not uniformly interpreted.This process was also limited by the data that wereavailable and retrievable from the medical records. Infor-mation was missing for some cases; specifically, the sourceof funding was unknown for a large portion of the pediatricambulatory visits in 2006.

The statistical software we used could abstract data onlyfor specific visits and the primary procedure during thevisit. These visits potentially could have included multipleprocedures and anesthetics that were counted as 1 visit.Sample size was limited in the pediatric population and,therefore, further data could not be reported because ofunacceptable standard errors. Also, the 1996 and 2006NSAS medical abstracts were not identical. For example,the “not-specified” field used in 2006 was not used in 1996,and thus “not specified” in 1996 was defined as no otherfield filled. Options for payment source were slightlydifferent in the 2 time periods, and therefore, comparisonscannot be made for this category.

In addition, sampling variables were not available forthe 1996 NSAS database and thus accurate standard errorscould not be calculated for 1996 data. This lack of samplingvariables limited the comparisons that we could makebetween the 2 time periods. Percentages do not add up to100% because all data represent estimates based on sam-pling rates and population size.

CONCLUSIONSThe rate of ambulatory anesthesia for children in the USincreased by �40% over a decade, partly because of a shiftin procedures from an inpatient to an outpatient setting.These databases are useful to health care policy makers,

educators, and administrators, as well as other partiesinvolved in health care organization and provision. Thistype of information is currently of particular importance inthis era of health care reform when, to make decisionsregarding health care spending and reform, data on utili-zation of all aspects of health care are needed from allgroups.

APPENDIXMedical Abstract Form of the National Survey of Ambula-tory Surgery, NSAS-5 (2-1-2006). (Adapted from US CensusBureau and US Department of Commerce. Available at:http://www.cdc.gov/nchs/data/hdasd/nsas_participant/nsas5.pdf.)

REFERENCES1. Pregler JL, Kapur PA. The development of ambulatory anes-

thesia and future challenges. Anesthesiol Clin North Am2003;21:207–28

2. Bian J, Morrisey MA. Free-standing ambulatory surgery cen-ters and hospital surgery volume. Inquiry 2007;44:200–10

3. Medicare Payment Advisory Commission (MedPAC). June 2008Healthcare Spending and the Medicare Program: A Data Book.Available at: http://www.medpac.gov/documents/Jun08DataBook_Entire_report.pdf. Accessed February 16, 2010

4. Medicare Payment Advisory Commission (MedPAC). Reportto the Congress: Medicare Payment Policy. Available at:http://www.medpac.gov/documents/Mar09_EntireReport.pdf. Accessed February 16, 2010

5. McLemore T, Lawrence L. Plan and operation of the NationalSurvey of Ambulatory Surgery. Vital Health Stat 11997;37:I–IV, 1–124

6. Cullen KA, Hall MJ, Golosinskiy A. Ambulatory surgery in theUnited States, 2006. Natl Health Stat Report 2009;11:1–25

7. Hall MJ, Lawrence L. Ambulatory surgery in the United States,1995. Adv Data 1997;296:1–15

8. DeFrances CJ, Lucas CA, Buie VC, Golosinskiy A. 2006 Na-tional Hospital Discharge Survey. Natl Health Stat Report2008;5:1–20

9. Graves EJ, Owings MF. 1996 summary: National HospitalDischarge Survey. Adv Data 1998;301:1–12

10. Emhardt JD, Saysana C, Sirichotvithyakorn P. Anesthetic con-siderations for pediatric outpatient surgery. Semin PediatrSurg 2004;13:210–21

11. Twersky RS. Educational protocols in ambulatory anesthesia.Ambul Surg 1997;5:117–9


REVIEW ARTICLE

CME

The Anesthetic Considerations of TracheobronchialForeign Bodies in Children: A Literature Review of12,979 CasesChristina W. Fidkowski, MD,* Hui Zheng, PhD,† and Paul G. Firth, MBChB*‡

Asphyxiation by an inhaled foreign body is a leading cause of accidental death among childrenyounger than 4 years. We analyzed the recent epidemiology of foreign body aspiration andreviewed the current trends in diagnosis and management. In this article, we discuss anestheticmanagement of bronchoscopy to remove objects. The reviewed articles total 12,979 pediatricbronchoscopies. Most aspirated foreign bodies are organic materials (81%, confidence interval[CI] � 77%–86%), nuts and seeds being the most common. The majority of foreign bodies (88%,CI � 85%–91%) lodge in the bronchial tree, with the remainder catching in the larynx or trachea.The incidence of right-sided foreign bodies (52%, CI � 48%–55%) is higher than that of left-sidedforeign bodies (33%, CI � 30%–37%). A small number of objects fragment and lodge in differentparts of the airways. Only 11% (CI � 8%–16%) of the foreign bodies were radio-opaque onradiograph, with chest radiographs being normal in 17% of children (CI � 13%–22%). Althoughrigid bronchoscopy is the traditional diagnostic “gold standard,” the use of computerizedtomography, virtual bronchoscopy, and flexible bronchoscopy is increasing. Reported mortalityduring bronchoscopy is 0.42%. Although asphyxia at presentation or initial emergency bronchos-copy causes some deaths, hypoxic cardiac arrest during retrieval of the object, bronchial rupture,and unspecified intraoperative complications in previously stable patients constitute the majorityof in-hospital fatalities. Major complications include severe laryngeal edema or bronchospasmrequiring tracheotomy or reintubation, pneumothorax, pneumomediastinum, cardiac arrest,tracheal or bronchial laceration, and hypoxic brain damage (0.96%). Aspiration of gastriccontents is not reported. Preoperative assessment should determine where the aspirated foreignbody has lodged, what was aspirated, and when the aspiration occurred (“what, where, when”).The choices of inhaled or IV induction, spontaneous or controlled ventilation, and inhaled or IVmaintenance may be individualized to the circumstances. Although several anesthetic tech-niques are effective for managing children with foreign body aspiration, there is no consensusfrom the literature as to which technique is optimal. An induction that maintains spontaneousventilation is commonly practiced to minimize the risk of converting a partial proximal obstructionto a complete obstruction. Controlled ventilation combined with IV drugs and paralysis allows forsuitable rigid bronchoscopy conditions and a consistent level of anesthesia. Close communica-tion between the anesthesiologist, bronchoscopist, and assistants is essential. (Anesth Analg2010;111:1016–25)

Aspiration of foreign bodies by children is a commonproblem around the world. Asphyxiation frominhaled foreign bodies is a leading cause of acci-

dental death among children younger than 4 years. Duringthe 19th century, treatment of foreign body aspiration by

purges, bleeding, and emetics were largely ineffective.Mortality was estimated at 23%. This rate plummeted withthe development of bronchoscopic techniques for the re-moval of these foreign bodies.1 In 1897, Gustav Killian, aGerman otolaryngologist, performed the first bronchos-copy using a rigid esophagoscope to successfully remove apig bone from a farmer’s right main bronchus.1,2 AlgernonCoolidge performed the first successful removal of a trachealforeign body in the United States at the Massachusetts Gen-eral Hospital in 1898.1 Shortly thereafter, Chevalier Jacksondeveloped the lighted bronchoscope and several special-ized instruments for the removal of foreign bodies.3 Hepioneered developments in the field and saved the lives ofhundreds of children who had aspirated objects.4 Whileearly clinicians used topical anesthesia, general anesthesiabecame commonplace for the removal of aspirated objectswith increased experience with the rigid bronchoscope andadvances in anesthetic delivery. The flexible bronchoscopewas introduced by Shigeto Ikeda in 1966,5 and the removalof an airway foreign body using this instrument wasreported in the 1970s.6

From the *Department of Anesthesia, Critical Care and Pain Medicine, andthe †Biostatistics Center, Massachusetts General Hospital, Boston; and the‡Department of Anesthesia, Massachusetts Eye and Ear Infirmary, Boston.

Christina W. Fidkowski is now affiliated with the Department of Anesthe-siology, Henry Ford Hospital, Detroit, Michigan.


Statistical and administrative support funded by the Department of Anes-thesia, Critical Care and Pain Medicine, Massachusetts General Hospital,Boston, and the Department of Anesthesia, Massachusetts Eye and EarInfirmary, Boston.


Address correspondence and reprint requests to Paul G. Firth, Departmentof Anesthesia, Critical Care and Pain Medicine, Massachusetts GeneralHospital, Boston, MA 02114. Address e-mail to [email protected].

Copyright © 2010 International Anesthesia Research SocietyDOI: 10.1213/ANE.0b013e3181ef3e9c


Over the years, the steady accumulation of clinicalreports has provided greater insight into the managementof foreign body aspiration in children. In part as a result,morbidity and mortality from foreign body aspiration havedrastically diminished. This article analyzes the recentepidemiology of foreign body aspiration, reviews the cur-rent trends in diagnosis and management, and discussesanesthetic management for bronchoscopy.

CURRENT TRENDS IN FOREIGN BODY ASPIRATIONA preliminary Medline search of the literature from 1950 to2009 yielded nearly 20,000 cases of foreign body aspirationin children. Before 2000, however, most case series weresmall, included both aspirated and ingested objects,mingled adult and pediatric patients, or were reported invarious styles in differing specialist journals that precludedmeta-analysis. In addition, anesthetic and surgical tech-niques have altered considerably in recent years, making adetailed review of older series less relevant. However,numerous large case series have been published recentlythat collectively allow for a clearer representation of theproblem of pediatric aspiration, as well as current trends inmanagement. A Medline search using the keywords foreignbody aspiration with limits of the year 2000 to present wasperformed on October 1, 2009. Analysis was limited tostudies (1) containing only patients with suspected orproven foreign body aspiration, (2) with �100 patients, (3)containing only children ages 18 years and younger, and (4)written in English. Of the 698 articles obtained, 33 met theinclusion criteria, of which 3 were excluded for containingduplicate patient data. The 30 articles reviewed report12,979 children with suspected foreign body aspiration, ofwhom 11,145 were found to have aspirated an object.7–36

Twenty-seven of these studies are retrospective analyses,and 3 studies are prospective analyses of all children withsuspected or actual foreign body aspiration.12,13,19 Thecases in these series occurred within the last 20 years in 21studies, 8,9,11,12,14–18,20,21,23–26,28–31,33,34 8 covered a largertime span dating back to the 1980s,10,13,19,22,27,32,35,36 and 1study covered a 30-year time frame starting in 1973.7

To obtain a robust estimate of the various rates (truepositive, gender, foreign body type and location, andradiographic outcome distributions), we applied a meta-analysis to the published data to account for the number ofpatients, the number of foreign body cases, and the numberof outcomes reported in these 30 articles.7–36 These meta-analyses use a Bayesian model to consider variations instudy design, inclusion and exclusion criteria, and thestudy population among the different reported studies.

Most patients with aspirated foreign bodies are childrenyounger than 3 years. Four series reported the median age,and 7 series reported the mean age of children withaspirated foreign bodies. The median and the mean ageranged from 1 to 2 years12,18,21,33 and from 2.1 to 3.8years,11,12,14,26,28,29,32 respectively. With the exception of 2Turkish studies with a high incidence of adolescent girlsaspirating headscarf pins,24,34 boys account for 61% (confi-dence interval [CI] � 59%–63%) of children with foreignbody aspiration.7,9,17,21,22,26–29,32,33

Most (81%, CI � 77%–86%) of the aspirated foreignbodies are organic materials.7,9–11,19–21,23,24,26–36 Nuts (es-pecially peanuts) and seeds (mainly sunflower and water-melon) are the most commonly aspirated foreign bodiesreported in almost all studies. The exception to this findingis an Italian series that found teeth to be the most com-monly aspirated objects (33/121).13 In adolescent Turkishfemales, headscarf pins are commonly aspirated.24,34 Aswas reported in 24 studies, the majority of foreign bodies(88%, CI � 85%–91%) lodge in the bronchial tree, with theremainder catching in the larynx or trachea.7,9,10,12–19,22,25–36 Ahigher incidence of right-sided foreign bodies (52%, CI �48%–55%) in comparison with left-sided foreign bodies (33%,CI � 30%–37%) was reported in all of these studies, with theexception being a small series in Israel.12 A small number ofobjects fragment and lodge in different parts of the airways.

A history of a witnessed choking event is highly sug-gestive of an acute aspiration. Data were available todetermine the sensitivity, specificity, positive predictivevalue, and negative predictive value of a witnessed event in10 studies (Table 1).7,8,11,12,16,20,25,31,34,35 Children with as-pirated foreign bodies typically present with the symptomsof coughing, dyspnea, wheezing, cyanosis, or stridor. Datawere available in 10 studies to determine the sensitivity andthe specificity of these presenting signs and symptoms(Table 2).8,12,16,20,23,25,29,31,34,35 A history of cough is highlysensitive for foreign body aspiration but is not very specific.On the other hand, a history of cyanosis or stridor is veryspecific for foreign body aspiration but is not very sensitive.Signs and symptoms typical in delayed presentations in-clude unilateral decreased breath sounds and rhonchi,persistent cough or wheezing, recurrent or nonresolvingpneumonia, or rarely, pneumothorax.

Table 1. Sensitivity, Specificity, PositivePredictive Value (PPV), and Negative PredictiveValue (NPV) of a Witnessed Aspiration Effect forForeign Body Aspiration

Sensitivity Specificity PPV NPV

Aydogan et al.(1887, 1493)7

93.2 45.1 86.5 63.6

Ciftci et al.(663, 563)11

91.1 46.0 90.5 47.9

Tomaske et al.(370, 221)35

74.7 53.7 70.5 58.8

Ayed et al.(235, 206)8

81.6 37.9 90.3 22.4

Tokar et al.(214, 152)34

84.9 87.1 94.2 70.1

Skoulakis et al.(210, 130)31

91.5 56.3 77.3 80.4

Kiyan et al.(207, 153)25

37.3 96.3 96.6 35.1

Erikci et al.(189, 127)16

58.3 87.1 90.2 50.5

Heyer et al.(160, 122)20

75.4 92.1 96.8 53.8

Cohen et al.(142, 61)12

83.6 32.1 48.1 72.2

Values are percentages.Data were available from 10 of the 30 studies that were reviewed. Study sizeis denoted (n, n) to represent the total number of patients and the number ofpatients with an aspirated foreign body, respectively.

Anesthetic Considerations of Tracheobronchial Foreign Bodies


Most stable patients had chest radiographs. As was re-ported in 20 studies, only 11% (CI � 8%–16%) of the foreignbodies were radio-opaque.7,9,12–17,19,20,23–25,27–29,32,34–36 Chestradiographs were normal in 17% (CI � 13%–22%) of childrenwith aspirated objects as were reported in 14 studies.9,11,12,17,22,24,25,27–29,32–35 The common radiographic abnor-malities included localized emphysema and air trapping,atelectasis, infiltrate, and mediastinal shift. Data were avail-able in 8 studies to calculate the sensitivity and specificity ofthese radiographic findings (Table 3).11,12,20,23,25,29,31,34 Pneu-mothorax and pneumomediastinum are less common find-ings on chest radiograph (range: 0.1%–3.7%), as was reportedin 7 studies.14,15,22,28,32,33

While rigid bronchoscopy was used solely for the re-moval of foreign bodies in most studies, both flexible andrigid bronchoscopies were used in 4 series.12,13,20,33 Aminority of foreign bodies were removed by flexible bron-choscopy in 3 of these studies (range: 4.1%–10.7%),12,13,20

whereas Tang et al33. reported successful removal byflexible bronchoscopy in 91.3% of children with foreignbody aspiration. For this study, local anesthesia with seda-tion was used during bronchoscopy. For children undergo-ing rigid bronchoscopy, general anesthesia was used in allstudies, and details regarding the anesthetic techniquewere provided in 12 studies. Both inhaled7,15,31,36 andIV13,19 inductions were reported. Similarly, anesthesia wasmaintained with either inhaled15,19,31,36 or IV20,22,25 drugsor a balanced anesthetic.13 Five studies reported the use ofneuromuscular blockers.7,9,15,19,22 Bittencourt et al.9 and

Hasdiraz et al.19 used paralysis as needed during theprocedure and attempted to maintain spontaneous ventila-tion when possible. On the other hand, Divisi et al.13

commented that spontaneous ventilation is not suitable forrigid bronchoscopy because of resultant oxygen desatura-tion and used a balance anesthetic with sevoflurane andremifentanil. Shivakumar et al.29 used jet ventilation toprevent oxygen desaturation. None of the authors com-mented on using drying drugs such as glycopyrolate beforebronchoscopy. Seven studies commented on using steroidsfor laryngeal edema, with the majority of those authorsfavoring steroid use only as needed,7,13,19,22,30 as opposedto routine administration.25,36 In 4 studies, antibiotics weregiven routinely preoperatively,19,25 postoperatively,36 or asa 5-day course,30 whereas authors in 3 studies favoredantibiotic administration only as needed for infection.7,13,24

Major iatrogenic complications were specified in 21studies with 9437 children with aspirated foreign bodies.The other 9 studies did not provide details or rates ofcomplications. These complications included severe laryn-geal edema or bronchospasm requiring tracheotomy orreintubation, pneumothorax, pneumomediastinum, cardiacarrest, tracheal or bronchial laceration, and hypoxic braindamage. These major complications occurred in 91 of these9437 children (0.96%) (Table 4). Of the 11 cardiac arreststhat were reported, 1 occurred after induction of anesthesiain a child who was hypoxic on admission, 5 occurredduring bronchoscopy because of hypoxia (3) or bleeding(2), and the remaining 5 were not specified. Other reported

Table 2. Sensitivity (Sens) and Specificity (Spec) of Symptoms for Foreign Body AspirationCough Dyspnea Wheeze Cyanosis Stridor

Sens Spec Sens Spec Sens Spec Sens Spec Sens SpecTomaske et al. (370, 221)35 87.8 45.0 57.9 73.2 39.4 74.5Ayed et al. (235, 206)8 80.1 34.5 30.1 65.5 16.5 65.5Tokar et al. (214, 152)34 94.1 32.3 27.6 66.1Skoulakis et al. (210, 130)31 82.3 53.8 24.6 85.0 5.4 100 11.5 98.8Kiyan et al. (207, 153)25 67.3 20.4 16.3 74.1 79.1 27.8 7.2 98.1Erikci et al. (189, 127)16 51.2 83.9 4.7 93.5 18.9 93.5Shivakumar et al. (165, 105)29 92.4 8.3 61.9 66.7 64.8 0 12.4 100 4.8 100Heyer et al. (160, 122)20 41.0 55.3 33.6 68.4Kadmon et al. (150, 80)23 51.3 12.9 18.8 72.9Cohen et al. (142, 61)67 93.4 28.4 14.8 92.6

Values are percentages.Data were available from 10 of the 30 studies that were reviewed to determine the sensitivity (Sens) and specificity (Spec) of the symptoms of cough, dyspnea,wheeze, cyanosis, and stridor for foreign body aspiration. Study size is denoted (n, n) to represent the total number of patients and the number of patients withan aspirated foreign body, respectively.

Table 3. Sensitivity (Sens) and Specificity (Spec) of Radiographic Findings for Foreign Body AspirationAir trapping Atelectasis Mediastinal shift Infiltrate

Sens Spec Sens Spec Sens Spec Sens SpecTokar et al. (214, 152)34 41.7 91.9 12.6 71.0 11.9 74.2Skoulakis et al. (210, 130)31 39.2 91.6 9.2 88.8 0 76.3Kiyan et al. (207, 153)25 63.8 79.6 8.0 94.4 4.4 94.4Shivakumar et al. (165, 105)29 49.5 80.0 22.9 83.3 3.8 41.7Heyer et al. (160, 122)20 62.3 97.4 8.2 97.4 20.5 97.4 18.9 84.2Kadmon et al. (150, 80)23 50.0 90.0 38.8 97.1Cohen et al. (142, 61)67 49.2 86.4 6.6 96.3 13.1 100 14.8 79.0

Values are percentages.Data were available from 8 of the 30 studies that were reviewed to determine the sensitivity (Sens) and specificity (Spec) of the radiographic findings of localizedair trapping, atelectasis, mediastinal shift, and infiltrate for foreign body aspiration. Study size is denoted (n, n) to represent the total the number of patients andthe number of patients with an aspirated foreign body, respectively.

REVIEW ARTICLE


serious complications included infection, bleeding, andfailed bronchoscopic removal that necessitated thora-cotomy or tracheotomy to remove the object (Table 4).

Mortality data were obtained from 26 articles with 43deaths among 10,236 children (0.42%) with aspirated for-eign bodies (Table 5). The remaining 4 articles did notprovide details of death rates. Twenty-five deaths occurredin the 5 largest series with 5927 children (0.42%).7,15,19,22,33

In 2003, Eren et al.15 reported 10 deaths in 1068 children(0.94%) undergoing rigid bronchoscopy for foreign bodyremoval under general anesthesia in Turkey. Seven died ofhypoxic arrest during bronchoscopy, 2 of bronchial rup-ture, and 1 of intractable bronchospasm. Shortly thereafter,their countrymen Aydogen et al. reported 4 deaths in 1493children (0.27%) with foreign body aspiration undergoingrigid bronchoscopy over a 31-year period.7 All 4 fatalities

were in children presenting with acute cyanosis and respi-ratory distress. Hasdiraz et al.19 reported 8 deaths in 911Turkish children (0.88%) with foreign body aspirationundergoing rigid bronchoscopy. Three children developedcardiac arrest from total tracheal obstruction, 2 had heartfailure and bronchopneumonia at the time of bronchoscopyand developed cardiac arrest postoperatively, 1 developedrespiratory arrest due to inhalation of cement powder, 1developed sepsis and respiratory failure after explosiverelease of purulent discharge from behind the foreign body,and 1 developed a respiratory arrest after a negativebronchoscopy and was found to have a tracheal foreignbody at autopsy. In 2008, Hui et al.22 reported 3 deathsamong 1428 children (0.21%) undergoing rigid bronchos-copy over a 22-year period in China. Two died after foreignbody displacement during bronchoscopy, and 1 died ofasphyxia during a delay before bronchoscopy. In 2009,Tang et al.33 reported no deaths among 1027 children inChina undergoing bronchoscopy for foreign body removal.In that series, 938 foreign bodies were removed by flexiblebronchoscopy, and 89 foreign bodies were removed byrigid bronchoscopy. Of the remaining 18 deaths in the 21other reports, 10 were due to irreversible cardiac arrest onadmission.9,11,27,32,36

DIAGNOSIS AND MANAGEMENTA suggestive history is important in diagnosing an aspi-rated object, because it is often difficult to make a definitivediagnosis on the basis of an abnormal physical examinationor radiological studies alone. The work-up of the stablepatient should include a chest radiograph to assess forother potential causes of symptoms, to identify a radio-opaque foreign body, or to detect the position of a foreignbody on the basis of localized emphysema and air-trapping,atelectasis, infiltrate, or mediastinal shift.37 The common ab-normality of unilateral hyperinflation seen on the chest radio-graph due to air trapping behind the foreign body is bestviewed at end expiration (Fig. 1). (Video 1; see SupplementalDigital Content 1, http://links.lww.com/AA/A169; see theAppendix for video legends). Although a decubitous view hasbeen suggested to look for air trapping in the dependent lungof small children who cannot cooperate with expiratory films,one study found this to be an unreliable technique.38

Neck radiographs may be helpful in managing upperaerodigestive tract foreign bodies. The alignment of flatobjects, such as coins, may suggest the location of an object(Fig. 2A, 2B). Tracheal objects tend to align in the sagittalplane, whereas esophageal objects tend to align in theanterior plane. An object that overlaps the boundaries ofthe airway on an anterior–posterior view is unlikely to beinside the airway. Lateral radiographs may show soft tissueswelling, loss of cervical lordosis, or an object posterior tothe trachea. Proximal esophageal objects can be removedwith a forceps under direct vision, with the laryngoscopeblade inserted into the esophagus to visualize the body andprotect the airway during removal of the object.

Thoracic computed tomography (CT) and virtualbronchoscopy—a reformatted 3-dimensional CT image thatgenerates intraluminal views of the airway to the sixth andseventh generation bronchi—are emerging as new modali-ties to diagnose tracheobronchial foreign bodies in children

Table 4. Morbidity Associated with Bronchoscopyfor the Removal of TracheobronchialForeign Bodies

Complication Total n

Major nonfatal complications (n � 91)Severe laryngeal edema or bronchospasm

requiring tracheotomy orreintubation15,19,22,25,27,30,32

43

Pneumothorax orpneumomediastinum7,11,15,18,19,22,30,32,33

27

Cardiac arrest11,15,25,27 11Hypoxic brain damage20,21 5Tracheal or bronchial laceration requiring

repair11,15,275

Other serious complications (n � 136)Infection13,19,21,26,32 58Failed bronchoscopy requiring thoracotomy

(27)7,8,11,13,15,19,36 or tracheotomy (10)7,1537

Bleeding15,19,27 29Thoracotomy (5)8,15,19 or tracheotomy

(7)7,32—not specified12

Major iatrogenic complications, as were specified in 21 studies, occurred in91 of the 9437 children with aspirated foreign bodies. Other seriouscomplications occurred in 136 of these 9437 children.

Table 5. Mortality Associated with Bronchoscopyfor the Removal of TracheobronchialForeign Bodies

Cause of death (n � 43) nCardiac/respiratory arrest 37

Hypoxic arrest at presentation7,9,11,22,27,32,36 15Arrest due to tracheal foreign body19,26 5Cardiac arrest during bronchoscopy, not

specified10,113

Postoperative arrest19,29 3Hypoxic arrest due to shifting foreign body22 2Rupture of puss under pressure behind

foreign body191

Respiratory arrest due to inhaled cementpowder19

1

Not specified15 7Bronchial rupture15 2Severe bronchospasm15 1Postoperative infection11 1Multiorgan failure21 1Not specified18 1

Deaths, as were specified in 26 studies, occurred in 43 of the 10,236children with aspirated foreign bodies.



(Fig. 3).39,40 CT and virtual bronchoscopy are more sensi-tive diagnostic modalities for foreign body aspiration incomparison with conventional chest radiography.41,42 Se-cretions, tumors, or other obstructive lesions can produce

false-positive findings. In a retrospective analysis, spiral CTcorrectly identified all 42 children with aspirated foreignbodies.41 In that study, 3 children had false-positive CTimages due to excess bronchial secretions, and 6 children

Figure 1. A, Chest radiograph on end inspiration of a patient with a delayed presentation of an aspirated foreign body aspiration. B, Chestradiograph on end expiration. Delayed emptying of the left lung suggests local air trapping. The foreign body was in the left bronchus. C, Theoffending object seen on rigid bronchoscopy. The airway edema (white-gray) can be seen around the black foreign body, with bubbles reflectingdelayed air release during expiration. A Fogarty catheter is passed beyond the object in preparation for dislodgement. Images courtesy of Dr.Dan Doody, Department of Surgery, Massachusetts General Hospital, Boston, Massachusetts.

Figure 2. A, The sagittal orientation of aproximal aerodigestive foreign body sug-gests an esophageal location. B, A lat-eral view can demonstrate a positionposterior to the esophagus. Images cour-tesy of Dr. Allan Goldstein, Department ofSurgery, Massachusetts General Hospi-tal, Boston, Massachusetts.

Figure 3. A, B, Computerized tomography scan of an aspirated soda can top, using a low-resolution pediatric protocol to minimize radiationexposure. The object was not seen on initial chest radiograph. A small aluminum object, although metal, has insufficient radiopacity for a plainchest radiograph, and the object did not produce major obstruction leading to overt pulmonary changes. A computed tomography (CT) scan hasa greater range of sensitivity. C, The offending object in the bronchus intermedius. (Images courtesy of Dr. Pallavi Sagar, Department ofRadiology, and Dr. David Lawlor, Department of Surgery, Massachusetts General Hospital, Boston, MA.)

REVIEW ARTICLE


had true negative scans. In 2 retrospective studies, virtualbronchoscopy correctly identified 11of 11 and 15 of 23children, respectively, with aspirated foreign bodies.42,43

No false-positive virtual bronchoscopies were reported inthose studies. The diagnostic utility of virtual bronchos-copy has also been shown prospectively.44,45 Haliloglu etal.45 demonstrated that virtual bronchoscopy findings cor-related with those of conventional bronchoscopy in 23children, of whom 7 had foreign body aspiration and 16 didnot have foreign body aspiration. In a prospective study of37 children with suspected foreign body aspiration, 16 hada positive virtual bronchoscopy, of whom 13 had a foreignbody found with conventional bronchoscopy, and 3 hadeither mucous plugs or a schwannoma found with conven-tional bronchoscopy.44 The remaining 21 patients had anegative bronchoscopy and were observed with improve-ment in their symptoms.44 These studies demonstrate thatCT and virtual bronchoscopy correctly identified all casesof foreign body aspiration. Therefore, some authors sug-gested that children with a negative CT and virtual bron-choscopy may not require conventional bronchoscopy as adefinitive work-up.44

A drawback of CT and virtual bronchoscopy is thepotential for excessive radiation exposure. A chest radio-graph exposes the child to 0.1 mSv of radiation, equivalentto several days of background environmental radiation.Although a high-resolution pediatric chest CT can involveup to 7 mSv of radiation, a lower-resolution scan protocolusing 1.5 mSv is usually sufficient to diagnose a foreignbody. Adequate 3-dimensional views can be subsequentlyformatted from this level of detail.

Further limitations include the cost and limited avail-ability of equipment and radiologists. In addition, CTexamination is limited to stable and cooperative children,because anesthesia in a remote location for a child with anunstable object that can potentially acutely obstruct theairway poses significant risks.

Although rigid bronchoscopy has traditionally been thedefinitive method to diagnose and remove tracheobron-chial foreign bodies, a diagnostic flexible bronchoscopyunder local anesthesia may be indicated for patients with-out a clear history or findings of aspiration.23,46–48 In aprospective study, children with convincing evidence offoreign body aspiration were examined with rigid bron-choscopy under general anesthesia, whereas others withless-suggestive findings underwent flexible bronchoscopywith local anesthesia.47 Of the 28 children who underwentrigid bronchoscopy, 23 (82%) had a foreign body aspiration.Of the 55 children who underwent flexible bronchoscopy,only 17 (34%) had a foreign body aspiration. Anotherprospective study found that 43 (84%) of 51 children whounderwent rigid bronchoscopy and only 7 (37%) of 19children undergoing flexible bronchoscopy had positivestudies for foreign body aspiration.48 Both studies found asignificant association of aspirated foreign bodies withunilateral decreased breath sounds, localized wheezing,and obstructive emphysema on chest radiograph.47,48

These authors recommended that children undergo rigidbronchoscopy only if they have acute asphyxiation, aradio-opaque foreign body, unilateral pulmonary signs, orobstructive emphysema. All other children should undergo

a diagnostic flexible bronchoscopy. When this algorithmwas applied retrospectively, the negative finding rate ofrigid bronchoscopy decreased from 18% to 4% and from16% to 6%, respectively.47,48 No adverse events were re-ported with flexible bronchoscopy.47,48 Therefore, diagnos-tic flexible bronchoscopy in selected children minimizes thepotential complications of rigid bronchoscopies. More re-cently, Kadmon et al.23 proposed a computer model basedon history, physical examination, and radiographic find-ings to calculate a score that predicts the likelihood offoreign body aspiration in children. They further suggestedan algorithm to observe a child, perform diagnostic flexiblebronchoscopy, or perform therapeutic rigid bronchoscopyon the basis of the calculated score. A prospective study iswarranted to determine the utility of this model.

In addition to aiding in the diagnosis of aspirated foreignbodies, flexible bronchoscopy is becoming more popular forthe removal of foreign bodies.33,49–52 In a large retrospectivestudy, a foreign body was successfully removed by flexiblebronchoscopy in 938 (91.3%) children.33 Flexible bronchos-copy is better suited for removing foreign bodies from distalairways and upper lobe bronchi, because of the smallerdiameter and greater flexibility in comparison with the rigidbronchoscope. Fewer instruments, however, are available foruse with the flexible bronchoscope to remove the foreignbodies. Rigid bronchoscopy continues to be used to removeaspirated foreign bodies because multiple extraction instru-ments are available and because it provides good visualiza-tion, controls the airway, and allows ventilation.

ANESTHETIC MANAGEMENT FORBRONCHOSCOPYAnesthetic considerations encompass preoperative assess-ment, management techniques for flexible or rigid bron-choscopy, and postbronchoscopic disposition.

Preoperative AssessmentThe preoperative assessment should determine where theaspirated foreign body has lodged, what was aspirated,and when the aspiration occurred. If the foreign body islocated in the trachea, the child is at risk for completeairway obstruction and should be taken urgently to theoperating room. Conversely, the risk of complete airwayobstruction is less if the object is firmly lodged beyond thecarina. It is important to determine the type of foreignbody: Organic materials can absorb fluid and swell, oilsfrom nuts cause localized inflammation, and sharp objectscan pierce the airway. The time since the aspiration shouldbe established because airway edema, granulation tissue,and infection may make retrieval more difficult with de-layed presentations. A recently aspirated object may moveto a different position with coughing.

The time of the last meal should be established to assessthe risk of aspiration. There are no reports of aspiration ofgastric contents in the literature surveyed, although fatalprogression of obstruction has been reported.7,9,11,22,27,32,36

In acute cases, therefore, the dangers of delayed removalappear to outweigh the risk of a full stomach in a well-conducted anesthetic. In urgent cases, the stomach can besuctioned through a large-bore gastric tube after induction butbefore the bronchoscope is inserted to minimize the risk of



gastric aspiration. In delayed presentations in which bron-choscopy is not urgent, a preanesthetic fast is appropriate.

The airway patency should be assessed. If the patient isin severe distress, urgent bronchoscopy should be per-formed. If the patient is stable, however, some authorssuggest that bronchoscopy may be performed during nor-mal daytime operating hours to ensure optimal conditionswith an experienced bronchoscopist and anesthesiologist.53

These authors found no increase in morbidity in stablepatients by delaying bronchoscopy for a suspected foreignbody until the next available elective daytime slot.53

Anesthetic Considerations for RigidBronchoscopyBecause surgeon and anesthesiologist share management of apotentially obstructed airway, clear communication and goodcooperation are essential. Before induction, a detailed anes-thetic and operative plan should be discussed. The 3 mainanesthetic issues involve the methods of induction, ventilationduring bronchoscopy, and maintenance of anesthesia.

The choice of induction is dominated by the consider-ation of converting a proximal partial obstruction into acomplete obstruction. The conversion from spontaneousnegative pressure breathing to positive pressure ventilationtheoretically risks dislodging an unstable proximal body,causing complete obstruction.54 Although hypoxic arrestduring the initial stages of bronchoscopy is a recognizedcause of death,10,11,19 the relative contributions of obstructionon initial presentation, during the induction of anesthesia, andfrom dislodgement during bronchoscopy, are unclear frompublished accounts. A survey of 838 pediatric anesthesiolo-gists found that the majority preferred an inhaled inductionwhen foreign bodies were present in the tracheobronchialtree.55 A cautious IV induction that maintains spontaneousventilation is also possible, although this was not an option inthat particular survey study. While the optimal method ofinduction is not definitively established, maintaining sponta-neous ventilation during the induction of a patient with aproximal foreign body is commonly practiced.

After induction of general anesthesia, the rigid broncho-scope is inserted through the glottic opening. The anesthesiacircuit is connected to the sideport of the bronchoscope toallow ventilation. Both spontaneous ventilation and con-trolled ventilation are feasible for removal of foreign bodies.Spontaneous ventilation around the bronchoscope may bemore suitable for removal of proximal bodies, during whichleakage around the scope may make effective positive pres-sure ventilation difficult. Manually closing the mouth andnose can diminish a large leak around the scope and improveventilation. Positive pressure ventilation down the broncho-scope, with intermittent apnea while manipulating the object,may be more suitable for distal retrieval. The use of opticalforceps allows for positive pressure ventilation to be main-tained while the foreign body is being manipulated so thatperiods of apnea can be minimized (Video 2; see Supplemen-tal Digital Content 2, http://links.lww.com/AA/A170; seethe Appendix for video legends). Because airway trauma andrupture are significant and potentially fatal complications, it isessential to avoid coughing and bucking secondary to theintense stimulation from a rigid bronchoscope deep in the

bronchial tree. Movement can be prevented with neuromus-cular blocking drugs9,54,56,57 or with a deep level of anesthesia.One study suggests that topicalization of the tracheobronchialmucosal using a rigid bronchoscope coated with local anes-thetic gel improves surgical conditions and more effectivelymaintains spontaneous ventilation while decreasing the dosesof anesthetics.58 Although the risk of positive pressure venti-lation causing distal air trapping by a ball-valve effect hasbeen suggested,59,60 there is no clear clinical evidence in theliterature surveyed to support this as a practical concern.

A retrospective review of 94 children with aspiratedforeign bodies detected no difference in adverse events onthe basis of the type of ventilation.61 However, 5 of 18children who were maintained on assisted ventilation and11 of 26 who were maintained on spontaneous ventilationwere switched to controlled ventilation. A prospectivestudy of 36 children with aspirated foreign bodies foundthat controlled ventilation is more effective than is sponta-neous ventilation.54 All children in the spontaneous venti-lation group were switched to either assisted or controlledventilation because of coughing and bucking. It is possible,however, that the necessity of switching from spontaneousto either assisted or controlled ventilation was due to aninadequate depth of anesthesia with inhaled drugs ratherthan an inherent problem with spontaneous ventilation.62

Larger prospective studies, with both inhaled and IVmaintenance techniques, are necessary to further evaluatewhether spontaneous or controlled ventilation is moreadvantageous. In a nonrandomized observational study,manual jet ventilation was shown to decrease the incidenceof intraoperative hypoxemia in comparison with manualcontrolled ventilation and spontaneous ventilation.63

Manual jet ventilation may better allow oxygenation andventilation of the unobstructed lung during manipulationof the foreign body because the jet ventilation catheter wasinserted separately from the bronchoscope.63

Halothane and sevoflurane are 2 volatile anesthetics thatare widely used in pediatric practice. Meretoja et al.64

compared sevoflurane with halothane in 120 childrenundergoing bronchoscopy, gastroscopy, or combinedprocedures. They reported a higher incidence of cardiacarrhythmias (nodal rhythm, bigeminy or ventricularectopy) in the halothane group (18/60 vs. 4/60). Batra etal.65 compared the 2 drugs in 44 children undergoingbronchoscopy specifically for foreign body removal andfound a higher incidence of cardiac arrhythmias in thehalothane group (7/22 vs. 2/22). When comparing halo-thane and sevoflurane for 62 pediatric bronchoscopies,Davidson66 found no differences in cord closure, desatura-tions, breath holding, or coughing.

Although inhaled drugs have traditionally been used forthe maintenance of anesthesia,54,59,61,67 total IV techniquesare becoming more popular in the pediatric popula-tion.56,62,68,69 A total IV anesthetic with propofol (200 to 400�g � kg�1 � min�1) and remifentanil (0.05 to 0.2 �g � kg�1 �min�1) infusions in combination with vocal cord topical-ization with lidocaine (1 mg � kg�1) allows for spontaneousventilation.62 Children younger than 3 years of age cantolerate higher doses of remifentanil and still maintainspontaneous ventilation in comparison with older chil-dren.70 An advantage of an IV anesthetic is that it provides

REVIEW ARTICLE


a constant level of anesthesia irrespective of ventilation. Bycontrast, hypoventilation and leaks around the rigid bron-choscope may produce an inadequate depth of inhaledanesthesia. Pollution of the operating room, due to thecombination of leaks around the rigid bronchoscope andhigh gas flows needed for ventilation, are additional draw-backs of inhalation anesthetics. Chen et al.63 showed that atotal IV technique with spontaneous ventilation was asso-ciated with a higher incidence of body movement, breathholding, and laryngospasm in comparison with an inhaledtechnique. However, the doses of IV propofol (100 to 150�g � kg�1 � min�1) and remifentanil (0.1 �g � kg�1 � min�1)were less than those previously described to provide anes-thesia and maintain spontaneous ventilation.

Dropping the foreign body during retrieval is a poten-tially life-threatening complication.71,72 The vocal cordsshould be well relaxed, either by residual topicalization,paralysis, or an adequate depth of anesthesia, before re-moval of the foreign body through the larynx. Droppingthe foreign body has a higher correlation with the experi-ence level of the bronchoscopist than with the mode ofventilation.72 If the object is dropped in the proximalairway and cannot immediately be removed, pushing itback into a bronchus can eliminate an obstruction. If abronchial body falls into the other bronchus, there ispotential for complete airway obstruction due to edemaand inflammation at the original site.71 In the setting of amarginal airway, optimization of other components ofventilation is essential. Ventilation may be impaired notonly by the object, but also proximally by upper-airway softtissue or cord closure, and distally by atelectasis afterprolonged intraoperative hypoventilation. Optimal headposition, open cords, reinflation of atelectatic segments,and slow prolonged breaths with adequate pressure canprovide ventilation past a partial obstruction. If ventilationis impossible, emergent efforts must be made to extract ormove the object. In severe cases of cardiopulmonary failuredue to foreign body obstruction, extracorporeal membraneoxygenation may facilitate foreign body removal and car-diopulmonary recovery.73

After the extraction of the foreign body and the removalof the rigid bronchoscope, the choice of ventilation duringemergence is influenced by pulmonary gas exchange andthe degree of airway edema. For uncomplicated cases,spontaneous ventilation assisted by mask ventilation asneeded may be adequate. Intubation during emergencemay be indicated for a marginal airway, pulmonary com-promise, or residual neuromuscular blockade.

Anesthetic Considerations forFlexible BronchoscopyFlexible bronchoscopy can be performed with local anes-thetic topicalization and sedation in both children andadults.47–49,74–77 IM meperidine and oral diazepam,75 IVmidazolam or fentanyl,75 and atropine and diazepam49,76

or sublingual codeine74 have been successfully used tosedate adolescents and adults. Topical lidocaine to thenasopharynx and larynx was combined with 0.1 to 0.3mg/kg rectal midazolam for 19 younger children.48 Aero-solized lidocaine in combination with an IM dose of eitheratropine (0.01 to 0.02 mg/kg) and diazepam (0.1 to 0.2

mg/kg) or midazolam (0.1 to 0.15 mg/kg) was used forsedation in 938 young children who had a foreign bodyremoved by flexible bronchoscopy.33 In that series, theflexible bronchoscope was inserted intranasally unless na-sal stenosis was present.33

In smaller children who are unable to cooperate, severaltechniques of general anesthesia have been reported. Abalanced anesthetic using IV propofol and sevoflurane withtopical lidocaine and oxymetazoline was used for 23 chil-dren ages 9 months to 16 years.50 The fiberoptic broncho-scope was then inserted through a T-piece on the child’sfacemask and advanced transnasally. In a series of 6children ages 1.2 to 5 years spontaneously breathing undersevoflurane anesthesia, the bronchoscope was insertedthrough a swivel adapter on a laryngeal mask airway.52

The foreign bodies were removed en bloc with the laryn-geal mask with no adverse events. Flexible bronchoscopythrough endotracheal tubes under general anesthesia isalso described in which the foreign body, bronchoscope,and endotracheal tube are removed en bloc.51 A standardpediatric bronchoscope (3.6 mm outer diameter) can beused with a size 4.5 or larger endotracheal tube, whereasstandard adult bronchoscopes (4.9 mm diameter) will fitinto size 2 or larger laryngeal mask.

Postoperative ConsiderationsEarly discharge after uncomplicated bronchoscopy is rea-sonable. In one study, 187 (65%) children were dischargedhome within 4 hours after rigid bronchoscopy.78 In anotherstudy, 82 (60.7%) children had a hospital stay �1 day.32

Prolonged pulmonary recovery may prevent early dis-charge. Predictive factors of prolonged recovery includedevidence of inflammation on preoperative radiographs,aggravation of pulmonary lesions on postoperative films,and a prolonged duration of bronchoscopy.28,79 Ciftci etal.11 found bronchoscopy time (57 � 2.9 minutes vs. 23 �1.2 minutes) to be prolonged in children with postoperativecomplications in comparison with those without complica-tions. Chen et al.63 found that postoperative hypoxemiawas associated with prolonged emergence from anesthesiaand with foreign bodies that were plant seeds.

