1.Special ArticleSurviving Sepsis Campaign: International guidelines for management of severe sepsis and septic shock: 2008 R. Phillip Dellinger, MD; Mitchell M. Levy, MD; Jean M. Carlet, MD; Julian Bion, MD; Margaret M. Parker, MD; Roman Jaeschke, MD; Konrad Reinhart, MD; Derek C. Angus, MD, MPH; Christian Brun-Buisson, MD; Richard Beale, MD; Thierry Calandra, MD, PhD; Jean-Francois Dhainaut, MD; Herwig Gerlach, MD; Maurene Harvey, RN; John J. Marini, MD; John Marshall, MD; Marco Ranieri, MD; Graham Ramsay, MD; Jonathan Sevransky, MD; B. Taylor Thompson, MD; Sean Townsend, MD; Jeffrey S. Vender, MD; Janice L. Zimmerman, MD; Jean-Louis Vincent, MD, PhD; for the International Surviving Sepsis Campaign Guidelines Committee Objective: To provide an update to the original Surviving Sepsis Campaignpressure is identied to be poorly responsive to uid and vasopressor therapy clinical management guidelines, Surviving Sepsis Campaign Guidelines for Man-(2C); recombinant activated protein C in patients with severe sepsis and clinical agement of Severe Sepsis and Septic Shock, published in 2004.assessment of high risk for death (2B except 2C for postoperative patients). In theDesign: Modied Delphi method with a consensus conference of 55 interna- absence of tissue hypoperfusion, coronary artery disease, or acute hemorrhage, tional experts, several subsequent meetings of subgroups and key individuals, target a hemoglobin of 79 g/dL (1B); a low tidal volume (1B) and limitation of teleconferences, and electronic-based discussion among subgroups and amonginspiratory plateau pressure strategy (1C) for acute lung injury (ALI)/acute respi- the entire committee. This process was conducted independently of any industryratory distress syndrome (ARDS); application of at least a minimal amount of funding.positive end-expiratory pressure in acute lung injury (1C); head of bed elevation inMethods: We used the Grades of Recommendation, Assessment, Development mechanically ventilated patients unless contraindicated (1B); avoiding routine use and Evaluation (GRADE) system to guide assessment of quality of evidence from of pulmonary artery catheters in ALI/ARDS (1A); to decrease days of mechanical high (A) to very low (D) and to determine the strength of recommendations. Aventilation and ICU length of stay, a conservative uid strategy for patients with strong recommendation (1) indicates that an interventions desirable effectsestablished ALI/ARDS who are not in shock (1C); protocols for weaning and clearly outweigh its undesirable effects (risk, burden, cost) or clearly do not. Weak sedation/analgesia (1B); using either intermittent bolus sedation or continuous recommendations (2) indicate that the tradeoff between desirable and undesirableinfusion sedation with daily interruptions or lightening (1B); avoidance of neuro- effects is less clear. The grade of strong or weak is considered of greater clinical muscular blockers, if at all possible (1B); institution of glycemic control (1B), importance than a difference in letter level of quality of evidence. In areas without targeting a blood glucose 65 mm Hg (1C); dobutamine inotropic therapy when cardiac Grades of Recommendation, Assessment, Development and Evaluation criteria; output remains low despite uid resuscitation and combined inotropic/vasopres-GRADE; guidelines; evidence-based medicine; Surviving Sepsis Campaign; sepsis sor therapy (1C); stress-dose steroid therapy given only in septic shock after bloodbundlesFrom Cooper University Hospital, Camden, NJ (RPD);Italy (MR); West Hertfordshire Health Trust, Hemel Hemp- Infection Society,* World Federation of Societies of Intensive Rhode Island Hospital, Providence, RI (MML); Hospital Saint- stead, UK (GR); The Johns Hopkins University School of and Critical Care Medicine.