CONCLUSIONSAspiration of a foreign body is a potentially lethal event.Although many deaths occur before arrival at the hospital,anesthesia and bronchoscopy to remove the offending itemare associated with considerable mortality and morbidity.Outcomes have improved over the years because of ad-vances in anesthesia and bronchoscopy. Although severalanesthetic techniques are effective for managing childrenwith foreign body aspiration, there is no consensus fromthe literature as to which technique is optimal. An induc-tion that maintains spontaneous ventilation is commonlypracticed to minimize the risk of converting a partialproximal obstruction to a complete obstruction. Controlledventilation combined with IV drugs and paralysis allowsfor suitable rigid bronchoscopy conditions and a consistentlevel of anesthesia. The use of CT and virtual bronchoscopyto diagnose foreign body aspiration and the use of flexiblebronchoscopy for the diagnosis and removal of foreignbodies may decrease the necessity for rigid bronchoscopy



under general anesthesia in patients with suspected foreignbody aspiration. As a result, morbidity and mortality inthese children may further decrease. Regardless of themanagement strategy, close cooperation within a skilledsurgical and anesthetic team is essential to avoid thepotential hazards of foreign body aspiration.

ACKNOWLEDGMENTSWe wish to thank the following for assistance in preparation ofthis manuscript: Daniel Doody, MD, Gennadiy Fuzaylov, MD,Allan Goldstein, MD, Kenan Haver, MD, David Lawlor, MD,and Pallavi Sagar, MD, all of the Massachusetts GeneralHospital; Cory Collins, DO, and Christopher Hartnick, MD,both of the Massachusetts Eye and Ear Infirmary; and all ofHarvard Medical School, Boston, Massachusetts.

APPENDIX: VIDEO CAPTIONSVideo 1. Rigid bronchoscopy down the left mainstem bronchus.Bubbles formed by release of trapped air can be seen duringspontaneous breathing.Video 2. An optical forceps is used to grasp and remove the objectvia rigid bronchoscopy.

REFERENCES1. Clerf LH. Historical aspects of foreign bodies in the air and

food passages. South Med J 1975;68:1449–542. Zaytoun GM, Rouadi PW, Baki DH. Endoscopic manage-

ment of foreign bodies in the tracheobronchial tree: predic-tive factors for complications. Otolaryngol Head Neck Surg2000;123:311– 6

3. Jackson C. The Life of Chevalier Jackson: An Autobiography.New York: Macmillan, 1938

4. Jackson C. Foreign Bodies in the Air and Food Passages.Charted Experiences in Cases from no. 631 to no. 1155 at theBronchoscopic Clinic. New Bedford: Reynolds the Printer, 1923

5. Miyazawa T, Miyazu Y, Kanoh K, Iwamoto Y. Flexible bron-choscopy historical perspective. In: edited by Beamis J, MathurP, Mehta A, eds. Interventional Pulmonary Medicine (1st ed.,chap. 3). New York: M. Dekker, 2004:33–48

6. Dikensoy O, Usalan C, Filiz A. Foreign body aspiration: clinicalutility of flexible bronchoscopy. Postgrad Med J 2002;78:399–403

7. Aydogan LB, Tuncer U, Soylu L, Kiroglu M, Ozsahinoglu C.Rigid bronchoscopy for the suspicion of foreign body in theairway. Int J Pediatr Otorhinolaryngol 2006;70:823–8

8. Ayed AK, Jafar AM, Owayed A. Foreign body aspiration inchildren: diagnosis and treatment. Pediatr Surg Int 2003;19:485–8

9. Bittencourt PF, Camargos PA, Scheinmann P, de Blic J. Foreignbody aspiration: clinical, radiological findings and factorsassociated with its late removal. Int J Pediatr Otorhinolaryngol2006;70:879–84

10. Brkic F, Delibegovic-Dedic S, Hajdarovic D. Bronchoscopicremoval of foreign bodies from children in Bosnia and Herze-govina: experience with 230 patients. Int J Pediatr Otorhino-laryngol 2001;60:193–6

11. Ciftci AO, Bingol-Kologlu M, Senocak ME, Tanyel FC, Buyuk-pamukcu N. Bronchoscopy for evaluation of foreign bodyaspiration in children. J Pediatr Surg 2003;38:1170–6

12. Cohen S, Avital A, Godfrey S, Gross M, Kerem E, Springer C.Suspected foreign body inhalation in children: what are theindications for bronchoscopy? J Pediatr 2009;155:276–80

13. Divisi D, Di Tommaso S, Garramone M, Di Francescantonio W,Crisci RM, Costa AM, Gravina GL, Crisci R. Foreign bodiesaspirated in children: role of bronchoscopy. Thorac CardiovascSurg 2007;55:249–52

14. Emir H, Tekant G, Besik C, Elicevik M, Senyuz OF, BuyukunalC, Sarimurat N, Yeker D. Bronchoscopic removal of tracheo-broncheal foreign bodies: value of patient history and timing.Pediatr Surg Int 2001;17:85–7

15. Eren S, Balci AE, Dikici B, Doblan M, Eren MN. Foreign bodyaspiration in children: experience of 1160 cases. Ann TropPaediatr 2003;23:31–7

16. Erikci V, Karacay S, Arikan A. Foreign body aspiration: afour-years experience. Ulus Travma Acil Cerrahi Derg2003;9:45–9

17. Girardi G, Contador AM, Castro-Rodríguez JA. Two newradiological findings to improve the diagnosis of bronchialforeign-body aspiration in children. Pediatr Pulmonol 2004;38:261–4

18. Gregori D, Salerni L, Scarinzi C, Morra B, Berchialla P, SnideroS, Corradetti R, Passali D, Klaus A, Isidor H, Gernot S, Jan B,Bernard B, Karchev T, Tzolov T, Ranko M, Lana K, Ivo S, MirkoT, Caye-Thomasen P, Anne P, Volker J, Onder G, Simasko N,Matilda C, Christopoulos I, PassA li GC, PassA li F, Damiani V,MieczysÅ,aw C, Dorin S, Gheorghe DC, Janka J, Miha Z, AlesG, Ales M, Lorenzo R, Javier C, Pontus S, Philippe P, Ahmed C,Metin OT, Ozden CA, Riza D, John G, Peter R, Rupert O.Foreign bodies in the upper airways causing complicationsand requiring hospitalization in children aged 0–14 years:results from the ESFBI study. Eur Arch Otorhinolaryngol2008;265:971–8

19. Hasdiraz L, Oguzkaya F, Bilgin M, Bicer C. Complications ofbronchoscopy for foreign body removal: experience in 1,035cases. Ann Saudi Med 2006;26:283–7

20. Heyer CM, Bollmeier ME, Rossler L, Nuesslein TG, Stephan V,Bauer TT, Rieger CH. Evaluation of clinical, radiologic, andlaboratory prebronchoscopy findings in children with sus-pected foreign body aspiration. J Pediatr Surg 2006;41:1882–8

21. Higuchi O, Adachi Y, Ichimaru T, Asai M, Kawasaki K. Foreignbody aspiration in children: a nationwide survey in Japan. IntJ Pediatr Otorhinolaryngol 2009;73:659–61

22. Hui H, Na L, Zhijun CJ, Fugao ZG, Yan S, Niankai ZK, JingjingCJ. Therapeutic experience from 1428 patients with pediatrictracheobronchial foreign body. J Pediatr Surg 2008;43:718–21

23. Kadmon G, Stern Y, Bron-Harlev E, Nahum E, Battat E,Schonfeld T. Computerized scoring system for the diagnosis offoreign body aspiration in children. Ann Otol Rhinol Laryngol2008;117:839–43

24. Karakoc F, Karadag B, Akbenlioglu C, Ersu R, Yildizeli B,Yuksel M, Dagli E. Foreign body aspiration: what is theoutcome? Pediatr Pulmonol 2002;34:30–6

25. Kiyan G, Gocmen B, Tugtepe H, Karakoc F, Dagli E, Dagli TE.Foreign body aspiration in children: the value of diagnosticcriteria. Int J Pediatr Otorhinolaryngol 2009;73:963–7

26. Mahafza T, Khader Y. Aspirated tracheobronchial foreign bodies:a Jordanian experience. Ear Nose Throat J 2007;86:107–10

27. Roda J, Nobre S, Pires J, Estevao MH, Felix M. Foreign bodiesin the airway: a quarter of a century’s experience. Rev PortPneumol 2008;14:787–802

28. Roh JL, Hong SJ. Lung recovery after rigid bronchoscopicremoval of tracheobronchial foreign bodies in children. IntJ Pediatr Otorhinolaryngol 2008;72:635–41

29. Shivakumar AM, Naik AS, Prashanth KB, Shetty KD, PraveenDS. Tracheobronchial foreign bodies. Indian J Pediatr 2003;70:793–7

30. Shubha AM, Das K. Tracheobronchial foreign bodies in infants.Int J Pediatr Otorhinolaryngol 2009;73:1385–9

31. Skoulakis CE, Doxas PG, Papadakis CE, Proimos E,Christodoulou P, Bizakis JG, Velegrakis GA, Mamoulakis D,Helidonis ES. Bronchoscopy for foreign body removal inchildren. A review and analysis of 210 cases. Int J PediatrOtorhinolaryngol 2000;53:143–8

32. Tan HK, Brown K, McGill T, Kenna MA, Lund DP, Healy GB.Airway foreign bodies (FB): a 10-year review. Int J PediatrOtorhinolaryngol 2000;56:91–9

33. Tang LF, Xu YC, Wang YS, Wang CF, Zhu GH, Bao XE, Lu MP,Chen LX, Chen ZM. Airway foreign body removal by flexiblebronchoscopy: experience with 1027 children during 2000–2008.World J Pediatr 2009;5:191–5

34. Tokar B, Ozkan R, Ilhan H. Tracheobronchial foreign bodies inchildren: importance of accurate history and plain chest radi-ography in delayed presentation. Clin Radiol 2004;59:609–15

35. Tomaske M, Gerber AC, Stocker S, Weiss M. Tracheobronchialforeign body aspiration in children—diagnostic value ofsymptoms and signs. Swiss Med Wkly 2006;136:533–8

REVIEW ARTICLE


36. Yadav SP, Singh J, Aggarwal N, Goel A. Airway foreign bodiesin children: experience of 132 cases. Singapore Med J 2007;48:850–3

37. Pinto A, Scaglione M, Pinto F, Guidi G, Pepe M, Del Prato B,Grassi R, Romano L. Tracheobronchial aspiration of foreignbodies: current indications for emergency plain chest radiog-raphy. Radiol Med 2006;1112:497–506

38. Assefa D, Amin N, Stringel G, Dozor AJ. Use of decubitusradiographs in the diagnosis of foreign body aspiration inyoung children. Pediatr Emerg Care 2007;23:154–7

39. Sodhi KS, Saxena AK, Singh M, Rao KL, Khandelwal N. CTvirtual bronchoscopy: new non invasive tool in pediatricpatients with foreign body aspiration. Indian J Pediatr2008;75:511–3

40. Veras TN, Hornburg G, Schner AM, Pinto LA. Use of virtualbronchoscopy in children with suspected foreign body aspira-tion. J Bras Pneumol 2009;35:937–41

41. Hong SJ, Goo HW, Roh JL. Utility of spiral and cine CT scansin pediatric patients suspected of aspirating radiolucent for-eign bodies. Otolaryngol Head Neck Surg 2008;138:576–80

42. Kosucu P, Ahmetoglu A, Koramaz I, Orhan F, Ozdemir O,Dinc H, Okten A, Gumele HR. Low-dose MDCT and virtualbronchoscopy in pediatric patients with foreign body aspira-tion. AJR Am J Roentgenol 2004;183:1771–7

43. Huang HJ, Fang HY, Chen HC, Wu CY, Cheng CY, Chang CL.Three-dimensional computed tomography for detection oftracheobronchial foreign body aspiration in children. PediatrSurg Int 2008;24:157–60

44. Adaletli I, Kurugoglu S, Ulus S, Ozer H, Elicevik M, Kantarci F,Mihmanli I, Akman C. Utilization of low-dose multidetectorCT and virtual bronchoscopy in children with suspectedforeign body aspiration. Pediatr Radiol 2007;37:33–40

45. Haliloglu M, Ciftci AO, Oto A, Gumus B, Tanyel FC, SenocakME, Buyukpamukcu N, Besim A. CT virtual bronchoscopy inthe evaluation of children with suspected foreign body aspi-ration. Eur J Radiol 2003;48:188–92

46. Even L, Heno N, Talmon Y, Samet E, Zonis Z, Kugelman A.Diagnostic evaluation of foreign body aspiration in children: aprospective study. J Pediatr Surg 2005;40:1122–7

47. Martinot A, Closset M, Marquette CH, Hue V, Deschildre A,Ramon P, Remy J, Leclerc F. Indications for flexible versusrigid bronchoscopy in children with suspected foreign-bodyaspiration. Am J Respir Crit Care Med 1997;155:1676–9

48. Righini CA, Morel N, Karkas A, Reyt E, Ferretti K, Pin I,Schmerber S. What is the diagnostic value of flexible bronchos-copy in the initial investigation of children with suspectedforeign body aspiration? Int J Pediatr Otorhinolaryngol2007;71:1383–90

49. Gencer M, Ceylan E, Koksal N. Extraction of pins from theairway with flexible bronchoscopy. Respiration 2007;74:674–9

50. Ramirez-Figueroa JL, Gochicoa-Rangel LG, Ramirez-San JuanDH, Vargas MH. Foreign body removal by flexible fiberopticbronchoscopy in infants and children. Pediatr Pulmonol2005;40:392–7

51. Swanson KL, Prakash UB, Midthun DE, Edell ES, Utz JP, McDou-gall JC, Brutinel WM. Flexible bronchoscopic management ofairway foreign bodies in children. Chest 2002;121:1695–1700

52. Yazbeck-Karam VG, Aouad MT, Baraka AS. Laryngeal maskairway for ventilation during diagnostic and interventional fibre-optic bronchoscopy in children. Paediatr Anaesth 2003;13:691–4

53. Mani N, Soma M, Massey S, Albert D, Bailey CM. Removal ofinhaled foreign bodies—middle of the night or the next morn-ing? Int J Pediatr Otorhinolaryngol 2009;73:1085–9

54. Soodan A, Pawar D, Subramanium R. Anesthesia for removalof inhaled foreign bodies in children. Paediatr Anaesth2004;14:947–52

55. Kain ZN, O’Connor TZ, Berde CB. Management of tracheo-bronchial and esophageal foreign bodies in children: a surveystudy. J Clin Anesth 1994;6:28–32

56. Sersar SI, Rizk WH, Bilal M, El Diasty MM, Eltantawy TA,Abdelhakam BB, Elgamal AM, Bieh AA. Inhaled foreignbodies: presentation, management and value of history andplain chest radiography in delayed presentation. OtolaryngolHead Neck Surg 2006;134:92–9

57. Soysal O, Kuzucu A, Ulutas H. Tracheobronchial foreign bodyaspiration: a continuing challenge. Otolaryngol Head NeckSurg 2006;135:223–6

58. Yu H, Yang XY, Liu B. EMLA cream coated on the rigidbronchoscope for tracheobronchial foreign body removal inchildren. Laryngoscope 2009;119:158–61

59. Baraka A. Bronchoscopic removal of inhaled foreign bodies inchildren. Br J Anaesth 1974;46:124–6

60. Farrell PT. Rigid bronchoscopy for foreign body removal:anaesthesia and ventilation. Paediatr Anaesth 2004;14:84–9

61. Litman RS, Ponnuri J, Trogan I. Anesthesia for tracheal orbronchial foreign body removal in children: an analysis ofninety-four cases. Anesth Analg 2000;91:1389–91

62. Buu NT, Ansermino M. Anesthesia for removal of inhaledforeign bodies in children. Paediatr Anaesth 2005;15:533

63. Chen LH, Zhang X, Li SQ, Liu YQ, Zhang TY, Wu JZ. The riskfactors for hypoxemia in children younger than 5 years oldundergoing rigid bronchoscopy for foreign body removal.Anesth Analg 2009;109:1079–84

64. Meretoja OA, Taivainen T, Raiha L, Korpela R, WirtavuoriK. Sevoflurane-nitrous oxide or halothane-nitrous oxide forpaediatric bronchoscopy and gastroscopy. Br J Anaesth1996;76:767–71

65. Batra YK, Mahajan R, Bangalia SK, Chari P, Rao KL.A comparison of halothane and sevoflurane for broncho-scopic removal of foreign bodies in children. Ann CardAnaesth 2004;7:137– 43

66. Davidson A. The correlation between bispectral index andairway reflexes with sevoflurane and halothane anaesthesia.Paediatr Anaesth 2004;14:241–6

67. Cohen SR, Geller KA. Anesthesia and pediatric endoscopy: thesurgeon’s view. Otolaryngol Clin North Am 1981;14:705–13

68. Maddali MM, Badur RS, Fernando MS, Alsajwani MJ. Totalcontralateral atelectasis following rigid bronchoscopy in achild with scarf pin aspiration. Paediatr Anaesth 2006;16:1095–7

69. Perrin G, Colt HG, Martin C, Mak MA, Dumon JF, Gouin F.Safety of interventional rigid bronchoscopy using intravenousanesthesia and spontaneous assisted ventilation. a prospectivestudy. Chest 1992;102:1526–30

70. Barker N, Lim J, Amari E, Malherbe S, Ansermino JM. Rela-tionship between age and spontaneous ventilation duringintravenous anesthesia in children. Paediatr Anaesth 2007;17:948–55

71. Kumar S, Saxena AK, Kumar M, Rautela RS, Gupta N, GoyalA. Anesthetic management during bronchoscopic removal of aunique, friable foreign body. Anesth Analg 2006;103:1596–7

72. Pawar DK. Dislodgement of bronchial foreign body duringretrieval in children. Paediatr Anaesth 2000;10:333–5

73. Brown KL, Shefler A, Cohen G, DeMunter C, Pigott N,Goldman AP. Near-fatal grape aspiration with complicatingacute lung injury successfully treated with extracorporealmembrane oxygenation. Pediatr Crit Care Med 2003;4:243–5

74. Mise K, Jurcev Savicevic A, Pavlov N, Jankovic S. Removal oftracheobronchial foreign bodies in adults using flexible bron-choscopy: experience 1995–2006. Surg Endosc 2009;23:1360–4

75. Tariq SM, George J, Srinivasan S. Inhaled foreign bodies inadolescents and adults. Monaldi Arch Chest Dis 2005;63:193–8

76. Uskul TB, Turker H, Arslan S, Slevi A, Kant A. Use offiberoptic bronchoscopy in endobronchial foreign body re-moval in adults. Turkish Resp J 2007;8:39–42

77. Wong KS, Lai SH, Lien R, Hsia SH. Retrieval of bronchialforeign body with central lumen using a flexible bronchoscope.Int J Pediatr Otorhinolaryngol 2002;62:253–6

78. Tomaske M, Gerber AC, Weiss M. Anesthesia and periinter-ventional morbidity of rigid bronchoscopy for tracheobron-chial foreign body diagnosis and removal. Paediatr Anaesth2006;16:123–9

79. Chung MK, Jeong HS, Ahn KM, Park SH, Cho JK, Son YI, BaekCH. Pulmonary recovery after rigid bronchoscopic retrieval ofairway foreign body. Laryngoscope 2007;117:303–7



Economics, Education, and Policy

Section Editor: Franklin Dexter

SPECIAL ARTICLE

An Optimistic Prognosis for the Clinical Utility ofLaboratory Test DataMing Zheng, PhD,* Palanikumar Ravindran, PhD,† Jianmei Wang, PhD,† Richard H. Epstein, MD,‡David P. Chen, PhD,* Atul J. Butte, MD, PhD,* and Gary Peltz, MD, PhD*

It is hoped that anesthesiologists and other clinicians will be able to increasingly rely uponlaboratory test data to improve the perioperative care of patients. However, it has beensuggested that in order for a laboratory test to have clinically useful diagnostic performancecharacteristics (sensitivity and specificity), its performance must be considerably better thanthose that have been evaluated in most etiologic or epidemiologic studies. This pessimismabout the clinical utility of laboratory tests is based upon the untested assumption thatlaboratory data are normally distributed within case and control populations.

We evaluated the data distribution for 700 commonly ordered laboratory tests, and foundthat the vast majority (99%) do not have a normal distribution. The deviation from normal wasmost pronounced at extreme values, which had a large quantitative effect on laboratory testperformance. At the sensitivity and specificity values required for diagnostic utility, theminimum required odds ratios for laboratory tests with a nonnormal data distribution weresignificantly smaller (by orders of magnitude) than for tests with a normal distribution.

By evaluating the effect that the data distribution has on laboratory test performance, wehave arrived at the more optimistic outlook that it is feasible to produce laboratory tests withdiagnostically useful performance characteristics. We also show that moderate errors in theclassification of outcome variables (e.g., death vs. survival at a specified end point) have asmall impact on test performance, which is of importance for outcomes research that usesanesthesia information management systems. Because these analyses typically seek to identifyfactors associated with an undesirable outcome, the data distributions of the independentvariables need to be considered when interpreting the odds ratios obtained from suchinvestigations. (Anesth Analg 2010;111:1026–35)

In contemporary clinical practice, a large number oflaboratory tests are performed to facilitate diagnosis, toassess disease progression or the effect of a therapy, and

for prognostication. With the advent of genomic science, aneven larger number of variables (mRNAs, proteins, andmetabolites) can be measured in blood or tissues in acost-effective way. For anesthesiologists, it is likely thatmany different metabolites, proteins, or neural variableswill be measured in the perioperative setting that couldprovide data useful for patient assessment and treatment.Genetic risk factors can be measured to identify individualsat increased risk for disease or perioperative complications,or for predicting surgical outcome. Also, analyzing joined

datasets from multiple institutions’ anesthesia informationmanagement systems (AIMS) through data-mining tech-niques could identify threshold parameters associated withundesirable outcomes or to assess risk.

There is reason to believe that additional data andgenetic risk factor measurements will improve our abilityto diagnose, stratify, and optimize the perioperative care ofour patients. For example, a quantitative simulation dem-onstrated how the use of pharmacogenetic information toindividualize drug dosage has the potential to significantlyimprove treatment outcome.1

Nevertheless, several publications have suggested thatnewly discovered laboratory tests would have limiteddiagnostic utility for individual patients. The argumentadvanced was that in order for a laboratory measurementto have clinically useful performance characteristics (sensi-tivity and specificity), the magnitude of the odds ratio (OR)for the test (Text Box) must be considerably higher thanthose seen in most etiologic or epidemiologic studies.2–4

For example, a hypothetical analysis by Ware2 indicatedthat an OR of 228 would be required for a test result to havesufficient diagnostic or predictive utility (80% specificityand 80% sensitivity).

If this level of certainty is truly required, it would bevirtually impossible to identify a genetic test with accept-able performance for the vast majority of clinical situations.This is because ORs for most of the genetic risk factors for

From *Stanford University, Stanford, California; †Roche Palo Alto LLC, PaloAlto, CA; and ‡Jefferson Medical College, Philadelphia, Pennsylvania.


Gary Peltz was partially supported by funding (7R56 GMO68885 to 05) fromthe NIGMS. Funding was provided by NLM T15 LM007033 (to David P.Chen) and by the Lucile Packard Foundation for Children’s Health (to AtulJ. Butte).


Address correspondence to Gary Peltz, MD, PhD, Stanford University, 800Welch Road, Room 213, Palo Alto, CA 94304. Address e-mail [email protected].

Copyright © 2010 International Anesthesia Research SocietyDOI: 10.1213/ANE.0b013e3181efff0c


quantitative traits or susceptibility to common diseases thathave been identified are usually much lower than theproposed threshold. For example, analysis of the 1967identified human single nucleotide polymorphisms with areported OR for a studied traita indicated that 67% haveORs �5, and 95% have ORs �30.

In a similar negative vein, Pepe et al.4 investigated therelationship between the OR and classification accuracy.They analyzed hypothetical data with a normal distributionand concluded that an OR of 74 was required to obtainclinically useful performance characteristics (79% sensitiv-ity and 79% specificity). Taken at face value, these 2analyses provide a pessimistic outlook for the utility oflaboratory tests, because they imply that the performanceof most laboratory tests will not be sufficient to have muchdiagnostic or predictive utility.

However, these analyses2–4 are based on a fundamentalassumption that laboratory data are normally distributedwithin case and control populations. They also assumethat currently used disease definitions and classificationschema, developed before the discovery of genetic riskfactors or novel laboratory tests, will continue to beapplied.

In this article, we demonstrate that the clinical utility oflaboratory data has a much better prognosis than thatsuggested by Ware and by Pepe et al. The vast majority oflaboratory results do not have a normal distribution withineither control or disease populations. As a consequence, theperformance characteristics (i.e., sensitivity and specificityfor disease diagnosis) of laboratory tests are substantiallyimproved over Gaussian assumptions. Thus, laboratorydata whose performance characteristics are within theusual range observed in epidemiologic or etiologic studiescan have substantial diagnostic utility. These consider-ations lead to a more optimistic assessment of the utility oflaboratory tests in improving patient care.

METHODSAnalysis of Clinical Laboratory Test DataAll available data for 700 clinical laboratory tests per-formed between 2000 and 2006 were retrieved from theStanford Translational Research Integrated Database(STRIDE) according to a protocol that was approved by theIRB. These tests were routinely performed at the Stanfordand Lucile Packard Children’s Hospitals, and includedpatients between 0 and 108 years of age. At least 1000measurements were available for each test (minimum 1001,maximum 2,090,227, median 3466). The Jarque–Bera test7

was used to evaluate the normality of the data distribu-tion in each test. The test was performed on the originaldata and on the logarithm-transformed data, which is atransformation commonly used to analyze data with asignificant rightward skewing. The larger of the 2 Pvalues obtained was reported. The Benjamini–Hochbergadjustment method was then used to adjust for multipletesting.9 Because the null hypothesis was that the distri-bution was indeed normal, the smaller the adjusted Pvalue, the greater the deviation from normal. When the

adjusted P value was smaller than 0.05, the null hypoth-esis was rejected at the 5% significance level (i.e., the datawere not normally distributed).

Biomarker Data AnalysisFor the 3 types of laboratory data studied in detail, deiden-tified data were obtained from the STRIDE. The Interna-tional Classification of Diseases, Clinical Modification(ICD-9), codes for each individual with an available labo-ratory value were evaluated to identify the control anddisease populations by using the STRIDE AnonymousPatient Cohort Discovery Tool that was developed by theStanford Center for Clinical Informatics.10 The hemoglobinA1C (immunoassay), CD19� cell counts (flow immunocy-tometry), and protein S activity (automated latex immuno-assay) were measured using standard protocols in theclinical laboratory at Stanford University.

The following ICD-9 codes were used to classify indi-viduals with available laboratory data: diabetes (250), lym-phoma (200,201), and coagulopathy (286). There were33,958 A1C measurements from 20,590 nondiabetic indi-viduals and 41,541 A1C measurements from 10,677 diabeticindividuals from the database. CD19� cell counts included17,706 measurements from 4498 individuals not diagnosedas having a lymphoma, and 1861 measurements from 541individuals with a lymphoma. Protein S activities included3701 measurements from 3385 individuals without a diag-nosed coagulopathy and 519 measurements from 420 indi-viduals with a diagnosed coagulopathy. When a test wasperformed more than once on an individual, the valueobtained on the initial visit was used. (However, resultswere insensitive to the substitution of lab values fromsubsequent visits.)

The A1C lab values were extensively right skewed;therefore, a log-transformation was applied to make thedata distribution more symmetric. Because values of 0 wereobtained for some CD19 and the protein S values, 1 wasadded to the original value before log transformation,because log(0) is undefined. All subsequent analyses wereperformed on the transformed data.

Additional information about the OR determinations fordifferent data distributions and for A1C and other lab testdata, and for the other simulation studies, is provided inthe online Supplementary Methods (see SupplementalDigital Content 1, http://links.lww.com/AA/A172) forthis journal.

RESULTSClinical Laboratory Data Usually Are notNormally DistributedEven if most of the population data appear to follow theshape of a normal distribution, the values at the tails(within the population extremes) may diverge markedlyfrom that predicted by the normal distribution. For in-stance, the measured serum concentration of a proteincould have a nonnormal distribution within the extremesof a population because of a limitation in syntheticcapacity, a biological feedback mechanism that limits itsmaximum concentration, or a threshold level of stimula-tion that may be required to initiate the synthesis of aahttp://www.genome.gov/gwastudies, accessed on April 10, 2010.

An Optimistic Prognosis for Clinical Utility of Lab Test


Text Box: OR Definitions and Data Distribution.

The OR compares the measured value for a risk factor within a population with a disease to that in a control populationwithout the disease. Traditionally, the relationship between a binary test result and a binary outcome is represented ina two-way table. For example, the Bispectral Index (BIS) is used to titrate the administration of drugs during generalanesthesia to minimize the risk of intraoperative awareness.5 This measure can be used to illustrate how a quantitativemeasure can be converted into a binary parameter (“BIS status”). Suppose analysis of an AIMS database that includesadverse outcome data leads to a putative test to predict the occurrence of intraoperative awareness if the BIS were morethan 70 for more than 5 min during the interval from intubation to the end of surgery. Such a test might be useful toguide the rapidity with which BIS increases should be treated as well as to identify patients who need postoperativefollow-up to assess and possibly treat the consequences of intraoperative awareness. Cases where the BIS was elevatedas such are characterized as BIS� and those where there were no such epochs as BIS-. Then, we can construct thefollowing two-way table:

Awareness AmnesiaBIS- n11 n12

BIS� n21 n22

We have a total of n individuals (n � n11�n12�n21�n22), where n11 of them are BIS- and have awareness during thesurgery and n12, n21 and n22 are similarly defined according to the row and column labels for the corresponding cellsin the table. Then, the odd of having awareness versus amnesia during the surgery in the BIS negative group is n11/n12.The odd of having awareness versus amnesia in the BIS positive group is n21/n22. The OR is then calculated as the ratioof these two odds: OR � (n11/n12)/(n21/n22) � n11n22/n21n12. If the criterion of the BIS being more than 70 for �5 minwere a good indicator of outcome, then n11 and n22 would be large, and n12 and n21 small, relative to n11 and n22. Asa consequence, a large OR is obtained. In contrast, a small OR (close to 1) indicates that the test result has a small effecton outcome probability, indicating that it has a poor performance.

Several other important measurements are frequently used to assess test performance. If we denote the individualshaving amnesia during the surgery as cases, and those with awareness as controls, then n11, n12, n21 and n22 are knownas true negative (TN), false negative (FN), false positive (FP) and true positive (TP), respectively. Specificity is theprobability that a control individual is correctly classified; and sensitivity is the probability that a case individual iscorrectly classified. Additionally, the positive predictive value and negative predictive value are the chances that anindividual predicted as case (or control) is actually a case or control, respectively. The following table illustrates howthese values are calculated:

Awareness AmnesiaBIS- n11 (TN) n12 (FN) NPV�TN/(TN�FN)

BIS� n21 (FP) n22 (TP) PPV�TP/(TP�FP)Specificity � Sensitivity �

TN/(TN � FP) TP/(TP�FN)

However, by focusing on different case and control populations, different methods can be used to calculate the OR,which can markedly affect its value. In the analysis by Ware,2 the OR is defined as the ratio of the frequency of an eventoccurring within the group with the disease (cases) whose risk factor value places them at the 90th percentile (FigureT1, left panel, arrow labeled Group 2) relative to the frequency of events within the control group whose risk factorvalue is at the 10th percentile (Figure T1, left panel, arrow labeled Group 1). The difference in the mean frequenciesbetween the case and control groups is used to calculate the OR for a risk factor. However, more conventionally, theOR is defined as the ratio of the odds of an event occurring in one group relative to the odds of it occurring in anothergroup where all values are at or above a given value (90th percentile) in one group (Figure T1, right panel, orange area),and the values that are at or below a given value (10th percentile) in the second group (Figure T1, right panel, blue area).The conventional method for OR calculation can be more easily and accurately determined.

The data distribution within a population also affects laboratory data performance. The data distribution forpopulations with a normal (Blue), a Double-Exponential (green) or a Cauchy (Red) distribution are shown in Figure T2.Note the differences at the extremes of the population distribution, where the Cauchy curve and the Double-Exponential curve do not tail-off as rapidly as does the normal curve. The maximum likelihood method is used to fitthe actual data to different types of distributions. The fitted curve can be overlaid to the histogram. Visual inspectionprovides an estimate of the goodness-of-fit of the fitted distribution and statistical tests (Kolmogorov-Smirnov test6 forgeneral distributions, and the Jarque-Bera7 or Shapiro-Wilk8 tests for normal distribution) can be used to rigorouslyassess the fit.

SPECIAL ARTICLE


disease-associated protein. Any of these effects wouldflatten the data distribution curve at the extremes ofcontrol or diseased populations. This is important be-cause calculation of ORs involves assessments at the tailsof the distribution, whether one follows the calculationmethod described by Ware or the more conventionalapproach (Text Box).

To assess the possibility that the prior estimates ofORs required for adequate test performance are overlyconservative because of deviations from the normaldistribution at the tails, we evaluated the data distribu-tion of all 700 clinical laboratory tests. A detailed ex-ample follows.

The A1C test (HbA1c, glycated hemoglobin, or glycosy-lated hemoglobin) is a commonly used laboratory test thatreflects the effectiveness of blood glucose regulation.11

Histograms of A1C lab values for 20,590 control (nondia-betic) individuals and 10,677 diabetic individuals reveal adeviation of these data from a normal distribution in eitherpopulation (Fig. 1). Although a superficial visual inspectionof the shape of the data distribution might seem to indicatea normal distribution, this is not a rigorous method for suchdeterminations. Therefore, we used the Jarque–Bera test toassess the normality of this data. The resulting P values arenearly zero for the case and control populations, indicatingthat these data have a highly nonnormal distribution,which is consistent with the shape of the histograms. Thesame data are also graphed as quantile–quantile plots,where the A1C lab values are plotted in relation to thepercentile for the theoretical normal distribution and de-partures from linearity indicate where the data do not havea normal distribution. These plots demonstrate that the

Figure T1: Definition of odds ratio using the Ware (Left) and Conventional Definitions (Right). According to the Ware Definition, (2) the oddsratio is defined as the ratio of the frequency of an event occurring within the group with the disease (cases) whose risk factor value places themat the 90th percentile (arrow labeled in Group 2) relative to the frequency of events within the control group whose risk factor value is at the10th percentile (arrow labeled in Group 1). The difference in the mean frequencies between the case and control groups is used to calculatethe odds ratio for a risk factor. According to the Conventional Definition, the odds ratio is defined as the ratio of the odds of an event occurringin one group relative to the odds of it occurring in another group where all values are at or above a given value (90th percentile) in one group(orange area), and the values that are at or below a given value (10th percentile) in the second group (blue area).

Figure T2: The data distribution for popu-lations with normal (Blue), Double-Exponential (green) or Cauchy (Red)distributions.



distribution of A1C lab values at the extremes deviatessignificantly from the normal distribution, especially in thecontrol population (Fig. 1).

We analyzed the raw data for all 700 laboratorytests with the same methodology, and found that 699(99.9%) tests were not normally distributed. After log-transformation, the data in 694 (99.1%) tests remained non-normally distributed. The serum transferrin level was the onlytest whose data had a normal distribution; only 5 other tests(B-type natriuretic peptide, 1-hour glucose tolerance test,glucostatin, hematocrit, and total protein) were normallydistributed after log-transformation. Thus, the data for nearlyall laboratory tests do not have a normal distribution.

A Nonnormal Distribution Significantly AltersLaboratory Data PerformanceBecause the ORs and laboratory data performance areassessed with data obtained from the extremes of a popu-lation (Text Box), the distribution of the data within thepopulation extremes has a large effect on its utility fordisease diagnosis. For example, if laboratory data weremore flatly distributed at the extremes, the data distribu-tion would resemble a double-exponential12 or a Cauchydistribution13 (Text Box). These two types of data distribu-tion may better fit the rate of decay of the probabilitydensity for the tails. Consequently, these distributionsbetter fit the actual data distribution at the extremes thandoes the normal distribution.

To investigate the potential implications of this effect,we plotted the OR as a function of the sensitivity andspecificity for lab test data with a normal or a Cauchy

distribution (Supplemental Figs. 1 [see Supplemental DigitalContent 2, http://links.lww.com/AA/A173 ] and 2 [see Supple-mental Digital Content 2, http://links.lww.com/AA/A174 ]; seeSupplementary Methods section for figure legends,http://links.lww.com/AA/A172). The dramatically differentshapes of these curves demonstrate that laboratory test per-formance (sensitivity and specificity) at a specified OR ismarkedly altered if the data are not normally distributed.

Furthermore, the effect of a nonnormal data distributionis especially pronounced under conditions in which highsensitivity and specificity are required. To illustrate this, weprepared a table showing the values of ORs calculated forclinically useful levels of sensitivity and specificity for datawith a normal, a double-exponential or a Cauchy distribu-tion (Tables 1 and 2). The OR definition applied clearlyimpacts laboratory test performance. In general, use of theWare definition increased the ORs that were required for alaboratory test to have clinically useful performance(80%–90% sensitivity and 80%–90% specificity) by �10-foldin relation to the conventional OR definition.

However, independent of which OR definition wasused, the data distribution within the extremes of a popu-lation had a very significant effect on test utility andperformance characteristics. A laboratory test with 80%sensitivity and 90% specificity require an OR (Ware defini-tion) of 231 for normally distributed data. However, usingthe same assumptions as Ware (the data distribution in thecase and control populations has the same shape), an OR of15 or 25 can provide the same performance for data with aCauchy distribution or a double-exponential distribution,

Figure 1. A1C lab values are not normallydistributed in control (nondiabetic) or dia-betic populations. Top panel, The histo-gram (left) and the quantile–quantile (QQ)plot (right) of the (log-transformed) A1C labvalues in control populations. Bottom pan-els, The histogram (left) and the QQ plot(right) of the (log-transformed) A1C labvalues in diabetic populations. In thehistograms, the red curve representsthe fitted normal curve. In the QQ plots,the red lines indicate data with a normaldistribution. The P values for the nullhypothesis that the data are normallydistributed are 0 for both the controland diabetic populations, indicatingthat the data have a highly nonnormaldistribution.

SPECIAL ARTICLE


respectively. The same trend was noted when the conven-tional OR definition was applied. More dramatically, an ORof 14,910 is required to achieve 90% sensitivity and 90%specificity for data with a normal distribution, and an OR of225 or 6 is required for data that has a double-exponentialor Cauchy distribution, respectively (Table 2).

At the sensitivity and specificity values required forclinical utility, the minimum required OR values for datawith a Cauchy distribution are substantially smaller thanthose for data with a normal distribution. The same resultswere consistently obtained when the shape of the datadistribution in the case and control groups were different(Supplementary Table 1; see Supplementary Methods,http://links.lww.com/AA/A172).

We further investigated whether the distribution ofactual laboratory data affects the OR that is required for alaboratory test to achieve a certain level of diagnosticperformance. Using a cutoff of 2.63 for the A1C lab values,this test would achieve 84% specificity and 71% sensitivityfor the diagnosis of diabetes in the populations evaluated.Using the actual A1C data, the calculated OR was 24. IfAIC data were normally distributed, the OR required toachieve this performance would have been �10-fold(271) higher.

To evaluate the impact that the data distribution had onthe required OR, we artificially shifted the distributioncurve for the A1C data in the case population to achieve thedesired performance characteristics. This enabled us todirectly compare the calculated ORs from the actual A1Cdata (after the hypothetical right shift) with those for datawith a normal distribution. By varying the extent of therightward shift in the diabetic population, any desired levelof performance could be achieved, which enabled theimpact of the shape of the distribution curve (especially atthe 2 tails) on the OR (conventional definition) to becharacterized (Table 3).

The required OR for the A1C test to achieve clinicallyuseful performance characteristics is significantly smallerthan if the data had a normal distribution. This analysisalso indicated that a double-exponential distribution couldreasonably approximate the properties of the A1C datadistribution at the 2 tails. Thus, the nonnormal data distri-bution at the extremes had a very significant effect on theperformance of a laboratory test; the ORs required toachieve clinically useful performance characteristics weresignificantly less than expected.

Nonnormal Data Distribution ImprovesA1C PerformanceWe next evaluated the effect that the data distribution hadon the performance of A1C laboratory data. To do this, wedetermined the number of individuals who would becorrectly classified as control (nondiabetic) or diabetic onthe basis of actual A1C laboratory data, and compared thatwith the expected results if the A1C data had a normaldistribution (Table 3). Using the indicated cutoff of 2.63(which produced an OR of 24.3) for the A1C data, wecorrectly identified 3439 more control (nondiabetic) indi-viduals by this result than if the data had a normaldistribution. In addition, 334 more diabetic individualswere correctly identified than if the A1C data were nor-mally distributed (Table 3). In other words, the nonnormaldistribution of the A1C laboratory data resulted in 13.7%fewer FP and 3.1% fewer FN classifications than if the datahad followed a normal distribution.

Table 1. Comparison of Odds Ratio ValuesRequired for Laboratory Data with a Normal,Double-Exponential, or Cauchy Distribution toAchieve the Indicated Performance Characteristics(Sensitivity and Specificity)

Odds ratio(Ware definition)

Sensitivity Specificity NormalDouble

exponential Cauchy80% 80% 75 25 1180% 90% 231 25 790% 80% 231 25 590% 90% 713 25 4

The odds ratios were calculated as described in the METHODS section usingthe Ware (2) definition. Of note, regardless of the odds ratio definitionused, the data distribution has a large impact on the required odds ratio forcertain performance characteristics: the odds ratio required for normallydistributed data may be 1 or 2 magnitudes higher than that required for datawith the double-exponential or Cauchy distribution.

Table 2. Comparison of Odds Ratio ValuesRequired for Laboratory Data with a Normal,Double-Exponential, or Cauchy Distribution toAchieve the Indicated Performance Characteristics(Sensitivity and Specificity)

Odds ratio(conventional definition)

Sensitivity Specificity NormalDouble

exponential Cauchy80% 80% 433 38 880% 90% 2415 100 1190% 80% 2415 100 1590% 90% 14,910 225 18

The odds ratios were calculated as described in the METHODS section usingthe conventional definition. Of note, regardless of the odds ratio definitionused, the data distribution has a large impact on the required odds ratio forcertain performance characteristics: the odds ratio required for normallydistributed data may be 1 or 2 magnitudes higher than that required for datawith the double-exponential or Cauchy distribution.

Table 3. Performance Characteristics (Specificityand Sensitivity) of A1C Laboratory Test Data

No. correctlyidentified Specificity

Control populationActual 17,319 84.1%Normal distribution 13,880 67.4%Improvement �3439 13.7%

No. correctlyidentified Sensitivity

Diabetic populationActual 7532 70.5%Normal distribution 7198 67.4%Improvement �334 3.1%

Using 2.63 as the diagnostic cutoff, which corresponds to an odds ratio of24.3, the number and percentage of individuals who were correctly classifiedamong the 20,590 control (nondiabetic) and 10,677 diabetic individuals areshown. This analysis was repeated using data with a normal distribution thathad the same number of individuals and odds ratios. The number ofindividuals who were correctly classified using the actual data was comparedwith that in the normally distributed data (improvement).



Generation of a receiver operating characteristics (ROC)curve is a useful procedure for assessing the overall per-formance of a test. By varying the cutoff for A1C data, weobtain different pairs of specificity and sensitivity. Insteadof focusing on a specific pair, we can plot all such pairs ina scatter plot. The resulting curve is an ROC curve, showingthe change of sensitivity with respect to the change inspecificity (Fig. 2A). An ROC curve close to the ideal point

of 100% specificity and 100% sensitivity indicates niceperformance, and a curve close to the diagonal line from 0%specificity and 100% sensitivity to 100% specificity and 0%sensitivity indicates a poor performance. The total areaunder the ROC curve is also an indicator of the perfor-mance, with a value of 1 indicating perfect classificationpower and a value of 0.5 demonstrating that the variableused for classification is irrelevant to the actual outcome.

Figure 2. Receiver operating characteristic (ROC) curves for (A) A1C, (B) CD19, (C) protein S actual, and (D) shifted protein S laboratory data.The ROC curve for the data distribution is shown in red, and the ROC curve for corresponding data with a normal distribution with the sameodds ratio is shown in green. A, The areas under the ROC curve for the A1C and for the normal distribution data are 0.84 and 0.74, respectively.The specificity and sensitivity for the 2 points (triangle) are shown in Table 2. B, For CD-19 data, the actual odds ratio is 12.6, and the areaunder the curve (AUC) is 0.77. The corresponding AUC for normally distributed data with the same odds ratio is only 0.69. C, For protein S data,the actual odds ratio is 2.5, and the AUC is 0.6. The corresponding AUC for normally distributed data with the same odds ratio is 0.57. D, Forprotein S data, the data for the case group were artificially shifted to the right to achieve the performance of 80% specificity and 80% sensitivity.The odds ratio for the transformed data is 38, and the AUC is 0.86. The corresponding AUC for normally distributed data with the same oddsratio is 0.77.

SPECIAL ARTICLE


The improved performance of the actual A1C data inrelation to that of the corresponding normal distributionis demonstrated by comparison of their ROC curves (Fig.2A); the area under the ROC curve for the A1C data(0.84) is �10% more than that of the normal distribution(0.74).

The performances of 2 other laboratory tests were simi-larly evaluated. The number of CD19� cells is a potentialbiomarker for the diagnosis of lymphoma, and protein Sactivity is a potential biomarker for coagulopathy. Thehistograms and the quantile–quantile plots indicate thatthe data for both of these biomarkers have a nonnormaldistribution (Supplementary Figs. 3 [see SupplementaryDigital Content 4, http://links.lww.com/AA/A175]and 4 [see Supplementary Digital Content 5,http://links.lww.com/AA/A176]; see Supplementary Methodsfor figure legends, http://links.lww.com/AA/A172). TheORs required for these 2 tests to achieve a useful level ofdiagnostic performance were also significantly smaller thanif they had a normal distribution (Table 4). The requiredORs were comparable in magnitude to those for datafollowing a double-exponential distribution, and weremuch larger than those for data following a Cauchy distri-bution. Therefore, the 2 tails of the distribution curves forthe data in the control and disease groups play an impor-tant role in determining the relationship between biomar-ker performance and OR, and this data distribution wasbetter approximated by a double-exponential distribution.

The ROC curves of CD19 and protein S were alsocompared with the ROC curves of a normally distributedcase and control population with the same OR. For theCD19 data, there was a �10% gain in the area under thecurve (AUC) for the actual CD19 data in relation to that ofthe data with a normal distribution and the same OR (Fig.2B). The performance of the actual data was superior to thatof the normally distributed data over most of the range. Theprotein S data did not have good diagnostic utility becausethe AUC of the ROC curve was close to 0.5. However, theROC curve of the actual protein S data had a gain in AUCin relation to that of the normally distributed data (Fig. 2C).

The distributions of the protein S data in the case andcontrol populations were close to each other, which madethe impact of data distribution hard to compare. Therefore,the protein S data in the case group was artificially shiftedto the right to create a performance of 80% specificity and80% sensitivity, to investigate the effect that the data

distribution had on laboratory test performance when thecase and control populations were better separated bythe test. In this case, there was a 10% gain in the AUC in theshifted data in relation to that of the normally distributeddata (Fig. 2D). These results further demonstrate that thedata distribution has a large impact on the performance oflaboratory test data.

Additional Factors Affecting the LaboratoryData PerformanceIn addition to the data distribution, other factors can alsosignificantly impact the performance of laboratory data.For example, lactic dehydrogenase (LDH) is a prognosticbiomarker for survival in adult leukemia–lymphomacases.14 However, the response variable (survival at 5years) could have an erroneous value in a small percentageof patients if death were due to a nonleukemia-associatedcause (e.g., an automobile accident) or because of a failurein mortality reporting. In the OR example presented in theintroductory section of this article, such misclassificationwould represent errors due to patients having awarenessunder anesthesia that was not discovered by the inves-tigators, or patients claiming to have had awareness butthe events recalled did not take place during a generalanesthetic.

In a simulation, we evaluated the impact of such mis-classification of the response variable on laboratory dataperformance. At 80% specificity, a 1% (or 10%) misclas-sification incidence decreased the average sensitivityfrom 74.5% to 73.7% (or 66.3%) (Table 5). The samedecrease occurred when the underlying data distributionwas assumed to be normal. This simulation indicates thatlaboratory data performance decreases as the measure-ment error in the binary dependent variable increases,but the decrease is not very large if the measurementerror is small.

In many perioperative medicine studies, data are oftencollected from multiple centers to achieve a desired sample

Table 4. Odds Ratios Required to Achieve theIndicated Performance Characteristics (Sensitivityand Specificity)

Odds ratio(conventional definition)

Sensitivity Specificity A1C CD-19 Protein S80% 80% 50 44 3880% 90% 165 139 8790% 80% 165 71 6190% 90% 332 235 129

The odds ratios were calculated using the conventional definition of the oddsratio and under the assumption that the data distribution had the same shapeas that of A1C, CD-19� cell count, or protein S activity. These distributionswere shifted in order to obtain the desired performance characteristics, as isdescribed in the METHODS section.

Table 5. Results of a Simulation Examining theImpact of Misclassified Samples (% Mislabeled orIntercenter Differences in Data DistributionParameters [Intercenter �] on BiomarkerSensitivity [at 80% Specificity])

Sensitivity

Actual Normal% Mislabeled

0 74.5% 74.5%1% 73.7%�0.1% 73.6%�0.5%3% 72.0%�0.2% 71.9%�0.5%5% 70.4%�0.2% 70.1%�0.5%7.5% 68.4%�0.3% 67.9%�0.5%10% 66.3%�0.3% 65.6%�0.5%

Intercenter �0 74.5% 74.5%10% 74.1%�0.2% 74.3%�0.2%30% 71.9%�1.1% 72.6%�0.9%50% 68.8%�2.1% 69.7%�2.0%75% 64.3%�3.3% 64.8%�3.6%100% 60.4%�4.0% 59.9%�4.7%

The results are shown for simulations using the actual A1C data or when itwas adjusted to have a normal distribution.



size. However, laboratories at different centers could pro-duce data with different statistical distribution parameters,which introduces a degree of heterogeneity into the pooleddata. Therefore, we evaluated the impact of intercenterdifferences on a simulation study examining A1C datadistributions obtained from 10 hypothetical centers. At 80%specificity, the sensitivity of the A1C data only decreasedfrom 74.5% (if all data were from a single center) to 74.1%if the SD in the distribution parameter across centers was10% of the control A1C data (Table 5); and the shape of thedata distribution curve did not affect the size of thedecrease in the sensitivity. Thus, combining data frommultiple centers had a small effect on laboratory dataperformance if the tests achieved a reasonable level ofstandardization across the centers. If care is taken tominimize the variation among different data centers, thebenefit from a larger sample size outweighs any decrease inlaboratory data performance resulting from a multicenterstudy.

DISCUSSIONThe effect that a nonnormal data distribution has onlaboratory test performance may at first appear to be a topicthat is of interest to statisticians rather than physicians.However, this finding has significant implications for sci-entists who are discovering and developing new diagnosticmarkers, and subsequently for physicians who will use theresults provided by the next generation of diagnostic teststo care for their patients. This information is also relevant toresearchers using retrospective AIMS data to develop pre-dictors of postoperative outcomes on the basis of quantita-tive data recorded during anesthesia.

First, this analysis demonstrates that a nonnormal dis-tribution of laboratory data enables laboratory tests withmoderate ORs (6–30) to provide useful diagnostic tools.Second, these ORs are within the range observed forlaboratory data identified in etiologic and epidemiologicstudies. This implies that contemporary genomic studieshave the potential to produce clinically useful diagnostictools. Third, when evaluated within the larger contextof populations that are affected by common diseases(prevalence �1% of the population), the improvement inlaboratory test performance due to the nonnormal datadistribution assumes substantial significance. Finally, theevaluation of the usefulness of a marker (i.e., the indepen-dent variable) in predicting an outcome (i.e., the dependentvariable) requires a determination of the probability distri-bution of that marker in the population. It is insufficient tomake assumptions that the marker follows a normal distri-bution without evaluating this formally, because the failureto do this may result in a useful test being inappropriatelydiscarded. This caveat applies equally to quantitative fac-tors identified retrospectively from data-mining analyses ofAIMS databases to be associated with an undesirableoutcome.

For example, it is estimated that 18.2 million people(6.3% of the population) in the United States had diabetesin 2002, and 5.2 million of these were undiagnosed cases.15

Let us imagine that a scientist has just discovered thatmeasurement of A1C could be a potential diagnosticmarker that could be used for long-term monitoring of

patients with diabetes. The scientist has just received alarge number of serum samples that were obtained fromcontrol and diabetic individuals. After the A1C values weremeasured, the test developer must establish a thresholdvalue that will enable physicians to determine whether anindividual has an abnormal test result. As is shown here,the nonnormal distribution of A1C laboratory values re-sulted in a 13.7% increase in specificity and a 3.1% increasein sensitivity for the diagnosis of diabetes on the basis ofA1C test results. Thus, for every 1 million diabetic indi-viduals that were evaluated by this newly developed test, ifthe analyst used threshold values that were based upon adouble-exponential distribution of A1C laboratory data inthe population, 31,000 more test results would be correctlyclassified as abnormal. Similarly, when the test results froma population of 1 million control individuals were ana-lyzed, the use of these cutoff values would avoid misclas-sifying 137,000 individuals as having an abnormal testresult.

However, diagnoses are not based solely upon a labo-ratory test result; clinical context and judgment will alwaysbe essential for the proper use of any risk stratification tool.The prevalence of a particular condition and the level ofpretest suspicion within the tested population (pretestprobability) will impact laboratory test performance andthe interpretation of test results. For many genomic teststhat an anesthesiologist might use, different clinical sce-narios might lead to requirements for different test perfor-mance characteristics (sensitivity and specificity).

For example, suppose that there was a genomic screen-ing test that predicted susceptibility to malignant hyper-thermia (MH). For this test, we would want a sensitivity of�99%, because this condition can be fatal and we do notwant to miss anyone. However, we could accept a specific-ity of only 50% (for every 2 patients who are resistant toMH, only 1 will be predicted correctly), because there areeffective alternative methods of providing anesthesia usingnon-MH triggering drugs. Alternatively, suppose anothergenetic test predicted whether a patient is likely to experi-ence prolonged sedation when given midazolam. For thistest, we do not need an extremely high sensitivity, but wanta higher level of specificity than for MH. We do not want todeprive too many patients of the anxiolytic benefits of thisdrug, and we can effectively reverse the effect of midazo-lam with an antagonist (flumazenil), if necessary. However,knowing whether a patient is at risk for prolonged sedationwould be helpful in assessing a patient who is slow toawaken after an anesthetic, and it could improve outcome.If the patient were at risk for excessive sedation, thepractitioner might more quickly reach a decision to admin-ister flumazenil than if the patient were not at risk. Avoid-ing the administration of flumazenil in situations in whichslow emergence is not likely due to midazolam is desirable,because provoking a seizure is a potential complication ofreversal. These 2 different scenarios indicate how verydifferent test performance characteristics can be required toaddress different clinical situations.

The results from this study are especially important fororganizations such as the Anesthesia Quality Institute

SPECIAL ARTICLE


(AQI)b and the Multicenter Perioperative Outcomes group(MPOG),c both of which have recently initiated efforts toimprove the quality of anesthesia care through the retro-spective analysis of anesthesia data. The finding that up toa 10% error in survival classification has a small impact ontest performance is fortunate. Typically, the cause of deathcannot be determined from mortality databases that arepublically available, and there is also incomplete reportingto these databases. If small reporting errors adverselyaffected test performance, misclassification of the cause ofdeath would call into question the interpretation of suchstudies. Also encouraging for AIMS research involvingcollaborative databases is that statistical variation in thedistributions of reported parameters among various con-tributing centers are not likely to have a major impact onthe analysis. However, AQI and MPOG need to considerthe implications of the data distribution of identified inde-pendent variables and report on these in their publishedreports. The data distributions of the independent variableswill also affect power analysis calculations (which usuallyare based on assumptions that the data are normallydistributed) for subsequent prospective studies, becausegroup sizes necessary to demonstrate a specified change inoutcome will be influenced.

The prognosis for laboratory data is also further im-proved when viewed through a future-oriented lens. First,although this analysis evaluated commonly available labo-ratory data, it has significant implications for newly emergingdiagnostic tests identified using contemporary transcriptional,proteomic, or metabolomic technologies. It is likely that thedata for these newly discovered tests would also have asimilar (nonnormal) distribution, and we also have reason tobe more optimistic about their diagnostic performance. Sec-ond, this analysis evaluated the performance of a singlelaboratory test. However, it is likely that many laboratory testswill be identified by contemporary genomic analyses.

As has been noted for disease-associated gene expres-sion signatures,16 it is likely that a combination of geneticrisk factors and laboratory test data will produce betterclinical stratification tools. Fortunately, many statisticallearning methods and modeling tools,17 including nonlin-ear mixed-effects modeling,18 have been developed toidentify optimal combinations for clinical stratification.

Also, this analysis evaluated test performance within thecontext of existing systems for disease classification. Cur-rently, very large and diverse groups of individuals areincluded within a disease grouping that is largely basedupon a collection of symptoms, which often are disparate,and sometimes incorporating various laboratory variables.However, genetic risk factors and laboratory data have thepotential to transform our diagnostic and stratificationpractices. In the not too distant future, patient groupings

may be redefined using genetic risk factors and laboratorytest measurements. When emerging laboratory test resultsand genetic data are incorporated into disease classificationcriteria, patients with common underlying predispositionsand pathogenesis will be similarly classified. The sensitiv-ity and specificity of the laboratory results that are used fordiagnosis and prognosis will then improve.

ACKNOWLEDGMENTSWe thank Dr. Gomathi Krishnan for retrieving the biomarkerdata, and Dr. Bob Lewis for helpful discussions.

REFERENCES1. Guo Y, Shafer S, Weller P, Usuka J, Peltz G. Pharmacogenomics

and drug development. Pharmacogenomics 2005;6:857–642. Ware JH. The limitations of risk factors as prognostic tools.

New Engl J Med 2006;355:2615–73. Wald NJ, Hackshaw AK, Frost CD. When can a risk factor be

used as a worthwhile screening test? BMJ 1999;319:1562–54. Pepe MS, Janes H, Longton G, Leisenring W, Newcomb P.

Limitations of the odds ratio in gauging the performance of adiagnostic, prognostic, or screening marker. Am J Epidemiol2004;159:882–90

5. Sigl JC, Chamoun NG. An introduction to bispectral analysisfor the electroencephalogram. J. Clin Monit 1994;10:392–404

6. Conover WJ. Practical Nonparametric Statistics. New York:John Wiley & Sons, 1971

7. Jarque C, Bera A. Efficient tests for normality, homoscedastic-ity and serial independence of regression residuals. Econ Lett1980;6:255–9

8. Royston P. An extension of Shapiro and Wilk’s W test fornormalit to large samples. Appl Stat 1982;31:115–24

9. Benjamini Y, Hochberg Y. Controlling the false discovery rate:a practical and powerful approach to multiple testing. J R StatSoc B 1995;57:289–300

10. Lowe HJ, Ferris TA, Hernandez PM, Weber SC. STRIDE—Anintegrated standards-based translational research informaticsplatform. AMIA Ann Symp Proc 2009;2009:391–5

11. Larsen ML, Horder M, Mogensen EF. Effect of long-termmonitoring of glycosylated hemoglobin levels in insulin-dependent diabetes mellitus. N Engl J Med 1990;323:1021–5

12. Norton RM. The double exponential distribution: using calcu-lus to find a maximum likelihood estimator. Am Stat1984;38:135–6

13. Cassela G, Berger R. Statistical Inference. Pacific Grove, CA:Duxbury Press, 2002

14. Tsukasaki K, Hermine O, Bazarbachi A, Ratner L, Ramos JC,Harrington W Jr, O’Mahony D, Janik JE, Bittencourt AL, TaylorGP, Yamaguchi K, Utsunomiya A, Tobinai K, Watanabe T.Definition, prognostic factors, treatment, and response criteriaof adult T-cell leukemia–lymphoma: a proposal from an inter-national consensus meeting. J Clin Oncol 2009;27:453–9

15. Centers for Disease Control and Prevention. National DiabetesFact Sheet: National Estimates on Diabetes. Atlanta, GA:Centers for Disease Control and Prevention, 2005

16. van de Vijver MJ, He YD, van’t Veer LJ, Dai H, Hart AA,Voskuil DW, Schreiber GJ, Peterse JL, Roberts C, Marton MJ,Parrish M, Atsma D, Witteveen A, Glas A, Delahaye L, van derVelde T, Bartelink H, Rodenhuis S, Rutgers ET, Friend SH,Bernards R. A gene-expression signature as a predictor ofsurvival in breast cancer. N Engl J Med 2002;347:1999–2009

17. Hastie T, Tibshirani R, Friedman J. The Elements of StatisticalLearning. New York: Springer, 2001

18. Lindstrom MJ, Bates DM. Nonlinear mixed effects models forrepeated measures data. Biometrics 1990;46:673–87

bhttp://www.aqihq.org/resources.aspx, last accessed June 2, 2010.chttp://mpog.med.umich.edu/, last accessed June 2, 2010.



General Article

A Comparison of Liver Function After Hepatectomywith Inflow Occlusion Between Sevoflurane andPropofol AnesthesiaJ. C. Song, MD,* Y. M. Sun, MD,* L. Q. Yang, MD,* M. Z. Zhang, MD,† Z. J. Lu, MD,*and W. F. Yu, MD*

BACKGROUND: In this study, we compared liver function tests after hepatectomy with inflowocclusion as a function of propofol versus sevoflurane anesthesia.METHODS: One hundred patients undergoing elective liver resection with inflow occlusion wererandomized into a sevoflurane group or a propofol group. General anesthesia was induced with3 �g/kg fentanyl, 0.2 mg/kg cisatracurium, and target-controlled infusion of propofol, set at aplasma target concentration of 4 to 6 �g/mL, or sevoflurane initially started at 8%. Anesthesiawas maintained with target-controlled infusion of propofol (2–4 �g/mL) or sevoflurane(1.5%–2.5%). The primary end point was postoperative liver injury assessed by peak values ofliver transaminases.RESULTS: Transaminase levels peaked between the first and the third postoperative day. Peakalanine aminotransferase was 504 and 571 U/L in the sevoflurane group and the propofol group,respectively. Peak aspartate aminotransferase was 435 U/L after sevoflurane and 581 U/L inthe propofol group. There were no significant differences in peak alanine aminotransferase orpeak aspartate aminotransferase between groups. Other liver function tests including bilirubinand alkaline phosphatase, and peak values of white blood cell counts and creatinine, were alsonot different between groups.CONCLUSIONS: Sevoflurane and propofol anesthetics resulted in similar patterns of liverfunction tests after hepatectomy with inflow occlusion. These data suggest that the 2anesthetics are equivalent in this clinical context. (Anesth Analg 2010;111:1036–41)

Inflow occlusion by clamping of the portal triad (Pringlemaneuver) is routinely used in many centers1–3 toprevent blood loss during liver transsection.4,5 How-

ever, the Pringle maneuver induces ischemic injury in theremnant liver, which is associated with increased morbid-ity and mortality.6 Diseased livers such as steatotic orfibrotic livers may be the most vulnerable to temporaryinterruption of blood flow.7–9

Although volatile anesthetics and propofol have beenstudied and compared in ischemia/reperfusion injuries inmany organ systems, few studies have compared volatileanesthetics and propofol for liver resection with inflowocclusion. In this study, we compared sevoflurane- andpropofol-based anesthetics for hepatectomy with inflowocclusion, using indices of postoperative liver function asthe major end point for comparison.

METHODSOne hundred consecutive ASA physical status I/II/IIIpatients undergoing elective liver resection with inflow

occlusion between July 2009 and December 2009 wereassessed for study eligibility. Exclusion criteria were age�18 years, additional ablation therapies (cryosurgery orradiofrequency ablation), prior liver resection for donation,or scheduled resection not requiring inflow occlusion.Enrolled patients were randomized at the beginning of theoperation into a sevoflurane group (inhaled anesthesiawith sevoflurane) or a propofol group (target-controlledinfusion of propofol). All other anesthetic and surgicalmanagement was the same. The randomization sequencewithout any stratification was generated by computer andsealed with consecutively numbered envelopes providingconcealment of random allocation. The study was ap-proved by our local institutional research ethics committee.Written informed patient consent was obtained from allparticipants.

All patients received oral midazolam (5.0 mg) and IMatropine (0.5 mg) as a premedication. Electrocardiogram,radial arterial blood pressure, arterial oxygen saturation,and bispectral index were monitored routinely. Epiduralcatheters were placed at T7-10 interspaces before anesthe-sia, and a test dose of lidocaine was injected. A bolus of 7 to10 mL ropivacaine 0.75% was administered epidurally afterinduction of anesthesia. In the propofol group, generalanesthesia was induced with 3 �g/kg fentanyl, followed bya target-controlled infusion of propofol, set at a plasmatarget concentration of 4 to 6 �g/mL, and 0.2 mg/kgcisatracurium. In the sevoflurane group, anesthesia wasinduced with 3 �g/kg fentanyl, sevoflurane initially startedat 8%, and 0.2 mg/kg cisatracurium. After tracheal intuba-tion, anesthesia was maintained with a target-controlledinfusion of propofol (in the propofol group, as above) or

From the *Department of Anesthesiology, Eastern Hepatobiliary SurgeryHospital, Second Military Medical University; and †Department of Anes-thesiology, Shanghai Children’s Medical Center, Shanghai Jiao Tong Uni-versity School of Medicine, Shanghai, China.


Supported by a grant from the National Natural Science Foundation ofChina (No. 30901394).


Address correspondence and reprint requests to W.F. Yu, MD, Departmentof Anesthesiology, Eastern Hepatobiliary Surgery Hospital, Second MilitaryMedical University, Changhai Rd., No. 225, Shanghai, China. Address e-mailto [email protected].

Copyright © 2010 International Anesthesia Research SocietyDOI: 10.1213/ANE.0b013e3181effda8


sevoflurane (1.5%–2.5%), fentanyl 1 to 2 �g/kg, and cisa-tracurium 5- to 10-mg boluses according to clinical needs.To keep arterial mean blood pressure at a target of 60 mmHg, we administered 2 to 5 �g � kg�1 � min�1 dopamine asindicated. Depth of anesthesia was determined with thebispectral index with a target range between 35 and 45during surgery. Epidural catheters were removed after theoperation. Intravenous analgesia (continuous IV infusion of1000 mg tramadol and 100 �g sufentanil over 48 hours) wascontinued in all patients postoperatively.

Surgical procedures were performed in a standardizedmanner under the supervision of 2 experienced hepatobili-ary surgeons. After mobilization of the liver, inflow occlu-sion was achieved by the tourniquet technique around theportal triad with a 4-mm Mersilene tape. During resections,a low central venous pressure (0–5 mm Hg) was main-tained. The length of time for continuous inflow occlusionwas determined by the surgeons. All patients received thesame chemotherapy regimen before and after surgery.

Each patient was followed for the entire hospitalization.The primary end point was postoperative hepatocyte injurydefined by peak alanine aminotransferase (ALT) and aspar-tate aminotransferase (AST) levels over 6 postoperative days.Additional end points were peak values of white blood cells,bilirubin, alkaline phosphatase, creatinine levels, and lengthof hospital stay. All outcome variables were measured beforeand 1, 3, and 6 days after surgery. Additionally, cirrhosis(yes/no) was defined by histologic evaluation.

Group sample size was calculated based on differencesin postoperative peak ALT concentration in a pilot study ofpatients who received propofol anesthesia (685 � 392 U/L)and sevoflurane anesthesia (487 � 308 U/L). The followingformula: n � 15.7/ES2 � 1, where ES � effect size �(difference between groups)/(mean of the SD betweengroups), with � � 0.05 and power � 0.8 was used todetermine that the study would be adequately poweredwith n � 50 per group.10

We compared peak transaminases (primary outcome)between groups using a linear regression model with peaktransaminases as the dependent and group allocation as theindependent variable (corresponding to a 2-sample t test).In addition, we adjusted these comparisons for baselinetransaminases and bilirubin levels, ischemic (Pringle) time,size of excised liver (cm3), and blood loss in multivariablelinear regression analyses (analysis of covariance). The othermeasurement data were also compared using a 2-sample ttest, and count data were compared using �2. Hepaticfunction in patients with cirrhosis might be more sensitiveto inflow occlusion than without cirrhosis. Therefore, weconducted a limited number of subgroup analyses to assessthe effect of cirrhosis (yes/no) on postoperative hepatocyteinjury (ALT/AST) using a 2-sample t test. All analyses wereconducted using SPSS 16.0 (SPSS Inc., Chicago, IL). Resultswere expressed as the mean � SD. A P value �0.05 wasconsidered to represent statistical significance.

RESULTSFifty patients were included in each group. Table 1 showsthe patient characteristics and baseline values of the out-come variables. Table 2 shows a summary of the importantintraoperative data. In both data sets (Tables 1 and 2), there

were no significant differences. Table 3 shows the hemo-dynamic data during surgery. Epidural anesthesia was notattempted in 2 patients in each group for a platelet count�80 � 109/L.

Table 1. Patient Characteristics and PreoperativeLaboratory Values

Sevoflurane group(n � 50)

Propofol group(n � 50)

Age (y) 48.5 (8.9) 51.4 (7.8)Body height (cm) 170.3 (3.7) 167.5 (10.3)Weight (kg) 70.3 (9.8) 68.9 (12.1)Gender, male/female 40/10 36/14Baseline ALT (U/L) 40 (13) 43 (26)Baseline AST (U/L) 36 (11) 46 (29)Baseline bilirubin (�mol/L) 14.6 (5.2) 13.5 (5.9)Baseline WBC count (�103/mL) 5.9 (1.8) 6.1 (2.1)Baseline creatinine (�mol/L) 74.6 (10.1) 70.3 (12.3)Baseline ALP (U/L) 92 (41) 103 (57)MELD scorea 6.85 (1.01) 6.94 (0.82)Cirrhosis, yes/nob 33/17 27/23Malignant/benign disease 45/5 43/7

Hepatocellular carcinomas 36 36Intrahepatic

cholangiocarcinoma9 7

Intrahepatic bile duct stone 2 3Angioleiomyolipoma of liver 3 4

Data are expressed as mean (SD).ALT � alanine aminotransferase; AST � aspartate aminotransferase; WBC �white blood cell; ALP � alkaline phosphatase.a MELD, the Model for End-Stage Liver Disease, consists of serum bilirubinand creatinine levels, international normalized ratio (INR) for prothrombintime, and etiology of liver disease. The formula for the MELD score is 3.8 �loge(bilirubin �mg/dL�) � 11.2 � loge(INR) � 9.6 � loge(creatinine�mg/dL�) � 6.4 � (etiology: 0 if cholestatic or alcoholic, 1 otherwise).b Cirrhosis (yes/no) was defined by histologic evaluation.

Table 2. Intraoperative Data from PatientsUndergoing Hepatectomy with Inflow Occlusion

Sevoflurane group(n � 50)

Propofol group(n � 50)

Pringle time (min) 21.4 (8.5) 18.4 (6.3)Operation time (min) 136.0 (38.4) 124.3 (29.1)Blood loss (mL) 302 (269) 291 (187)Bispectral index 40.6 (5.2) 39.1 (6.1)Size of excised liver (cm3) 474.1 (450.4) 527.7 (398.9)Major/minor resection

Major resection �3segments

22 23

Minor resection �3segments

28 27

Data are mean (SD).

Table 3. Intraoperative Hemodynamic DataMean arterial bloodpressure (mm Hg) Heart rate (bpm)

Sevofluranegroup

(n � 50)

Propofolgroup

(n � 50)

Sevofluranegroup

(n � 50)

Propofolgroup

(n � 50)T0 94.9 (9.9) 97.1 (13.2) 86.0 (18.7) 83.7 (14.1)T1 76.7 (10.2) 80.3 (11.4) 74.6 (13.6) 74.2 (9.4)T2 76.2 (10.9) 79.9 (10.9) 72.6 (9.1) 69.5 (8.6)T3 82.7 (11.8) 80.7 (10.9) 74.9 (10.8) 69.7 (11.5)T4 80.0 (9.9) 84.6 (9.7) 70.3 (9.6) 66.2 (8.9)

Data are mean (SD).T0 � baseline; T1 � 5th min after anesthetic induction; T2 � 5th min beforePringle; T3 � middle moment of Pringle; T4 � 5th min after Pringle.


No patient died in this study. Major complicationsincluded sepsis (2 patients in each group), bleeding (2patients in each group), and biloma (1 patient in thesevoflurane group). The mean hospital stay was 2 daysshorter in the propofol group (14 vs 16 days) but withoutstatistical significance. The degree of ischemia and reper-fusion injury of the liver was assessed by postoperativepeak serum ALT and AST levels. Transaminase levelspeaked between the first and the third postoperative day.The sevoflurane group had slightly lower peak ALT andAST levels than the propofol group, but these were notstatistically different (P 0.05) (Table 4). Unadjusted andadjusted results in multivariable linear regression analyses(analysis of covariance) were almost identical. Unadjustedresults are presented in Table 4. Other liver function testssuch as bilirubin and alkaline phosphatase, as well as peakwhite blood cell levels and creatinine, were not differentbetween groups (P 0.05). The results of subgroup analy-ses comparing cirrhotic and noncirrhotic patients are pre-sented in Table 5. Although the serum levels of ALT/ASTwere higher in cirrhotic patients than in noncirrhotic pa-tients, there were no significant differences between thesesubgroups (P 0.05). Figures 1 and 2 show ALT and ASTlevels over 6 postoperative days, respectively.

DISCUSSIONWe compared the effect of volatile anesthetics with propo-fol anesthetics on liver function in patients undergoingliver resection with inflow occlusion. There were no signifi-cant differences in postoperative liver function as measuredby serial transaminase levels, or in clinical outcomes in the2 groups.

Numerous strategies have been designed to reduceischemia/reperfusion injury after liver resection. Two pro-tective strategies to prevent ischemic-reperfusion injuryhave been clinically accepted: ischemic preconditioning4,9

and intermittent clamping11of the portal triad. Both proce-dures require a surgical intervention and prolong the overalltime of the surgical procedure. Hence, a pharmacological ap-proach not requiring additional surgical procedures may bea more attractive alternative than the established surgicalstrategies. In this study, we wanted to address whether theanesthetic affects postoperative hepatic function in patientsundergoing elective liver resection with inflow occlusion.

Many factors contribute to hepatic injury and outcomesafter liver resection. Among these, patient characteristicssuch as the severity of liver disease and surgical complica-tions are common contributors to poor postoperative out-come. In this study, the patient characteristics, baselinevalues of the outcome variables, and the data of surgery-related events showed that the 2 patient groups were trulyequivalent (Tables 1 and 2). In addition, surgery-relatedfactors (Pringle time, size of excised liver, and blood loss)and baseline transaminases and bilirubin levels were con-sidered in multivariable linear regression analyses.

The effect of sevoflurane on hepatic function has beenexamined in several studies. The study conducted by Ebertand Arain12 suggested that sevoflurane, but not propofol,was associated with increased liver injury in patientsundergoing hepatectomy without inflow occlusion, but theliver function tests in sevoflurane-exposed patients were inthe upper limits of normal. The study by Beck-Schimmer etal.,13 however, suggested that sevoflurane preconditioningwas protective against ischemia/reperfusion injury duringliver resection. Sevoflurane preconditioning was shown toprevent hepatic injury, defined by transaminase levels, andimprove clinical outcome. In the volatile preconditioninggroup, the expression of inducible nitric oxide synthaseupon reperfusion significantly increased compared withthe baseline value, which points to a possible protectiverole of nitric oxide in pharmacological preconditioning. In arat hepatic ischemia/reperfusion injury model, clinically

Table 4. Postoperative Laboratory Data and Length of Hospital Stay

Sevofluranegroup

(n � 50)

Propofolgroup

(n � 50)

95% confidence intervalof the difference

(equal variances assumed)

Lower UpperPeak ALT (U/L) 504 (295) 571 (460) �159 293Peak AST (U/L) 435 (275) 581 (494) �92 383Peak bilirubin (�mol/L) 25.2 (9.3) 33.3 (28.7) �4.3 20.6Peak ALP (U/L) 121 (35) 144 (83) �16 61Peak WBC (�103/mL) 13.1 (2.7) 14.6 (4.6) �0.6 3.7Peak creatinine (�mol/L) 70.3 (11.0) 66.8 (11.7) �9.7 2.6Hospital stay (d) 16.1 (4.8) 14.0 (2.9) �4.6 0.4

Data are mean (SD).ALT � alanine aminotransferase; AST � aspartate aminotransferase; ALP � alkaline phosphatase; WBC � white blood cell.

Table 5. The Peak Values of ALT/AST of the Subgroups over 6 Postoperative DaysSevoflurane group Propofol group

Cirrhosis(n � 33)

Noncirrhosis(n � 17) P value

Cirrhosis(n � 27)

Noncirrhosis(n � 23) P value

Peak ALT (U/L) 558 (310) 398 (245) 0.215 675 (554) 449 (309) 0.231Peak AST (U/L) 482 (289) 342 (232) 0.218 672 (581) 474 (310) 0.326

Data are mean (SD).ALT � alanine aminotransferase; AST � aspartate aminotransferase.

Liver Function Tests are Similar After Hepatectomy with Propofol or Sevoflurane


relevant concentrations of sevoflurane given before, dur-ing, and after hepatic ischemia protected the liver againstischemia/reperfusion injury. Increased hepatic adenosinetriphosphate and energy levels decreased hepatocyte in-jury, and the hepatic tissue blood flow almost completelyrecovered after ischemia/reperfusion in the sevofluranegroup.14 The study by Imai et al.15 suggested that volatile

anesthetics may protect the fasted liver from early,neutrophil-independent, ischemia/reperfusion injury byacting during the reperfusion phase. Sevoflurane reducedhepatic oxygen consumption and attenuated lactate dehydro-genase release during reperfusion. Sevoflurane precondition-ing may provide a new and easily applicable therapeuticoption to protect the liver in hepatectomy.

Figure 1. Alanine aminotransferase (ALT) levelsover 6 postoperative days. T0 � baseline; T1 � firstpostoperative day; T2 � third postoperative day;T3 � sixth postoperative day; sevoflur � sevoflu-rane; CI � confidence interval. Each bar representsthe mean � SD.

Figure 2. Aspartate aminotransferase (AST) levelsover 6 postoperative days. T0 � baseline; T1 � firstpostoperative day; T2 � third postoperative day;T3 � sixth postoperative day; sevoflur � sevoflu-rane; CI � confidence interval. Each bar representsthe mean � SD.


Propofol is an IV sedative-hypnotic drug frequentlyused in anesthesia and intensive care that not only inhibitslipid peroxidation, but also enhances the cellular antioxi-dant defense system in several tissues including brain,liver, and heart.16,17 Propofol pretreatment attenuatesischemia/reperfusion–induced intestinal epithelial apopto-sis, which might be attributable to its antioxidant propertymodulating the ceramide pathway.18 Propofol improvessurvival of liver cells in rat hepatocyte suspensions exposedto oxidant injury by a free radical generator 2,2-azobis(2-amidinopropane) dihydrochloride (AAPH).19 Propofolprotects hepatic L02 cells from hydrogen peroxide–inducedapoptosis, partly through activating the MEK-ERK path-way and further suppressing Bad and Bax expression.20

The study by Shimono et al., however, suggested thatpropofol did not have a protective effect againsthypoxia/reoxygenation injury in rat liver slices.21 Theeffects of pharmacological preconditioning with propo-fol in patients undergoing liver resection with inflowocclusion remain to be determined.

In our study, patients with biopsy-proven cirrhosis didnot have worse postoperative liver dysfunction than thosewithout cirrhosis, likely because the cirrhotic patients hadmild disease, indicated by the Model for End-Stage LiverDisease score. In addition, the period of ischemia was short;longer ischemic stress may have unmasked a differencebetween cirrhotic and noncirrhotic patients. Previous stud-ies suggest that patients with early-stage cirrhosis cantolerate as much as 60 to 75 minutes of inflow occlusionduring hepatectomy without serious postoperative liverdecompensation.22–24

Our study had several limitations. It is not possible todetermine from our results whether sevoflurane andpropofol are protective in the setting of hepaticischemia/reperfusion injury. Second, patients were notsystematically assessed for hepatic fibrosis or steatosis,both risk factors for liver dysfunction after liver ischemia.In conclusion, our data suggest that sevoflurane or propo-fol can be used safely in patients undergoing hepatectomywith inflow occlusion, and that there is no advantage of onedrug over the other.

AUTHOR CONTRIBUTIONSJCS helped design and conduct the study, analyze the data,and write the manuscript. This author has seen the originalstudy data, reviewed the analysis of the data, approved thefinal manuscript, and is the author responsible for archiv-ing the study files. YMS helped design and conduct thestudy, analyze the data, and write the manuscript. Thisauthor has seen the original study data, reviewed theanalysis of the data, and approved the final manuscript.LQY helped conduct the study and analyze the data. Thisauthor has seen the original study data, reviewed theanalysis of the data, and approved the final manuscript.MZZ helped conduct the study. This author has seen theoriginal study data and approved the final manuscript. ZJLhelped conduct the study. This author has seen the originalstudy data and approved the final manuscript. WFY helpeddesign the study. This author has seen the original studydata, reviewed the analysis of the data, and approved thefinal manuscript.

REFERENCES1. Gozzetti G, Mazziotti A, Grazi GL, Jovine E, Gallucci A,

Gruttadauria S, Frena A, Morganti M, Ercolani G, Masetti M,Pierangeli F. Liver resection without blood transfusion. Br JSurg 1995;82:1105–10

2. Kooby DA, Stockman J, Ben-Porat L, Gonen M, Jarnagin WR,Dematteo RP, Tuorto S, Wuest D, Blumgart LH, Fong Y.Influence of transfusions on perioperative and long-term out-come in patients following hepatic resection for colorectalmetastases. Ann Surg 2003;237:860–70

3. van der Bilt JD, Livestro DP, Borren A, van Hillegersberg R,Borel Rinkes IH. European survey on the application ofvascular clamping in liver surgery. Dig Surg 2007;24:423–35

4. Clavien PA, Yadav S, Sindram D, Bentley RC. Protectiveeffects of ischemic preconditioning for liver resection per-formed under inflow occlusion in humans. Ann Surg2000;232:155– 62

5. Lesurtel M, Selzner M, Petrowsky H, McCormack L, ClavienPA. How should transection of the liver be performed? Aprospective randomized study in 100 consecutive patients:comparing four different transection strategies. Ann Surg2005;242:814–23

6. Clavien PA, Petrowsky H, DeOliveira ML, Graf R. Strategiesfor safer liver surgery and partial liver transplantation. N EnglJ Med 2007;356:1545–59

7. Wei AC, Tung-Ping Poon R, Fan ST, Wong J. Risk factors forperioperative morbidity and mortality after extended hepatec-tomy for hepatocellular carcinoma. Br J Surg 2003;90:33–41

8. Coelho JC, Claus CM, Machuca TN, Sobottka WH, GoncalvesCG. Liver resection: 10-year experience from a single institu-tion. Arq Gastroenterol 2004;41:229–33

9. Clavien PA, Selzner M, Rudiger HA, Graf R, Kadry Z, RoussonV, Jochum W. A prospective randomized study in 100 consecu-tive patients undergoing major liver resection with versuswithout ischemic preconditioning. Ann Surg 2003;238:843–52

10. Lerman J. Study design in clinical research: sample size esti-mation and power analysis. Can J Anaesth 1996;43:184–91

11. Petrowsky H, McCormack L, Trujillo M, Selzner M, Jochum W,Clavien PA. A prospective, randomized, controlled trial com-paring intermittent portal triad clamping versus ischemicpreconditioning with continuous clamping for major liverresection. Ann Surg 2006;244:921–30

12. Ebert TJ, Arain SR. Renal responses to low-flow desflurane,sevoflurane, and propofol in patients. Anesthesiology 2000;93:1401–6

13. Beck-Schimmer B, Breitenstein S, Urech S, De Conno E, Wit-tlinger M, Puhan M, Jochum W, Spahn DR, Graf R, Clavien PA.A randomized controlled trial on pharmacological precondi-tioning in liver surgery using a volatile anesthetic. Ann Surg2008;248:909–18

14. Bedirli N, Ofluoglu E, Kerem M, Utebey G, Alper M, YilmazerD, Bedirli A, Ozlu O, Pasaoglu H. Hepatic energy metabolismand the differential protective effects of sevoflurane and isoflu-rane anesthesia in a rat hepatic ischemia-reperfusion injurymodel. Anesth Analg 2008;106:830–7

15. Imai M, Kon S, Inaba H. Effects of halothane, isoflurane andsevoflurane on ischemia-reperfusion injury in the perfusedliver of fasted rats. Acta Anasthesiol Scand 1996;40:1242– 8

16. De La Cruz JP, Sedeno G, Carmona JA, Sanchez de la Cuesta F.The in vitro effects of propofol on tissular oxidative stress inthe rat. Anesth Analg 1998;87:1141–6

17. Kharasch ED, Armstrong AS, Gunn K, Artru A, Cox K, KarolMD. Clinical sevoflurane metabolism and disposition. II. Therole of cytochrome P450 2E1 in fluoride and hexafluoroisopro-panol formation. Anesthesiology 1995;82:1379–88

18. Liu KX, Chen SQ, Huang WQ, Li YS, Irwin MG, Xia Z.Propofol pretreatment reduces ceramide production andattenuates intestinal mucosal apoptosis induced by intesti-nal ischemia/reperfusion in rats. Anesth Analg 2008;107:1884 –91

19. Navapurkar VU, Skepper JN, Jones JG, Menon DK. Propofolpreserves the viability of isolated rat hepatocyte suspensionsunder an oxidant stress. Anesth Analg 1998;87:1152–7

Liver Function Tests are Similar After Hepatectomy with Propofol or Sevoflurane


20. Wang H, Xue Z, Wang Q, Feng X, Shen Z. Propofol protectshepatic L02 cells from hydrogen peroxide-induced apoptosisvia activation of extracellular signal-regulated kinases path-way. Anesth Analg 2008;107:534–40

21. Shimono H, Goromaru T, Kadota Y, Tsurumaru T, Kanmura Y.Propofol displays no protective effect against hypoxia/reoxygenation injury in rat liver slices. Anesth Analg2003;97:442–8

22. Kim YI, Kobayashi M, Aramaki M, Nakashima K, Mitarai Y,Yoshida T. “Early-stage” cirrhotic liver can withstand 75minutes of inflow occlusion during resection. Hepatogastroen-terology 1994;41:355–8

23. Kim YI, Nakashima K, Tada I, Kawano K, Kobayashi M.Prolonged normothermic ischaemia of human cirrhotic liverduring hepatectomy: a preliminary report. Br J Surg1993;80:1566 –70

24. Jeppsson B. Liver resection: prolonged inflow occlusion inhuman cirrhotic livers. HPB Surg 1996;10:123–5


Pain Medicine

Section Editor: Spencer S. Liu

The Costs and Benefits of Extending the Role of theAcute Pain Service on Clinical Outcomes After MajorElective SurgeryAnna Lee, PhD, Simon K. C. Chan, MBBS, Phoon Ping Chen, MBBS, Tony Gin, MD,Angel S. C. Lau, MPhil, and Chun Hung Chiu, MPhil

BACKGROUND: Acute pain services have received widespread acceptance and formal supportfrom institutions and organizations, but available evidence on their costs and benefits is scarce.Although there is good agreement on the provision of acute pain services after many majorsurgical procedures, there are other procedures for which the benefits are unclear. Data arerequired to justify any expansion of acute pain services. In this randomized, controlled clinicaltrial we compared the costs and effects of acute pain service care on clinical outcomes withconventional pain management on the ward. Patients included in the trial were considered bytheir anesthesiologist to have either arm be suitable for the procedure.METHODS: Four hundred twenty-three patients undergoing major elective surgery were random-ized either to an anesthesiologist-led, nurse-based acute pain service group with patient-controlled analgesia or to a control group with IM or IV boluses of opioid analgesia. Both groupswere treated with medications to treat opioid-related adverse effects and received the usual carefrom health professionals assigned to the ward. The main outcome measures were quality ofrecovery scores, pain intensity measures, global measure of treatment effectiveness, and overallpain treatment cost. Cost-effectiveness acceptability curves were drawn to detect a difference inthe joint cost-effect relationship between groups.RESULTS: There was no difference in quality of recovery score on postoperative day 1 betweentreatment and control groups (mean difference, 0; 95% confidence interval [CI], �0.7 to 0.7; P �0.94) or in the rate of improvement in quality of recovery score (mean difference, �0.1; 95% CI,�0.4 to 0.1; P � 0.34). The proportion of patients with 1 or more days of highly effective painmanagement was higher in the acute pain service group than in the control group (86% vs. 75%;P � 0.01). Costs were higher in the acute pain service group (mean difference, US$46; 95% CI,$44 to $48 per patient; P � 0.001). A cost-effectiveness acceptability curve showed that theacute pain service was more cost effective than was control for providing highly effective painmanagement if the decision maker was willing to pay more than US$546 per patient per 1 daywith highly effective treatment.CONCLUSION: In extending the role of the acute pain service to a specific group of major surgicalprocedures, the acute pain service was likely to be cost effective. (Anesth Analg 2010;111:1042–50)

Approximately 234 million major surgical proceduresare performed every year worldwide, and most ofthese will require some form of pain management.1

Although patient-controlled analgesia (PCA) and epiduralanalgesia provide better postoperative analgesia than doesintermittent IM analgesia,2 these techniques require specialcare and monitoring from acute pain service (APS) teams. InHong Kong, it is routine practice for APS,3 rather than wardnurses, to oversee PCA because of a lack of clinical nursespecialists.