** Participation and endorsement Joseph, Paris, France (JMC); Birmingham University, Bir- Medicine, Baltimore, MD (JS); Massachusetts General Hos- by the German Sepsis Society and the Latin American Sepsis mingham, UK (JB); SUNY at Stony Brook, Stony Brook, NY pital, Boston, MA (BTT); Rhode Island Hospital, Providence, RI Institute. *Sponsor of 2004 guidelines. **Sponsors of 2008 (MMP); McMaster University, Hamilton, Ontario, Canada (RJ);(ST); Evanston Northwestern Healthcare, Evanston, IL (JSV);guidelines who did not participate formally in revision pro- Friedrich-Schiller-University of Jena, Jena, Germany (KR); The Methodist Hospital, Houston, TX (JLZ); Erasme University cess. Members of the 2008 SSC Guidelines Committee are University of Pittsburgh, Pittsburgh, PA (DCA); Hopital HenriHospital, Brussels, Belgium (JLV). listed in Appendix I. Appendix J provides author disclosureSponsoring organizations: American Association of Crit-information. Mondor, Crteil, France (CBB); Guys and St Thomas Hos-ical-Care Nurses,* American College of Chest Physicians,*Also published in Intensive Care Medicine (January pital Trust, London, UK (RB); Centre Hospitalier Universitaire2008). Vaudois, Lausanne, Switzerland (TC); French Agency for American College of Emergency Physicians,* Canadian Crit-ical Care Society, European Society of Clinical Microbiology For information regarding this article, E-mail: Evaluation of Research and Higher Education, Paris, FranceDellinger-Phil@CooperHealth.eduand Infectious Diseases,* European Society of Intensive Care (JFD); Vivantes-Klinikum Neukoelin, Berlin, Germany (HG);Medicine,* European Respiratory Society,* International Sep- Copyright 2007 by the Society of Critical Care Consultants in Critical Care, Inc, Glenbrook, NV (MH); Univer- sis Forum,* Japanese Association for Acute Medicine, Jap-Medicine sity of Minnesota, St. Paul, MN (JJM); St. Michaels Hospital, anese Society of Intensive Care Medicine; Society of Critical Toronto, Ontario, Canada (JM); Universit di Torino, Torino, Care Medicine,* Society of Hospital Medicine,** Surgical DOI: 10.1097/01.CCM.0000298158.12101.41 296 Crit Care Med 2008 Vol. 36, No. 1

2. Severe sepsis (acute organ dys- Table 1. Determination of the quality of evidencefunction secondary to infec- Underlying methodologytion) and septic shock (severe A. RCTsepsis plus hypotension not re-B. Downgraded RCT or upgraded observational studies versed with uid resuscitation) are major C. Well-done observational studies healthcare problems, affecting millions ofD. Case series or expert opinion individuals around the world each year, Factors that may decrease the strength of evidence 1. Poor quality of planning and implementation of available RCTs, suggesting high likelihood of killing one in four (and often more), andbias increasing in incidence (15). Similar to 2. Inconsistency of results (including problems with subgroup analyses) polytrauma, acute myocardial infarction,3. Indirectness of evidence (differing population, intervention, control, outcomes, comparison) or stroke, the speed and appropriateness4. Imprecision of results 5. High likelihood of reporting bias of therapy administered in the initial Main factors that may increase the strength of evidence hours after severe sepsis develops are1. Large magnitude of effect (direct evidence, RR 2 with no plausible confounders) likely to inuence outcome. In 2004, an 2. Very large magnitude of effect with RR 5 and no threats to validity (by two levels) international group of experts in the di- 3. Dose-response gradient agnosis and management of infection andRCT, randomized controlled trial; RR, relative risk. sepsis, representing 11 organizations, published the rst internationally ac- cepted guidelines that the bedside clini- Table 2. Factors determining strong vs. weak recommendation cian could use to improve outcomes in What Should Be ConsideredRecommended Process severe sepsis and septic shock (6, 7). These guidelines represented phase II ofQuality of evidenceThe lower the quality of evidence, the less likely a strong the Surviving Sepsis Campaign (SSC), an recommendation international effort to increase awarenessRelative importance of the If values and preferences vary widely, a strong and improve outcomes in severe sepsis.outcomesrecommendation becomes less likely Baseline risks of outcomes The higher the risk, the greater the magnitude of benet Joined by additional organizations, the Magnitude of relative risk,Larger relative risk reductions or larger increases in group met again in 2006 and 2007 to including benets, harms, and relative risk of harm make a strong recommendation update the guidelines document using aburdenmore or less likely, respectively new evidence-based methodology system Absolute magnitude of the effect The larger the absolute benets and harms, the greater or for assessing quality of evidence and lesser likelihood, respectively, of a strong strength of recommendations (8 11).recommendation These recommendations are intendedPrecision