In some countries, APS is considered to be a standard ofcare,4–6 but little is known about its economic benefits. Oursystematic review of 10 economic evaluations of APSprograms (involving 14,774 patients) also showed thatthere was insufficient evidence to draw conclusions aboutthe cost effectiveness and cost–benefit ratio of APS.7 Theoverall quality of published economic evaluations was poorbecause most studies were limited to partial economicanalyses.7

A limitation of the latest guidelines, “Acute Pain Man-agement: Scientific Evidence,”8 is that they generalized theevidence and did not present data on a specific procedure.9

There is usually good agreement on the provision of APSafter many major surgical procedures (e.g., with upperabdominal laparotomy). One problem with assessing thecosts and benefits of an established APS program is that itwould be unreasonable to revert to standard postoperativeward care for these patients, making a randomized trial ofAPS or no APS difficult. However, there are some surgicalprocedures for which the benefits of APS are unclear.4 This

From the Department of Anaesthesia and Intensive Care, The ChineseUniversity of Hong Kong, Prince of Wales Hospital, Shatin, NT, Hong Kong.


The work described in this paper was fully supported by a grant from theCentral Policy Unit of the Government of HKSAR and the Research GrantsCouncil of the HKSAR, China (project reference: CUHK4004-PPR20051).

Address correspondence to Anna Lee, PhD, Department of Anaesthesia andIntensive Care, The Chinese University of Hong Kong, Prince of WalesHospital, Shatin, NT, Hong Kong. Address e-mail to [email protected].

Copyright © 2010 International Anesthesia Research SocietyDOI: 10.1213/ANE.0b013e3181ed1317


may be because of new surgical techniques that maydecrease postoperative pain (laparoscopic assisted proce-dures), or patients expected to have significant postopera-tive pain but had not been offered APS in the past (e.g.,cardiac surgery). Although all anesthesiologists would con-sider using APS in these patients, mixed views were heldby staff on the desirability and necessity of providingAPS for them. Thus, data are required to justify anyexpansion of acute pain services to these patients. Be-cause they had not traditionally been receiving APS, itwas possible to randomize these patients to APS orstandard ward care.

Therefore, we performed prospective cost-effectivenessanalyses alongside a randomized controlled trial of APScare versus conventional pain management on the ward inpatients undergoing major elective surgery, for whom theattending anesthesiologist was uncertain about the benefitsof APS care. Our objective was to compare the costs andeffects of APS care on clinical outcomes with conventionalpain management on the ward in patients undergoingmajor elective surgery from the perspective of the HospitalAuthority (a government body funding public health ser-vices in Hong Kong).

METHODSThe study was conducted at the Prince of Wales Hospital inHong Kong, a large university teaching hospital. The studywas approved by the local Clinical Research Ethics Commit-tee. The study was registered in the Centre for Clinical TrialsClinical Registry of The Chinese University of Hong Kong(trial no. CUHK_CCT00105) on February 28, 2006, available athttp://www.cct.cuhk.edu.hk/registry/publictriallist.aspx. Aftergiving written informed consent, adult patients were en-rolled from April 8, 2006, to February 12, 2009. The APS isstaffed by 2.2 anesthesiologists and 1 nurse full-timeequivalent to provide the service continuously 24 h/day,serving an average of 17 patients per day (unpublisheddata for 2009).

PatientsGroups of patients were identified that could possiblybenefit from APS, but for whom APS care was not oftenused in the past. Inclusion criteria were adults, ages 18years or older and undergoing major surgery (such aslaparoscopic hemicolectomy, open reduction and internalfixation, total abdominal hysterectomy) or complex majorsurgery (e.g., coronary artery bypass graft, discectomy) asis defined by the Hong Kong Government Gazette.10 Pa-tients recruited to this study were only those identified bythe attending anesthesiologist as being candidates whomay or may not benefit from APS, in comparison withconventional ward pain service (CWPS). All patients hadgeneral anesthesia with no additional regional anesthesia.Patients were excluded if they were younger than 18 yearsof age, undergoing emergency or obstetric surgery, had ahistory of cognitive impairment or preoperative opioid use,or were unable to give consent.

The principal investigator (Anna Lee) generated therandom allocation sequence using a computer and was notinvolved in the data collection process. Two investigators(Angel S. C. Lau, and Chun Hung Chiu) were responsible

for randomizing eligible patients within 24 hours of meet-ing the study criteria to either APS or CWPS care, by usinga sealed opaque envelope containing a computer-generatedrandom treatment allocation.

Treatment ProceduresPatients randomized to the APS group received IV mor-phine PCA with or without supplementary oral analgesics,and medications to treat opioid-related adverse effects.They were seen by an APS nurse or an anesthesiologist orboth once daily (normal practice). The APS team wasinformed if any of the following occurred: inadequate paincontrol (persistent pain score �3 of 10), oxygen desatura-tion (Spo2 �90%), bradypnoea (respiratory rate �10/min),hypotension (systolic blood pressure �90 mm Hg), uncon-trolled nausea and vomiting with parenteral antiemetic,severe pruritus if uncontrolled with chlorpheniramine 5 mgq8h prn IM/IV, or difficulty awakening the patient. Al-though APS did offer acute pain techniques such as periph-eral nerve blocks and epidural PCA, these techniques werenot used in this study. Patients randomized to the CWPSgroup received opioid (pethidine, morphine, tramadol), non-opioid (diclofenac, paracetamol/phenyltoloxamine, paraceta-mol) analgesics by IM, IV boluses and/or oral routes, andmedications to treat opioid-related adverse effects prescribedby surgeons. All patients received the usual care from sur-geons and nursing professionals assigned to the surgical ward(and intensive care unit for postoperative care if undergoingcoronary artery bypass graft). The patient, attending healthprofessionals, and research staff were not blinded to thetreatment assignment.

Data CollectionAll surgical patients were admitted as inpatients on the daybefore surgery. Patients were interviewed by an investiga-tor (Angel S. C. Lau or Chun Hung Chiu) on 3 consecutivedays after their surgery with a standardized questionnaire.We collected data on patient’s demographics, ASA PhysicalStatus, and length of hospital stay. We measured painintensity (worse pain, average pain in the last 24 hours, andcurrent pain), pain at rest, pain during movement and paininterference with daily activities (walking ability, mood,sleep, relations with others, and ability to concentrate), andratings on a numeric rating scale (NRS) using the modifiedBrief Pain Inventory questionnaire11 on days 1 to 3 aftersurgery. A global measure of treatment effectiveness wasmeasured by asking patients, “How effective do you thinkthe treatment for pain was?” using a 5-point Likert scale(0 � poor, 1 � fair, 2 � good, 3 � very good, 4 � excellent).12

The 9-item Quality of Recovery (QOR) score13 was col-lected to measure the patient’s health-related quality of lifeafter anesthesia and surgery on a daily basis with a scorefrom 0 to 18. The frequency, severity, and distress ofopioid-related side effects (nausea, vomiting, difficulty inconcentrating, drowsiness or difficulty staying awake, feel-ing confused, and feeling of general fatigue or weakness)were totaled daily into an adverse effect score (from 0 to60). These specific symptoms from the opioid-relatedsymptom distress scale14 were chosen because they werethought to have a negative effect on patient’s daily activi-ties and recovery after surgery.


All calculated direct costs related to postoperative painmanagement were based on the first 3 days after surgery.From the patient’s drug chart, we recorded the type, dose,and frequency of analgesic drugs and the drugs used totreat opioid-related side effects. The medication costs wereestimated from the unit costs of the hospital pharmacy. ThePCA costs, obtained from the hospital administration,included the cost for infusion pump, IV tubing sets, car-tridges, catheters, batteries, syringes, needles, swabs, dress-ings, saline, and morphine. From the patient’s APS record,the staff cost was calculated using the total nursing andanesthesiologist time spent for each patient. The nursingand anesthesiologists’ staff salaries, obtained from thehospital administration, were based on the midpoint of therelevant pay scale. The ward nursing cost for APS andCWPS groups were assumed to be the same as a previousstudy at our hospital,15 showing that there was no signifi-cant difference in total ward nursing time (communication,documentation, administration of drug, and observations)for patients receiving PCA or IM opioid injection. The totalpostoperative pain management cost was a total of the costsfor analgesic drugs, drugs to treat opioid-related sideeffects, PCA, and APS staffing. At the time of reporting thestudy results (October 31, 2009), 1 US$ � HK$7.75. All costsare reported in U.S. dollars.

Outcome MeasuresThe primary outcome measure was QOR scores. Painintensity (mean pain ratings for worst pain, average pain inthe last 24 hours, and current pain), pain intensity at rest,pain intensity during movement, global measure of treat-ment effectiveness, and overall pain treatment cost out-comes were also measured. These outcome measures wereused in defining the incremental cost-effectiveness ratiosand incremental net benefits, the appropriate measuresof reporting results from a cost-effectiveness analysis.16 Spe-cifically, the effectiveness of the intervention for cost-effectiveness analysis purposes was expressed as the numberof pain-free days at rest,17 pain-free days with movement,17

and days with highly effective treatment. A pain-free daywas defined as having a NRS �3 on a 0 to 10 scale. Theanalgesic effectiveness was 3 if the patient had 3 pain-freedays, and 0 if the patient did not experience NRS �3 at all.The number of days with highly effective treatment was 3if the patient rated his or her global measure of effective-ness as excellent or very good on all 3 days, and 0 if thepatient did not have days with excellent or very good ratings.

Other secondary end points included pain interferenceduring daily activities, adverse effect score, in-hospital mor-tality, and length of hospital stay. A pain-free interference daywas defined as having a mean NRS of 0 with pain interferenceon daily activities scales ranging from 0 to 10. Theinterference-free effect was 3 if the patient had 3 pain-freeinterference days, and 0 if the patient experienced painthat interfered with daily activities on all 3 days. Thelength of stay was not included in the cost-effectivenessanalysis because we believed that it was a weak outcomemeasure. The delay in patient discharge from hospitalwas often not due to pain or analgesia-related side effectsbut due to postoperative rehabilitation plans, level of

social support, and surgeon’s postoperative treatmentpreferences.

Statistical AnalysisWe calculated the sample size using QOR as the primaryoutcome because this was a more patient-centered outcomethan were pain ratings and cost. We calculated that asample size of 522 would provide 80% power to detecta small to moderate effect size (0.25) between groups usinga 2-sample t test (EAST 5.2, Cytel Software Corporation,Cambridge, Massachusetts), allowing for interim analyses.Two interim analyses were planned after 174 and 348patients had completed their participation in the studyusing the O’Brien–Fleming stopping rules, with ad prioriboundaries of P � 0.0002 and P � 0.0121 to reject the nullhypothesis (efficacy boundary, if large treatment differ-ences appear before the end of the study), and P � 0.9659and P � 0.3444 to accept the null hypothesis (futilityboundary, if there is little chance of finding a significantdifference between groups).

The primary analyses were performed on a modifiedintention-to-treat basis (i.e., patients were analyzed accord-ing to their randomized allocated groups but were ex-cluded from the analysis if they did not adequately adhereto the protocol after randomization.). We used the 2-samplet test, �2 test, Fisher exact test, and Mann–Whitney U test tocompare baseline characteristics. The mean difference isdefined as the APS outcome measure minus CWPS outcomemeasure. For the QOR, pain intensity and interference out-comes, global measure of effectiveness, and adverse effectscore, we used multilevel regression models18,19 to assess theintervention effect on the change between the measure-ments taken on the first to third day after surgery. Anadvantage of using a multilevel regression model over arepeated-measure analysis of variance is that it can accountfor complex covariance structure and accommodate incom-plete data.19

Given the expected large variability of cost data, thestudy was underpowered to test the economic hypothesisthat APS would be more cost effective than would CWPS.However, in the absence of sufficient power to test theeconomic hypotheses, there have been methodological ad-vances in examining sampling uncertainty for incrementalcost-effectiveness ratios, with emphasis on the likelihoodthat the intervention represents good value for the costrather than on economic hypothesis testing.20,21

We assumed that APS was cost effective if the extra costof an extra gain in effect was less than the decision maker’swillingness to pay (WTP) for it.22 For example, if themaximum WTP was set at $200, APS would be costeffective if the incremental cost-effectiveness ratio (ratio ofthe extra cost to extra benefit, i.e., �C/�E) was less than$200. The 95% confidence interval (95% CI) around theincremental cost-effectiveness ratio was estimated usingthe Fieller method. Because there is uncertainty on the WTPvalue and the true estimate of the incremental cost-effectiveness ratio, the cost-effectiveness acceptability curvewas constructed from net benefit (NB) regressions.23 Underthe NB framework, each subject’s NB is computed from theobserved data as WTP � effecti � costi, where effecti andcosti are the data for the ith person’s effect and cost,

Cost Effectiveness of Acute Pain Service


respectively, and WTP is a willingness-to-pay number thatmust be specified.24 In its simplest form, the NB regressioninvolves fitting the following linear regression model:

NBi � �0 � �TX TXi � �i,

where TXi is the ith person’s treatment indicator (TXi � 1 forAPS and 0 for CWPS) and �i is a stochastic error term.24 Theequation is fitted several times, each time with a differentvalue of WTP value.24 To generate a cost-effectiveness accept-ability curve, we used the probability that �TX �0 for the yaxis and WTP for the x axis.24 In other words, the cost-acceptability curves showed the probability that APS wasmore cost effective than was CWPS for a range of values thatdecision makers might be willing to pay for 1 day gained ofbeneficial effect. All analyses were performed with STATAsoftware version 10.0 (StataCorp, College Station, Texas), andcost-effectiveness analysis was performed using the macro“iprogs” available from the University of Pennsylvania(www.uphs.upenn.edu/dgimhsr/stat-cicer.htm; accessed April28, 2010). A 2-sided P � 0.05 was considered statisticallysignificant. We declared the trial to be positive if more than 1of the outcomes, but not all, were significant after correctingfor multiplicity using the Holm stepwise approach (correctedoverall critical P value was 0.0167).25

RESULTSThe trial was stopped at the second interim analysis, after 402patients had completed the study, on the basis of slower-than-anticipated accrual rate and a prespecified futility stoppingrule. On the basis of 398 patients who had complete day 1QOR data (195 in the APS group and 203 in the CWPS group),comparison of the day 1 QOR via a 2-sample t test (P � 0.43)crossed the a priori futility boundary for early stopping withacceptance of the null hypothesis of no difference betweengroups. At interim analysis, it was calculated that if the studyhad continued to the planned enrollment of 522, the probabil-ity of demonstrating a difference in day 1 QOR betweentreatment groups was �1% under the alternative hypothesison the basis of the observed unadjusted day 1 QOR treatmentgroup differences.

Study PopulationOf the 470 surgical patients screened for the study, 422 metthe study criteria and were randomized. Two hundred ninepatients were allocated to the APS group, and 213 to theCWPS group (Fig. 1). Ten patients from each group with-drew from the study after randomization. Although thegender and type of surgery distributions in the withdrawalgroup were similar to those of patients who completed thestudy (Fisher exact test P � 1.00 and �2 test P � 0.16,respectively), the patients who withdrew were older thanthose who completed the study (mean [SD], 58 [8] for thosewho withdrew and 52 [12] for those who completed;2-sample t test P � 0.02). The baseline characteristics atenrollment were similar for age, gender, type and magni-tude of surgery, ASA Physical Status, and length of stay inthe intensive care unit between APS and CWPS patients(Table 1). Patients in the APS group had longer duration ofanesthesia than did those in the CWPS group (meandifference, 26 minutes; 95% confidence interval [CI], 10 to

42, 2-sample t test P � 0.01). Although the proportion ofadmissions to the intensive care unit after surgery wassimilar (�2 test P � 0.12) between the 2 groups, thepredicted risk of death from the Acute Physiologic andChronic Health Evaluation II score26 was higher in the APSgroup (Mann–Whitney U test, P � 0.01). The median [IQR]duration of PCA with morphine was 29.5 [18 to 43] hours.The median [IQR] time that anesthesiologists (includingPCA set-up time in the recovery room) and pain nursesspent caring for the patients in the APS group were 31 [23to 39] and 16 [8 to 16] minutes, respectively.

OutcomesThe point estimates on the first day after surgery and themean change over 3 days for QOR, pain intensity, interfer-ence with daily activity from pain, global measure oftreatment effectiveness, and adverse effect outcomes areshown in Table 2. There were no differences betweengroups for outcomes on the first day of surgery: QOR (P �0.94), pain intensity (P � 0.31), and pain on movement (P �0.17). However, APS patients had lower pain scores at rest,less interference with daily activities because of pain, andbetter treatment effectiveness than did CWPS patients onthe first day after surgery (Table 2). The rate of improve-ment in QOR scores (P � 0.34), daily rate reductions in painintensity (P � 0.20), and pain during movement (P � 0.07)between the 2 groups were similar. The APS group hadsignificantly smaller daily reductions in pain scores at rest

Figure 1. Patient flow through clinical trial. APS � acute pain service;CWPS � conventional ward pain service.


and interference with daily activities than did the CWPSgroup over the first 3 days after surgery (Table 2).

The incidence of moderate to severe pain (NRS �3 on a 0to 10 scale) at rest and on movement after major surgery in the2 groups is shown in Table 3. The incidences of “poor”treatment effectiveness on the first day after surgery in theAPS and CWPS groups were 0.5% (95% CI, 0.1 to 2.9) and4.8% (95% CI, 2.5 to 8.8), respectively. The proportion ofpatients with 1 or more days of highly effective pain manage-ment (i.e., treatment effectiveness rated as very good andexcellent) was higher in the APS group than in the controlgroup (86% vs. 75%; absolute risk difference 11%; 95% CI, 3%to 20%; �2 test P � 0.01; Fig. 2). This is equivalent to a numberneeded to treat of 9 (95% CI, 5 to 33). There were no significant

Table 1. Patient Characteristics at Enrollment

CharacteristicAcute pain service group

(N � 209)Conventional ward pain service group

(N � 213) P valueAge, mean (SD), years 53 (12.3) 52 (11.3) 0.41Women, no. (%) 106 (50.7) 108 (50.7) 1.00Type of surgery, no. (%)a

Orthopedic 32 (15.3) 36 (15.9)Gynecology 47 (22.5) 50 (23.5) 0.64Cardiothoracic 80 (38.3) 69 (32.4)General/other 50 (23.9) 58 (27.2)

Magnitude of surgery, no. (%)Major 123 (58.9) 140 (65.7) 0.13Ultramajor 86 (41.1) 73 (34.3)

ASA physical status grade, no. (%)b

I 62 (29.8) 78 (40.0)II 78 (37.5) 81 (38.4) 0.37III 61 (29.3) 45 (21.3)IV 7 (3.4) 7 (3.3)

Duration of anesthesia, mean(SD), minutes

225 (88.1) 199 (79.6) �0.01

Admission to ICU, no. (%) 64 (30.6) 51 (23.9) 0.12APACHE II score, median (IQR)c 12 (10–16) 11 (8–13) �0.01Length of ICU stay, median (IQR),

hours22 (19–23) 21 (18–22) 0.17

SD, standard deviation; ASA, American Society of Anesthesiologists; ICU, intensive care unit; APACHE II, Acute Physiology and Chronic Health Evaluation II; IQR,interquartile range.a Because of rounding, percentages may not all total 100.b American Society of Anesthesiologists’ physical status could not be determined for 3 patients (1 in acute pain service group and 2 in conventional ward painservice group).c APACHE II score ranges from 0 to 71, with higher scores indicating higher probability of death.

Table 2. Point Estimate on First Day After Surgery and Mean Change Over the First 3 Days AfterSurgery by Group, and Mean Difference Between Groups (95% Confidence Intervals) for Primaryand Secondary Outcomes

Acute pain service Conventional ward pain service Mean difference between groups

Day 1estimate

Mean change over3 days

Day 1estimate


Day 1estimate


Primary outcomeQORa 11.2 (10.7 to 11.7) 1.4 (1.2 to 1.6) 11.2 (10.7 to 11.7) 1.5 (1.3 to 1.7) 0 (�0.7 to 0.7) �0.1 (�0.4 to 0.1)

Secondary outcomesPain intensity 5.3 (4.9 to 5.6) �0.7 (�0.9 to �0.6) 5.5 (5.2 to 5.9) �0.9 (�1.0 to �0.7) �0.3 (�0.8 to 0.2) 0.1 (�0.1 to 0.3)Pain at rest 2.3 (1.9 to 2.8) �0.2 (�0.4 to �0.1) 3.2 (2.8 to 3.6) �0.6 (�0.7 to �0.4) �0.9 (�1.4 to �0.3)* 0.3 (0.1 to 0.5)*Pain on movement 4.5 (4.0 to 4.9) �0.5 (�0.7 to �0.3) 4.9 (4.5 to 5.4) �0.7 (�0.9 to �0.5) �0.5 (�1.1 to 0.2) 0.2 (0 to 0.5)Interference 3.0 (2.5 to 3.5) �0.4 (�0.6 to �0.2) 3.9 (3.4 to 4.4) �0.7 (�0.9 to �0.5) �0.9 (�1.6 to �0.2)* 0.3 (0.1 to 0.6)*Global treatment

effectiveness3.0 (2.8 to 3.2) �0.1 (�0.1 to 0) 2.4 (2.2 to 2.6) 0.1 (0.1 to 0.2) 0.6 (0.3 to 0.9)† �0.2 (�0.3 to �0.1)†

Adverse effectsb 13.9 (12.0 to 15.7) �2.9 (�3.6 to �2.2) 16.3 (14.5 to 18.1) �4.0 (�4.6 to �3.3) �2.4 (�5.0 to 0.1) 1.0 (0.0 to 2.0)‡

QOR, Quality of Recovery Score.a Higher QOR scores represent better recovery after anesthesia and surgery; higher global measure of effectiveness scores represent better effectiveness of painintervention.b Higher adverse effect scores represent worse experience with opioid-related side effects.* P � 0.01; † P � 0.001; ‡ P � 0.04.

Table 3. Incidence (%; 95% Confidence Interval)of Moderate to Severe Pain at Rest and onMovement After Major Surgery by Groups

Acutepain service

Conventionalward

pain serviceP

valueDay 1

At rest 25.8 (20.1 to 32.4) 34.7 (28.4 to 41.5) 0.06On movement 56.1 (49.0 to 63.0) 57.0 (50.1 to 63.7) 0.86

Day 2At rest 12.4 (8.3 to 18.0) 20.8 (15.7 to 27.0) 0.03On movement 41.9 (34.9 to 49.2) 45.1 (38.3 to 52.1) 0.53

Day 3At rest 19.7 (14.6 to 26.0) 14.9 (10.6 to 20.5) 0.22On movement 37.2 (30.5 to 44.5) 33.8 (27.6 to 40.7) 0.50



differences in the other pain-free outcomes (Table 4) betweenthe 2 groups. The mean duration of hospital stay (days) wassimilar between the APS and CWPS groups (12 [11] vs. 10 [12],respectively; 2-sample t test P � 0.13).

Adverse EventsOne patient in the APS group had respiratory depressiondue to PCA with morphine (incidence 0.5%, 95% CI, 0.1% to2.8%) and required IV 0.8 mg naloxone treatment. Duringthe study, 1 patient in the APS group died after coronaryartery bypass surgery in the intensive care unit because ofuncontrolled bleeding from the surgical site.

The risk of opioid-related side effects at any time duringthe follow-up was similar between groups (41% in the APSgroup and 40% in the CWPS group; absolute risk difference1%, 95% CI, �9% to 12%; �2 test P � 0.76). However, the

severity of opioid-related side effects on the first day aftersurgery tended to be less in the APS group than in the CWPSgroup (Table 2; P � 0.06). Overall, both groups experienced 1day of no opioid-related side effects (Table 4).

CostsAs was expected, the costs of analgesia, medications to treatopioid-related side effects, and APS staff costs were signifi-cantly higher in the APS group than in the CWPS group(Table 5). The mean difference in the total cost of paintreatment was US$46 (95% CI, $44 to $48) per patient (P �0.001). Because there was no significant extra day gainedfor being pain free at rest, pain free during movement, noopioid-related side effects, and no interference with dailyactivity measures in the APS group over the CWPS group,incremental cost-effectiveness ratios were not estimated.The incremental cost-effectiveness ratio for costs per 1 daywith highly effective treatment gained was US$151 (95% CI,$87 to $546) per patient. Decision makers who are willing topay less than US$87 per patient per 1 day with highlyeffective treatment can be 95% confident that the APSrepresents bad value; between US$87 and US$546 perpatient per 1 day with highly effective treatment, thedecision maker cannot be 95% confident that the 2 inter-ventions differ in value; for those willing to pay more thanUS$546 per patient per 1 day with highly effective treat-ment, they can be 95% confident that the APS representsgood value in comparison with CWPS (Fig. 3).

DISCUSSIONWe conducted a cost-effectiveness analysis alongside arandomized controlled trial of APS versus CWPS. Previousstudies included in systematic reviews7,27,28 examining theeffect of APS on postoperative outcomes were likely to bebiased because the studies were observational in design(before–after studies, matched comparisons). In this trial of422 patients, there were no significant differences in the

Figure 2. The number of days experiencing a highly effectivetreatment (“very good” and “excellent” ratings) in the acute painservice and conventional ward pain service groups. There was asignificant difference between the 2 groups (P � 0.01).

Table 4. Number of Outcome-Free Days (Median, IQR) During the First 3 Days After Surgery

Outcome

Acute pain servicegroup

(n � 199)

Conventional ward pain servicegroup

(n � 203) P valuePain freea 1 (0 to 2) 1 (0 to 2) 0.62Pain free at resta 3 (2 to 3) 3 (2 to 3) 0.63Pain free on movementa 2 (0 to 3) 2 (1 to 3) 0.69Pain-free interference on daily activitiesa 0 (0 to 1) 0 (0 to 1) 0.80Free of opioid-related side effectsb 1 (0 to 1) 1 (0 to 1) 0.81

IQR, interquartile range.a An outcome-free day was defined as numeric rating �3 on a 0 to 10 scale.b Defined as adverse effect score � 0.

Table 5. Mean Pain Treatment Use and Costs (in US$) per Patient

Cost categorya

Acute pain servicegroup

(n � 199)

Conventional ward pain servicegroup

(n � 203)Mean difference

(95% CI) P valueAnalgesia 18.74 1.21 17.53 (17.00 to 18.05) �0.001Medications to treat side effects 2.19 0.94 1.25 (0.07 to 2.42) 0.04Acute pain service staff costs 27.34 0.50 26.84 (25.63 to 28.05) �0.001Total cost of pain treatment 48.27 2.65 45.62 (43.52 to 47.71) �0.001a One patient in the acute pain service group did not receive the intervention because of lack of staff to set up the IV patient-controlled analgesia in the recoveryroom. Three patients in the conventional ward pain service group received acute pain service intervention after initial pain treatment was inadequate. Refer toMETHODS section in text for costing methodology.


QOR score on the first day after surgery or in the rate ofimprovement in the QOR score between the 2 groups. It ispossible that the 9-item QOR instrument lacked responsive-ness and discrimination when compared with the 40-itemQOR instrument.29 Nevertheless, we found lower painscores at rest, less interference with daily activities becauseof pain, and better global measure of treatment effective-ness for the APS group on the first day after surgery. Theseresults suggest that there are some initial benefits associ-ated with APS over CWPS in the early postoperativeperiod, but that they disappear on the second and third dayafter surgery. Although we did not measure patients’ locusof control in pain, we believe that PCA afforded greaterperceived control over pain relief30 and that this benefitextended past the first day after surgery. Not surprisingly,APS was associated with additional costs, mainly from staffcosts, which were 57% of the total cost of pain treatment.

Although up to one third of patients in this trial expe-rienced moderate to severe pain at rest on the first day aftersurgery, the mean difference in pain at rest scores (�0.9,95% CI, �1.4 to �0.3) between groups was statistically andclinically significant (28% reduction) if one considers a 20%reduction to represent a minimum clinically importantdifference.31 Half of our patients experienced moderate tosevere pain during movement on the first day after surgery.This incidence appears to be high in comparison with theresults of a meta-analysis of 33 studies by Dolin et al. (95%CI, 25% to 40%).2 We found differences between the 2groups for pain at rest, but not pain during movement onthe first day after surgery. These conflicting findings maysuggest that many patients are unable to distinguish be-tween pain at rest and pain during movement, because thismay be influenced by coughing, need for physiotherapy,and dressing changes.2 Therefore, our results for painduring movement require cautious interpretation.

Pain interference with daily activities—such as walkingability, mood, sleep, relations with others, and ability toconcentrate—are increasingly being used in conjunctionwith pain intensity outcomes. This outcome represents thephysical and mental functions affected by pain during therecovery process from anesthesia and surgery. We showed

that APS was associated with less pain interference on dailyactivities on the first day of surgery, but this effect oversubsequent days was less than that in the CWPS group andmade no difference to the quality of recovery.

During the trial, 1 patient died from surgical complica-tions. The risk of respiratory depression associated withPCA in the APS group (0.5%) was similar to that reportedin Werner et al.’s systematic review.27 We found no evidenceto support APS in preventing or reducing the incidence ofopioid-related side effects. However, in comparison with theCWPS group, there was some weak evidence to suggestthat APS was associated with milder opioid-related sideeffects.

We used a global measure of treatment effectivenessoutcome for our cost-effectiveness analysis. The globalmeasure of treatment effectiveness allows patients to bal-ance the unpleasantness or inconvenience of the painservice intervention, the personal meaningfulness of im-provements in pain and function, and the unpleasantnessand meaning of any opioid-related side effects.8 Weshowed that for every 9 patients treated with the APS, 1would experience 1 or more days of highly effective paintreatment. A previous study12 showed that the globalmeasure of treatment effectiveness is valid and can provideestimates of analgesic efficacy equivalent to pain relief.

When interpreting whether APS is cost effective, it isimportant to ask how much decision makers are willing topay for APS to be highly effective rather than spend fundson improving analgesic techniques per se for being painfree at rest or during movement. Our trial suggests that wecan be certain that APS is cost effective when the WTP foran extra day gained from a highly effective treatment wasmore than US$546 per patient. This estimate is higher thanthe value of what �90% of Canadian patients were willingto pay, which was no more than US$245* for APS toprovide PCA treatment.32 The differences in results may bedue to differences in median household income, patientpopulation, and the WTP methodology. It is likely that ouranesthesiologist-led nurse-based APS is cost effective be-cause funding of APS reflects public preferences for acutepain relief and perceived aversion to major complicationsrelated to inadequate pain relief.33,34

This study has several limitations that need to beconsidered. We expanded the coverage of APS care to aspecific group of major surgical procedures for which thebenefits of PCA were uncertain. The IV PCA mode ac-counted for 92% of the pain techniques given by the APS in2009 at our hospital. It could be argued that this studycompared IV PCA to IM analgesia, rather than a compari-son between APS and CWPS care. However, we believethat our study is the latter because we chose to focus on thehealth service infrastructure providing the postoperativepain management rather than the pain technique per se.

In the design of economic analysis alongside clinicaltrials, it is recommended that the study should be morenaturalistic to increase the external validity of the economicresults for which the objective is to determine the value for

*Conversion of CAN$200 in 1997 to US$245 at 2008 value fromhttp://eppi.ioe.ac.uk/costconversion/default.aspx. Accessed April 26,2010.

Figure 3. Cost-effectiveness acceptability curve showing the probabilitythat acute pain service (APS) is cost effective for a range of decisionmakers’ maximum willingness to pay (U.S. dollars) for 1 day with highlyeffective treatment gained. The observed mean incremental cost-effectiveness ratio of US$151 per patient (x axis) corresponds to a 50%probability of the acute pain service being cost effective (y axis).



money (i.e., lowest cost per unit benefit).20,35 That is,pragmatic trials evaluate effectiveness in comparison withstandard care in “real-world” patient populations andpractice settings, and although internal validity may belower than a standard randomized controlled trial, they aremore suitable for collection of health economic data.20,35

Although PCA is a common device used by APS, manypatients (63%) who are not cared for by an anesthesiologst-based service also receive PCA.6 Given the same availabil-ity of oral/IM analgesia and PCA technology, a dedicatedAPS will prescribe and use these pain techniques differ-ently than will non-APS physicians.36 For example, if sur-geons prescribed a regular parenteral opioid to patients on thefirst day after surgery, we would expect pain relief outcomesto be similar to those of the APS group. However, becausethere were significant differences in pain outcomes betweenthe groups on the first day after surgery, it is likely thatpatients were prescribed PRN analgesia in the CWPS group.Although surgeons want to maintain an active role in post-operative pain management,37 extending the role of APS islikely to be more cost effective than the CWPS by surgeons.

The results of this trial are not generalizable to differentorganizational structures of APS. Our APS is ananesthesiologist-led, nurse-based model that is the typicalmodel found in Hong Kong and Australian hospitals. Inmany centers, the anesthesiologist in the APS may onlyattend ward rounds once or twice a week, with most caregiven by nurses who specialize in pain medicine. Thiswould likely reduce the staffing costs associated with ananesthesiologist-led, nurse-based APS program. An eco-nomic analysis of a nursed-based APS33 concluded that itwas cost effective. However, we believe that there is stilluncertainty about its cost effectiveness given that a cost-effectiveness acceptability curve was not drawn. We esti-mate that, if we changed our APS model to a nurse-basedAPS, we would reduce the overall pain cost by 25%.However, we are uncertain about what correspondingchanges in clinical benefits would arise.

The nature of the pain service intervention meant thatpatients, health care professionals and the research assistantscould not be blinded. Therefore, we cannot exclude thepossibility of measurement bias and the Hawthorne effect ofthe APS visits in this trial. However, we believe that selectionbias and confounding would be minimal in this randomizedcontrolled trial. Because the pain score measured every 4hours was not part of routine postoperative care on the wardat the time of the study, we collected the daily pain scoresfrom patients; however, this may be prone to recall bias. Weused multiple primary outcomes measures to adequatelycapture the complexity of the pain experience and how it maybe modified by pain management interventions consistentwith best practices in pain research.8 Furthermore, we mini-mized the risk of making a false positive conclusion about theclinical benefits of APS using methods recommended by theIMMPACT (Initiative on Methods, Measurement, and PainAssessment in Clinical Trials) guidelines.25

Improving postoperative pain management by the useof APS remains a public health challenge.38 Many APS facestructural, political, cultural, educational, emotional, andphysical/technological challenges despite considerable ef-forts by APS members.38 Many surgeons are satisfied with

the service provided by the APS but many want to main-tain an active role in postoperative pain management.37 Wehave shown that the anesthesiologist-led, nurse-based APSis likely to be more cost effective than is the CWPS.

RECUSE NOTETony Gin is section editor of Anesthetic Clinical Pharmacologyfor the Journal. This manuscript was handled by Spencer S.Liu, section editor of Pain Medicine, and Dr. Gin was notinvolved in any way with the editorial process or decision.

REFERENCES1. Weiser TG, Regenbogen SE, Thompson KD, Haynes AB,

Lipsitz SR, Berry WR, Gawande AA. An estimation of theglobal volume of surgery: a modelling strategy based onavailable data. Lancet 2008;372:139–44

2. Dolin SJ, Cashman JN, Bland JM. Effectiveness of acute post-operative pain management: I. Evidence from published data.Br J Anaesth 2002;89:409–23

3. Cheung CW, Ying CL, Lee LH, Tsang SF, Tsui SL, Irwin MG.An audit of postoperative intravenous patient-controlled anal-gesia with morphine: evolution over the last decade. Eur J Pain2009;13:464–71

4. Hung CT, Lau LL, Chan CK, Chow B, Chui PT, Ho B, KungMC, Lui J, Hui T, Ho E, Chan SF, Chen PP. Acute pain servicesin Hong Kong: facilities, volume, and quality. Hong Kong Med J2002;8:196–201

5. McDonnell A, Nicholl J, Read SM. Acute Pain Teams inEngland: current provision and their role in postoperative painmanagement. J Clin Nurs 2003;12:387–93

6. Miaskowski C, Crews J, Ready LB, Paul SM, Ginsberg B.Anesthesia-based pain services improve the quality of postop-erative pain management. Pain 1999;80:23–9

7. Lee A, Chan S, Chen PP, Gin T, Lau AS. Economic evaluationsof acute pain service programs: a systematic review. Clin J Pain2007;23:726–33

8. Australian and New Zealand College of Anaesthetists andFaculty of Pain Med. Acute Pain Management: ScientificEvidence. 3nd ed. Melbourne: Australian and New ZealandCollege of Anaesthetists, 2010

9. Schug SA, Kehlet H, Bonnet F, Camu F, Fischer B, Joshi G,Neugebauer EAM, Rawal N, Simanski C. Procedure specificpain management after surgery—“PROSPECT.” Acute Pain2007;9:55–7

10. Hong Kong Government. Hong Kong Government Gazette2003;13/2003(Special Suppl 4, Annex IV): D4668–704

11. Mendoza TR, Chen C, Brugger A, Hubbard R, Snabes M,Palmer SN, Zhang Q, Cleeland CS. The utility and validity ofthe modified brief pain inventory in a multiple-dose postop-erative analgesic trial. Clin J Pain 2004;20:357–62

12. Collins SL, Edwards J, Moore RA, Smith LA, McQuay HJ.Seeking a simple measure of analgesia for mega-trials: is asingle global assessment good enough? Pain 2001;91:189–94

13. Myles PS, Hunt JO, Nightingale CE, Fletcher H, Beh T, Tanil D,Nagy A, Rubinstein A, Ponsford JL. Development and psycho-metric testing of a quality of recovery score after generalanesthesia and surgery in adults. Anesth Analg 1999;88:83–90

14. Apfelbaum JL, Gan TJ, Zhao S, Hanna DB, Chen C. Reliabilityand validity of the perioperative opioid-related symptomdistress scale. Anesth Analg 2004;99:699–709

15. Chang AM, Ip WY, Cheung TH. Patient-controlled analgesiaversus conventional intramuscular injection: a cost effective-ness analysis. J Adv Nurs 2004;46:531–41

16. Hoch JS, Dewa CS. A clinician’s guide to correct cost-effectiveness analysis: think incremental not average. CanJ Psychiatry 2008;53:267–74

17. Bartha E, Carlsson P, Kalman S. Evaluation of costs andeffects of epidural analgesia and patient-controlled intrave-nous analgesia after major abdominal surgery. Br J Anaesth2006;96:111–7

18. Singer JD, Willet JB. Applied Longitudinal Data Analysis. NewYork: Oxford University Press, 2003


19. Fitzmaurice GM, Ravichandran C. A primer in longitudinaldata analysis. Circulation 2008;118:2005–10

20. Marshall DA, Hux M. Design and analysis issues for economicanalysis alongside clinical trials. Med Care 2009;47:S14–20

21. Glick HA, Doshi JA, Sonnad SS, Polsky D. Economic evalua-tion in clinical trials. New York: Oxford University Press, 2007

22. Hoch JS, Rockx MA, Krahn AD. Using the net benefit regres-sion framework to construct cost-effectiveness acceptabilitycurves: an example using data from a trial of external looprecorders versus Holter monitoring for ambulatory monitoringof “community acquired” syncope. BMC Health Serv Res2006;6:68

23. Hoch JS, Briggs AH, Willan AR. Something old, somethingnew, something borrowed, something blue: a framework forthe marriage of health econometrics and cost-effectivenessanalysis. Health Econ 2002;11:415–30

24. Hoch JS. Improving efficiency and value in palliative care withnet benefit regression: an introduction to a simple method forcost-effectiveness analysis with person-level data. J Pain Symp-tom Manage 2009;38:54–61

25. Turk DC, Dworkin RH, McDermott MP, Bellamy N, Burke LB,Chandler JM, Cleeland CS, Cowan P, Dimitrova R, Farrar JT,Hertz S, Heyse JF, Iyengar S, Jadad AR, Jay GW, Jermano JA,Katz NP, Manning DC, Martin S, Max MB, McGrath P,McQuay HJ, Quessy S, Rappaport BA, Revicki DA, RothmanM, Stauffer JW, Svensson O, White RE, Witter J. Analyzingmultiple endpoints in clinical trials of pain treatments:IMMPACT recommendations. Pain 2008;139:485–93

26. Knaus WA, Draper EA, Wagner DP, Zimmerman JE. APACHEII: a severity of disease classification system. Crit Care Med1985;13:818–29

27. Werner MU, Soholm L, Rotboll-Nielsen P, Kehlet H. Does anacute pain service improve postoperative outcome? AnesthAnalg 2002;95:1361–72

28. McDonnell A, Nicholl J, Read SM. Acute pain teams and themanagement of postoperative pain: a systematic review andmeta-analysis. J Adv Nurs 2003;41:261–73

29. Myles PS, Weitkamp B, Jones K, Melick J, Hensen S. Validityand reliability of a postoperative quality of recovery score: theQoR-40. Br J Anaesth 2000;84:11–5

30. Chumbley GM, Hall GM, Salmon P. Patient-controlled analge-sia: an assessment by 200 patients. Anaesthesia 1998;53:216–21

31. Cepeda MS, Africano JM, Polo R, Alcala R, Carr DB. Whatdecline in pain intensity is meaningful to patients with acutepain? Pain 2003;105:151–7

32. Badner NH, Komar WE, Craen RA. Patient attitudes regardingpca and associated costs. Can J Anaesth 1997;44:255–8

33. Stadler M, Schlander M, Braeckman M, Nguyen T, BoogaertsJG. A cost-utility and cost-effectiveness analysis of an acutepain service. J Clin Anesth 2004;16:159–67

34. Macario A, Fleisher LA. Is there value in obtaining a patient’swillingness to pay for a particular anesthetic intervention?Anesthesiology 2006;104:906–9

35. Ramsey S, Willke R, Briggs A, Brown R, Buxton M, Chawla A,Cook J, Glick H, Liljas B, Petitti D, Reed S. Good researchpractices for cost-effectiveness analysis alongside clinical trials:the ISPOR RCT-CEA Task Force report. Value Health2005;8:521–33

36. Stacey BR, Rudy TE, Nelhaus D. Management of patient-controlled analgesia: a comparison of primary surgeons and adedicated pain service. Anesth Analg 1997;85:130–4

37. Chan SK, Chui PT, Lee A, Lai PB, Li TY, Gin T. Surgeons’attitudes and perception of an acute pain service. Hong KongMed J 2008;14:342–7

38. Powell AE, Davies HT, Bannister J, Macrae WA. Challenge ofimproving postoperative pain management: case studies ofthree acute pain services in the UK National Health Service.Br J Anaesth 2009;102:824–31



Pain Mechanisms

Section Editor: Tony L. Yaksh/Quinn H. Hogan

The Effect of Ketamine Anesthesia on the ImmuneFunction of Mice with Postoperative SepticemiaTetsuya Takahashi, MD, PhD,* Manabu Kinoshita, MD, PhD,† Satoshi Shono, MD,† Yoshiko Habu, PhD,†Takahiro Ogura, MD,* Shuhji Seki, MD, PhD,† and Tomiei Kazama, MD, PhD*

BACKGROUND: It is unknown how ketamine anesthesia immunologically affects the outcome ofpatients with postoperative septicemia. We investigated the effects of ketamine anesthesia onmice with an Escherichia coli or lipopolysaccharide (LPS) challenge after laparotomy, focusing onphagocytosis by liver macrophages (Kupffer cells) and cytokine production.METHODS: C57BL/6 mice received ketamine or sevoflurane anesthesia during laparotomy, whichwas followed by an E. coli or LPS challenge; thereafter, mouse survival rates and cytokine secretionswere examined. The effects of a �-adrenoceptor antagonist, nadolol, on ketamine anesthesia werealso assessed to clarify the mechanisms of ketamine-induced immunosuppressive effects.RESULTS: Ketamine anesthesia increased the mouse survival rate after LPS challenge afterlaparotomy compared with sevoflurane anesthesia, whereas such an effect of ketamine was notobserved after E. coli challenge. Ketamine suppressed tumor necrosis factor (TNF) and interferon(IFN)-� secretion after LPS and E. coli challenge. When bacterial growth was inhibited using anantibiotic, ketamine anesthesia effectively improved mouse survival after E. coli challengecompared with sevoflurane anesthesia. Neutralization of TNF also improved survival anddecreased IFN-� secretion after bacterial challenge in antibiotic-treated mice with sevofluraneanesthesia, suggesting that ketamine’s suppression of TNF may improve survival. Ketamine alsosuppressed in vivo phagocytosis of microspheres by Kupffer cells in LPS-challenged mice.Concomitant use of nadolol with an anesthetic dose of ketamine did not restore TNF suppressionin LPS-challenged mice, suggesting a mechanism independent of the �-adrenergic pathway.However, it restored TNF secretion under low-dose ketamine (10% anesthetic dose). In contrast,nadolol restored the decrease in phagocytosis by Kupffer cells, which was induced by theanesthetic dose of ketamine via the �-adrenergic pathway, suggesting distinct mechanisms.CONCLUSION: Ketamine suppresses TNF production and phagocytosis by Kupffer cells/macrophages.Therefore, unless bacterial growth is well controlled (by an antibiotic), postoperative infection mightnot improve despite reduction of the inflammatory response. (Anesth Analg 2010;111:1051–8)

Anesthesia affects the immune functions of patients.The decision to use one or a combination of anes-thetics might affect both patient outcome and prog-

nosis.1 A combination of anesthetics also might haveaffected the tumor metastasis in experimental animals.2

Ketamine is an anesthetic commonly used for patients andlaboratory animals with septicemia or trauma. Ketaminereduces proinflammatory cytokine production, and therebyimproves the survival rate in lipopolysaccharide (LPS)-induced septicemia models.3–5 Whereas ketamine reportedlyalso improves the survival rate of animals after E. coli infec-tion,4 other reports have demonstrated that ketamine does notaffect the outcome of sepsis using not only C3H/HeN mice

but also C3H/HeJ mice initiated by cecal ligation and punc-ture.6 Moreover, it has been reported that ketamineincreases mouse mortality after cecal ligation and punc-ture.7 In addition, studies of human polymorphonuclearleukocytes have demonstrated that ketamine attenuatesnot only proinflammatory cytokine secretion but alsophagocytosis and bactericidal activity in vitro.8 –11 Theseequivocal results may be attributed in part to the differ-ent septic models used. Nonetheless, a detailed immu-nological analysis of ketamine anesthesia is required tofully understand the ramifications of its clinical use.

Sepsis is defined as a systemic inflammatory responsesyndrome (SIRS) induced by infection.12 If SIRS occurs inthe absence of massive bacterial infection, suppressinginflammatory responses and cytokine production mayincrease host survival.13,14 In contrast, when the inflamma-tory response is insufficient and production of proinflam-matory cytokines is inadequate, the host cannot eliminatethe invading bacteria.15–17 Thus, inflammatory responsesinduced in the presence and absence of bacterial infectionmust be considered separately to fully understand theprocesses that are critical to survival.

Ketamine stimulates the �-adrenergic pathway insympathetic nerves and induces the production of cat-echolamines.18,19 �-Adrenoceptor agonists such as adrena-line and isoproterenol (catecholamine) suppress LPS-induced

From the *Department of Anesthesiology, National Defense Medical Col-lege, 3-2 Namiki Tokorozawa Saitama Japan 359-8513, †Department ofImmunology and Microbiology, National Defense Medical College, 3-2Namiki Tokorozawa Saitama Japan 359-8513.


This work was supported in part by a grant-in-aid for Special Research Program(Host stress responses to internal and external factors) from the NationalDefense Medical College (to S.S.).

Disclosure: The authors report no conflict of interest.

Address correspondence to Shuhji Seki, MD, Department of Immunologyand Microbiology, National Defense Medical College, 3-2 Namiki, To-korozawa, 359-8613 Japan. Address e-mail to [email protected]

Copyright © 2010 International Anesthesia Research SocietyDOI: 10.1213/ANE.0b013e3181ed12fc


tumor necrosis factor (TNF) production20–22 and natural killer(NK) cell activity.23–25 Therefore, ketamine-induced sympa-thetic nerve stimulation may impair the host immunesystem. Nevertheless, there is only limited understandingof how ketamine anesthesia affects cytokine productionand phagocytosis by macrophages and liver Kupffer cellsand how the ketamine-stimulated �-adrenergic pathwayaffects host immune functions. We hypothesized thatketamine anesthesia during surgery may change theproinflammatory cytokine response against LPS andbacteria challenges and may also affect phagocytosis ofbacteria by Kupffer cells. Based on this hypothesis, wefound the immune-suppressive effect of ketamine onKupffer cells and its possible mechanism.

METHODSAnimals and RegentsAll experiments were approved by the National DefenseMedical College Institution Animal Care and Use Commit-tee. Male C57BL/6 mice (10 weeks old, 25 g) were obtainedfrom SLC, Inc. (Shizuoka, Japan). Escherichia coli strain B(ATCC11303, Sigma-Aldrich, St. Louis, MO) and LPS (E.coli 0111: B4, Sigma-Aldrich) were used for experiments.Ketamine (preservative-free; Daiichi Sankyo Propharma,

Kanagawa, Japan) and sevoflurane (Maruishi, Osaka, Ja-pan) were used as anesthetics. �-Galactosylceramide(�-GalCer) was kindly provided by Pharmaceutical Re-search Laboratory, Kirin Brewery, Takasaki, Japan.

Ketamine Anesthesia and Surgical InterventionMice were injected intraperitoneally (IP) with an anestheticdose of ketamine (100 mg/kg/0.5 mL) or phosphate buffersodium (PBS) (0.5 mL) 10 minutes before laparotomy.During laparotomy, mice were anesthetized with approxi-mately 0.2% to 0.5% sevoflurane (ketamine group, n � 25)or approximately 2% to 3% sevoflurane (ketamine plussevoflurane group, n � 25) for 5 minutes after ketamineinjection or approximately 2% to 3% sevoflurane for 5minutes (sevoflurane group, n � 25) in room air via avaporizer (Fig. 1A). During laparotomy, we must some-times change the concentration of sevoflurane to regulatethe movement and respiratory conditions of mice duringsurgical maneuvers, as described in our previous study.2

The 3-cm midline incision was made for the laparotomyand was closed in layers using 4 to 0 silk sutures. Ananesthetic dose of ketamine alone cannot satisfactorilysedate mouse movement during laparotomy. Also, ket-amine is not clinically used without other anesthetics. We

Figure 1. Experimental design of ketamine anesthe-sia in mice.

Ketamine Anesthesia and Postoperative Infection


studied the effects of anesthetic methods using ketaminebut not the direct pharmacological effect of ketamine per se.Therefore, a small dose of sevoflurane inhalant (approxi-mately 0.2%–0.5%) was administered to ketamine-treatedmice during the laparotomy to obtain the appropriatesedative state.

E. coli or LPS ChallengeImmediately after wound closure, mice were challenged IPwith a lethal (1 � 109 colony-forming units (CFU)/mouse)or sublethal (1 � 108 CFU/mouse) dose of E. coli (n � 25 ineach group) (Fig. 1B). Antibiotic-treated mice were alsochallenged IP with 2 � 108 CFU/mouse of E. coli immedi-ately after laparotomy (Fig. 1C). Similarly, mice werechallenged IP with LPS (200 �g/mouse) after laparotomy(Fig. 1A, n � 25 in each group).

Antibiotic Treatment (Double andSingle Injections)Tail vein injections of cefazolin sodium (50 mg/kg) wereadministered to mice in both the ketamine (with approxi-mately 0.2%– 0.5% sevoflurane) and sevoflurane (ap-proximately 2%–3%) groups, one immediately before thelaparotomy and the second 6 hours after the E. colichallenge (n � 20 in each group). A single injection withcefazolin sodium (50 mg/kg) was also given to ketamine-anesthetized mice just before the laparotomy but notafter bacterial challenge (Fig. 1C, n � 25).

Neutralization of TNF with Antibiotic TreatmentAnti-TNF neutralizing antibody (0.5 mg/mouse, MP6-XT3;BD Pharmingen) or rat IgG was injected via IV 1 hourbefore E. coli challenge (2 � 108 CFU/mouse) after lapa-rotomy. The mice received double or single injections withcefazolin sodium, as described above (Fig. 1D, n � 20 ineach group).

Pretreatment with NadololNadolol is a nonselective �-adrenoceptor antagonist. Micewere injected with nadolol (10 mg/kg/0.2 mL) or 0.2 mLPBS via the tail vein 5 minutes before ketamine/PBStreatment. The mice were then laparotomized, followed byLPS challenge (Fig. 1E, n � 5 in each group). To preciselyevaluate the relationship between ketamine and nadolol,some mice did not receive laparotomy, thereby avoidingsevoflurane treatment. Ten minutes after ketamine/PBStreatment, these mice were similarly challenged IP withLPS (n � 5 in each group). To examine the affect of nadololon mice treated with a low dose of ketamine, mice werealso injected IP with 10 mg/kg ketamine (10% of theanesthetic induction dose) 5 minutes after the nadolol/PBStreatment (n � 5 in each group).

Cytokine Measurements and Bacterial CountsBlood samples of individual mice were obtained from theretro-orbital sinus immediately before anesthesia adminis-tration and at 1, 3, 6, 12, 24, and 48 hours after E. coli or LPSchallenge. Sera and culture supernatants of mononuclearcells (MNCs) were detected using Enzyme-Linked Immuno-Sorbent Assay (ELISA) kits (TNF and interferon [IFN]-�,BD Pharmingen, San Diego, CA; IL-12, Endogen, Woburn,MA). Livers were aseptically removed and homogenized in

PBS. Homogenates were serially diluted 10-fold with PBS,plated on brain heart infusion agar, and then incubated at37°C for 24 hours for colony counting.

Isolation of Peripheral Blood Leukocytes, Liver,and Spleen MNCs, and In Vitro LPS StimulationBlood samples were obtained from mice by cardiac punc-ture, and leukocytes were separated by hemolysis. LiverMNCs were prepared as previously described.13,26,27 Inbrief, liver specimens were minced, placed in 0.05% colla-genase solution (Wako, Osaka, Japan), shaken for 20 min-utes in a 37°C water bath, and passed through steel mesh.After mixing in 33% Percoll, the samples were centrifugedat 500g for 20 minutes at room temperature. Liver MNCswere obtained after the red blood cells were lysed. Spleno-cytes were also obtained after the spleen was passedthrough a mesh and the red blood cells were lysed. Thecells (5 � 105 per well in 200 �L Roswell Park MemorialInstitute [RPMI] 1640 supplemented with 10% fetal bovineserum in 96 well flat–bottomed plates) were incubated withketamine (1 � 10�6 M) at 37°C, in 5% CO2 for 1 hour.Subsequently, the cells were incubated with LPS (200ng/mL) for 2 hours. Three individual experiments wereperformed, with three or four samples per each group.

In Vivo Phagocytosis of Microspheres byKupffer CellsMice were injected with 20 �L of Fluoresbrite YG (FITC)Carboxylate Microspheres (75-nm diameter; PolysciencesEurope, Eppelheim, Germany) (hereafter called FITC-microspheres) via tail veins, immediately after LPS/PBSchallenge (n � 5 in each group). One hour later, liver MNCswere isolated. After incubation for 10 minutes with anFc-blocker (2.4 G2; Pharmingen) to prevent nonspecificbinding, the MNCs were labeled with PE-conjugated anti-CD11b mAb and PC5-conjugated anti-Gr-1 mAb. Kupffercells stain positively with CD11b but are negative for Gr-1staining.15 CD11b�Gr-1� Kupffer cells were analyzed for invivo phagocytosis by flow cytometry (EPICS XL, Coulter,Miami, FL).

Statistical AnalysisThe data are presented as means � SE. Statistical analysiswas performed using Graphpad-Prism software, version4.00 (Graphpad Software, Inc., San Diego, CA). Survivalfrequencies were compared using log rank tests. Changesin serum cytokine concentrations were compared usingtwo-way analysis of variance (ANOVA). Other cytokinelevels between groups or among several groups werecompared using the Student t test or one-way ANOVAfollowed by the Newman-Keuls test. The survival ofantibiotic-treated mice was evaluated using Fisher exacttest. The level of significance was set at P � 0.05. Samplesizes were chosen based on data in other papers in thisfield.3,5,28–30

RESULTSKetamine Anesthesia Improves Survival AfterLaparotomy with LPS Challenge and SuppressesElevation of Serum Proinflammatory CytokinesBoth ketamine-treated groups (anesthetized either with0.2%– 0.5% sevoflurane or 2%–3% sevoflurane) showed


significantly higher survival rates and greater suppres-sion of serum peaks of TNF, interleukin (IL)-12, andIFN-� compared with the sevoflurane (approximately2%–3%) group (Fig. 2). However, the two ketamine-treated groups showed similar survival rates and serumcytokine levels, suggesting that ketamine increasesmouse survival and suppresses proinflammatory cyto-kine secretion. The differences observed between themice treated with ketamine/low-dose sevoflurane (ket-amine group) versus high-dose sevoflurane alone(sevoflurane group) might not have been due to the useof a lower dose of sevoflurane but due to the use ofketamine. Therefore, the affect of ketamine anesthesia onthe ketamine (with low-dose sevoflurane) and sevoflu-rane groups was compared in further experiments.

Ketamine Anesthesia Does Not IncreaseSurvival After Laparotomy with E. coliChallenge Despite Suppression of SerumTNF and IFN-�Mice in the ketamine group did not survive at higher ratesafter laparotomy with E. coli (1 � 108 CFU/mouse) chal-lenge compared with those in the sevoflurane group (Fig.3), despite significant suppression of TNF and IFN-� (Table1). Ketamine anesthesia also failed to increase survival afterlaparotomy with lethal E. coli challenge (1 � 109

CFU/mouse) (Fig. 3), despite suppression of TNF andIFN-� (Table 1).

Ketamine Anesthesia Plus Antibiotic TherapyIncreases Survival After Laparotomy withE. coli ChallengeUnlike LPS challenge, bacterial growth and proliferationcould affect the survival of E. coli–challenged mice afterketamine treatment. Because perioperative septic patientsare usually treated with antibiotics, E. coli–challenged micewere given two injections of cefazolin. In addition, TNFwas depleted in mice receiving sevoflurane anesthesia andcefazolin injections to determine if TNF levels decreased byketamine might explain the increased survival of cefazolin-treated mice in postoperative infection. Concomitant use ofcefazolin with ketamine anesthesia significantly increasedsurvival after E. coli challenge with suppression of TNF andIFN-� compared with sevoflurane anesthesia alone (Fig. 4,Table 2). The TNF-depleted mice also showed a markedincrease in survival after postlaparotomy E. coli challengeafter sevoflurane anesthesia (Fig. 4) with significant sup-pression of IFN-� (Table 2).

Suppression of TNF Decreases Survival inE. coli–Challenged Mice Under InsufficientRegulation of Bacterial GrowthTNF-neutralizing antibodies (Ab) was administered tosingle cefazolin-treated mice before laparotomy (but notafter E. coli challenge) followed by sevoflurane anesthesia.

Figure 2. The effect of ketamine anesthesia onsurvival (A), serum tumor necrosis factor (TNF) (B),interleukin (IL)-12 (C), and interferon (IFN)-� (D)levels in mice after lipopolysaccharide (LPS) chal-lenge after laparotomy. Mice were anesthetizedwith ketamine �0.2% to 0.5% sevoflurane (ket-amine group), ketamine �2% to 3% sevoflurane(sevoflurane � ketamine group) or 2% to 3%sevoflurane (sevoflurane group). Data are themeans � SE, n � 25 in each group. *P � 0.01,†P � 0.05 vs other groups.

Figure 3. The effects of ketamine anesthesia on mouse survivalafter E. coli challenge after laparotomy. Mice received ketamineanesthesia (�0.2%–0.5% sevoflurane) or sevoflurane (2%–3%) an-esthesia. Thereafter, they were challenged IP with sublethal (1 �108 colony-forming units (CFU)/mouse) or lethal (1 � 109

CFU/mouse) doses of E. coli after laparotomy; n � 25 in each group.

Table 1. Serum TNF and IFN-� Levels in MiceAfter E. coli Challenge Following Laparotomy

Sevoflurane KetamineSerum TNF at 1 h

1 � 108 CFU 119 � 14 66 � 2*1 � 109 CFU 404 � 53 146 � 19*

Serum IFN-� at 6 h1 � 108 CFU 396 � 63 145 � 32*1 � 109 CFU 1103 � 181 530 � 51*

TNF � tumor necrosis factor; IFN � interferon.Sera were obtained from the mice 1 hour and 6 hours after E. coli challengeto measure serum TNF and IFN-�, respectively. Data are means � SE (pg/mL)from 25 mice in each group. * P � 0.01 vs sevoflurane.



Most survived the initial 36-hour period but eventuallydied 7 days after infection (Table 3). Bacteria were thencounted in the livers of TNF-depleted mice treated withcefazolin (one or two injections) or without cefazolin 24hours after E. coli challenge. A single injection of cefazolindid not effectively suppress bacterial growth (counts) in theliver compared with double injections (single; 1.1 � 0.4versus double; 0.1 � 0.1 � 106 CFU, P � 0.05, n � 6 in eachgroup), although growth was significantly inhibited inantibiotic-treated mice (without antibiotic injection; 5.1 �0.3 � 106 CFU, P � 0.01, n � 6). Because ketaminesuppressed TNF secretion, survival after E. coli challengewas examined in the mice receiving ketamine anesthesiathat were treated with a single injection of cefazolin.During the initial 36 hours, these mice exhibited a rate of

survival comparable to the mice with two cefazolin injec-tions but showed significantly lower survival at 7 days,although both antibiotic treatments increased mouse sur-vival after bacterial challenge (Table 4). These findingssuggest that unless bacterial growth/proliferation is wellcontrolled (by repeated antibiotic injections), suppressionof TNF (by neutralizing Ab or ketamine) may adverselyaffect surgical patients with bacterial infections.

Ketamine Suppresses LPS-Induced TNFProduction by Cultured Lymphocytes In VitroCoculturing with ketamine significantly suppressed TNFproduction from blood leukocytes, liver, or spleen MNCsstimulated by LPS in vitro, suggesting that ketamine di-rectly suppresses LPS-induced TNF production by thesecells (Table 5).

Ketamine Attenuates In Vivo Phagocytic Activityof Kupffer Cells in LPS-Challenged MiceKetamine treatment significantly decreased phagocytosisby Kupffer cells in mice challenged with PBS (Table 6).

Figure 4. The effect of ketamine anesthesia and neutralizing tumornecrosis factor (TNF) � antibiotic therapy on mouse survival aftersublethal E. coli challenge after laparotomy. Mice received ketamineanesthesia, sevoflurane anesthesia, or sevoflurane anesthesia afterTNF neutralization. They were then treated twice with cefazolin andchallenged IP with 2 � 108 E. coli colony-forming units (CFU)/mouseafter laparotomy; n � 20 in each group, *P � 0.01.

Table 2. Serum TNF and IFN-� Levels in MiceAfter Antibiotic-Treated E. coli ChallengeFollowing Laparotomy

SevofluraneSevoflurane �Anti-TNF Ab Ketamine

Serum TNF 172.0 � 23.2* Undetected 58.6 � 2.2*Serum IFN-� 649.3 � 86.7* 124.5 � 13.1 146.7 � 18.1

TNF � tumor necrosis factor; IFN � interferon; Ab � antibodies.Cefazolin (double injections)-treated mice were injected IV with anti-TNFneutralizing Ab or rat IgG, as a control, 1 hour before E. coli challenge. Serawere obtained from the mice at 1 hour and 6 hour to measure serum TNF andIFN-�, respectively. Data are means � SE (pg/mL) from 20 mice in eachgroup. * P � 0.01 vs other groups.

Table 3. The Effect of Neutralizing Anti-TNFAntibody on Survival After E. coli Challenge ofMice Treated by a Single Injection of Antibiotic

Survival

Sevoflurane Sevoflurane � Anti-TNF-Ab36 h 15/20 17/2072 h (3 d) 13/20 4/20*168 h (7 d) 12/20 3/20*

TNF � tumor necrosis factor; Ab � antibodies.Mice were injected IV with anti-TNF neutralizing antibody or rat IgG 1 hourbefore E. coli challenge following laparotomy. Cefazolin was injected onceimmediately before laparotomy. * P � 0.05 vs sevoflurane.

Table 4. The Effect of Antibiotic Treatment onSurvival After E. coli Challenge FollowingLaparotomy in Ketamine-Anesthetized Mice

Survival

Without injection Single injection Double injections36 h 10/25* 20/25 21/2572 h (3d) 5/25* 17/25 20/25168 h (7d) 4/25* 12/25§ 19/25

Ketamine-anesthetized mice were treated with a single injection of cefazolin,double injections, or without injection. * P � 0.01 vs other groups, § P � 0.05vs double injections.

Table 5. The Effect of In Vitro KetamineTreatment on LPS-Stimulated TNF Production

Ketamine

Without ketamine With ketamine (10�6 M)Circulating leukocytes 240 � 20 111 � 20*Liver MNCs 548 � 31 278 � 20*Spleen MNCs 200 � 12 82 � 5*

LPS � lipopolysaccharide; MNCs � mononuclear cells.Data are means � SE (pg/mL) from three individual experiments with twosamples per each group. * P � 0.05 vs without ketamine.

Table 6. The Effect of Ketamine Treatmenton Phagocytosis by Kupffer Cells inLPS-Challenged MiceWithout laparotomy

Treatment PBS Ketamine PBS KetamineChallenge PBS PBS LPS LPSPhagocytosis

(%)13.6 � 1 7.5 � 0.8* 20.7 � 1.1* 14.02 � 1.1

LaparotomyAnesthesia Sevoflurane Ketamine Sevoflurane KetamineChallenge PBS PBS LPS LPSPhagocytosis

(%)14.9 � 0.5 9.4 � 0.8* 21.4 � 0.7* 14.6 � 0.6

PBS � phosphate buffer sodium; LPS � lipopolysaccharide.Proportion of microsphere phagocytosis by Kupffer cells are shown asmeans � SE from 5 mice in each group. * P � 0.05 vs other groups.


Although LPS challenge increased phagocytosis by Kupffercells, ketamine treatment also significantly decreased theirphagocytosis (Table 6). Although laparotomy alone did notsignificantly affect the phagocytic activity by Kupffer cells,ketamine anesthesia significantly decreased the phagocyto-sis after laparotomy (Table 6).

The Effect of Nadolol Pretreatment onKetamine-Suppressed Proinflammatory CytokineResponse or Phagocytosis in LPS-Challenged MiceAlthough the anesthetic dose of ketamine (100 mg/kg)markedly suppressed serum TNF levels 1 hour after LPSchallenge in mice (without laparotomy), nadolol pretreat-ment did not abolish the ketamine-induced suppression ofTNF (Fig. 5). A low dose of ketamine (10 mg/kg, 10% of theanesthetic dose) also suppressed serum TNF levels afterLPS challenge; however, nadolol pretreatment significantlyrestored TNF levels in mice (Fig. 5). Further examination ofLPS-induced TNF secretion after laparotomy revealed thatlaparotomy obviously suppressed LPS-induced TNF eleva-tion in sevoflurane-anesthetized mice (Table 7, Fig. 5).Many investigators have also demonstrated that surgical

stress such as surgery or traumatic injury induces hypore-sponsiveness of TNF secretion to LPS, in particular, theearly phase after surgery,31–34 but TNF suppression byketamine anesthesia (100 mg/kg) did not significantlydiffer in mice with (Table 7) or without laparotomy (Fig. 5).

Phagocytosis by Kupffer cells was examined in LPS-challenged mice (without laparotomy) that received ananesthetic dose of ketamine. Unlike the TNF response,nadolol almost completely restored the ketamine-suppressed phagocytic activity of Kupffer cells (Table 8).This restoration by nadolol was also observed in laparoto-mized mice (Table 8).

DISCUSSIONKetamine anesthesia improved survival in LPS-challengedmice but not in mice challenged with viable E. coli. Never-theless, the survival rate after E. coli challenge wasimproved by the concomitant use of an antibiotic andketamine anesthesia, suggesting that inhibition of bacterialgrowth critically influences the ketamine-induced benefi-cial effect on survival.

Since the septic condition induced by LPS is merely aresult of SIRS without infection, suppression of proinflam-matory cytokines by ketamine improves mouse survival. Incontrast, septicemia after E. coli challenge is a result ofinfection accompanied by SIRS. TNF and IFN-� are impor-tant for the elimination of invading bacteria, although theyalso induce shock and organ injury.13,14,16 The suppressionof proinflammatory cytokines by ketamine may therebyattenuate bacterial elimination. Ketamine also decreasedthe phagocytic activity of Kupffer cells, indicating furtherimpairment of bacterial clearance. These findings mayexplain why ketamine (without antibiotic) could not im-prove the survival of E. coli–infected mice. A subanestheticdose of ketamine reportedly suppressed LPS-stimulatedTNF production by peripheral blood MNCs in patients,11

suggesting that ketamine, even at a low dose, suppressesLPS-induced TNF production in humans and mice.

IL-12 and �-GalCer, a synthetic NKT cell ligand, were usedto determine whether ketamine directly affects IFN-� produc-tion from NK and/or NKT cells. Exogenous IL-12 directlyinduces IFN-� production by mouse NKT cells.35–37 �-GalCeralso directly stimulates IFN-� production from NKT cells andsubsequently from NK cells.2,28,38–41 Ketamine anesthesia didnot suppress serum IFN-� levels after IL-12 or �-GalCerchallenge in mice (data not shown), suggesting that the IFN-�

Figure 5. The effect of nadolol on serum tumor necrosis factor (TNF)levels after lipopolysaccharide (LPS) challenge in mice given theanesthetic or low dose of ketamine. After pretreatment with nadololor phosphate buffer sodium (PBS), mice were treated with ketamine(100 mg kg�1 � anesthetic induction dose; 10 mg kg�1 � 10%anesthetic induction dose) or PBS followed by LPS challenge. Serawere obtained from the mice 1 hour after LPS challenge. Data arethe means � SE from five mice in each group. §P � 0.05, *P � 0.01vs other groups.

Table 7. The Effect of Nadolol on the Serum TNFLevels After LPS Challenge Following LaparotomyPretreatment PBS PBS Nadolol Nadolol

Anesthesia Sevoflurane Ketamine Sevoflurane KetamineSerum TNF

levels at 1 h989 � 74 248 � 47* 1020 � 109 239 � 34*

TNF � tumor necrosis factor; LPS � lipopolysaccharide;Data are means � SE (pg/mL) from 5 mice in each group. * P � 0.01 vssevoflurane.

Table 8. The Effect of Nadolol on Phagocytosisby Kupffer Cells in With or Without Ketamine-Treated MiceWithout laparotomy

Pretreatment PBS PBS Nadolol NadololAnesthesia PBS Ketamine PBS KetaminePhagocytosis (%) 14.5 � 0.7 8.4 � 0.5* 15.8 � 1.3 15.8 � 1.5

LaparotomyPretreatment PBS PBS Nadolol NadololAnesthesia Sevoflurane Ketamine Sevoflurane KetaminePhagocytosis (%) 15.8 � 0.9 9.6 � 0.8* 16.7 � 1.0 15.0 � 0.9

PBS � phosphate buffer sodium.Proportion of microsphere phagocytosis by Kupffer cells are shown asmeans � SE from 5 mice in each group. * P � 0.05 vs other groups.



producing capacities of NK/NKT cells are not directly af-fected by ketamine anesthesia but are suppressed primarilythrough reduced IL-12 production from macrophages/Kupffer cells.

Sevoflurane anesthesia for 15 to 30 minutes reportedlysuppresses LPS-induced TNF secretion in mice.42,43 In thisstudy, sevoflurane was only administered for 5 minutes.The TNF elevation observed in the ketamine plus high-dosesevoflurane group was similar to that seen after LPSchallenge to the ketamine plus low-dose sevoflurane group,suggesting that ketamine-induced suppression of TNF isindependent of sevoflurane dose.

Nadolol did not inhibit suppression of TNF secretion byan anesthetic dose of ketamine (100 mg/kg), although itsignificantly restored TNF secretion under low-dose ket-amine (10 mg/kg). Low-dose ketamine also reportedlyincreased serum adenosine levels (0.4 �M) and therebyinhibited secretion of proinflammatory cytokines after LPSchallenge,5 as evidenced by use of an adenosine receptorantagonist. However, this study did not examine how theanesthetic dose of ketamine affects serum adenosine levelsor proinflammatory cytokine levels. Propranolol (another�-adrenoceptor blocker) inhibited the secretion of adeno-sine from the kidneys after electric stimulation of theperiarterial sympathetic nerve.44 Therefore, adenosine and,presumably, adrenaline appear to be responsible for thereduction of TNF production by activated-macrophagesafter a low dose of ketamine. However, the direct effect(non-�-adrenergic pathway) of an anesthetic dose of ket-amine may overcome or overwhelm the inhibitory effectvia the �-adrenergic pathway by nadolol (as seen withlow-dose ketamine) and further suppress TNF production.The addition of ketamine to cultures consistently suppressedLPS-induced TNF production from isolated MNCs in vitro,suggesting a direct suppression of TNF by ketamine.

The results of this study also demonstrated for the firsttime that ketamine anesthesia significantly inhibited in vivophagocytic activity by Kupffer cells. Previous in vitro studieshad shown that ketamine did not inhibit phagocytosis byeither polymorphonuclear leukocytes or macrophages at clini-cal concentrations, although it could suppress at higher con-centrations.9,10,45 Interestingly, inhibiting the �-adrenergicpathway by nadolol restored ketamine-suppressed phagocy-tosis by Kupffer cells. The anesthetic dose of ketamine mightdirectly suppress TNF secretion in LPS-challenged mice inde-pendent of the �-adrenergic pathway, while also indirec-tly suppressing phagocytosis by Kupffer cells via the�-adrenergic pathway. Inhibiting the �-adrenergic pathwayby nadolol may be effective in ketamine anesthesia for septicpatients because it might abrogate the enhanced catechol-amine effect. However, when nadolol was administered to E.coli–challenged mice under ketamine anesthesia without us-ing an antibiotic, survival was not improved (unpublishedobservations). Nadolol may induce heart failure and severehypotension in mice with septicemia under ketamine anes-thesia and thereby mask restoration of the phagocytic func-tion of macrophages.

Ketamine anesthesia may be preferred for septic patientsbecause of its suppressive effects of proinflammatory cyto-kines. However, ketamine also might decrease bacterial

phagocytosis by macrophages/Kupffer cells, which pre-sumably enhance bacterial growth/proliferation in periop-erative septic patients. Therefore, antibiotic treatment couldbe important for septic patients who received ketamineanesthesia. It should also be noted that ketamine reportedlyinjures Kupffer cells and endothelial cells and may affectliver functions.46

REFERENCES1. Exadaktylos AK, Buggy DJ, Moriarty DC, Mascha E, Sessler DI.

Can anesthetic technique for primary breast cancer surgery affectrecurrence or metastasis? Anesthesiology 2006;105:660–4

2. Wada H, Seki S, Takahashi T, Kawarabayashi N, Higuchi H,Habu Y, Sugahara S, Kazama T. Combined spinal and generalanesthesia attenuates liver metastasis by preserving TH1/TH2cytokine balance. Anesthesiology 2007;106:499–506

3. Taniguchi T, Takemoto Y, Kanakura H, Kidani Y, YamamotoK. The dose-related effects of ketamine on mortality andcytokine responses to endotoxin-induced shock in rats. AnesthAnalg 2003;97:1769–72

4. Shaked G, Czeiger D, Dukhno O, Levy I, Artru AA, Shapira Y,Douvdevani A. Ketamine improves survival and suppressesIL-6 and TNFalpha production in a model of Gram-negativebacterial sepsis in rats. Resuscitation 2004;62:237–42

5. Mazar J, Rogachev B, Shaked G, Ziv NY, Czeiger D, Chaimo-vitz C, Zlotnik M, Mukmenev I, Byk G, Douvdevani A.Involvement of adenosine in the antinflammatory action ofketamine. Anesthesiology 2005;102:1174–81

6. Imai T, Takahashi K, Masuo F, Goto F. Anaesthesia affectsoutcome of sepsis in mice. Can J Anaesth 1998;45:360–6

7. Hansbrough JF, Zapata-Sirvent RL, Bartle EJ, Anderson JK, ElliottL, Mansour MA, Carter, WH. Alterations in splenic lymphocytesubpopulations and increased mortality from sepsis followinganesthesia in mice. Anesthesiology 1985;63:267–73

8. Krumholz W, Endrass J, Hempelmann G. Inhibition of phago-cytosis and killing of bacteria by anaesthetic agents in vitro.Br J Anaesth 1995;75:66–70

9. Heller A, Heller S, Blecken S, Urbaschek R, Koch T. Effects ofintravenous anesthetics on bacterial elimination in humanblood in vitro. Acta Anaesthesiol Scand 1998;42:518–26

10. Nishina K, Akamatsu H, Mikawa K, Shiga M, Maekawa N,Obara H, Niwa Y. The inhibitory effects of thiopental, mida-zolam, and ketamine on human neutrophil functions. AnesthAnalg 1998;86:159–65

11. Beilin B, Rusabrov Y, Shapira Y, Roytblat L, Greemberg L,Yardeni IZ, Bessler H. Low-dose ketamine affects immuneresponses in humans during the early postoperative period.Br J Anaesth 2007;99:522–7

12. Am College of Chest Physicians/Society of Critical Care MedConsensus Conference: definitions for sepsis and organ failureand guidelines for the use of innovative therapies in sepsis.Crit Care Med 1992;20:864–74

13. Ogasawara K, Takeda K, Hashimoto W, Satoh M, Okuyama R,Yanai N, Obinata M, Kumagai K, Takada H, Hiraide H, Seki S.Involvement of NK1� T cells and their IFN-gamma produc-tion in the generalized Shwartzman reaction. J Immunol1998;160:3522–7

14. Sato K, Kinoshita M, Motegi A, Habu Y, Takayama E,Nonoyama S, Hiraide H, Seki S. Critical role of the liver CD8�CD122� T cells in the generalized Shwartzman reaction ofmice. Eur J Immunol 2005;35:593–602

15. Kinoshita M, Uchida T, Nakashima H, Ono S, Seki S, HiraideH. Opposite effects of enhanced tumor necrosis factor-alphaproduction from Kupffer cells by gadolinium chloride on liverinjury/mortality in endotoxemia of normal and partially hepa-tectomized mice. Shock 2005;23:65–72

16. Kinoshita M, Seki S, Ono S, Shinomiya N, Hiraide H. Paradoxicaleffect of IL-18 therapy on the severe and mild Escherichia coliinfections in burn-injured mice. Ann Surg 2004;240:313–20

17. Watanabe H, Numata K, Ito T, Takagi K, Matsukawa A. Innateimmune response in Th1- and Th2-dominant mouse strains.Shock 2004;22:460–6


18. Zsigmond EK, Kelsch RC, Kothary SP. Rise in plasma free-norepinephrine during anesthetic induction with ketamine.Behav Neuropsychiatry 1974;6:81–4

19. Clanachan AS, McGrath JC. Effects of ketamine on the periph-eral autonomic nervous system of the rat. Br J Pharmacol1976;58:247–52

20. Severn A, Rapson NT, Hunter CA, Liew FY. Regulation oftumor necrosis factor production by adrenaline and beta-adrenergic agonists. J Immunol 1992;148:3441–5

21. Spengler RN, Chensue SW, Giacherio DA, Blenk N, Kunkel SL.Endogenous norepinephrine regulates tumor necrosis factor-alpha production from macrophages in vitro. J Immunol1994;152:3024–31

22. Verhoeckx KC, Doornbos RP, van der Greef J, Witkamp RF,Rodenburg RJ. Inhibitory effects of the beta-adrenergic recep-tor agonist zilpaterol on the LPS-induced production of TNF-alpha in vitro and in vivo. J Vet Pharmacol Ther 2005;28:531–7

23. Schedlowski M, Hosch W, Oberbeck R, Benschop RJ, Jacobs R,Raab HR, Schmidt RE. Catecholamines modulate human NKcell circulation and function via spleen-independent beta2-adrenergic mechanisms. J Immunol 1996;156:93–9

24. Shakhar G, Ben-Eliyahu S. In vivo beta-adrenergic stimulationsuppresses natural killer activity and compromises resistanceto tumor metastasis in rats. J Immunol 1998;160:3251–8

25. Kappel M, Poulsen TD, Galbo H, Pedersen BK. Effects ofelevated plasma noradrenaline concentration on the immunesystem in humans. Eur J Appl Physiol Occup Physiol1998;79:93–8

26. Dobashi H, Seki S, Habu Y, Ohkawa T, Takeshita S, Hiraide H,Sekine I. Activation of mouse liver natural killer cells andNK1.1(�) T cells by bacterial superantigen-primed Kupffercells. Hepatology 1999;30:430–6

27. Habu Y, Seki S, Takayama E, Ohkawa T, Koike Y, Ami K, MajimaT, Hiraide H. The mechanism of a defective IFN-gamma responseto bacterial toxins in an atopic dermatitis model, NC/Nga mice,and the therapeutic effect of IFN-gamma, IL-12, or IL-18 ondermatitis. J Immunol 2001;166:5439–47

28. Inui T, Nakagawa R, Ohkura S, Habu Y, Koike Y, Motoki K,Kuranaga N, Fukasawa M, Shinomiya N, Seki S. Age-associated augmentation of the synthetic ligand- mediatedfunction of mouse NK1.1 ag(�) T cells: their cytokine produc-tion and hepatotoxicity in vivo and in vitro. J Immunol2002;169:6127–32

29. Gurfinkel R, Czeiger D, Douvdevani A, Shapira Y, Artru AA,Sufaro Y, Mazar J, Shaked G. Ketamine improves survival inburn injury followed by sepsis in rats. Anesth Analg2006;103:396–402

30. Shibakawa YS, Sasaki Y, Goshima Y, Echigo N, Kamiya Y,Kurahashi K, Yamada Y, Andoh T. Effects of ketamine andpropofol on inflammatory responses of primary glial cellcultures stimulated with lipopolysaccharide. Br J Anaesth2005;95:803–10

31. Majetschak M, Flach R, Kreuzfelder E, Jennissen V, HeukampT, Neudeck F, Schmit-Neuerburg KP, Obertacke U, Schade FU.The extent of traumatic damage determines a graded depres-sion of the endotoxin responsiveness of peripheral bloodmononuclear cells from patients with blunt injuries. Crit CareMed 1999;27:313–8

32. Ogata M, Okamoto K, Kohriyama K, Kawasaki T, Itoh H,Shigematsu A. Role of interleukin-10 on hyporesponsivenessof endotoxin during surgery. Crit Care Med 2000;28:3166–70

33. Kawasaki T, Ogata M, Kawasaki C, Tomihisa, T, Okamoto K,Shigematsu A. Surgical stress induces endotoxin hyporespon-siveness and an early decrease of monocyte mCD14 andHLA-DR expression during surgery. Anesth Analg 2001;92:1322–6

34. Lahat N, Rahat MA, Brod V, Cohen S, Weber G, Kinarty A,Bitterman H. Abdominal surgery reduces the ability of ratspleen cells to synthesize and secrete active tumour necrosisfactor-alpha (TNF-alpha) by a multilevel regulation. Clin ExpImmunol 1999;115:19–25

35. Seki S, Hashimoto W, Ogasawara K, Satoh M, Watanabe H,Habu Y, Hiraide H, Takeda K. Antimetastatic effect of NK1� Tcells on experimental haematogenous tumour metastases inthe liver and lungs of mice. Immunology 1997;92:561–6

36. Hashimoto W, Takeda K, Anzai R, Ogasawara K, Sakihara H,Sugiura K, Seki S, Kumagai K. Cytotoxic NK1.1 Ag� alphabeta T cells with intermediate TCR induced in the liver of miceby IL-12. J Immunol 1995;154:4333–40

37. Habu Y, Uchida T, Inui T, Nakashima H, Fukasawa M, Seki S.Enhancement of the synthetic ligand-mediated function ofliver NK1.1Ag� T cells in mice by interleukin-12 pretreatment.Immunology 2004;113:35–43

38. Nakagawa R, Nagafune I, Tazunoki Y, Ehara H, Tomura H,Iijima R, Motoki K, Kamishohara M, Seki S. Mechanisms of theantimetastatic effect in the liver and of the hepatocyte injuryinduced by alpha-galactosylceramide in mice. J Immunol2001;166:6578–84

39. Carnaud C, Lee D, Donnars O, Park SH, Beavis A, Koezuka Y,Bendelac A. Cutting edge: Cross-talk between cells of theinnate immune system: NKT cells rapidly activate NK cells.J Immunol 1999;163:4647–50

40. Nakagawa R, Inui T, Nagafune I, Tazunoki Y, Motoki K,Yamauchi A, Hirashima M, Habu Y, Nakashima H, Seki S.Essential role of bystander cytotoxic CD122�CD8� T cells forthe antitumor immunity induced in the liver of mice byalpha-galactosylceramide. J Immunol 2004;172:6550–7

41. Kawabata T, Kinoshita M, Inatsu A, Habu Y, Nakashima H,Shinomiya N, Seki S. Functional alterations of liver innateimmunity of mice with aging in response to CpG-oligodeoxynucleotide. Hepatology 2008;48:1586–97

42. Kidani Y, Taniguchi T, Kanakura H, Takemoto Y, Tsuda K,Yamamoto K. Sevoflurane pretreatment inhibits endotoxin-induced shock in rats. Anesth Analg 2005;101:1152–6

43. Hofstetter C, Boost KA, Flondor M, Basagan-Mogol E, Betz C,Homann M, Muhl H, Pfeilschifter J, Zwissler B. Anti-inflammatory effects of sevoflurane and mild hypothermia inendotoxemic rats. Acta Anaesthesiol Scand 2007;51:893–9

44. Mi Z, Jackson EK. Effects of alpha- and beta-adrenoceptorblockade on purine secretion induced by sympathetic nervestimulation in the rat kidney. J Pharmacol Exp Ther1999;288:295–301

45. Davidson JA, Boom SJ, Pearsall FJ, Zhang P, Ramsay G.Comparison of the effects of four i.v. anaesthetic agents onpolymorphonuclear leucocyte function. Br J Anaesth1995;74:315–8

46. Thompson JS, Brown SA, Khurdayan V, Zeynalzadedan A,Sullivan PG, Scheff SW. Early effects of tribromoethanol,ketamine/xylazine, pentobarbitol, and isoflurane anesthesiaon hepatic and lymphoid tissue in ICR mice. Comp Med2002;52:63–7



Regional Anesthesia

Section Editor: Terese T. Horlocker

Estimation and Pharmacodynamic Consequences ofthe Minimum Effective Anesthetic Volumes forMedian and Ulnar Nerve Blocks: A Randomized,Double-Blind, Controlled Comparison BetweenUltrasound and Nerve Stimulation GuidanceMatthieu Ponrouch, MD,* Nicolas Bouic, MD,* Sophie Bringuier, PharmD, PhD,†Philippe Biboulet, MD,* Olivier Choquet, MD,* Michele Kassim, MD,* Nathalie Bernard, MD, MSc,*and Xavier Capdevila, MD, PhD‡

BACKGROUND: Nerve stimulation and ultrasound guidance are the most popular techniques forperipheral nerve blocks. However, the minimum effective anesthetic volume (MEAV) in selectednerves for both techniques and the consequences of decreasing the local anesthetic volume on thepharmacodynamic characteristics of nerve block remain unstudied. We designed a randomized,double-blind controlled comparison between neurostimulation and ultrasound guidance to estimatethe MEAV of 1.5% mepivacaine and pharmacodynamics in median and ulnar nerve blocks.METHODS: Patients scheduled for carpal tunnel release were randomized to ultrasound guidance(UG) or neurostimulation (NS) groups. A step-up/step-down study model (Dixon method) was used todetermine the MEAV with nonprobability sequential dosing based on the outcome of the previouspatient. The starting dose of 1.5% mepivacaine was 13 and 11 mL for median and ulnar nerves atthe humeral canal. Block success/failure resulted in a decrease/increase of 2 mL. A blindedphysician assessed sensory blockade at 2-minute intervals for 20 minutes. Block onset time andduration were noted.RESULTS: The MEAV50 (SD) of the median nerve was lower in the UG group 2 (0.1) mL (95%confidence interval [CI] � [1, 96] to [2, 04]) than in the NS group 4 (3.8) mL (95% CI � [2, 4] to [5,6]) (P � 0.017). There was no difference for the ulnar nerve between UG group 2 (0.1) mL (95% CI �[1, 96] to [2, 04]) and NS group 2.4 (0.6) mL (95% CI � [2, 1] to [2, 7]). The duration of sensoryblockade was significantly correlated to local anesthetic volume, but onset time was not modified.CONCLUSION: Ultrasound guidance selectively provided a 50% reduction in the MEAV of mepiva-caine 1.5% for median nerve sensory blockade in comparison with neurostimulation. Decreasing thelocal anesthetic volume can decrease sensory block duration but not onset time. (Anesth Analg2010;111:1059–64)

Nerve stimulation is an indirect technique of nerveidentification but is still one of the most populartechniques for peripheral nerve blocks. The suc-

cess rate is 91% to 98%, depending on the trials. Ultrasoundguidance may be of benefit for peripheral nerve blocks.1–3

Ultrasound guidance permits a dynamic vision of nerves,

vessels, muscles, and needle movements and allows thevolume distribution of local anesthetic to be controlled.4

The local anesthetic volume injected near a nerve is a factordetermining the rate of successful nerve block.5,6 Thevolume and concentration affect the absorption of local anes-thetic.7–9 The possibility of decreasing local anesthetic vol-umes for peripheral nerve blocks is a relevant question.Several studies have reported that multiple neurostimulationsfor locating nerves permitted reduction of the local anestheticvolume.10–16 The study of Casati et al.6 and the systematicreview of Koscielniak-Nielsen2 found that decreasing the localanesthetic volume seems possible using ultrasound guidance.O’Donnell and Iohom17 recently reported in a descriptivestudy that 1 mL of 2% lidocaine on each nerve componentduring an ultrasound-guided axillary block was sufficient topromote complete anesthetic block. However, we have onlysparse data supporting the consequences of such low volumeson the pharmacodynamic parameters of sensory blockade.18

Furthermore, some authors have claimed that it is man-datory to develop studies comparing neurostimulationwith ultrasound.2,6,19 –21

From the *Department of Anesthesiology and Critical Care, MontpellierUniversity Hospital, Montpellier, France; †Department of Anesthesiologyand Critical Care Medicine, Lapeyronie University Hospital, and Epidemi-ology and Clinical Research Department, Arnaud de Villeneuve UniversityHospital Montpellier, France; and ‡Department of Anesthesiology andCritical Care, Montpellier I University and Montpellier University Hospital;Institut National de la Sante et de la Recherche Medicale, Montpellier,France.