of the estimates of theThe greater the precision, the more likely a strong effects recommendation to provide guidance for the clinician car-CostsThe higher the cost of treatment, the less likely a strong ing for a patient with severe sepsis or recommendation septic shock. Recommendations from these guidelines cannot replace the clini- cians decision-making capability when he or she is provided with a patientsthis dysfunction has varied somewhat from dated search into 2007 (see following meth- unique set of clinical variables. Most of one severe sepsis research study to another.ods and rules). these recommendations are appropriate An example of typical thresholds identica- The 2001 guidelines were coordinated for the severe sepsis patient in both the tion of severe sepsis is shown in Table 2 by the International Sepsis Forum; the intensive care unit (ICU) and non-ICU (13). Sepsis-induced hypotension is dened2004 guidelines were funded by unre- settings. In fact, the committee believes as a systolic blood pressure (SBP) 90 mmstricted educational grants from industry that currently, the greatest outcome im-Hg or mean arterial pressure 70 mm Hg and administered through the Society of provement can be made through educa-or a SBP decrease 40 mm Hg or 2 SDCritical Care Medicine (SCCM), the Eu- tion and process change for those caringbelow normal for age in the absence ofropean Society of Intensive Care Medi- for severe sepsis patients in the non-ICU other causes of hypotension. Septic shock cine (ESICM), and the International Sep- setting and across the spectrum of acuteis dened as sepsis-induced hypotension sis Forum. Two of the SSC administering care. It should also be noted that re-persisting despite adequate uid resuscita- organizations receive unrestricted indus- source limitations in some institutions tion. Sepsis-induced tissue hypoperfusion try funding to support SSC activities (ES- and countries may prevent physiciansis dened as either septic shock, an elevated ICM and SCCM), but none of this funding from accomplishing particular recom-lactate, or oliguria. was used to support the 2006/2007 com- mendations. The current clinical practice guidelinesmittee meetings. build on the rst and second editions fromIt is important to distinguish between 2001 (discussed subsequently) and 2004 (6,the process of guidelines revision and the METHODS 7, 14). The 2001 publication incorporated a SSC. The SSC is partially funded by un- MEDLINE search for clinical trials in the restricted educational industry grants, Sepsis is dened as infection plus sys- preceding 10 yrs, supplemented by a man-including those from Edwards Life- temic manifestations of infection (Tableual search of other relevant journals (14). Sciences, Eli Lilly and Company, and 1) (12). Severe sepsis is dened as sepsisThe 2004 publication incorporated the ev- Philips Medical Systems. SSC also re- plus sepsis-induced organ dysfunction oridence available through the end of 2003. ceived funding from the Coalition for tissue hypoperfusion. The threshold for The current publication is based on an up-Critical Care Excellence of the Society ofCrit Care Med 2008 Vol. 36, No. 1 297 3. Critical Care Medicine. The great major- ic-related search or recent trials. Quality all discussions and deliberations among ity of industry funding has come from Eliof evidence was judged by predened the guidelines committee members as to Lilly and Company. Grades of Recommendation, Assessment, grading decisions. Subsequently, the SSC Current industry funding for the SSC Development and Evaluation (GRADE)authors used written material prepared is directed to the performance improve-criteria (discussed subsequently). Signif-by the GRADE group and conferred with ment initiative. No industry funding was icant education of committee membersGRADE group members who were avail- used in the guidelines revision process. on the GRADE approach was performed able at the rst committee meeting and For both the 2004 and the 2006/2007via e-mail before the rst committeesubsequent nominal group meetings. efforts, there were no members of themeeting and at the rst meeting. RulesGRADE representatives were also used as committee from industry, no industry were distributed concerning assessing a resource throughout subgroup deliber- input into guidelines development, and the body of evidence, and GRADE experts ation. no industry presence at any of the were available for questions throughout The GRADE system is based on a se- meetings. Industry awareness or com- the process. Subgroups agreed electroni-quential assessment of the quality of ev- ment on the recommendations was notcally on draft proposals that were pre- idence, followed by assessment of the bal- allowed. No member of the guidelinesented to committee meetings for generalance between benets vs. risks, burden, committee received any honoraria for discussion. In January 2006, the entire and cost and, based on the preceding, de- any role in the 2004 or 2006/2007group met during the 35th SCCM Critical velopment and grading of a management guidelines process. The committee con- Care Congress in San Francisco, Califor-recommendations (9 11). Keeping the rat- sidered the issue of recusement of indi- nia. The results of that discussion wereing of quality of evidence and strength of vidual committee members during de-incorporated into the next version of rec-recommendation explicitly separate consti- liberation and decision making in areasommendations and again discussed usingtutes a crucial and dening feature of the where committee members had either electronic mail. Recommendations were GRADE approach. This system classies nancial or academic competing inter-nalized during nominal group meetingsquality of evidence as high (grade A), mod- ests; however, consensus as to thresh- (composed of a subset of the committeeerate (grade B), low (grade C), or very low old for exclusion could not be reached.members) at the 2007 SCCM (Orlando, (grade D). Randomized trials begin as high- Alternatively, the committee agreed to FL) and 2007 International Symposiumquality evidence but may be downgraded ensure full disclosure and transparencyon Intensive Care and Emergency Medi- due to limitations in implementation, in- of all committee members potentialcine (Brussels) meetings with recircula-consistency or imprecision of the results, conicts at time of publication. (Seetion of deliberations and decisions to theindirectness of the evidence, and possible disclosures at the end of this docu- entire group for comment or approval. Atreporting bias (Table 1). Examples of indi- ment.) the discretion of the chair and following rectness of the evidence include population The guidelines process included aadequate discussion, competing propos-studied, interventions used, outcomes mea- modied Delphi method, a consensus als for wording of recommendations or sured, and how these relate to the question conference, several subsequent meetingsassigning strength of evidence were re- of interest. Observational (nonrandomized) of subgroups and key individuals, tele-solved by formal voting. On occasions,studies begin as low-quality evidence, but conferences and electronic-based discus- voting was performed to give the com- the quality level may be upgraded on the sions among subgroups and members of mittee a sense of distribution of opinionsbasis of large magnitude of effect. An exam- the entire committee, and two follow-upto facilitate additional discussion. Theple of this is the quality of evidence for early nominal group meetings in 2007.manuscript was edited for style and formadministration of antibiotics. Subgroups were formed, each chargedby the writing committee with nal ap-The GRADE system classies recom- with updating recommendations in spe-proval by section leads for their respec- mendations as strong (grade 1) or weak cic areas, including corticosteroids, tive group assignment and then by the (grade 2). The grade of strong or weak is blood products, activated protein C, renal entire committee. considered of greater clinical importance replacement therapy, antibiotics, source The development of guidelines and than a difference in letter level of quality control, and glucose control. Each sub-grading of recommendations for the 2004 of evidence. The committee assessed group was responsible for updating the guideline development process werewhether the desirable effects of adher- evidence (into 2007, with major addi-based on a system proposed by Sackett ence will outweigh the undesirable ef- tional elements of information incorpo-(15) in 1989, during one of the rstfects, and the strength of a recommenda- rated into the evolving manuscript American College of Chest Physicianstion reects the groups degree of throughout 2006 and 2007). A separate(ACCP) conferences