The study has been presented in part at the American Society of Anesthe-siologists meeting, New Orleans, Louisiana, October 17–21, 2009.

Address correspondence to Prof. Xavier Capdevila, Department of Anesthe-siology, Lapeyronie University Hospital, Route de Ganges, 34295 Montpel-lier Cedex 5, France. Address e-mail to [email protected].



Therefore, we conducted a prospective, randomized,double-blind controlled study to determine the minimumeffective anesthetic volume (MEAV) necessary to achievemedian and ulnar nerve blocks, by using neurostimulationor ultrasound guidance. We tested the hypothesis thatultrasound guidance reduces MEAV by 20%. The second-ary end point provides information on the influence of thelocal anesthetic volume on the pharmacodynamic charac-teristics of the nerve block.

METHODSAfter obtaining ethics committee approval (Comite deprotection des personnes Sud Mediterannee 3) and writteninformed consent, ASA physical status I–III patients ages 18to 90 years scheduled to undergo ambulatory endoscopic oropen pit carpal tunnel release surgery were recruited forthis randomized controlled study. Patients who did notcooperate and those who had psychological disorders orlinguistic difficulties that might interfere with sensoryblockade were excluded. Medical exclusion criteria werecoagulopathies, known allergy to the trial drugs, infectionat the puncture site, a body mass index �40 or �19 kg/m2,diabetes mellitus or known neuropathies, patients whoreceived opiates for chronic pain, and cardiac conductionproblems (third-degree atrioventricular block).

Patients were included in the ultrasound guidance (UG)group or neurostimulation (NS) group using a random listat the preanesthetic consultation. On the day of surgery,patients were premedicated with 1 mg/kg hydroxyzine,and 500 mL of saline at the rate of 4 to 6 mL/kg/h wasinfused with an IV 20-G catheter on the contralateralforearm. Patients were monitored using a noninvasivearterial blood pressure measurement, a continuous electro-cardiogram, and pulse oximetry. A high-concentration oxy-gen mask at 6 l/minute was put on the patient’s face. Thepatients were sedated with 0.05 �g/kg IV sufentanil. Theoperating arm was positioned at 80° abduction and externalrotation. The blocks of the median and ulnar nerves weredone at the junction between the upper and middle thirdsof the arm (i.e., humeral canal) with a 15 mg/mL mepiva-caine solution. All blocks were always placed by 1 of thesame 2 investigators (Michele Kassim or MatthieuPonrouch), who had substantial expertise in regional anes-thesia techniques. Patients were blinded to the technique atthe beginning of the procedure by the passage of theultrasound probe in the NS group and a single stimulationat 1 mA of the median nerve in the UG group. In the 2groups the patient could not see the screen of the ultra-sound device. The anesthesiologist in charge of the blockprocedure turned off the ultrasound machine in the NSgroup. The ultrasound probe was placed at the beginningof the nerve stimulation procedure. A conventional asepticprocedure was used for peripheral nerve blocks. The anes-thesiologist wore a mask, cap, and gloves. The puncturesite was prepared with an alcohol povidone–iodine solu-tion, and surrounding areas were disinfected. The probewas covered with a film-type sterile Tegaderm®.

In the NS group, nerve location was made with a 50-mm22-G needle (Uniplex nanoLine Facet�, Pajunk©, Germany)and a nerve stimulator (MultiStim Sensor�, Pajunk©, Ger-many) initially set at pulse duration 0.1 ms, intensity 1.5

mA, and stimulation frequency 2 Hz. Nerve blocks wereachieved as has been previously demonstrated5 at thehumeral canal by a palmar flexion of the first 3 fingers anda carpal pronation for the median nerve, and a flexion ofthe fifth and fourth fingers, and a thumb adduction andflexion of carpi ulnaris for the ulnar nerve. The procedurestarted at the median nerve. The puncture site was locatedat the humeral canal immediately above the brachial artery.The needle was inserted tangentially to the skin untilminimum stimulus intensity between 0.4 and 0.6 mA wasachieved. The starting dose of 1.5% mepivacaine was 13and 11 mL for median and ulnar nerves. The predefinedlocal anesthetic volume was injected after an aspiration testrepeated between each bolus of 2 mL, until the finalvolume. The needle was withdrawn without leaving theskin, and the intensity of stimulation was increased again to1.5 mA. The needle was redirected medially and posteriorlyand the same procedure was performed to locate and blockthe ulnar nerve separately.

In the UG group, localization of both nerves was per-formed with a 50-mm 22-G needle (Uniplex nanoLineFacet�, Pajunk©, Germany) and an ultrasound machine(Logic E�, GE Healthcare©) with a linear probe set at afrequency of 12 MHz. After analysis of different anatomicalelements, the probe was positioned perpendicularly to theskin to obtain a cross-section of the humeral canal. Thenerves were visualized in their short axis. The needle wasinserted at the lateral end of the probe to keep it in theplane of the sonogram. The needle bevel and shaft wereviewed throughout the approach to the selected nerve. Thepredefined local anesthetic volume was injected after anaspiration test repeated between each bolus of 2 mL, untilthe final volume. The injection was slow and at lowpressure. The absence of intraneural injection was con-trolled by ultrasound. After a unique needle puncture,needle repositioning was allowed to optimize the distribu-tion of local anesthetic around each nerve. A circumferen-tial spread was required without exceeding the definedvolume. For both groups the volume of 1 mL was consid-ered the lowest possible volume.

At the end of the last nerve injection, the sensoryblockade was tested every 2 minutes for 20 minutes by anobserver blinded to the technique and the volume injected.The sensory block was assessed by the patient’s ability todistinguish hot and cold and to discriminate a light touch inthe center of the skin area innervated by each nerve: thethenar eminence and the anterior surface of the distal endof the index finger for the median nerve, and the hypothe-nar eminence and the anterior surface of the distal end ofthe fifth finger for the ulnar nerve. A comparison with thecontralateral area was used to evaluate sensory blockade. Avalue of 0 was noted if the sensation was the same on bothsides; 1 in case of decreased sensation in the anesthetizedhand and 2 in case of no sensation (complete sensoryblock). When the sum of the variables was equal to 4 within20 minutes, the block was defined as complete. A sum �4was considered a failure. Adverse events (i.e., paresthesia,pain during injection, intravascular injection, and cardio-vascular and neurologic events) were noted during theprocedure and until the end of the sensory block. The

Minimum Effective Anesthetic Volumes for Median and Ulnar Nerve Blocks


sensory blockade was assessed on arrival in the postanes-thetic care unit, then every 15 minutes until the end of theblock (score returned to 0).

The definition of sensory onset time was the time fromperformance of the block to a value of 4 in sensory blockadein the selected areas, and the definition of duration ofsensory blockade was the time from performance of theblock to a value of 0 for the sensory block. Motor blockadewas not noted because it was not necessary for surgery, andwe wanted to avoid interfering with the blinded evaluationof the sensory blockade. If the block was ineffective, thesurgeon performed a wrist infiltration with 8 mL of 15mg/mL mepivacaine. In case of failure, the patient wasanesthetized with sufentanil 0.2 �g/kg and propofol in atarget-controlled infusion, and a laryngeal mask airwaywas inserted. Postoperative analgesia was provided by 1 gparacetamol every 6 hours for 48 hours. Patients weredischarged home in the evening of the day of surgery.

The primary hypothesis was to determine the MEAVnecessary to achieve median and ulnar nerve blocks usingneurostimulation or ultrasound guidance. We tested thehypothesis that ultrasound guidance reduces MEAV by20%. We compared the MEAV measured by the Dixonmethod for both selected nerves in each group. Secondaryend points of the study examined the pharmacodynamiccharacteristics of all successful nerve blocks (onset time andduration of blocks) related to the local anesthetic volumeinjected. A flowchart summarizing the design of both partsof the study is presented in Figure 1.

StatisticsThe primary end point chosen for the 2 techniques was a20% decrease in the MEAV in the UG group. The descrip-tive study of Carles et al. reported that the mean volumerequired to block the ulnar nerve successfully using a nerve

stimulator was 11 mL and 13 mL for the ulnar and mediannerves.5 We used these values as the initial volumes. Adifference of 2 mL in this volume in the UG group wasconsidered clinically significant. The Dixon method wasused.22 The MEAV50 corresponds to the MEAV that suc-cessfully anesthetized 50% of the patients. Accepting an �risk of 5% and a power (1–�) of 80%, 12 subjects in eachgroup were necessary for a significant difference. It hasbeen demonstrated that beyond 11 patients, the signifi-cance of the results increases with the number of patientsincluded.22 Considering these factors, the number of patientsenrolled was arbitrarily set at 42. Patients were divided into 2groups, 21 in the NS group and 21 in the UG group. Thedistribution was made according to a computer-randomizedlist in clusters of 6 patients. Statistical analysis was per-formed using SAS� version 9 software (SAS Institute, Cary,North Carolina). Comparison between qualitative variableswas performed by �2 test or Fisher’s exact test, if necessary.Comparisons of quantitative variables were performedusing the Student’s t test or Mann–Whitney Wilcoxon test,if necessary. The relationship between quantitative vari-ables was evaluated using Spearman’s correlation coeffi-cients. Correlation of the volume of 1.5% mepivacaine inrelation to the sensory onset time and duration of sensoryblockade was done by linear regression. P � 0.05 wasconsidered significant.

RESULTSForty-two patients were included in the study, 21 in the NSgroup and 21 in the UG group. There were no significantdifferences in the patients’ characteristics in both groups(Table 1). The MEAV50 (SD) of the median nerve was lowerin the UG group 2 (0.1) mL (95% confidence interval [CI] �[1, 96] to [2, 04]) than in NS group 4 (3.8) mL (95% CI �[2, 4] to [5, 6]) (P � 0.017). There was no difference for theulnar nerve between UG group 2 (0.1) mL (95% CI � [1, 96]to [2, 04]) and NS group 2.4 (0.6) mL (95% CI � [2, 1] to[2, 7]) (Fig. 2). With both nerve localization techniques,there was a positive significant correlation between dura-tion of sensory blockade and the volume of mepivacaine1.5% injected for both ulnar and median nerves. For bothnerves, the duration of sensory block decreased when lesslocal anesthetic was injected. In some patients a volume of1 mL allowed a sensory blockade for �60 minutes. Con-versely, onset times of sensory blockade of the median andulnar nerves were not significantly correlated to the local

Figure 1. Flowchart summarizing the design of both parts of thestudy. A, Comparison of the MEAV measured by the Dixon method forboth selected nerves in each group. B, Pharmacodynamics ofsuccessful nerve blocks (onset time and duration of blocks) relatedto local anesthetic volume. MEAV � minimum effective anestheticvolume for 50% of the patients; BMI � body mass index; UG �ultrasound guidance; NS � nerve stimulation.

Table 1. Patient Characteristics of theNeurostimulation (NS) and Ultrasound Guidance(UG) Groups (n � 42)

NS group(n � 21)

UG group(n � 21)

Age (years) 56 (17) 55 (17)Weight (kg) 70 (22.1) 67.5 (20.2)Height (cm) 164 (36) 167 (36)BMI (kg/m2) 26 (6.9) 25.1 (6.9)Sex (male/female) 8/13 6/15ASA physical status I/II/III 11/10/0 9/11/1

Data are presented as mean � SD. P value �0.05. ASA � American Societyof Anesthesiology; BMI � body mass index. There were no significantdifferences between groups.


anesthetic volume injected (Fig. 3). For 1 patient in the UGgroup, an unplanned subcutaneous infiltration of localanesthetic in the musculocutaneous nerve skin area wasnecessary for surgical incision. Fifteen blocks showed anegative response 20 minutes after block placement. Afterrecording the presence of a negative response to theup-and-down sequence, these patients received a wristinfiltration with 8 mL of 15 mg/mL mepivacaine. No

general anesthesia was required. No adverse events werenoted in either group.

DISCUSSIONWe report that ultrasound guidance can reach an MEAV50

value lower than neurostimulation for the median nervebut not for the ulnar nerve. In addition, the reduction inlocal anesthetic volume caused a decrease in the duration of

Figure 2. Minimum effective anesthetic volume(MEAV)50 measured by the Dixon up-and-downmethod for the median (A) and ulnar (B) nerves. A,P value � 0.017. B, P value � 0.441. The volumeX0 was chosen at 13 mL for the median nerve and11 mL for the ulnar nerve. The interval betweenvolumes in 2 different patients was set at 2 mL.The next volume X1 was determined by the per-formance of the previous volume, X0. If X0 waseffective, X1 � X0 – 2 mL; if X0 was ineffective,X1 � X0 � 2 mL. UG � ultrasound guidance;NS � nerve stimulation.

Figure 3. A, B, Significant correlationsbetween local anesthetic volume and du-ration of sensory blockade for medianand ulnar nerves (*P � 0.03). C, D, Nocorrelation between local anesthetic vol-ume and onset time for median and ulnarnerves (P � 0.71).



sensory blockade but did not modify the onset time of theblock for both nerves. Ultrasound guidance can reduce thelocal anesthetic volume for a selected nerve block incomparison with neurostimulation. Our results confirmdata reported in those previous studies.1–3,6,21,23 It is inter-esting to note that if we apply the same methodology forthe calculation of MEAV for both guidance methods,neurostimulation allows the same volume reduction asdoes ultrasound guidance for some isolated nerves. For theulnar nerve, the results show that the MEAV50 is compa-rable for both techniques of nerve location. For the femoralnerve, Casati et al.6 reported that the MEAV50 was signifi-cantly lower in an UG group (15 mL) than in a NS group (26mL). The same authors reported in another study that theMEAV50 using neurostimulation was significantly de-creased in a multiple-neurostimulations and multiple-injections group (14 mL) than in a single-neurostimulationand single-injection group (23 mL) for a femoral nerveblock.24 In other words, a precise identification of nervesreduces MEAV50. Ultrasound guidance permits direct vi-sualization of the spread of local anesthetic around and incontact with the nerve.20

The location of the median nerve and its anatomicalrelationship into the humeral canal may explain the differ-ence in the MEAV50 in the UG group in comparison withthe NS group. The median nerve is close to the brachialartery.1,25,26 The circumferential distribution of a smalllocal anesthetic volume around the median nerve is avail-able with ultrasound guidance through needle redirectionsand visualization of the local anesthetic spread. The inti-mate relationship of the median nerve and the brachialartery, by using neurostimulation, suggests that the circum-ferential spread around the nerve seems unpredictable anddifficult to obtain with a low volume of local anesthetic.Conversely, the ulnar nerve is located farther from theartery, allowing that needle movements are not alwaysnecessary to obtain a circumferential distribution of localanesthetic around the ulnar nerve.25,26 This may explainwhy there was no difference in the MEAV50 for the ulnarnerve in both guidance groups. Our randomized compara-tive study reports that, depending on the nerves andsurrounding anatomical structure, ultrasound guidanceallowed a smaller volume than did neurostimulation. TheMEAV obtained was lower by 50% in the UG group. Theseresults confirm the results in the Danelli et al. study.27

These authors compared neurostimulation and ultrasoundguidance for sciatic nerve blocks using a subgluteal ap-proach. They reported a 37% decrease of MEAV in the UGgroup.

The decrease in volumes of local anesthetic leads tochanges in some of the nerve block pharmacodynamiccharacteristics. When the volume decreases, the onset timeof the sensory blockade does not vary. Conversely, theduration of nerve sensory blockade is closely correlated tothe local anesthetic volume. Serradell et al., using 20 to 36mL of local anesthetic, did not report a difference in theduration of the axillary block, corresponding to the rightside of the correlation curve in Figure 3.13 In a recent studyof 20 volunteers scheduled for sciatic nerve block withultrasound guidance, Latzke et al.18 reported that a localanesthetic volume of 0.10 mL/mm cross-sectional nerve

area had no effect on sensory onset time, whereas theduration of sensory block was shorter. Marhofer et al.23

reported that ultrasound can reduce the local anestheticvolume and onset time of the femoral sensory blockade.However, the authors did not report significant correlationsbetween the local anesthetic volume and onset time of theblock. Onset time and quality of the blocks also depend onthe nerve and approach.28,29

The spread of local anesthetic in contact with the nerveis crucial for obtaining the MEAV. It is conceptualized asthe minimum circumferential area of nerve exposed toMEAV associated with the surface of the nerve. Eichen-berger et al.21 studied the concept of MEAV (in millilitersper millimeter) on the basis of the diameter of the nerve.They reported a MEAV of 0.11 mL/mm for mepivacaine1% for ulnar nerve block at the proximal forearm. This isinteresting but does not predict MEAV values in other partsof the same nerve. The variability of the structure of a singlenerve from the brachial plexus to its distal part explains thedifferent expected MEAV values. Moayeri et al.30 reportedfrom a cadaveric study that the nonneural tissue composi-tion of the nerves within the epineurium increased fromtheir proximal to distal parts. The neuronal tissue/nonneural tissue ratio increases from 1:1 to 1:2. Thesestudies demonstrated that for the same dose of localanesthetic (volume and concentration), the onset time de-pends on the quality of nerve localization using ultrasoundguidance or neurostimulation, the nerve studied, and thelocal anesthetic spread close to the nerve.25,26

Some limitations of our study deserve comment. Theclinical relevance of a MEAV50 value might be questioned,because in the method used, about 50% of the studiedpatients had an incomplete nerve block. Nonetheless, theED50 concept has been used to determine volume–effectrelationships in the field of peripheral nerve blocks.6,22,24

On the other hand, we reported a correlation betweenvolume of local anesthetic solution and duration of theblocks. However, the study was not designed for a com-parison in block duration for fixed high or low volumes oflocal anesthetics. Further comparative randomized studiesare necessary to determine whether the volume of localanesthetic solution really influences the duration of nerveblocks.

This double-blind, randomized, clinical trial showsthat the use of ultrasound guidance was effective inreducing the local anesthetic volume for the mediannerve. Neurostimulation can provide similar results forthe ulnar nerve. The ability to visualize anatomical struc-tures, to control the position of the needletip, and to controlthe circumferential spread of local anesthetic may partlyexplain these results. The decrease in volume helps tohighlight a strong correlation between local anestheticvolume and duration of the nerve block; the onset timedoes not vary with the local anesthetic volume. The precisecontrol of these variables and knowledge of variations ofpharmacodynamic characteristics based on the local anes-thetic volume injected, concentration, and approach of theblock should assist practitioners to adapt the procedure todifferent clinical situations.


REFERENCES1. Abrahams MS, Aziz MF, Fu RF, Horn JL. Ultrasound guidance

compared with electrical neurostimulation for peripheralnerve block: a systematic review and meta-analysis of random-ized controlled trials. Br J Anaesth 2009;102:408–17

2. Koscielniak-Nielsen ZJ. Ultrasound-guided peripheral nerveblocks: what are the benefits? Acta Anaesthesiol Scand 2008;52:727–37

3. Marhofer P, Chan VW. Ultrasound-guided regional anesthesia:current concepts and future trends. Anesth Analg 2007;104:1265–9

4. Retzl G, Kapral S, Greher M, Mauritz W. Ultrasonographicfindings of the axillary part of the brachial plexus. AnesthAnalg 2001;92:1271–5

5. Carles M, Pulcini A, Macchi P, Duflos P, Raucoules-Aime M,Grimaud D. An evaluation of the brachial plexus block at thehumeral canal using a neurostimulator (1417 patients): theefficacy, safety, and predictive criteria of failure. Anesth Analg2001;92:194–8

6. Casati A, Baciarello M, Di Cianni S, Danelli G, De Marco G,Leone S, Rossi M, Fanelli G. Effects of ultrasound guidance onthe minimum effective anaesthetic volume required to blockthe femoral nerve. Br J Anaesth 2007;98:823–7

7. Brown DL, Ransom DM, Hall JA, Leicht CH, Schroeder DR,Offord KP. Regional anesthesia and local anesthetic-inducedsystemic toxicity: seizure frequency and accompanying cardio-vascular changes. Anesth Analg 1995;81:321–8

8. Becker DE, Reed KL. Essentials of local anesthetic pharmacol-ogy. Anesth Prog 2006;53:98–108

9. Mather LE, Copeland SE, Ladd LA. Acute toxicity of localanesthetics: underlying pharmacokinetic and pharmacody-namic concepts. Reg Anesth Pain Med 2005;30:553–66

10. Casati A, Danelli G, Baciarello M, Corradi M, Leone S, DiCianni S, Fanelli G. A prospective, randomized comparisonbetween ultrasound and nerve stimulation guidance for mul-tiple injection axillary brachial plexus block. Anesthesiology2007;106:992–6

11. Sia S, Bartoli M, Lepri A, Marchini O, Ponsecchi P. Multiple-injection axillary brachial plexus block: a comparison of twomethods of nerve localization–nerve stimulation versus pares-thesia. Anesth Analg 2000;91:647–51

12. Sia S, Lepri A, Ponzecchi P. Axillary brachial plexus blockusing peripheral nerve stimulator: a comparison betweendouble- and triple-injection techniques. Reg Anesth Pain Med2001;26:499–503

13. Serradell A, Herrero R, Villanueva JA, Santos JA, Moncho JM,Masdeu J. Comparison of three different volumes of mepiva-caine in axillary plexus block using multiple nerve stimulation.Br J Anaesth 2003;91:519–24

14. Inberg P, Annila I, Annila P. Double-injection method usingperipheral nerve stimulator is superior to single injection inaxillary plexus block. Reg Anesth Pain Med 1999;24:509–13

15. Deleuze A, Gentili ME, Marret E, Lamonerie L, Bonnet F. Acomparison of a single-stimulation lateral infraclavicularplexus block with a triple-stimulation axillary block. RegAnesth Pain Med 2003;28:89–94

16. Benhamou D. Axillary plexus block using multiple nervestimulation: a European view. Reg Anesth Pain Med2001;26:495–8

17. O’Donnell BD, Iohom G. An estimation of the minimumeffective anesthetic volume of 2% lidocaine in ultrasound-guided axillary brachial plexus block. Anesthesiology 2009;111:25–9

18. Latzke D, Marhofer P, Zeitlinger M, Machata A, Neumann F,Lackner E, Kettner SC. Minimal local anaesthetic volumes forsciatic nerve block: evaluation of ED 99 in volunteers. Br JAnaesth 2009;104:239–44

19. Macaire P, Singelyn F, Narchi P, Paqueron X. Ultrasound- ornerve stimulation-guided wrist blocks for carpal tunnel re-lease: a randomized prospective comparative study. RegAnesth Pain Med 2008;33:363–8

20. Hadzic A, Dewaele S, Gandhi K, Santos A. Volume and dose oflocal anesthetic necessary to block the axillary brachial plexususing ultrasound guidance. Anesthesiology 2009;111:8–9

21. Eichenberger U, Stockli S, Marhofer P, Huber G, Willimann P,Kettner SC, Pleiner J, Curatolo M, Kapral S. Minimal localanesthetic volume for peripheral nerve block: a newultrasound-guided, nerve dimension-based method. RegAnesth Pain Med 2009;34:242–6

22. Dixon WJ. Staircase bioassay: the up-and-down method. Neu-rosci Biobehav Rev 1991;15:47–50

23. Marhofer P, Schrogendorfer K, Koinig H, Kapral S, WeinstablC, Mayer N. Ultrasonographic guidance improves sensoryblock and onset time of three-in-one blocks. Anesth Analg1997;85:854–7

24. Casati A, Fanelli G, Beccaria P, Magistris L, Albertin A, Torri G.The effects of single or multiple injections on the volume of0.5% ropivacaine required for femoral nerve blockade. AnesthAnalg 2001;93:183–6

25. Sites BD, Neal JM, Chan V. Ultrasound in regional anesthesia:where should the “focus” be set? Reg Anesth Pain Med2009;34:531–3

26. van Geffen GJ, Moayeri N, Bruhn J, Scheffer GJ, Chan VW,Groen GJ. Correlation between ultrasound imaging, cross-sectional anatomy, and histology of the brachial plexus: areview. Reg Anesth Pain Med 2009;34:490–7

27. Danelli G, Ghisi D, Fanelli A, Ortu A, Moschini E, Berti M,Ziegler S, Fanelli G. The effects of ultrasound guidance andneurostimulation on the minimum effective anesthetic volumeof mepivacaine 1.5% required to block the sciatic nerve usingthe subgluteal approach. Anesth Analg 2009;109:1674–8

28. Taboada Muniz M, Rodriguez J, Bermudez M, Valino C,Blanco N, Amor M, Aguirre P, Masid A, Cortes J, Alvarez J,Atanassoff PG. Low volume and high concentration of localanesthetic is more efficacious than high volume and lowconcentration in Labat’s sciatic nerve block: a prospective,randomized comparison. Anesth Analg 2008;107:2085–8

29. Casati A, Fanelli G, Borghi B, Torri G. Ropivacaine or 2%mepivacaine for lower limb peripheral nerve blocks. Studygroup on orthopedic anesthesia of the Italian Society of Anes-thesia, Analgesia, and Intensive Care. Anesthesiology 1999;90:1047–52

30. Moayeri N, Bigeleisen PE, Groen GJ. Quantitative architectureof the brachial plexus and surrounding compartments, andtheir possible significance for plexus blocks. Anesthesiology2008;108:299–304



BRIEF REPORT

Pulse-Oximetric Measurement of Prilocaine-InducedMethemoglobinemia in Regional AnesthesiaPeter Soeding, MD,* Matthias Deppe,* and Hartmut Gehring, MD, PhD*

BACKGROUND: The Masimo Radical 7® is a new pulse CO oximeter designed to measuremethemoglobin. The device has not been evaluated in a clinical setting.METHODS: In this prospective observational study we compared the arterial methemoglobinlevels and the corresponding pulse CO-oximetric values of the Radical 7® in regional anesthesiawith prilocaine.RESULTS: We analyzed 360 data pairs with methemoglobin values up to 6.6%. The mean biasand limits (�1.96 SD) of the device were 0.27% (�1.33%).CONCLUSION: We found a high degree of agreement in measurement of methemoglobinbetween the 2 methods. (Anesth Analg 2010;111:1065–8)

Prilocaine, in comparison with all other local anesthet-ics, has the lowest direct systemic toxicity, but may leadto an increased formation of methemoglobin

(MetHb).1,2 Whereas in healthy individuals higher concentra-tions of MetHb are usually well tolerated, it may endangeroxygen supply in patients with diminished cardiopulmonaryreserves or anemia.3–6 Because of their 2-wavelength technol-ogy, conventional pulse oximeters are not able to identify theextent of dyshemoglobinemias.7–9 The optical analysis of anew pulse CO oximeter (Masimo, Radical 7®)) is based onabsorbance measurements at several wavelengths. Conse-quently, dyshemoglobinemias might also be recorded by apulse oximeter with sufficient precision. This has been dem-onstrated in 1 preclinical trial10 and 2 case reports.11,12

This prospective study is the first to evaluate the efficacyof this pulse CO oximeter in a clinical setting. We hypoth-esized that the results for MetHb are the same for the pulseoximetric method and arterial blood sample measurementusing a CO oximeter.

METHODSAfter approval by the ethics committee and writteninformed consent, we investigated 40 patients, physicalstatus ASA I–III, having orthopedic surgery: 20 patientsreceived an interscalene plexus block with 30 mL prilo-caine 1% (i.e., 300 mg) and 20 patients a combinedfemoral–sciatic nerve blockade with 2 � 30 mL prilocaine1% (i.e., 600 mg in all). All blocks were performed usinga nerve stimulator (Stimuplex HNS 11, Braun, Germany).Injection was only performed if a contraction of indicatormuscles could be demonstrated at 0.3 to 0.5 mA (stimu-lus duration: 0.1 ms).

The pulse CO oximeter Radical 7® is a device manufac-tured by Masimo Corp. (Irvine, California), which in addi-tion to oxygen saturation (Spo2 in %) can also measure

MetHb by pulse oximetry (SpMet%).10 These values couldbe read using a laptop and stored after conversion into anExcel spreadsheet. The continuous monitoring of patientswith the Radical 7® was started before the onset of regionalanesthesia.

At the times 0, 15, 30, 60, 120, 180, 240, 300, and 360minutes after the first injection of prilocaine, 2 mL ofarterial blood was taken and immediately analyzed in ablood gas analyzer (ABL-625, Radiometer America, Copen-hagen, Denmark) with an integrated CO oximeter as areference device. The data for MetHb content (cMetHb) andfor Sao2 were compared with the corresponding valuesobtained with the Radical 7®.

For statistical analysis using the Software Package forSocial Sciences (SPSS) 15.0 for Windows, we averaged the 9repeated measurements for each of the 40 patients accord-ing to Bland and Altman.13

A probability of P � 0.05 was considered statisticallysignificant.

RESULTSOne patient after interscalene block and 1 patient aftercombined femoral–sciatic nerve block required general

From the *Department of Anesthesiology, University Clinic of Schleswig—Holstein, Campus Luebeck, Luebeck, Germany.


Parts of the work were presented at the 2008 American Society of Anesthe-siologists annual meeting in Baltimore, Maryland, and at the 2008 EuropeanSociety of Anesthesiologists annual meeting in Copenhagen, Denmark.

Address correspondence to Peter Soeding, MD, Department of Anesthesiology,University Clinic of Schleswig—Holstein, Campus Luebeck, Ratzeburger Allee160, D-23,538 Luebeck, Germany. Address e-mail to [email protected].


Figure 1. Methemoglobin (MetHb) levels after regional anesthesiawith prilocaine: plots show mean MetHb levels after 300 mgprilocaine for interscalene block (ISB) and after 600 mg prilocaine forcombined femoral–sciatic nerve block (FSNB).


anesthesia for block failure (success rate 95%). Figure 1shows MetHb levels over time for both types of blocks.Peak levels were reached for interscalene blocks after 120minutes and for combined femoral–sciatic nerve blocksafter 5 hours (Table 1).

The statistical agreement of the MetHb measurementbetween the laboratory method as a reference methodand the pulse oximetric measurement using the Radical

7® for all 40 patients is shown in Table 2 and Figure 2.According to Bland–Altman analysis (Fig. 2), the biaswas 0.27%, and the 95% confidence limits (�1.96 sd)1.33%. With the increasing rise in MetHb, there is a cleargap between the values reported for functional oxygensaturation by the Radical 7® and the values reported bythe CO oximeter in the blood gas analyzer device (Fig. 3).The Radical 7® displays Spo2 readings that run in parallel

Table 1. Methemoglobin Levels and Success Rate After Regional Anesthesia with Prilocaine forInterscalene or Femoral–Sciatic Nerve Blocks

Nerve blockade No. of patients Prilocaine Success rate

Methemoglobin

Peak mean � SD

(range) Time to peakInterscalene 20 300 mg (30 mL) 95% (19/20) 2.3 � 0.8% (1.1–4.9) 120 minutesFemoral–sciatic 20 600 mg (2 � 30 mL) 95% (19/20) 4.1 � 1.5% (2.0–6.6) 300 minutes

Table 2. Interclass Statistical Comparison of SpMet (%) Versus cMetHb (%) for Each Subject and forPooled Data

Regression analysis Bland–Altman

n Slope y intercept SEE Bias Limits (�1.96 SD) rPooled data 360 1.19 �0.14 0.61 0.27 1.33 0.95Subject

1 9 1.09 �0.35 0.21 �0.12 0.47 0.992 9 1.12 �0.32 0.21 �0.01 0.48 0.983 9 0.84 0.36 0.41 0.13 0.79 0.844 9 1.30 �0.86 0.27 �0.42 0.59 0.945 9 1.05 �0.004 0.54 0.21 1.05 0.986 9 0.79 0.61 0.68 0.37 1.28 0.47 9 1.34 �0.68 0.15 �0.08 0.64 0.998 9 1.21 �0.04 0.18 0.2 0.39 0.969 9 1.24 �0.17 0.10 0.04 0.23 0.9610 9 1.20 0.53 0.82 1.16 1.76 0.9611 9 1.27 �1.02 0.96 0.04 2.18 0.9512 9 1.26 �0.64 0.52 �0.08 1.09 0.9313 9 1.17 0.24 1.04 0.74 2.05 0.9214 9 1.19 �0.40 0.24 0.13 0.68 0.9915 9 1.11 �0.21 0.07 �0.01 0.23 0.9916 9 1.16 �0.31 0.53 0.2 1.15 0.9817 9 1.18 �0.34 0.13 �0.12 0.3 0.9818 9 1.21 �0.44 0.47 0.11 1.07 0.9719 9 0.61 1.27 0.72 0.61 1.43 0.5120 9 1.38 �0.39 0.75 0.89 2.09 0.9721 9 1.15 0.78 0.94 1.34 1.93 0.9622 9 1.43 �0.08 0.34 0.77 0.9 0.9623 9 1.09 �0.34 0.16 �0.18 0.36 0.9924 9 0.88 0.25 0.38 0.06 0.73 0.8225 9 1.26 �0.49 0.37 0.1 0.92 0.9726 9 1.18 �0.12 0.19 0.36 0.66 0.9927 9 1.81 �0.73 0.37 0.4 1.15 0.9528 9 0.89 0.46 0.94 0.31 1.76 0.4429 9 0.59 0.23 0.19 �0.12 0.4 0.5730 9 0.84 0.70 0.44 0.4 0.87 0.8531 9 0.65 0.87 0.94 0.38 1.79 0.2832 9 1.19 �0.37 0.20 �0.04 0.46 0.9733 9 1.12 0.18 0.31 0.38 0.61 0.9334 9 1.20 0.03 0.41 0.52 0.98 0.9835 9 1.21 �0.45 0.41 0.16 1.03 0.9836 9 1.03 �0.05 0.44 0 0.82 0.9037 9 1.22 �0.28 0.24 0.14 0.57 0.9838 9 1.46 0.13 0.77 1.26 1.77 0.9239 9 0.62 0.74 0.42 0.02 0.95 0.7540 9 1.26 �0.40 0.38 0.49 1.37 0.99

SpMet � Radical 7® measurement of methemoglobin; cMetHb% � CO oximeter measurement of methemoglobin; n � number of data points for each subject;SEE � SE of the estimate; r � correlation coefficient.

BRIEF REPORT


with and slightly higher than the values for fractionalsaturation (Fig. 4).

DISCUSSIONIn this prospective study we evaluated a new pulse Cooximeter (Radical 7®, Masimo, Inc.) after regional anesthe-sia with high doses of prilocaine. We compared the pulseoximetric method for noninvasive monitoring of MetHbwith direct arterial measurements by using a referenceprocedure. A high degree of agreement between the 2methods could be shown for MetHb values up to 6.6%(mean correlation coefficient � 0.95, bias � 0.27, 95%limits � �1.33).

In comparison with lidocaine and mepivacaine, prilo-caine has the advantage of far less cardiac or centralnervous system toxicity.14,15 Nevertheless, prilocaine is notavailable in much of the world. The presence of acquiredmethemoglobinemia is often assumed.3,16,17 It is triggerednot only by prilocaine but also by many other drugs,3,18–21

in particular by benzocaine.17 The symptoms are nonspe-cific and often unrecognized.3,9 Methemoglobinemia isassociated with the reduction in the fractional oxygensaturation. Concentrations �15% are usually well toleratedby healthy individuals, but in patients with anemia orcardiopulmonary diseases, clinical symptoms can occurwhen the concentration exceeds 8%.3 Despite administra-tion of the recommended threshold dose of prilocaine, itwas only possible to assess MetHb values up to 6.6%.Further investigations are necessary to analyze the accu-racy of the pulse oximeter at higher levels.

Dyshemoglobinemias cannot be identified by conven-tional pulse oximeters, because their measurements, basedon 2 wavelengths of light absorption, only allow therecording of oxyhemoglobin and desoxyhemoglobin.7,22

The pulse CO oximeter Radical 7® measures the lightabsorption of 8 different wavelengths. In a preclinical studyin healthy volunteers, Barker et al. evaluated a predecessorof the pulse oximeter that we used.10 The results (bias 0%,sd � 0.45%) differ only slightly from the data presented inour clinical study.

Although the Radical 7®, according to the manufactu-rer’s information, only displays the functional oxygensaturation,23 the data collected for the Spo2 display duringthe study period tend to follow the fractional rather thanthe functional oxygen saturation (Figs. 3 and 4). Thisdifference might be a result of the analysis algorithm of thedevice.

In conclusion, we found a high degree of agreement inmeasurement of MetHb with a CO oximeter and a nonin-vasive and readily available pulse-oximetric procedure.This may facilitate early diagnosis and treatment, whennecessary, of dyshemoglobinemia.

Figure 2. Bland and Altman analysis for repeated measurements:bias plot of the difference of pulse-oximeter estimate of methe-moglobin (MetHb) (SpMet [%]) and cMetHb% versus the averageof SpMet and cMetHb. Averaged data of the 9 repeated measure-ments. Lines show values of bias (mean of the differences)and �1.96 SD. cMetHb% � CO-oximeter measurement ofmethemoglobin.

Figure 3. Regression analysis: scatter plot of the functional satura-tion measured by Radical 7® (SPO2) and the functional saturationmeasured by the CO oximeter (SAO2) versus cMetHb%. Averaged dataof the 9 repeated measurements of each of the 40 patients.cMetHb% � CO oximeter measurement of methemoglobin. SEE � SE

of the estimate

Figure 4. Regression analysis: scatter plot of the functional satura-tion measured by Radical 7® (SPO2) and the fractional saturationmeasured by the CO oximeter (O2Hb) versus cMetHb%. Averageddata of the 9 repeated measurements of each of the 40 patients.cMetHb% � CO oximeter measurement of methemoglobin. SEE � SE

of the estimate

Pulse-Oximetric Measurement of Methemoglobin


DISCLOSUREFinancial support for the work: Peter Soeding and HartmutGehring are employees of the University Clinic of Schleswig—Holstein, Campus Luebeck, Luebeck, Germany. Device wasprovided by Masimo, Inc., Germany.