on the use of anti-condence in that assessment. A strong search was performed for each clearly de-thrombotic therapies. The revised guide-recommendation in favor of an interven- ned question. The committee chair lines recommendations are based on thetion reects that the desirable effects of worked with subgroup heads to identify GRADE system, a structured system for adherence to a recommendation (bene- pertinent search terms that always in- rating quality of evidence and gradingcial health outcomes, less burden on staff cluded, at a minimum, sepsis, severe sep-strength of recommendation in clinicaland patients, and cost savings) will sis, septic shock, and sepsis syndrome practice (8 11). The SSC Steering Com- clearly outweigh the undesirable effects crossed against the general topic area ofmittee and individual authors collabo-(harms, more burden, and greater costs). the subgroup as well as pertinent keyrated with GRADE representatives to ap- A weak recommendation in favor of an words of the specic question posed. All ply the GRADE system to the SSC intervention indicates that the desirable questions of the previous guidelines pub-guidelines revision process. The mem- effects of adherence to a recommendation lications were searched, as were pertinent bers of GRADE group were directly in- probably will outweigh the undesirable new questions generated by general top-volved, either in person or via e-mail, ineffects, but the panel is not condent298 Crit Care Med 2008 Vol. 36, No. 1 4. about these tradeoffs either because Table 3. Initial resuscitation and infection issues some of the evidence is low quality (and Strength of recommendation and quality of evidence have been assessed using the GRADE criteria, thus there remains uncertainty regarding presented in parentheses after each guideline the benets and risks) or the benets and Indicates a strong recommendation, or we recommend downsides are closely balanced. While the Indicates a weak recommendation, or we suggest degree of condence is a continuum andInitial resuscitation (rst 6 hrs) there is no precise threshold between a Begin resuscitation immediately in patients with hypotension or elevated serum lactate 4 strong and a weak recommendation, the mmol/L; do not delay pending ICU admission (1C) Resuscitation goals (1C) presence of important concerns about CVP 812 mm Hga one or more of the preceding factors Mean arterial pressure 65 mm Hg makes a weak recommendation more Urine output 0.5 mL kg 1 hr 1 likely. A strong recommendation is Central venous (superior vena cava) oxygen saturation 70% or mixed venous 65% If venous oxygen saturation target is not achieved (2C) worded as we recommend and a weakConsider further uid recommendation as we suggest.Transfuse packed red blood cells if required to hematocrit of 30% and/or The implications of calling a recom- Start dobutamine infusion, maximum 20 g kg 1 min 1 mendation strong are that most well-Diagnosis informed patients would accept that inter- Obtain appropriate cultures before starting antibiotics provided this does not signicantly delay antimicrobial administration (1C) vention and that most clinicians should useObtain two or more BCs it in most situations. There may be circum-One or more BCs should be percutaneous stances in which a strong recommendation One BC from each vascular access device in place 48 hrs cannot or should not be followed for anCulture other sites as clinically indicated individual patient because of that patients Perform imaging studies promptly to conrm and sample any source of infection, if safe to do so (1C) Antibiotic therapy preferences or clinical characteristics that Begin intravenous antibiotics as early as possible and always within the rst hour of make the recommendation less applicable.recognizing severe sepsis (1D) and septic shock (1B) Being a strong recommendation does not Broad-spectrum: one or more agents active against likely bacterial/fungal pathogens and with automatically imply standard of care. For good penetration into presumed source (1B) example, the strong recommendation for Reassess antimicrobial regimen daily to optimize efcacy, prevent resistance, avoid toxicity, administering antibiotics within 1 hr of theand minimize costs (1C) Consider combination therapy in Pseudomonas infections (2D) diagnosis of severe sepsis, although desir- Consider combination empiric therapy in neutropenic