REFERENCES1. Scott DB, Owen JA, Richmond J. Methaemoglobinaemia due to

prilocaine. Lancet 1964;3:728–92. Wright RO, Lewander WJ, Woolf AD. Methemoglobinemia:

etiology, pharmacology, and clinical management. Ann EmergMed 1999;34: 646–56

3. Ash-Bernal R, Wise R, Wright SM. Acquiredmethemoglobinemia—A retrospective series of 138 cases at 2teaching hospitals. Med 2004;83:265–73

4. Bellamy MC, Hopkins PM, Halsall PJ, Ellis FR. A study intoincidence of methemoglobinaemia after “three-in-one” blockwith prilocaine. Anaesthesia 1992;47:1084–5

5. Knobeloch L, Goldring J, LeMay W, Anderson H. Three casesof methemoglobinemia associated with dental anesthesia. WisDent Assoc J 1994;70:34–5

6. Kreeftenberg HG, Braams R, Nauta P. Methemoglobinemiaafter low-dose prilocaine in an adult patient receiving barbitu-rate comedication. Anesth Analg 2007;104:459–60

7. Reynolds KJ, Palayiwa E, Moyle JT, Sykes MK, Hahn CE. Theeffect of dyshemoglobins on pulse oximetry: part I. theoreticalapproach and part II. experimental results using an in vitro testsystem. J Clin Monit 1993;9:81–90

8. Barker SJ, Tremper KK, Hyatt J. Effects of methemoglobinemiaon pulse oximetry and mixed venous oximetry. Anesthesiol-ogy 1989;70:112–7

9. Yang JJ, Lin N, Lv R, Sun J, Zhao F, Zhang J, Xu JG.Methemoglobinemia misdiagnosed as ruptured ectopic preg-nancy. Acta Anesth Scand 2005;49:586–8

10. Barker SJ, Curry J, Redford D, Morgan S. Measurement ofcarboxyhemoglobin and methemoglobin by pulse oximetry.Anesthesiology 2006;105:892–7

11. Annabi EH, Barker SJ. Severe methemoglobinemia detected bypulse oximetry. Anesth Analg 2009;108:898–9

12. Macknet M, Kimball-Jones P, Applegate R. Benzocaine in-duced methemoglobinemia after TEE. Resp Care 2007;52:2007Open Forum Abstracts [e-abstracts]

13. Bland JM, Altman DG. Agreement between methods of mea-surement with multiple observations per individual. J Biop-harm Stat 2007;17:571–82

14. Zink W, Graf BM. Toxikologie der Lokalanasthetika. Anaes-thesist 2003;52:1102–23

15. Scott DB, Jerson. PJR, Braid DP, Ortengren B, Frisch P. Factorsaffecting plasma levels of lignocaine and prilocaine. Br JAnaesth 1972;44:1040–9

16. Weinberg GL. Banning benzocaine: of bananas, bureaucrats,and blue men. Anesth Analg 2009;108:699–701

17. Guay J. Methemoglobinemia related to local anesthetics: asummary of 242 episodes. Anesth Analg 2009;108:837–45

18. Anderson ST, Hajduczek J, Barker SJ. Benzocaine-induced met-hemoglobinemia in an adult. Anesth Analg 1988;67:1099–101

19. Rehman HU. Methemoglobinemia. West J Med 2001;175:193–620. Fung HT, Lai CH, Wong OF, Lam KK, Kam CW. Two cases of

methemoglobinemia following zoplicone ingestion. Clin Toxi-col 2008;46:167–70

21. White CD, Weiss LD. Varying presentations of methemoglo-binemia: two cases. J Emerg Med Suppl 1991;1:45–9

22. Rieder HU, Frei FJ, Zbinden AM, Thomson DA. Pulse oximetryin methaemoglobinaemia. Failure to detect low oxygen satu-ration. Anaesthesia 1989;44:326–7

23. Radical-7 color display signal extraction pulse co-oximeterwith Rainbow Technology. Operator‘s manual. Irvine, CA:Masimo Corporation, 2007

BRIEF REPORT


CASE REPORT

Symptomatic Axillary Hematoma After Ultrasound-Guided Infraclavicular Block in a Patient withUndiagnosed Upper Extremity Mycotic AneurysmsDave Gleeton, MD, Simon Levesque, MD, FRCPC, Claude A. Trepanier, MD, FRCPC,Jean-Luc Gariepy, MD, FRCPC, Jean Brassard, MD, FRCPC, and Nicolas Dion, MD, FRCPC

We present a case of axillary hematoma complicating an ultrasound-guided infraclavicularblock in a patient with undiagnosed mycotic aneurysms of the peripheral arteries. Mycoticaneurysm is a rare medical condition with well-identified risk factors. When performingregional anesthesia in patients with these risk factors, clinicians should have a high degree ofsuspicion about the possible existence of vascular anomalies. A preprocedure Doppler study of theblock area and real-time guidance of the needle using ultrasound may be useful. (Anesth Analg2010;111:1069–71)

Ultrasound guidance can detect variations in normalanatomy, thus allowing the anesthesiologist totailor the technique to the patient-specific condi-

tion.1–3 Despite these advantages, complications still occurduring ultrasound-guided peripheral nerve blocks.4 Wepresent a case in which the operator using ultrasound failedto detect mycotic aneurysms of the upper extremity arter-ies, leading to a large hematoma of the axillary region afteran ultrasound-guided infraclavicular block.

CASE DESCRIPTIONA 44-year-old women, ASA physical status III, was admittedin our hospital with a diagnosis of infected olecranon bursa ofthe right elbow. Her recent medical history included a mitralvalve replacement for mitral endocarditis (Streptococcus viri-dans) 4 months previously. Her medical history revealed IVdrug abuse, recurrent cellulitis of the right upper limb, highblood pressure, Type 2 diabetes, gastroesophageal reflux, andmild asthma. Her preoperative medication included warfarin,rosuvastatin, metformin, diltiazem, acetylsalicylic acid,budesonide/formoterol, salbutamol, and pantoprazole. Be-cause surgery to drain the bursa was planned, warfarin wasdiscontinued 24 hours before surgery and the patient wasgiven 5 mg IV of vitamin K after which the internationalnormalized ratio was 1.3. While she awaited surgery, prophy-lactic IV heparin was started and stopped 7 hours beforesurgery. No other coagulation test was done immediatelybefore surgery.

At her arrival in the operating room, standard monitor-ing and an IV line were started. Because the patient did notshow any sign of systemic infection (normal white bloodcell count and temperature), an anesthetic technique con-sisting of an ultrasound-guided infraclavicular block was

offered and accepted by the patient. After sedation withmidazolam 2 mg, standard skin asepsis was accomplishedwith 2% wt/vol chlorhexidine gluconate and 70% vol/volisopropyl alcohol and a sterile sheath was used to cover theultrasound probe. A 5- to 12-MHz linear probe was posi-tioned in a parasagittal plane, medial to the coracoidprocess and adjusted to give a transverse view of theaxillary artery using an ultrasound device (Zone Ultra;Zonare Medical Systems, Mountain View, CA). Using anin-plane technique, an 8.89-cm 20-gauge Tuohy needle (B.Braun, Bethlehem, PA) was advanced to the posterior sideof the axillary artery until a fascial click was perceived and30 mL of mepivacaine 1.5% was injected slowly aftermultiple negative aspirations. After documentation of suc-cessful sensory block of the upper arm, surgery wasperformed using a tourniquet (pressure of 250 mm Hg).The surgeon opened and drained the olecranon bursa, andthe procedure lasted 10 minutes. The patient was thendirected to the postanesthesia care unit for a 30-minuteobservation period and was subsequently discharged to theward. With agreement of the surgeon, IV heparin wasresumed 2 hours after the end of the surgery and warfarinwas restarted on the first postoperative day. The patientwas followed daily and discharged from the hospital 6 dayslater, after a therapeutic level of anticoagulation (interna-tional normalized ratio 2.6) had been obtained. At that time,the patient complained of pain in the right shoulder but aphysical examination did not reveal any abnormality.

Two weeks later, the patient consulted at emergencyroom for severe pain in her right shoulder. A physicalexamination demonstrated a significant swelling of theanterior part of the shoulder, the axilla, and the upper partof the right arm. There was no redness or discoloration ofthe skin. A neurological examination of the upper arm didnot demonstrate any motor or sensory deficit but the rangeof movement of the right shoulder was limited by severepain. The radial pulse was present at the right wrist. Asuperficial sonogram of the upper part of the right armshowed a solid hyperechogenic mass of 7.8 � 3.5 � 4.7 cmcontaining a small amount of liquid. A Doppler study didnot reveal any flow in this mass. A computed tomographicangiography showed a 5.4-cm hematoma located at theaxillo-humeral junction of the axillary artery within which

From the Departement d’Anesthesie-Reanimation, Centre de Recherche duCHA, Unite de Recherche en Traumatologie-Urgence-Soins Intensifs, Hopi-tal de l’Enfant-Jesus, Universite Laval, Quebec, Canada.


Supported by intramural department sources.


Address correspondence and reprint requests to Dave Gleeton, MD, Depart-ment of Anesthesiology, Hopital de l’Enfant-Jesus, Universite Laval, 1401,18eme rue, Quebec, QC, G1J 1Z4, Canada. Address e-mail to [email protected].



was a 20-mm pool of contrast dye suggesting an aneurysmjust underneath the wound left by a recent cutaneouspuncture below the right clavicle (consistent with thepuncture site of the infraclavicular block). A second aneu-rysm was suspected in the deltoid area. Digital angiogra-phy confirmed the presence of 2 aneurysms, 1 near thehumeral neck and the other in the prescapular region (5and 13 mm). These were supplied by a right lateral branchof the dorsoscapular artery and the right thoracoacromialartery, respectively. Supraselective catheterization of thefeeding artery and embolization with liquid adhesive glue(Indermil; Tyco, Norwalk, CT) and lipiodol was performed.Control angiogram confirmed the occlusion of both aneu-rysms (Fig. 1). The patient was discharged from the hospi-

tal the following day without any complication. Threefollow-up visits revealed a complete regression of theswelling and the pain. One month later, the patient had nosequel.

DISCUSSIONMycotic aneurysms are defined as an infectious break in thewall of an artery with formation of a blind saccularoutpouching that is contiguous with the arterial lumen,usually in an area of bifurcation or narrowing. The aorta,peripheral arteries, cerebral arteries, and visceral arteriesare involved in descending order of frequency.5 In thepre–antibiotic era, �85% of mycotic aneurysms were asso-ciated with bacterial endocarditis. Currently, the majorityof mycotic aneurysms occur in IV drug users or after invasivemedical procedures. Depressed host immunity secondary tosystemic disease (diabetes, cirrhosis, collagen vascular dis-ease) and corticosteroid therapy are also contributing fac-tors.6 Only 15 mycotic aneurysms of the subclavian arteryhave been reported since 1923.7 The diagnosis is oftendifficult because of the insidious nature of the disease. Pain,erythema, palpable mass, or ischemia distal to the affectedarea is sometimes present.6 Computed tomographic an-giography is the imaging modality of choice for evaluationof mycotic aneurysms but Doppler sonography has bothgood sensitivity and specificity for detection of mycoticaneurysms located in peripheral arteries.8 The usual treat-ment is surgical excision but endovascular techniques havealso been reported.9

One of the main advantages of ultrasound-guided nerveblocks is the possibility of a real-time visualization of theneedle, nerve, and surrounding structures, notably thevessels. It has been demonstrated that ultrasound guidancediminishes the rate of vascular puncture during infracla-vicular nerve block compared with a nerve stimulationtechnique.10 Moreover, it sometimes allows the detection ofabnormal anatomy, thus offering the possibility of makingadjustments to the anesthetic technique planned.1,3 How-ever, complications have not always been prevented by theuse of ultrasound guidance to perform nerve blockade.4,11

The location of the mycotic aneurysms just below the scarleft by the needle on the skin most likely suggests that inour case 1 of the 2 preexisting mycotic aneurysms waspunctured by the needle during the technique. This acci-dental puncture during the course of an apparently uncom-plicated technique can occur in 2 different situations. First,it is possible that the needle was not adequately visualizedby the anesthesiologist (one of the most frequent errors ofclinicians performing ultrasound-guided nerve block12)and could have punctured the mycotic aneurysms justoutside the ultrasound visualization plane. It is also pos-sible that the mycotic aneurysms were located within theultrasound visualization plane but were not recognizedand diagnosed, leading to a witnessed but unrecognizedpuncture during the technique.

The use of color Doppler study during sonographicexamination results in a characteristic yin-yang sign in thepresence of mycotic aneurysms (Fig. 2), thus greatly en-hancing the sensibility of ultrasound to detect mycoticaneurysms.5 The blood stasis inside the mycotic aneurysmsproduces an unusual gray ultrasound image of arterial

Figure 1. A, Angiography of the shoulder area before the emboliza-tion; black arrow � 13-mm mycotic aneurysm; white arrow � 5-mmmycotic aneurysm. B, Angiography performed after the embolization;black arrow � previous site of the 13-mm mycotic aneurysm showingcomplete exclusion of the aneurysm.

CASE REPORT


lumen that might complicate its recognition when Doppleris not used. In our case, a preprocedure Doppler study wasnot performed before the technique. Preprocedure diagno-sis of the mycotic aneurysms in the needle path could haveresulted in modification of the anesthetic technique (choiceof another block site, general anesthesia, use of smallerneedle, etc.) that could have prevented the occurrence ofthe complication. However, it should be kept in mind thatultrasound is a relatively recent technology in the anesthe-siology field. Because of its rare use for diagnostic pur-poses, certain rare medical conditions such as the onereported herein may still go unrecognized.

In conclusion, this case report illustrates that the use ofultrasound cannot completely eliminate the occurrence of

complications. Despite the advantages of this technology, ahigh degree of suspicion should be maintained to minimizethe risk of complications.13 Finally, IV drug users orpatients with a history of endocarditis are at a high risk ofmycotic aneurysms and may benefit from a formal prepro-cedure examination before a peripheral nerve block.14

AUTHOR CONTRIBUTIONSAll authors helped with manuscript preparation.

REFERENCES1. Duggan E, Brull R, Lai J, Abbas S. Ultrasound-guided brachial

plexus block in a patient with multiple glomangiomatosis. RegAnesth Pain Med 2008;33:70–3

2. Kessler J, Gray AT. Sonography of scalene muscle anomaliesfor brachial plexus block. Reg Anesth Pain Med 2007;32:172–3

3. Sites BD, Spence BC, Gallagher JD, Beach ML. On the edge ofthe ultrasound screen: regional anesthesiologists diagnosingnonneural pathology. Reg Anesth Pain Med 2006;31:555–62

4. Zetlaoui PJ, Labbe JP, Benhamou D. Ultrasound guidance foraxillary plexus block does not prevent intravascular injection.Anesthesiology 2008;108:761

5. Lee W, Mossop PF, Little AF, Fitt GJ, Vrazas JI, Hoang JK,Hennessy OF. Infected (mycotic) aneurysms: spectrum ofimaging appearances and management. Radiographics2008;28:1853– 68

6. Kearney RA, Eisen HJ, Wolf JE. Nonvalvular infections of thecardiovascular system. Ann Intern Med 1994;121:219–30

7. Tsao JW, Marder SR, Goldstone J, Bloom AI. Presentation,diagnosis, and management of arterial mycotic pseudoaneu-rysms in injection drug users. Ann Vasc Surg 2002;16:652–62

8. Coughlin BF, Paushter DM. Peripheral pseudoaneurysms:evaluation with duplex US. Radiology 1988;168:339–42

9. Leon LR, Psalms SB, Labropoulos N, Mills JL. Infected upperextremity aneurysms: a review. Eur J Vasc Endovasc Surg2008;35:320–31

10. Maalouf D, Gordon M, Paroli L, Tong-Ngork S. Ultrasound-guidance vs. nerve stimulation for the infraclavicular blockadeof the brachial plexus: a comparison of the vascular puncturerate. Reg Anesth Pain Med 2006;30:A46

11. Loubert C, Williams SR, Helie F, Arcand G. Complicationduring ultrasound-guided regional block: accidental intra-vascular injection of local anesthetic. Anesthesiology2008;108:759 – 60

12. Sites BD, Spence BC, Gallagher JD, Wiley CW, Bertrand ML,Blike GT. Characterizing novice behaviour associated withlearning ultrasound-guided peripheral regional anesthesia.Reg Anesth Pain Med 2007;32:107–15

13. Benitez PR, Newell MA. Vascular trauma in drug abuse:patterns of injury. Ann Vasc Surg 1986;1:175–81

14. Manickam BP, Perlas A, Chan VW, Brull R. The role of apreprocedure systematic sonographic survey in ultrasound-guided regional anesthesia. Reg Anesth Pain Med2008;33:566 –70

Figure 2. Sonograms show a 5.5-cm complex lesion. On the colorDoppler image, the hypoechoic center has turbulent flow (“ying-yang”sign), a finding indicative of a patent aneurysm lumen. The thick,heterogeneous, hypoechoic rind is attributable to hematoma andinflammatory tissue. (Reprinted with permission from Lee et al.5 [LeeW-K, Mossop PJ, Little AF, et al. Infected (mycotic) aneurysms:spectrum of imaging appearances and management. Radiographics2008;28:1853–68].)

Undiagnosed Upper Extremity Mycotic Aneurysms


Infraclavicular Brachial PlexusBlock for Regional Anaesthesia ofthe Lower Arm

Ki Jinn Chin, Mandeep Singh, VeerabadranVelayutham, and Victor Chee

BACKGROUND: Several approaches exist to produce localanaesthetic blockade of the brachial plexus. It is not clearwhich is the technique of choice for providing surgical anaes-thesia of the lower arm although infraclavicular blockade (ICB)has several purported advantages. We therefore performed asystematic review of ICB compared to the other brachial plexusblocks (BPBs).OBJECTIVES: To evaluate the efficacy and safety of ICBcompared to other BPBs in providing regional anaesthesia ofthe lower arm.SEARCH STRATEGY: We searched CENTRAL (The CochraneLibrary 2008, Issue 3), MEDLINE (1950 to September 22nd2008) and EMBASE (1980 to September 22nd 2008). We alsosearched conference proceedings (from 2004 to 2008) andthe www.clinicaltrials.gov registry. No language restriction wasapplied.SELECTION CRITERIA: We included any randomized con-trolled trials (RCTs) that compared ICB with other BPBs as thesole anaesthetic techniques for surgery on the lower arm.DATA COLLECTION AND ANALYSIS: The primary outcomewas adequate surgical anaesthesia within 30 minutes of blockcompletion. Secondary outcomes included sensory block ofindividual nerves, tourniquet pain, onset time of sensoryblockade, block performance time, block-associated pain andcomplications related to the block.MAIN RESULTS: We identified 15 studies with 1020 partici-pants, of whom 510 received ICB and 510 received otherBPBs. The control group intervention was the axillary block in10 studies, mid-humeral block in two studies, supraclavicularblock in two studies and parascalene block in one study. Threestudies employed ultrasound-guided ICB. The risk of failedsurgical anaesthesia and of complications were low and similarfor ICB and all other BPBs. Tourniquet pain was less likely withICB (risk ratio (RR) 0.47, 95% CI 0.24 to 0.92, P � 0.03).When compared to a single-injection axillary block, ICB wasbetter at providing complete sensory block of the musculocu-taneous nerve (RR for failure 0.46, 95% CI 0.27 to 0.60, P �0.0001) and the axillary nerve (RR of failure 0.37, 95% CI 0.24to 0.58, P � 0.0001). ICB was faster to perform thanmultiple-injection axillary (mean difference (MD) �2.7 min,95% CI �4.2 to �1.1, P � 0.0006) or midhumeral blocks (MD�4.8 min, 95% CI �6.0 to �3.6, P � 0.00001) but this wasoffset by a longer sensory block onset time (MD 3.9 min, 95%CI 3.2 to 4.5, P � 0.00001).AUTHORS’ CONCLUSIONS: ICB is a safe and simple tech-nique for providing surgical anaesthesia of the lower arm, withan efficacy comparable to other BPBs. The advantages of ICBinclude a lower likelihood of tourniquet pain during surgery, andmore reliable blockade of the musculocutaneous and axillarynerves when compared to a single-injection axillary block. Theefficacy of ICB is likely to be improved if adequate time isallowed for block onset (at least 30 minutes) and if a volume ofat least 40 ml is injected. Since publication of many of thetrials included in this review, it has become clear that a distalposterior cord motor response is the appropriate endpoint forelectrostimulation-guided ICB; we recommend it be used in all

future comparative studies. There is also a need for additionalRCTs comparing ultrasound-guided ICB with other BPBs.

Chin KJ, Singh M, Velayutham V, Chee V. Infraclavicularbrachial plexus block for regional anaesthesia of the lower armpublished, in the Cochrane Database Syst Rev 2010, Issue 2. Art.No.: CD005487.DOI: 10.1002/14651858.CD005487.pub2

Heated Humidification Versus Heatand Moisture Exchangers forVentilated Adults and Children

Margaret Kelly, Donna Gillies, David A. Todd, andCatherine Lockwood

BACKGROUND: Humidification by artificial means must beprovided when the upper airway is bypassed during mechanicalventilation. Heated humidification (HH) and heat and moistureexchangers (HMEs) are the most commonly used types ofartificial humidification in this situation.OBJECTIVES: To determine whether HHs or HMES are moreeffective in preventing mortality and other complications inpeople who are mechanically ventilated.SEARCH STRATEGY: We searched the Cochrane CentralRegister of Controlled Trials (The Cochrane Library 2010, Issue4) and MEDLINE, EMBASE and CINAHL (January, 2010) toidentify relevant randomized controlled trials.SELECTION CRITERIA: We included randomized controlled tri-als comparing HMEs to HHs in mechanically ventilated adults andchildren. We included randomized crossover studies.DATA COLLECTION AND ANALYSIS: We assessed the qualityof each study and extracted the relevant data. Where appro-priate, results from relevant studies were meta-analyzed forindividual outcomes.MAIN RESULTS: We included 33 trials with 2833 participants;25 studies were parallel group design (n � 2710) and 8crossover design (n � 123). Only 3 included studies reporteddata for infants or children. There was no overall effect onartificial airway occlusion, mortality, pneumonia, or respiratorycomplications; however, the PaCO2 and minute ventilationwere increased when HMEs were compared to HHs and bodytemperature was lower. The cost of HMEs was lower in allstudies that reported this outcome. There was some evidencethat hydrophobic HMEs may reduce the risk of pneumonia andthat blockages of artificial airways may be increased with theuse of HMEs in certain subgroups of patients.AUTHORS’ CONCLUSIONS: There is little evidence of anoverall difference between HMEs and HHs. However, hydropho-bic HMEs may reduce the risk of pneumonia and the use of anHMEs may increase artificial airway occlusion in certain sub-groups of patients. Therefore, HMEs may not be suitable forpatients with limited respiratory reserve or prone to airwayblockage. Further research is needed relating to hydrophobicversus hygroscopic HMEs and the use of HMEs in the pediatricand neonatal populations. As the design of HMEs evolves,evaluation of new generation HMEs will also need to beundertaken.

Kelly M, Gillies D, Todd DA, Lockwood C. Heated humidifi-cation versus heat and moisture exchangers for ventilatedadults and children. Cochrane Database Syst Rev 2010, Issue4. Art. No.: CD004711.DOI: 10.1002/14651858.CD004711.pub2.

COCHRANE CORNER


Infraclavicular Brachial PlexusBlock for Regional Anaesthesia ofthe Lower Arm

Ki Jinn Chin, Mandeep Singh, VeerabadranVelayutham, and Victor Chee

BACKGROUND: Several approaches exist to produce localanaesthetic blockade of the brachial plexus. It is not clearwhich is the technique of choice for providing surgical anaes-thesia of the lower arm although infraclavicular blockade (ICB)has several purported advantages. We therefore performed asystematic review of ICB compared to the other brachial plexusblocks (BPBs).OBJECTIVES: To evaluate the efficacy and safety of ICBcompared to other BPBs in providing regional anaesthesia ofthe lower arm.SEARCH STRATEGY: We searched CENTRAL (The CochraneLibrary 2008, Issue 3), MEDLINE (1950 to September 22nd2008) and EMBASE (1980 to September 22nd 2008). We alsosearched conference proceedings (from 2004 to 2008) andthe www.clinicaltrials.gov registry. No language restriction wasapplied.SELECTION CRITERIA: We included any randomized con-trolled trials (RCTs) that compared ICB with other BPBs as thesole anaesthetic techniques for surgery on the lower arm.DATA COLLECTION AND ANALYSIS: The primary outcomewas adequate surgical anaesthesia within 30 minutes of blockcompletion. Secondary outcomes included sensory block ofindividual nerves, tourniquet pain, onset time of sensoryblockade, block performance time, block-associated pain andcomplications related to the block.MAIN RESULTS: We identified 15 studies with 1020 partici-pants, of whom 510 received ICB and 510 received otherBPBs. The control group intervention was the axillary block in10 studies, mid-humeral block in two studies, supraclavicularblock in two studies and parascalene block in one study. Threestudies employed ultrasound-guided ICB. The risk of failedsurgical anaesthesia and of complications were low and similarfor ICB and all other BPBs. Tourniquet pain was less likely withICB (risk ratio (RR) 0.47, 95% CI 0.24 to 0.92, P � 0.03).When compared to a single-injection axillary block, ICB wasbetter at providing complete sensory block of the musculocu-taneous nerve (RR for failure 0.46, 95% CI 0.27 to 0.60, P �0.0001) and the axillary nerve (RR of failure 0.37, 95% CI 0.24to 0.58, P � 0.0001). ICB was faster to perform thanmultiple-injection axillary (mean difference (MD) �2.7 min,95% CI �4.2 to �1.1, P � 0.0006) or midhumeral blocks (MD�4.8 min, 95% CI �6.0 to �3.6, P � 0.00001) but this wasoffset by a longer sensory block onset time (MD 3.9 min, 95%CI 3.2 to 4.5, P � 0.00001).AUTHORS’ CONCLUSIONS: ICB is a safe and simple tech-nique for providing surgical anaesthesia of the lower arm, withan efficacy comparable to other BPBs. The advantages of ICBinclude a lower likelihood of tourniquet pain during surgery, andmore reliable blockade of the musculocutaneous and axillarynerves when compared to a single-injection axillary block. Theefficacy of ICB is likely to be improved if adequate time isallowed for block onset (at least 30 minutes) and if a volume ofat least 40 ml is injected. Since publication of many of thetrials included in this review, it has become clear that a distalposterior cord motor response is the appropriate endpoint forelectrostimulation-guided ICB; we recommend it be used in all

future comparative studies. There is also a need for additionalRCTs comparing ultrasound-guided ICB with other BPBs.

Chin KJ, Singh M, Velayutham V, Chee V. Infraclavicularbrachial plexus block for regional anaesthesia of the lower armpublished, in the Cochrane Database Syst Rev 2010, Issue 2. Art.No.: CD005487.DOI: 10.1002/14651858.CD005487.pub2

Heated Humidification Versus Heatand Moisture Exchangers forVentilated Adults and Children

Margaret Kelly, Donna Gillies, David A. Todd, andCatherine Lockwood

BACKGROUND: Humidification by artificial means must beprovided when the upper airway is bypassed during mechanicalventilation. Heated humidification (HH) and heat and moistureexchangers (HMEs) are the most commonly used types ofartificial humidification in this situation.OBJECTIVES: To determine whether HHs or HMES are moreeffective in preventing mortality and other complications inpeople who are mechanically ventilated.SEARCH STRATEGY: We searched the Cochrane CentralRegister of Controlled Trials (The Cochrane Library 2010, Issue4) and MEDLINE, EMBASE and CINAHL (January, 2010) toidentify relevant randomized controlled trials.SELECTION CRITERIA: We included randomized controlled tri-als comparing HMEs to HHs in mechanically ventilated adults andchildren. We included randomized crossover studies.DATA COLLECTION AND ANALYSIS: We assessed the qualityof each study and extracted the relevant data. Where appro-priate, results from relevant studies were meta-analyzed forindividual outcomes.MAIN RESULTS: We included 33 trials with 2833 participants;25 studies were parallel group design (n � 2710) and 8crossover design (n � 123). Only 3 included studies reporteddata for infants or children. There was no overall effect onartificial airway occlusion, mortality, pneumonia, or respiratorycomplications; however, the PaCO2 and minute ventilationwere increased when HMEs were compared to HHs and bodytemperature was lower. The cost of HMEs was lower in allstudies that reported this outcome. There was some evidencethat hydrophobic HMEs may reduce the risk of pneumonia andthat blockages of artificial airways may be increased with theuse of HMEs in certain subgroups of patients.AUTHORS’ CONCLUSIONS: There is little evidence of anoverall difference between HMEs and HHs. However, hydropho-bic HMEs may reduce the risk of pneumonia and the use of anHMEs may increase artificial airway occlusion in certain sub-groups of patients. Therefore, HMEs may not be suitable forpatients with limited respiratory reserve or prone to airwayblockage. Further research is needed relating to hydrophobicversus hygroscopic HMEs and the use of HMEs in the pediatricand neonatal populations. As the design of HMEs evolves,evaluation of new generation HMEs will also need to beundertaken.

Kelly M, Gillies D, Todd DA, Lockwood C. Heated humidifi-cation versus heat and moisture exchangers for ventilatedadults and children. Cochrane Database Syst Rev 2010, Issue4. Art. No.: CD004711.DOI: 10.1002/14651858.CD004711.pub2.

COCHRANE CORNER


Airway Management: Standardization,Simplicity, and Daily Practice Are theKeys to Success

To the Editor

The editorial by Hung and Murphy,1 although inter-esting, contains potentially misleading statements.Some have opined that careful evaluation of the

airway as part of a preplanned strategy may lead toimproved outcome.2 However, it is potentially dangerousto suggest that the choice of the technique and hence thechoice of the equipment is or should be influenced primar-ily by different circumstances. Advocating such an ap-proach would result in multiple strategies using manydifferent and, at times, unfamiliar airway devices.

Only a few studies have focused on effective airwaymanagement and, in these, outcome was investigated un-der the prevailing clinical conditions.3–5 A common char-acteristic of each of these studies was a limitation oftechniques and devices and that deviation from the pre-defined algorithm was recorded infrequently.

A key factor regarding safety recorded by highly reliableorganizations such as aviation is a standardized process.6

However, the “recommendations” of Hung and Murphythat airway management is primarily “context sensitive”would guide us in the wrong direction. There are only afew situations wherein we might deviate from our difficultairway guidelines. For example, it is very unlikely that youcan perform an awake intubation in an uncooperativepatient. Importantly, such situations should be managed bythe most experienced physicians. However, if deviationfrom, rather than adherence to, the guidelines is currentpractice, the guidelines should be modified accordingly.

A second point is the missing commitment to fiberopticintubation (“unwritten truth”). If the authors mean thatthere are no prospective randomized studies showing theeffectiveness of fiberoptic intubation, they are correct. Un-fortunately, there are no prospective randomized studiesdemonstrating the effectiveness (not efficacy) of most tech-niques used in daily practice. However, it is generallyagreed among airway management practitioners and rec-ommended by many anesthesia societies that fiberopticintubation should be used for management of the antici-pated difficult airway.7 So, questioning this technique isprobably potentially misleading for the average anesthesi-ologist in practice.

Regarding airway management, the message should be:Standardization, simplicity, and daily practice are the keysto success.

Thomas Heidegger, MDDepartment of Anesthesia

Spitalregion Rheintal Werdenberg SarganserlandWalenstadt, Switzerland

[email protected]

REFERENCES1. Hung O, Murphy M. Context-sensitive airway management.

Anesth Analg 2010;110:982–32. American Society of Anesthesiologists Task Force on Manage-

ment of the Difficult Airway (2003). Practice guidelines formanagement of the difficult airway: an updated report by theAmerican Society of Anesthesiologists Task Force on Manage-ment of the Difficult Airway. Anesthesiology 2003;98:1269–77

3. Hopkins A. Measuring the Quality of Medical Care. 1st ed.Oxford: Royal College of Physicians of London, 1990

4. Combes X, Le Roux B, Suen P, Dumerat M, Motamed C, SauvatS, Duvaldestin P, Dhonneur G. Unanticipated difficult airway inanesthetized patients: prospective validation of a managementalgorithm. Anesthesiology 2004;100:1146–50

5. Heidegger T, Gerig HJ, Ulrich B, Kreienbuhl G. Validation of asimple algorithm for tracheal intubation: daily practice is thekey to success in emergencies—an analysis of 13,248 intuba-tions. Anesth Analg 2001;92:517–22

6. Reason J. Human error: models and management. BMJ200;320:768–70

7. Heidegger T, Gerig HJ, Henderson JJ. Strategies and algorithmsfor management of the difficult airway. Best Pract Res ClinAnaesthesiol 2005;19:661–74

DOI: 10.1213/ANE.0b013e3181ec312a

In ResponseNo one disagrees that the fundamental goal of airway man-agement is oxygenation and ventilation, not devices and tools.A systematic approach to airway management includes air-way evaluation, selection of an appropriate course of actionlikely to succeed (Plan A) and preparation for failure (i.e., PlanB, Plan C, etc.). This approach is, and must be, consistent withthe “context.” In other words, one must accept that airwayspresent themselves in a variety of forms, circumstances, andlocations, to a panoply of health care providers with varyingskill sets. Although there is a recommended “strategy,” thereare varying tactics or techniques one may use. The tactic ortechnique is virtually always dependent on the circumstancesand skill set of the airway practitioner.

Apart from the oxygenator of a bypass pump, cliniciansbasically use only 4 methods of ventilation and oxygenation:through a bag-mask, an extraglottic device (e.g., a laryngealmask airway), a tracheal tube, or a surgical airway. As statedin our editorial, “context-sensitive airway management im-plies that managing a difficult or failed airway should bedriven by the principles of ‘gas exchange’ and not be ‘device-dependent.’”1 In addition, we also stated that “clinicians mustbe trained to understand the basic principles of airwaymanagement using basic techniques and learn how to applythese techniques properly in an appropriate environment.”These statements were structured to inform clinicians thatthey must learn basic airway techniques and apply themproperly. Nowhere in the editorial did we advocate that “…an approach would result in multiple strategies using manydifferent and at times, unfamiliar airway devices.”

Dr. Heidegger correctly states that awake tracheal intu-bation is considered to be the most prudent approach in apatient with an anticipated difficult airway.2 But, in addi-tion to the flexible bronchoscope, awake intubation can beperformed safely and effectively by many techniques, in-cluding the rigid fiberoptic laryngoscopes, videolaryngo-scopes, and Macintosh laryngoscope with the Eschmann


LETTERS TO THE EDITORSection Editor: Lawrence Saidman

Airway Management: Standardization,Simplicity, and Daily Practice Are theKeys to Success

To the Editor

The editorial by Hung and Murphy,1 although inter-esting, contains potentially misleading statements.Some have opined that careful evaluation of the

airway as part of a preplanned strategy may lead toimproved outcome.2 However, it is potentially dangerousto suggest that the choice of the technique and hence thechoice of the equipment is or should be influenced primar-ily by different circumstances. Advocating such an ap-proach would result in multiple strategies using manydifferent and, at times, unfamiliar airway devices.

Only a few studies have focused on effective airwaymanagement and, in these, outcome was investigated un-der the prevailing clinical conditions.3–5 A common char-acteristic of each of these studies was a limitation oftechniques and devices and that deviation from the pre-defined algorithm was recorded infrequently.

A key factor regarding safety recorded by highly reliableorganizations such as aviation is a standardized process.6

However, the “recommendations” of Hung and Murphythat airway management is primarily “context sensitive”would guide us in the wrong direction. There are only afew situations wherein we might deviate from our difficultairway guidelines. For example, it is very unlikely that youcan perform an awake intubation in an uncooperativepatient. Importantly, such situations should be managed bythe most experienced physicians. However, if deviationfrom, rather than adherence to, the guidelines is currentpractice, the guidelines should be modified accordingly.

A second point is the missing commitment to fiberopticintubation (“unwritten truth”). If the authors mean thatthere are no prospective randomized studies showing theeffectiveness of fiberoptic intubation, they are correct. Un-fortunately, there are no prospective randomized studiesdemonstrating the effectiveness (not efficacy) of most tech-niques used in daily practice. However, it is generallyagreed among airway management practitioners and rec-ommended by many anesthesia societies that fiberopticintubation should be used for management of the antici-pated difficult airway.7 So, questioning this technique isprobably potentially misleading for the average anesthesi-ologist in practice.

Regarding airway management, the message should be:Standardization, simplicity, and daily practice are the keysto success.

Thomas Heidegger, MDDepartment of Anesthesia

Spitalregion Rheintal Werdenberg SarganserlandWalenstadt, Switzerland

[email protected]


Anesth Analg 2010;110:982–32. American Society of Anesthesiologists Task Force on Manage-

ment of the Difficult Airway (2003). Practice guidelines formanagement of the difficult airway: an updated report by theAmerican Society of Anesthesiologists Task Force on Manage-ment of the Difficult Airway. Anesthesiology 2003;98:1269–77

3. Hopkins A. Measuring the Quality of Medical Care. 1st ed.Oxford: Royal College of Physicians of London, 1990

4. Combes X, Le Roux B, Suen P, Dumerat M, Motamed C, SauvatS, Duvaldestin P, Dhonneur G. Unanticipated difficult airway inanesthetized patients: prospective validation of a managementalgorithm. Anesthesiology 2004;100:1146–50

5. Heidegger T, Gerig HJ, Ulrich B, Kreienbuhl G. Validation of asimple algorithm for tracheal intubation: daily practice is thekey to success in emergencies—an analysis of 13,248 intuba-tions. Anesth Analg 2001;92:517–22

6. Reason J. Human error: models and management. BMJ200;320:768–70

7. Heidegger T, Gerig HJ, Henderson JJ. Strategies and algorithmsfor management of the difficult airway. Best Pract Res ClinAnaesthesiol 2005;19:661–74

DOI: 10.1213/ANE.0b013e3181ec312a

In ResponseNo one disagrees that the fundamental goal of airway man-agement is oxygenation and ventilation, not devices and tools.A systematic approach to airway management includes air-way evaluation, selection of an appropriate course of actionlikely to succeed (Plan A) and preparation for failure (i.e., PlanB, Plan C, etc.). This approach is, and must be, consistent withthe “context.” In other words, one must accept that airwayspresent themselves in a variety of forms, circumstances, andlocations, to a panoply of health care providers with varyingskill sets. Although there is a recommended “strategy,” thereare varying tactics or techniques one may use. The tactic ortechnique is virtually always dependent on the circumstancesand skill set of the airway practitioner.

Apart from the oxygenator of a bypass pump, cliniciansbasically use only 4 methods of ventilation and oxygenation:through a bag-mask, an extraglottic device (e.g., a laryngealmask airway), a tracheal tube, or a surgical airway. As statedin our editorial, “context-sensitive airway management im-plies that managing a difficult or failed airway should bedriven by the principles of ‘gas exchange’ and not be ‘device-dependent.’”1 In addition, we also stated that “clinicians mustbe trained to understand the basic principles of airwaymanagement using basic techniques and learn how to applythese techniques properly in an appropriate environment.”These statements were structured to inform clinicians thatthey must learn basic airway techniques and apply themproperly. Nowhere in the editorial did we advocate that “…an approach would result in multiple strategies using manydifferent and at times, unfamiliar airway devices.”

Dr. Heidegger correctly states that awake tracheal intu-bation is considered to be the most prudent approach in apatient with an anticipated difficult airway.2 But, in addi-tion to the flexible bronchoscope, awake intubation can beperformed safely and effectively by many techniques, in-cluding the rigid fiberoptic laryngoscopes, videolaryngo-scopes, and Macintosh laryngoscope with the Eschmann


LETTERS TO THE EDITORSection Editor: Lawrence Saidman

Tracheal Introducer. Dr. Heidegger also correctly statesthat “. . . it is generally agreed among airway managementpractitioners and recommended by many anesthesia soci-eties that fiberoptic intubation should be used for manage-ment of the anticipated difficult airway.” However, itwould be grossly incorrect and perhaps dangerous for Dr.Heidegger to imply that “awake fiberoptic intubationshould be standardized in the management of the antici-pated difficult airway.” In an uncooperative patient, in thepresence of blood, in emergency situations, or in an envi-ronment with limited resources, fiberoptic intubationwould be difficult, if not impossible.

Having taught airway management to thousands ofpractitioners, we recognize that there is tremendous vari-ability within individual skill sets. Because of this variabil-ity, the principles of airway management and the strategyfor managing a difficult airway have been well elucidatedby guidelines promulgated by various societies includingthe American Society of Anesthesiologists.3,4 It is critical forall practitioners to recognize that these guidelines are notrecipes to be rigidly followed. Rather, as stated clearly inthe ASA Practice Guidelines for Management of the Diffi-cult Airway,3,4 “. . . these recommendations may be adopted,modified, or rejected according to clinical needs and constraints”(context-sensitive). Furthermore, “Practice guidelines arenot intended as standards or absolute requirements. Theuse of practice guidelines cannot guarantee any specificoutcome. Practice guidelines are subject to revision aswarranted by the evolution of medical knowledge, technol-ogy, and practice.”3,4

Orlando Hung, MDProfessor Anesthesiology, Surgery, and Pharmacology

Dalhousie UniversityQueen Elizabeth II Health Sciences

Halifax, Nova Scotia, [email protected]

Michael Murphy, MDProfessor and Chair Anesthesiology

Professor Emergency MedicineDalhousie University

District Chief AnesthesiologyCapital District Health Authority

Queen Elizabeth II Health SciencesHalifax, Nova Scotia, Canada

[email protected]


Anesth Analg 2010;110:982–32. Heidegger T. Airway management: standardization, simplicity,

and daily practice are the keys to success. Anesth Analg2010;111:1073

3. Practice guidelines for management of the difficult airway: areport by the American Society of Anesthesiologists Task Forceon Management of the Difficult Airway. Anesthesiology1993;78:597–602

4. American Society of Anesthesiologists Task Force on Manage-ment of the Difficult Airway. Practice guidelines for manage-ment of the difficult airway: an updated report by the AmericanSociety of Anesthesiologists Task Force on Management of theDifficult Airway. Anesthesiology 2003;98:1269–77

DOI: 10.1213/ANE.0b013e3181ec3153

Achieving Full Risk Disclosure inPediatric Anesthesia Research

To the Editor

Hong et al.’s1 study using fluoroscopy on 73 ASAphysical status I children, ages 1 to 5 years, toassess ropivacaine–radiopaque dye solution spread

in caudal injections for postorchiopexy pain raises a vexingquestion: Were all known risks fully disclosed to theirsubjects’ parents? The absence of details about fluoroscopyequipment, radiation reduction measures, radiation doses,and subject radiation risks suggests that they were not.

Strauss and Kaste2 stated (and Linet et al.3 concurred)that the ALARA concept means radiation should be “AsLow As Reasonably Achievable.” Children, they noted,“might be as much as 10 times more radiosensitive thanadults.” Cohen4 compared 2 fluoroscopic machines andnoted that “the radiologist can greatly increase patientexposure by merely altering settings over which they haveimmediate control” and cited exposure variations of 3486%and 4479% on the basis of such factors. Petterson et al.5

found a 2.23 testicular cancer relative risk for children withundescended testes operated on before age 13. Presumably,their risk is augmented by radiation exposure.

Hong et al. experimented to answer a clinical question.Their study conferred no additional benefit to subjects, onlya potential of harm. Science alone benefited. Beecher6

warned researchers away from such practices. In researchas in clinical practice, Primum Non Nocere still applies.