patients (2D) able, is not currently standard of care asCombination therapy 35 days and de-escalation following susceptibilities (2D) veried by current practice (M Levy, per- Duration of therapy typically limited to 710 days; longer if response is slow or there are sonal communication, from rst 8,000 pa-undrainable foci of infection or immunologic deciencies (1D) tients entered internationally into the SSC Stop antimicrobial therapy if cause is found to be noninfectious (1D) Source identication and control performance improvement database). The A specic anatomic site of infection should be established as rapidly as possible (1C) and implication of a weak recommendation is within rst 6 hrs of presentation (1D) that although a majority of well-informed Formally evaluate patient for a focus of infection amenable to source control measures (e.g. patients would accept it (but a substantial abscess drainage, tissue debridement) (1C) proportion would not), clinicians should Implement source control measures as soon as possible following successful initial consider its use according to particular cir- resuscitation (1C) (exception: infected pancreatic necrosis, where surgical intervention is best delayed) (2B) cumstance. Choose source control measure with maximum efcacy and minimal physiologic upset (1D) Differences of opinion among commit- Remove intravascular access devices if potentially infected (1C) tee members about interpretation of evi- dence, wording of proposals, or strength of GRADE, Grades of Recommendation, Assessment, Development and Evaluation; ICU, intensive recommendations were resolved using a care unit; CVP, central venous pressure; BC, blood culture. a specically developed set of rules. We will A higher target CVP of 1215 mm Hg is recommended in the presence of mechanical ventilation describe this process in detail in a separate or preexisting decreased ventricular compliance. publication. In summary, the main ap- proach for converting diverse opinions into a recommendation was as follows: 1) tobreadth of review. The entire review grouptense focused discussion (nominal group) give a recommendation a direction (for or (together with their parent organizations with broader review and monitoring using against the given action), a majority ofas required) participated in the larger, iter-the Delphi process. votes were to be in favor of that direction,ative, modied Delphi process. The smaller Note: Refer to Tables 35 for con- with 20% preferring the opposite direc- working group meetings, which took placedensed adult recommendations. tion (there was a neutral vote allowed as in person, functioned as the nominal well); 2) to call a given recommendationgroups. If a clear consensus could not be strong rather than weak, 70% strong obtained by polling within the nominalI. MANAGEMENT OF SEVERE votes were required; 3) if 70% of votes group meetings, the larger group was spe- SEPSIS indicated strong preference, the recom- cically asked to use the polling process. mendation was assigned a weak category of This was only required for corticosteroidsA. Initial Resuscitation strength. We used a combination of modi-and glycemic control. The larger group had ed Delphi process and nominal (expert) the opportunity to review all outputs. In 1. We recommend the protocolized re- group techniques to ensure both depth and this way the entire review combined in-suscitation of a patient with sepsis-Crit Care Med 2008 Vol. 36, No. 1 299 5. Table 4. Hemodynamic support and adjunctive therapy gets to be equivalent. Either intermittent or continuous measurements of oxygen Strength of recommendation and quality of evidence have been assessed using the GRADE criteria, saturation were judged to be acceptable. presented in parentheses after each guideline. Indicates a strong recommendation, or we recommendAlthough blood lactate concentrationIndicates a weak recommendation, or we suggest may lack precision as a measure of tissue Fluid therapy metabolic status, elevated levels in sepsis Fluid-resuscitate using crystalloids or colloids (1B) support aggressive resuscitation. In me- Target a CVP of 8 mm Hg ( 12 mm Hg if mechanically ventilated) (1C) Use a uid challenge technique while associated with a hemodynamic improvement (1D) chanically ventilated patients or patients Give uid challenges of 1000 mL of crystalloids or 300500 mL of colloids over 30 mins. Morewith known