Vincent J. Kopp, MDMichael G. Danekas, MD

Division of Pediatric AnesthesiaDepartment of Anesthesiology

School of MedicineUniversity of North Carolina at Chapel Hill

Chapel Hill, North [email protected]

REFERENCES1. Hong J-Y, Han SW, Kim WO, Cho JS, Kil HJ. A comparison of

high volume/low concentration and low volume/high concen-tration ropivacaine in caudal analgesia for pediatric orchiopexy.Anesth Analg 2009;109:1073–8

2. Strauss KJ, Kaste SC. The ALARA concept in pediatric interven-tional fluoroscopic imaging: striving to keep radiation doses aslow as possible during fluoroscopy of pediatric patients. Awhite paper executive summary. AJR 2006;187:818–9

3. Linet MS, Kim KP, Rajaraman P. Children’s exposure to diag-nostic medical radiation and cancer risk: epidemiologic anddosimetric considerations. Pediatr Radiol 2009;39(Suppl 1):S4–26;Epub 2008 Dec 16

4. Cohen M. Are we doing enough to minimize fluoroscopicradiation exposure in children? Pediatr Radiol 2007;37:1020–4

5. Petterson A, Richardi L, Nordensklod, Kaijser M, Akre O. Ageat surgery for undescended testis and risk of testicular cancer.N Engl J Med 2007;356:1835–41

6. Beecher HK. Ethics and clinical research. N Engl J Med1966;274:367–72

DOI: 10.1213/ANE.0b013e3181ed17ff

LETTERS TO THE EDITOR


Tracheal Introducer. Dr. Heidegger also correctly statesthat “. . . it is generally agreed among airway managementpractitioners and recommended by many anesthesia soci-eties that fiberoptic intubation should be used for manage-ment of the anticipated difficult airway.” However, itwould be grossly incorrect and perhaps dangerous for Dr.Heidegger to imply that “awake fiberoptic intubationshould be standardized in the management of the antici-pated difficult airway.” In an uncooperative patient, in thepresence of blood, in emergency situations, or in an envi-ronment with limited resources, fiberoptic intubationwould be difficult, if not impossible.

Having taught airway management to thousands ofpractitioners, we recognize that there is tremendous vari-ability within individual skill sets. Because of this variabil-ity, the principles of airway management and the strategyfor managing a difficult airway have been well elucidatedby guidelines promulgated by various societies includingthe American Society of Anesthesiologists.3,4 It is critical forall practitioners to recognize that these guidelines are notrecipes to be rigidly followed. Rather, as stated clearly inthe ASA Practice Guidelines for Management of the Diffi-cult Airway,3,4 “. . . these recommendations may be adopted,modified, or rejected according to clinical needs and constraints”(context-sensitive). Furthermore, “Practice guidelines arenot intended as standards or absolute requirements. Theuse of practice guidelines cannot guarantee any specificoutcome. Practice guidelines are subject to revision aswarranted by the evolution of medical knowledge, technol-ogy, and practice.”3,4

Orlando Hung, MDProfessor Anesthesiology, Surgery, and Pharmacology

Dalhousie UniversityQueen Elizabeth II Health Sciences

Halifax, Nova Scotia, [email protected]

Michael Murphy, MDProfessor and Chair Anesthesiology

Professor Emergency MedicineDalhousie University

District Chief AnesthesiologyCapital District Health Authority

Queen Elizabeth II Health SciencesHalifax, Nova Scotia, Canada

[email protected]


Anesth Analg 2010;110:982–32. Heidegger T. Airway management: standardization, simplicity,

and daily practice are the keys to success. Anesth Analg2010;111:1073

3. Practice guidelines for management of the difficult airway: areport by the American Society of Anesthesiologists Task Forceon Management of the Difficult Airway. Anesthesiology1993;78:597–602

4. American Society of Anesthesiologists Task Force on Manage-ment of the Difficult Airway. Practice guidelines for manage-ment of the difficult airway: an updated report by the AmericanSociety of Anesthesiologists Task Force on Management of theDifficult Airway. Anesthesiology 2003;98:1269–77

DOI: 10.1213/ANE.0b013e3181ec3153

Achieving Full Risk Disclosure inPediatric Anesthesia Research

To the Editor

Hong et al.’s1 study using fluoroscopy on 73 ASAphysical status I children, ages 1 to 5 years, toassess ropivacaine–radiopaque dye solution spread

in caudal injections for postorchiopexy pain raises a vexingquestion: Were all known risks fully disclosed to theirsubjects’ parents? The absence of details about fluoroscopyequipment, radiation reduction measures, radiation doses,and subject radiation risks suggests that they were not.

Strauss and Kaste2 stated (and Linet et al.3 concurred)that the ALARA concept means radiation should be “AsLow As Reasonably Achievable.” Children, they noted,“might be as much as 10 times more radiosensitive thanadults.” Cohen4 compared 2 fluoroscopic machines andnoted that “the radiologist can greatly increase patientexposure by merely altering settings over which they haveimmediate control” and cited exposure variations of 3486%and 4479% on the basis of such factors. Petterson et al.5

found a 2.23 testicular cancer relative risk for children withundescended testes operated on before age 13. Presumably,their risk is augmented by radiation exposure.

Hong et al. experimented to answer a clinical question.Their study conferred no additional benefit to subjects, onlya potential of harm. Science alone benefited. Beecher6

warned researchers away from such practices. In researchas in clinical practice, Primum Non Nocere still applies.

Vincent J. Kopp, MDMichael G. Danekas, MD

Division of Pediatric AnesthesiaDepartment of Anesthesiology

School of MedicineUniversity of North Carolina at Chapel Hill

Chapel Hill, North [email protected]

REFERENCES1. Hong J-Y, Han SW, Kim WO, Cho JS, Kil HJ. A comparison of


2. Strauss KJ, Kaste SC. The ALARA concept in pediatric interven-tional fluoroscopic imaging: striving to keep radiation doses aslow as possible during fluoroscopy of pediatric patients. Awhite paper executive summary. AJR 2006;187:818–9

3. Linet MS, Kim KP, Rajaraman P. Children’s exposure to diag-nostic medical radiation and cancer risk: epidemiologic anddosimetric considerations. Pediatr Radiol 2009;39(Suppl 1):S4–26;Epub 2008 Dec 16

4. Cohen M. Are we doing enough to minimize fluoroscopicradiation exposure in children? Pediatr Radiol 2007;37:1020–4

5. Petterson A, Richardi L, Nordensklod, Kaijser M, Akre O. Ageat surgery for undescended testis and risk of testicular cancer.N Engl J Med 2007;356:1835–41

6. Beecher HK. Ethics and clinical research. N Engl J Med1966;274:367–72

DOI: 10.1213/ANE.0b013e3181ed17ff



In ResponseWe agree with Kopp et al.1 that the ALARA (as low asreasonably achievable) concept must be considered inpediatric fluoroscopic imaging. Of course, in our study2 weobtained consent from the parents of the children afterexplaining details of the study and radiation exposure. Thegenital area was protected from radiation by a lead plateduring fluoroscopy using a new pulsed unit from Phillips(Eindhoven, The Netherlands). The study by Cohen3 de-scribed possible radiation exposure variations as much as3486% and 4479% with the maximum magnification, great-est pulse rate, intensifier high, grid in, and greatest inten-sifier exposure rate. We believe that the exposure variationswere not as great because we did not alter the basalmagnification setting and used the lowest pulse rate, gridout, and lowest intensifier exposure rate. The exposurefrequency to obtain 1 image (exposure rate) was 1 or 2, andoverall duration of fluoroscopy was �30 seconds. How-ever, because the average estimated entrance surface dose(ESD) and dose area product increases with age, the ESDalso increases in children in comparison with adults4; weinvestigated the ESD in 5 randomly selected children in ourprevious study.5 The ESD was 0.05 to 0.26 mSv, which wasless than the results of a United Kingdom nationwidesurvey in 2005.6 Although our procedures did not divergefrom the ALARA concept when we substituted the proce-dure to the flow diagram for managing patient dose byStrauss and Kaste,7 we believe that the radiation exposureshould be strictly limited in children.

Hae K. Kil, MDJeong Y. Hong, MD

Won O. Kim, MDDepartment of Anesthesiology and Pain Medicine

Yonsei University College of MedicineSeoul, South Korea

[email protected]

REFERENCES1. Kopp VJ, Danekas MG. Achieving full risk disclosure in pedi-

atric anesthesia research. Anesth Analg 2010;111:10742. Hong JY, Han SW, Kim WO, Cho JS, Kil HK. A comparison of


3. Cohen M. Are we doing enough to minimize fluoroscopicradiation exposure in children? Pediatr Radiol 2007;37:1020 – 4

4. Linet MS, Kim KP, Rajaraman R. Children’s exposure to diagnosticmedical radiation and cancer risk: epidemiologic and dosimetricconsiderations. Pediatr Radiol 2009;39(Suppl 1):S4–26

5. Shin SK, Hong JY, Kim WO, Koo BN, Kim JE, Kil HK. Ultra-sound evaluation of the acral area and comparison of sacralinterspinous and hiatal approach for caudal block in children.Anesthesiology 2009;111:1135–40

6. Hart D, Hillier MC, Wall BF. National reference doses forcommon radiographic, fluoroscopic and dental X-ray examina-tions in the UK. Br J Radiol 2009;82:1–12

7. Strauss KJ, Kaste SC. The ALARA (as low as reasonablyachievable) concept in pediatric interventional and fluoroscopicimaging: striving to keep radiation doses as low as possibleduring fluoroscopy of pediatric patients. A white paper execu-tive summary. Pediatr Radiol 2006;36(Suppl 2):110–2

DOI: 10.1213/ANE.0b013e3181ed1811

Pulse Dye Densitometry andIndocyanine Green PlasmaDisappearance: The Issue of“Normal” Values

To the Editor

Reekers et al.1 investigated normal values of plasmadisappearance rate of indocyanine green (PDR-ICG)measured by pulse dye densitometry in a population

of ASA physical status I and II patients, assumed free ofliver disease. They also evaluated the noninvasive mea-surement of PDR-ICG using a nose and finger probe andcompared these results with those using invasive arterialblood measurements.

Reekers et al. confirmed that, as has been shown before,PDR-ICG is adequately measured noninvasively by pulsedye densitometry.2,3 In addition, they found a mean PDR-ICG value of 23.1%/min (SD � 7.9%/min), which is in linewith previous studies.4–6 However, they found that PDR-ICG varied widely with a range from 9.7%/min to43.2%/min. This is in contrast to earlier publications inwhich PDR-ICG ranged from 18.7%/min to 30.1%/min.4–6

As explained by the authors, a PDR-ICG �18%/min isgenerally considered to indicate impaired hepatic function.5,7

Although the authors argue that only patients without overtliver pathology were examined, there is a lack of objectiveevidence of normal hepatic function in the entire populationstudied. Biochemical liver function tests were available inonly 22% of the patients. The authors recognize this limitationand argue that the patients studied are representative of apopulation with normal liver function based on the absence ofany clinical evidence of hepatic dysfunction. This is a rationalargument, but the clinical evidence for normal liver function isnot a surrogate for laboratory testing because patients withelevated liver enzymes do not necessarily present clinicalsigns of hepatic dysfunction.8 Because PDR-ICG is verysensitive for hepatic dysfunction, changes in PDR-ICG caneven precede changes in serum bilirubin, as the authorsstated.1 Given the fact that the authors ultimately want toargue for a testing threshold to identify normal liver function,the study would have been more compelling if all patientshad received a careful hepatic testing.

Unfortunately, Reekers et al. also did not present any datadescribing hemodynamics of their patients. This is, however,of utmost importance, because PDR-ICG is, as is ICG liverextraction capacity, primarily dependent on splanchnic bloodflow and total plasma volume.2,9 It is also not clear whyReekers et al. studied both awake and anesthetized patients.There are clearly potential differences between the groupswith respect to hemodynamics, which may influence theobtained PDR-ICG values. Reekers et al. argue that the resultsin the anesthetized patients were not different from the awakepatients when analyzed separately so that the data werepooled. However, it is well known that induction of anesthe-sia with propofol results in a significant decrease in cardiacoutput decreasing splanchnic blood flow and subsequentlyimpairing PDR-ICG, even in a relatively young group of ASAphysical status I and II patients.10

Letters to the Editor


In conclusion, we believe the relatively wide and lowranged PDR-ICG “normal” values found by Reekers et al.might be the result of nonhomogeneity of the patient groupwith nonconstant hemodynamics and thus should not lead toa premature rejection of a well-established method.

Jaap J. Vos, BScThomas W. L. Scheeren, MD, PhD

Gotz J. K. Wietasch, MD, PhDDepartment of Anesthesiology

University Medical Center GroningenUniversity of Groningen

Groningen, The [email protected]

REFERENCES1. Reekers M, Simon MJ, Boer F, Mooren RA, van Kleef JW,

Dahan A, Vuyk J. Pulse dye densitometry and indocyaninegreen plasma disappearance in ASA physical status I-II pa-tients. Anesth Analg 2010;110:466–72

2. von Spiegel T, Scholz M, Wietasch G, Hering R, Allen SJ, WoodP, Hoeft A. Perioperative monitoring of indocyanine greenclearance and plasma disappearance rate in patients undergo-ing liver transplantation. Anaesthesist 2002;51:359–66

3. Sakka SG, Reinhart K, Meier-Hellmann A. Comparison ofinvasive and noninvasive measurements of indocyanine greenplasma disappearance rate in critically ill patients with me-chanical ventilation and stable hemodynamics. Intensive CareMed 2000;26:1553–6

4. Rowell LB, Blackmon JR, Bruce RA. Indocyanine green clear-ance and estimated hepatic blood flow during mild to maximalexercise in upright man. J Clin Invest 1964;43:1677–90

5. Hori T, Iida T, Yagi S, Taniguchi K, Yamamoto C, Mizuno S,Yamagiwa K, Isaji S, Uemoto S. K(ICG) value, a reliablereal-time estimator of graft function, accurately predicts out-comes in adult living-donor liver transplantation. LiverTranspl 2006;12:605–13

6. de Liguori Carino N, O’Reilly DA, Dajani K, Ghaneh P, PostonGJ, Wu AV. Perioperative use of the LiMON method ofindocyanine green elimination measurement for the predictionand early detection of post-hepatectomy liver failure. EurJ Surg Oncol 2009;35:957–62

7. Kuntz H, Schregel W. Indocyanine green: evaluation of liverfunction—application in intensive care medicine. In: Lewis F,Pfeiffer U, eds. Practical Applications of Fiberoptics in CriticalCare Monitoring. 2nd ed. New York: Springer, 1990:57–62

8. Hultcrantz R, Glaumann H, Lindberg G, Nilsson LH. Liverinvestigation in 149 asymptomatic patients with moderatelyelevated activities of serum aminotransferases. Scand J Gastro-enterol 1986;21:109–13

9. Caesar J, Shaldon S, Chiandussi L, Guevara L, Sherlock S. Theuse of indocyanine green in the measurement of hepatic bloodflow and as a test of hepatic function. Clin Sci 1961;21:43–57

10. Lange H, Stephan H, Rieke H, Kellermann M, Sonntag H, BircherJ. Hepatic and extrahepatic disposition of propofol in patientsundergoing coronary bypass surgery. Br J Anaesth 1990;64:563–70

DOI: 10.1213/ANE.0b013e3181ef35ba

In ResponseWe agree with Vos et al.1 that the absence of liver enzymemeasurements in some of our patients may be a drawbackin our study. However, biochemical hepatic function test-ing may not offer the key information on hepatocellulardysfunction2 as Vos et al. suggest. Hultcrantz et al.3 describedthat chronically elevated liver enzymes without symptoms orphysical signs of liver disease correspond with various formsof liver disease preferably in the presence of a positive historyon alcohol consumption, drug abuse, hepatitis, or obesity. All

patients displaying these factors were excluded from ourstudy population, thus removing those patients that mayexhibit elevated hepatic enzymes in the absence of physicalsigns or symptoms of liver disease.

Regarding the importance of hemodynamics and indocya-nine green plasma disappearance rate (ICG-PDR), we origi-nally intended to use the noninvasive method of ICG mea-surement, pulse dye densitometry, to measure cardiac outputin our study population. To validate the transcutaneousmethod versus intraarterial measurement of ICG, we per-formed simultaneous measurements in a subpopulation. Inthis subpopulation, we had to conclude that, for individualmeasurement of cardiac output, the transcutaneous measure-ment of ICG by pulse dye densitometry is not accurateenough.4 For the purpose of this discussion, the measure-ments of cardiac output versus the ICG-PDR value in thepatients in whom we performed arterial blood sampling isshown in Figure 1. The open diamonds represent the patientsreceiving a propofol induction. Inspection of Figure 1 leads tothe conclusion that the absence or presence of propofol didnot induce significant hemodynamic changes in this other-wise healthy population nor did it affect ICG-PDR.

We thus maintain our conclusion4 that ICG-PDR valuesin a population without clinical signs of liver failure rangewell below 18% min�1, cited as the cutoff value for hepaticfailure and propagated as criterion for clinical intervention.This cutoff value needs to be reconsidered as has beensuggested elsewhere.5 We agree with the reviewers thatfurther studies are needed and that the final word on thissubject has not yet been written.

Marije Reekers, MDFred Boer, MD, PhDJaap Vuyk, MD, PhD

Department of AnesthesiologyLeiden University Medical Centre

Leiden, The [email protected]

Figure 1. Representation of cardiac output measurements based onarterial blood indocyanine green (ICG) concentrations versus ICG–plasma disappearance rate (PDR) values determined in patients whounderwent simultaneous arterial ICG sampling. The circles representthe measurements in awake subjects; the diamonds represent themeasurements during propofol induction.



REFERENCES1. Vos JJ, Scheeren TWL, Wietasch GJK. Pulse dye densitometry

and indocyanine green plasma disappearance: the issue of“normal” values. Anesth Analg 2010;111:1075–6

2. Sakka SG. Assessing liver function. Curr Opin Crit Care2007;13:207–14

3. Hultcrantz H, Glaumann H, Lindberg G, Nilsson LH. Liverinvestigation in 149 asymptomatic patients with moderatelyelevated activities of serum aminotransferases. Scand J Gastro-enterol 1986;21:109–13

4. Reekers M, Simon MJ, Boer F, Mooren FA, van Kleef JW, Dahan A,Vuyk J. Cardiovascular monitoring by pulse dye densitometry orarterial indocyanine green dilution. Anesth Analg 2009;109:441–6

5. Merle U, Sieg O, Stremmel W, Encke J, Eisenbach C. Sensitivityand specificity of plasma disappearance rate of indocyanine greenas a prognostic indicator in acute liver failure BMC Gastroenterol-ogy 2009;9:91

DOI: 10.1213/ANE.0b013e3181ef35e7

Serotonin Syndrome in the PerioperativePeriod: Role of Tramadol

To the Editor

Altman and Jahangiri1 describe an important com-plication associated with certain psychiatric aswell as other serotonergic medications used in

the perioperative period. However, in addition to thevarious drugs mentioned by the authors, an importantdrug also implicated in this setting is tramadol. It is acentrally acting analgesic frequently used for treatingmoderate to severe postoperative pain, especially inthird world countries. Tramadol, a weak agonist at the�-opioid receptor, also has a non-opioid mechanism ofaction that includes release of serotonin and inhibition ofreuptake of norepinephrine and is therefore likely tocontribute to the development of serotonin syndrome.2

Satinder Gombar, MDNidhi Bhatia, MD

Department of Anaesthesia & Intensive CareGovernment Medical College and Hospital

Chandigarh, [email protected]

REFERENCES1. Altman CS, Jahangiri MF. Serotonin syndrome in the perioper-

ative period. Anesth Analg 2010;110:526–82. Takeshita J, Litzinger MH. Serotonin syndrome associated with

tramadol. Prim Care Companion J Clin Psychiatry 2009;11:273DOI: 10.1213/ANE.0b013e3181eb02e8

Pressure-Rated Needleless AccessConnectors Slow IV Flow Rate

To the Editor

Our hospital recently added pressure-rated needlelessaccess connectors (PNACs) to all of our IV fluid setsto facilitate postoperative withdrawal of blood and

medication administration. Shortly thereafter, some of ouranesthesiologists noted decreased flow rate and asked justhow much fluid administration was diminished by these

devices. We sought to answer this question by measuring IVflow rates through 2 nearly identical IV setups. The onlydifference between the 2 systems was the presence(“PNAC setup”) or absence (“control setup”) of a PNACa

on the distal aspect of the IV tubingb (Fig. 1). Althoughour results agree with that of the PNAC manufacturerc for22-gauge catheter sets, we observed a substantially decreasedgravity-driven flow rate when the device was used with 14-and 16-gauge catheters (Table 1); device performance is notreported by the manufacturer for either of these large-borecatheters. Admittedly, these observations merely illustrate awell-known fact: flow rate is a function of various parame-ters,1–4 yet they also underscore the importance of criticallyevaluating modifications to IV fluid sets before wholesaleadoption. Our aim is not to discourage the use of PNACs, but

aMaxPlus® Clear (Maximus Medical, Ontario, CA).bLifeShield® Bifurcated Blood Set (Hospira, Morgan Hill, CA) and ClearlinkExtension Set (Baxter, Deerfield, IL).cAvailable at: http://www.maximusmedical.com/pdf/ML3084FlowRatePerformance.pdf. Accessed May 27, 2010.

Figure 1. Testing the effect of a pressure-rated needleless accessconnector (PNAC) on IV flow rate. Two identical saline bags with IVtubing of equal length were primed, suspended 7 feet above thecatheter tips, then allowed to simultaneously drain to gravity. Theonly difference between the 2 setups was the presence of a PNACon the distal aspect of the IV tubing attached to 1 bag but not theother. The time required for complete drainage of each saline bagwas measured by means of a stopwatch; flow rate was calculatedby dividing the volume of saline drained by time elapsed.

Table 1. The Effect of Pressure-Rated NeedlelessAccess Connectors on Gravity-Driven Flow RateThrough Catheters of Various Internal DiametersCatheter

parameterControl setup,

mL/minPNAC setup,

mL/minDifference, mL/min

(% change)22 gaugea 40.8 40.0 �0.8 (�2%)20 gaugea 54.5 56.8 �2.3 (�4%)16 gaugeb 214 121 �93 (�44%)14 gaugeb 223 138 �85 (�38%)

PNAC � pressure-rated needleless access connector.a Introcan Safety® IV Catheter (B. Braun Medical Inc., Bethlehem, PA).b CATHLON® IV Catheter (Smiths Medical North America, Dublin, OH).









DOI: 10.1213/ANE.0b013e3181ef35e7


To the Editor










To the Editor









mL/minPNAC setup,












DOI: 10.1213/ANE.0b013e3181ef35e7


To the Editor










To the Editor









mL/minPNAC setup,






rather to raise awareness of the flow impedance caused bythese (and other) IV devices.

ACKNOWLEDGMENTSLawrence J. Saidman, MD, is acknowledged for his guidance inthe preparation of this letter.

Jorge A. Caballero, MDStanford University School of Medicine

Stanford, CaliforniaFrain Rivera, MD

Joshua Edwards, MDJohn G. Brock-Utne, MD, PhD

Department of AnesthesiaStanford University School of Medicine

Stanford, [email protected]

REFERENCES1. Brown N, Duttchen KM, Caveno JW. An evaluation of flow

rates of normal saline through peripheral and central venouscatheters. American Society of Anesthesiologists Annual Meet-ing, Orlando. Anesthesiology 2008:A1484

2. Andersen HW, Benumof JL, Trousdale FR, Ozaki GT. Increasingthe functional gauge on the side port of large catheter sheathintroducers. Anesthesiology 1982;56:57–9

3. Benumof JL, Trousdale FR, Alfery DD, Ozaki GT. Largercatheter sheath introducers and their side port functional gauge.Anesth Analg 1981;60:216–7

4. Benumof JL, Wyte SR, Rogers SN. A large catheter sheathintroducer with an increased side-port functional gauge. CritCare Med 1983;11:660–2

DOI: 10.1213/ANE.0b013e3181f0948c

Anatomic Snuffbox RadialArtery CannulationTo the Editor

Failure of radial artery cannulation at the wrist,although uncommon, is not infrequent and is sec-ondary to vasospasm, hematoma formation, and

intimal dissection or thickening. We describe an alterna-tive successful ultrasound-guided approach to dorsalradial artery cannulation at the “anatomic snuffbox.” A55-year-old woman, with a medical history significantfor diabetes mellitus, coronary artery disease (postcoro-nary artery bypass surgery), and end-stage renal diseasereceiving hemodialysis via a left arm arteriovenousfistula, was scheduled for renal transplant. After induc-tion of general anesthesia, several attempts to percuta-neously insert a right radial artery catheter failed be-cause of hematoma formation. Cannulation of thebrachial and axillary arteries of the right arm wasconsidered but might have precluded future surgicalfistula formation. We decided not to cannulate the ulnarartery because of the radial artery hematoma. The radialartery was, however, palpable as the dorsal radial arteryin the anatomic snuffbox. Using ultrasound guidance,the diameter of the dorsal radial artery was measured at2.2 mm and a 22-gauge (0.9-mm outer diameter) cannulawas inserted into the snuffbox radial artery.

In 1982, Pyles et al.1 published their clinical experi-ence with cannulation of the dorsal radial artery. The

cannulation site in the anatomic snuffbox is distal to thedivision of the radial artery that provides collateral flow to thehand through the superficial palmar (volar) arch (Fig. 1).2

Cannulation distal to that separation would be expected toreduce the potential for digital ischemia. The dorsal radialartery does provide arterial flow to the deep palmar arch andcannulation of the same may be preferable to a proximal site,which could interfere with both the deep and superficialpalmar arch flows. For successful cannulation, we recom-mend semiprone position of hand (Fig. 2) and small-sizedcatheters. To conclude, cannulation of the dorsal radial arteryis a viable alternative to the radial artery at the wrist and otherpossible arterial access sites.

Krishnaprasad Deepika, MDDhamodaran Palaniappan, MD

Figure 1. Arterial anatomy of the hand.

Figure 2. Semiprone position of the hand for cannulation.



rather to raise awareness of the flow impedance caused bythese (and other) IV devices.

ACKNOWLEDGMENTSLawrence J. Saidman, MD, is acknowledged for his guidance inthe preparation of this letter.

Jorge A. Caballero, MDStanford University School of Medicine

Stanford, CaliforniaFrain Rivera, MD

Joshua Edwards, MDJohn G. Brock-Utne, MD, PhD

Department of AnesthesiaStanford University School of Medicine

Stanford, [email protected]

REFERENCES1. Brown N, Duttchen KM, Caveno JW. An evaluation of flow

rates of normal saline through peripheral and central venouscatheters. American Society of Anesthesiologists Annual Meet-ing, Orlando. Anesthesiology 2008:A1484

2. Andersen HW, Benumof JL, Trousdale FR, Ozaki GT. Increasingthe functional gauge on the side port of large catheter sheathintroducers. Anesthesiology 1982;56:57–9

3. Benumof JL, Trousdale FR, Alfery DD, Ozaki GT. Largercatheter sheath introducers and their side port functional gauge.Anesth Analg 1981;60:216–7

4. Benumof JL, Wyte SR, Rogers SN. A large catheter sheathintroducer with an increased side-port functional gauge. CritCare Med 1983;11:660–2

DOI: 10.1213/ANE.0b013e3181f0948c

Anatomic Snuffbox RadialArtery CannulationTo the Editor

Failure of radial artery cannulation at the wrist,although uncommon, is not infrequent and is sec-ondary to vasospasm, hematoma formation, and

intimal dissection or thickening. We describe an alterna-tive successful ultrasound-guided approach to dorsalradial artery cannulation at the “anatomic snuffbox.” A55-year-old woman, with a medical history significantfor diabetes mellitus, coronary artery disease (postcoro-nary artery bypass surgery), and end-stage renal diseasereceiving hemodialysis via a left arm arteriovenousfistula, was scheduled for renal transplant. After induc-tion of general anesthesia, several attempts to percuta-neously insert a right radial artery catheter failed be-cause of hematoma formation. Cannulation of thebrachial and axillary arteries of the right arm wasconsidered but might have precluded future surgicalfistula formation. We decided not to cannulate the ulnarartery because of the radial artery hematoma. The radialartery was, however, palpable as the dorsal radial arteryin the anatomic snuffbox. Using ultrasound guidance,the diameter of the dorsal radial artery was measured at2.2 mm and a 22-gauge (0.9-mm outer diameter) cannulawas inserted into the snuffbox radial artery.

In 1982, Pyles et al.1 published their clinical experi-ence with cannulation of the dorsal radial artery. The

cannulation site in the anatomic snuffbox is distal to thedivision of the radial artery that provides collateral flow to thehand through the superficial palmar (volar) arch (Fig. 1).2

Cannulation distal to that separation would be expected toreduce the potential for digital ischemia. The dorsal radialartery does provide arterial flow to the deep palmar arch andcannulation of the same may be preferable to a proximal site,which could interfere with both the deep and superficialpalmar arch flows. For successful cannulation, we recom-mend semiprone position of hand (Fig. 2) and small-sizedcatheters. To conclude, cannulation of the dorsal radial arteryis a viable alternative to the radial artery at the wrist and otherpossible arterial access sites.

Krishnaprasad Deepika, MDDhamodaran Palaniappan, MD

Figure 1. Arterial anatomy of the hand.

Figure 2. Semiprone position of the hand for cannulation.



Thomas Fuhrman, MDBruce Saltzman, MD

Department of Anesthesiology, Perioperative Medicine andPain Management

Jackson Memorial HospitalUniversity of Miami

Miami, [email protected]

REFERENCES1. Pyles ST, Scher KS, Vega ET, Harrah JD, Rubis LJ. Cannulation

of the dorsal radial artery: a new technique. Anesth Analg1982;61:876–8

2. Gray H, Lewis WH. Gray’s Anatomy of the Human Body. 20th USed. Philadelphia: Lea & Febiger, originally published in 1918

DOI: 10.1213/ANE.0b013e3181ef343a

Correction to the Case Report“Emergency Interventional LungAssist for Pulmonary Hypertension”Anesth Analg 2009;109:382–5To the Editor

We recently described our experience of the anes-thetic management of a patient using an emer-gency lung assist device.1 Although the above

report focused on the anesthetic considerations for that case, it

has been brought to our attention that some of the data thatwe reported have been included in a series of 4 casespublished earlier this year.2

We were not aware of this publication at the time ourmanuscript was accepted for publication and therefore wedid not cite it in our case report. Therefore, our communi-cation was inaccurate inasmuch as it was not the secondtime it has been used in an emergency. We would empha-size that the use of the device as described by Strueber etal.2 is a novel application. Using the Novalung for rightventricular support mandates central placement as de-scribed, regardless of the size of the patient.

Helen Holtby, MB, BS, FRCPCDirector of Cardiac Anesthesia

Hospital for Sick ChildrenToronto, Ontario, [email protected]

REFERENCES1. Taylor K, Holtby H. Emergency interventional lung assist for

pulmonary hypertension. Anesth Analg 2009;109:382–52. Strueber M, Hoeper MM, Fischer S, Cypel M, Warnecke G,

Gottlieb J, Pierre A, Welte T, Haverich A, Simon AR, KeshavjeeS. Bridge to thoracic organ transplantation in patients withpulmonary arterial hypertension using a pumpless lung assistdevice. Am J Transplant 2009;9:853–7

DOI: 10.1213/ANE.0b013e3181ec2bea



An Introductory Curriculum forUltrasound-Guided Regional Anesthesia

Brian A. Pollard Vincent W. S. Chan; illustrations byDiana Kryski; photographs by Michele Dalgarno. Toronto:B. A. Pollard, 2010. ISBN: 978-0-7727-8735-4. $95.00.

Forty years after Dr. Alon Winnie described the paresthe-sia technique for interscalene blockade, evolving tech-

nology enables clinicians to see and differentiate anatomicstructures in real time. The growing interest in ultrasound-guided regional anesthesia (UGRA) has been an exciting,landmark development in the practice of regional anesthe-sia and pain medicine. The authors’ of An IntroductoryCurriculum for Ultrasound-Guided Regional Anesthesia, Drs.Pollard and Chan, are well-known and accomplished lead-ers in this emerging field.

With this new technology comes responsibility. Spiriteddebate now surrounds the definition of core competencies,training requirements, competency assessment, institu-tional certification for UGRA practice, and strategies forquality improvement in UGRA. In writing these introduc-tory curriculums on UGRA, the authors have appropriatelyside-stepped these issues and focused instead on a moreimportant founding principle for their text—patient safety.

As is stated by the authors, the goal in writing this bookwas “to create an educational curriculum that will permit aself-directed foundational path for clinicians, from noviceto expert, community and academic-based, to build onexisting regional anesthesia expertise by integrating essen-tial ultrasound techniques into daily practice.” They accom-plish this task through nine chapters in five well-organizedsections.

Section 1 (39 pages) is the largest and constitutes almosthalf of the text. Starting with physics as applied to ultra-sonography, this section then confronts the fundamentalsof ultrasound scanning and needle techniques. Sections 2, 3,and 4 categorize nerve blocks into introductory, intermedi-ate, and advanced designations. Chapters within thesesections cover the most commonly performed ultrasound-guided nerve blocks. Finally, section 5 describes the use ofUGRA to assist with conventionally performed epiduraland subarachnoid blocks.

An ubiquitous debate continues among practitioners asto what constitutes an “introductory” block versus an“advanced” block, which can be seen as a potential weak-ness in this presentation. For example, many clinicians maynot find a selective block of the radial nerve “easier” than aconventional ultrasound-guided approach to the brachialplexus.

Each chapter ends with a salient summary, as well as 4to 6 key references for the self-directed learner. The funda-mental learning points of the chapter are then listed asuseful “knowledge keys.” Esthetics were not overlooked inthe production of the text, which presents superior illustra-tions and an excellent collection of images. Often, imagesand illustrations are combined into impressive and thoughtful

schematics detailing phenomena such as reverberation arti-fact. The high-quality images and illustrations are a majorstrength of the book.

Each chapter reads easily, yet with only 81 pages andlacking references, this is not meant to be a comprehensivetextbook on the topic. Rather, it nicely complements exist-ing resources, and as such represents one of the finestpresentations on basic UGRA concepts available to practi-tioners. The emphasis on competency with imaging isappropriate and is reflected in the fact that the bulk of thetext material addresses this skill set. Specific clinical recom-mendations, such as type of local anesthetic and volumeselection, are not emphasized.

As interest in the use of UGRA in clinical practice andeducation continues to expand, efforts to optimize patientsafety and minimize risk should be at the forefront. Al-though no evidence exists to support this claim, one couldreasonably hypothesize that poor UGRA technique mayincrease the risk of associated complications. To that end,this text is an essential resource for UGRA educators andwould be an excellent addition to all libraries on regionalanesthesia.

Jonathan C. Beathe, MDGregory A. Liguori, MDWeill Medical College of Cornell UniversityHospital for Special SurgeryNew York, New [email protected]

Visionaries and Dreamers: TheStory of Founding Fathers ofAnesthesiology in Israel

Gabrial M. Gurman. Beer-Sheva, Israel: Ben-Gurion Uni-versity of the Negev Press, 2008. ISBN: 978-9-6534-2963-5. 228 pages. $16.53.

Visionaries and Dreamers is a series of accounts aboutpioneers of Israeli anesthesiology. The author, Gabriel

M. Gurman, MD, is a Professor Emeritus of Anesthesiologyand Critical Care at Ben-Gurion University of the Negev,and both Hebrew and English translations were edited byLior Granot. Dr. Gurman provides the details of the lives ofinnovative anesthesiologists through a series of interviewsand is able to capture compelling stories of adversity andeventual triumphs. As the lives of these 12 anesthesiolo-gists are described, the reader is pulled into what seemslike an almost “sacred circle of leaders.” As an anesthesi-ologist, it was helpful to see how the book breaks down theelements of their leadership, and sheds light on each ofthese qualities separately: the importance of building avision, the value of being bold and unconstrained byestablished rules, and having the foresight to build a solid,trustworthy team. These are invaluable lessons that wouldhelp the professional careers of all aspiring anesthesiologists.


BOOK, MULTIMEDIA, AND MEETING REVIEWSSection Editor: Paul F. White

An Introductory Curriculum forUltrasound-Guided Regional Anesthesia

Brian A. Pollard Vincent W. S. Chan; illustrations byDiana Kryski; photographs by Michele Dalgarno. Toronto:B. A. Pollard, 2010. ISBN: 978-0-7727-8735-4. $95.00.

Forty years after Dr. Alon Winnie described the paresthe-sia technique for interscalene blockade, evolving tech-

nology enables clinicians to see and differentiate anatomicstructures in real time. The growing interest in ultrasound-guided regional anesthesia (UGRA) has been an exciting,landmark development in the practice of regional anesthe-sia and pain medicine. The authors’ of An IntroductoryCurriculum for Ultrasound-Guided Regional Anesthesia, Drs.Pollard and Chan, are well-known and accomplished lead-ers in this emerging field.

With this new technology comes responsibility. Spiriteddebate now surrounds the definition of core competencies,training requirements, competency assessment, institu-tional certification for UGRA practice, and strategies forquality improvement in UGRA. In writing these introduc-tory curriculums on UGRA, the authors have appropriatelyside-stepped these issues and focused instead on a moreimportant founding principle for their text—patient safety.

As is stated by the authors, the goal in writing this bookwas “to create an educational curriculum that will permit aself-directed foundational path for clinicians, from noviceto expert, community and academic-based, to build onexisting regional anesthesia expertise by integrating essen-tial ultrasound techniques into daily practice.” They accom-plish this task through nine chapters in five well-organizedsections.

Section 1 (39 pages) is the largest and constitutes almosthalf of the text. Starting with physics as applied to ultra-sonography, this section then confronts the fundamentalsof ultrasound scanning and needle techniques. Sections 2, 3,and 4 categorize nerve blocks into introductory, intermedi-ate, and advanced designations. Chapters within thesesections cover the most commonly performed ultrasound-guided nerve blocks. Finally, section 5 describes the use ofUGRA to assist with conventionally performed epiduraland subarachnoid blocks.

An ubiquitous debate continues among practitioners asto what constitutes an “introductory” block versus an“advanced” block, which can be seen as a potential weak-ness in this presentation. For example, many clinicians maynot find a selective block of the radial nerve “easier” than aconventional ultrasound-guided approach to the brachialplexus.

Each chapter ends with a salient summary, as well as 4to 6 key references for the self-directed learner. The funda-mental learning points of the chapter are then listed asuseful “knowledge keys.” Esthetics were not overlooked inthe production of the text, which presents superior illustra-tions and an excellent collection of images. Often, imagesand illustrations are combined into impressive and thoughtful

schematics detailing phenomena such as reverberation arti-fact. The high-quality images and illustrations are a majorstrength of the book.

Each chapter reads easily, yet with only 81 pages andlacking references, this is not meant to be a comprehensivetextbook on the topic. Rather, it nicely complements exist-ing resources, and as such represents one of the finestpresentations on basic UGRA concepts available to practi-tioners. The emphasis on competency with imaging isappropriate and is reflected in the fact that the bulk of thetext material addresses this skill set. Specific clinical recom-mendations, such as type of local anesthetic and volumeselection, are not emphasized.

As interest in the use of UGRA in clinical practice andeducation continues to expand, efforts to optimize patientsafety and minimize risk should be at the forefront. Al-though no evidence exists to support this claim, one couldreasonably hypothesize that poor UGRA technique mayincrease the risk of associated complications. To that end,this text is an essential resource for UGRA educators andwould be an excellent addition to all libraries on regionalanesthesia.

Jonathan C. Beathe, MDGregory A. Liguori, MDWeill Medical College of Cornell UniversityHospital for Special SurgeryNew York, New [email protected]

Visionaries and Dreamers: TheStory of Founding Fathers ofAnesthesiology in Israel

Gabrial M. Gurman. Beer-Sheva, Israel: Ben-Gurion Uni-versity of the Negev Press, 2008. ISBN: 978-9-6534-2963-5. 228 pages. $16.53.

Visionaries and Dreamers is a series of accounts aboutpioneers of Israeli anesthesiology. The author, Gabriel

M. Gurman, MD, is a Professor Emeritus of Anesthesiologyand Critical Care at Ben-Gurion University of the Negev,and both Hebrew and English translations were edited byLior Granot. Dr. Gurman provides the details of the lives ofinnovative anesthesiologists through a series of interviewsand is able to capture compelling stories of adversity andeventual triumphs. As the lives of these 12 anesthesiolo-gists are described, the reader is pulled into what seemslike an almost “sacred circle of leaders.” As an anesthesi-ologist, it was helpful to see how the book breaks down theelements of their leadership, and sheds light on each ofthese qualities separately: the importance of building avision, the value of being bold and unconstrained byestablished rules, and having the foresight to build a solid,trustworthy team. These are invaluable lessons that wouldhelp the professional careers of all aspiring anesthesiologists.


BOOK, MULTIMEDIA, AND MEETING REVIEWSSection Editor: Paul F. White

Dr. Gurman began his career with a generation ofphysicians who had already survived a horrific time. Manyof Dr. Gurman’s friends and colleagues narrowly escapedthe front lines of the extermination camps of World War IIthrough unimaginable acts of courage. These physicianswere in some cases the only survivors of their entirefamilies. Most of them migrated to the new state of Israelshortly after its establishment in1948. In the beginning, thescarcity of anesthesiologists in the new State of Israel meantthat one anesthesiologist might cover an entire hospital.This is epitomized in the accounts of Dr. Thomas Gesztes,who had his own incarcerated hernia repaired under localanesthesia, and who afterwards anesthetized his patientswhile taking call duties the same night as his own surgery.This is an example of the ultimate “ambulatory” surgeryexperience! Each one of these anesthesiologists was instru-mental in building departments (and societies) of anesthe-sia from “the ground up.”

When Dr. Gurman migrated to Israel in 1972, anesthesiaclinical practices were still in a rapid flux from whatseemed archaic practices. The transformation was nothingshort of miraculous, and clinical anesthesia care is nowconsidered among the most advanced specialties in thefield of medicine. In his book, Dr. Gurman interrupts hisstorytelling periodically to offer vignettes and historicalcommentaries on the various personalities and their per-sonal and professional stories. Thus, the action of the bookdoes not proceed chronologically; instead, it moves backand forth in time, depending on which colleague he ismemorializing. This strategy of tying present-day individualsto events in the past is very helpful to those readers who maynot be very familiar with the geography and history of Israel.

There are many stories of success and significant ad-vances in clinical care in the field of anesthesiology, most ofwhich were achieved through hard work, perseverance,dedication, and an unwillingness to succumb to the “statusquo.” However, anesthesia was far from ideal in Israel,especially from an academic and professional standpoint.This is illustrated by Dr. Shamay Cotev, head of the firstintensive care unit (ICU) in Israel, who is quoted in thebook as having said, “In retrospect, I wouldn’t have goneinto anesthesia.” This is perhaps the only part of this brightand optimistic book that touches on the harsh reality thatmuch more work is still needed—in anesthesia and inmedicine. Currently, �2% of the graduates of medicalschools in Israel choose to pursue postgraduate training inanesthesiology!

In this well-written book, Dr. Gurman takes the readerinto the middle of Israeli anesthesia history. The tremen-dous strides of the past 50 years are apparent, but thesummit is yet to be achieved. Dr. Gurman’s book providesus with direction regarding where we need to be in thefuture. This book is dedicated to the 60th anniversary of theState of Israel, and is recommended reading for young andold anesthesiologists not only for its historical value, butmore importantly, for the optimistic glimpse it providesinto the future of anesthesiology, medicine, and humanity.

Sorin J. Brull, MDJoseph A. Cartwright, MDMayo Clinic College of MedicineJacksonville, [email protected]

Books, Multimedia, and Meeting Reviews


Documents

Anesthesia Analgesia Oct_2010