preexisting decreased ven-rapid and larger volumes may be required in sepsis-induced tissue hypoperfusion (1D) tricular compliance, a higher target cen- Rate of uid administration should be reduced if cardiac lling pressures increase withouttral venous pressure of 1215 mm Hg isconcurrent hemodynamic improvement (1D)recommended to account for the imped- Vasopressors Maintain MAP 65 mm Hg (1C) iment to lling (17). Similar consider- Norepinephrine and dopamine centrally administered are the initial vasopressors of choice (1C)ation may be warranted in circumstancesEpinephrine, phenylephrine, or vasopressin should not be administered as the initial of increased abdominal pressure or dia-vasopressor in septic shock (2C). Vasopressin 0.03 units/min may be subsequently added tostolic dysfunction (18). Elevated centralnorepinephrine with anticipation of an effect equivalent to norepinephrine alone venous pressures may also be seen withUse epinephrine as the rst alternative agent in septic shock when blood pressure is poorly preexisting clinically signicant pulmo-responsive to norepinephrine or dopamine (2B). Do not use low-dose dopamine for renal protection (1A)nary artery hypertension. Although the In patients requiring vasopressors, insert an arterial catheter as soon as practical (1D) cause of tachycardia in septic patients Inotropic therapy may be multifactorial, a decrease in ele- Use dobutamine in patients with myocardial dysfunction as supported by elevated cardiac vated pulse rate with uid resuscitation islling pressures and low cardiac output (1C) often a useful marker of improving intra- Do not increase cardiac index to predetermined supranormal levels (1B) Steroidsvascular lling. Recently published ob-Consider intravenous hydrocortisone for adult septic shock when hypotension responds poorlyservational studies have demonstrated anto adequate uid resuscitation and vasopressors (2C) association between good clinical out-ACTH stimulation test is not recommended to identify the subset of adults with septic shockcome in septic shock and MAP 65 mmwho should receive hydrocortisone (2B) Hg as well as central venous oxygen sat-Hydrocortisone is preferred to dexamethasone (2B)Fludrocortisone (50 g orally once a day) may be included if an alternative to hydrocortisone uration (ScvO2, measured in superioris being used that lacks signicant mineralocorticoid activity. Fludrocortisone if optional if vena cava, either intermittently or con-hydrocortisone is used (2C)tinuously) of 70% (19). Many recentSteroid therapy may be weaned once vasopressors are no longer required (2D)studies support the value of early proto- Hydrocortisone dose should be 300 mg/day (1A) colized resuscitation in severe sepsis and Do not use corticosteroids to treat sepsis in the absence of shock unless the patients sepsis-induced tissue hypoperfusion (20 endocrine or corticosteroid history warrants it (1D) Recombinant human activated protein C 25). Studies of patients with shock indi-Consider rhAPC in adult patients with sepsis-induced organ dysfunction with clinical cate that mixed venous oxygen saturationassessment of high risk of death (typically APACHE II 25 or multiple organ failure) if there (SVO2) runs 57% lower than central ve-are no contraindications (2B, 2C for postoperative patients).nous oxygen saturation (ScvO2) (26) and Adult patients with severe sepsis and low risk of death (typically, APACHE II 20 or one that an early goal-directed resuscitationorgan failure) should not receive rhAPC (1A) protocol can be established in a nonre-GRADE, Grades of Recommendation, Assessment, Development and Evaluation; CVP, centralsearch general practice venue (27). venous pressure; MAP, mean arterial pressure; ACTH, adrenocorticotropic hormone; rhAPC, recom-There are recognized limitations to binant human activated protein C; APACHE, Acute Physiology and Chronic Health Evaluation. ventricular lling pressure estimates as surrogates for uid resuscitation (28, 29). However, measurement of central venousinduced shock, dened as tissue hypo- Urine output 0.5 mLkg 1hr 1 pressure is currently the most readily ob-perfusion (hypotension persisting Central venous (superior vena tainable target for uid resuscitation.after initial uid challenge or blood cava) or mixed venous oxygen sat-There may be advantages to targetinglactate concentration 4 mmol/L).uration 70% or 65%, respec-uid resuscitation to ow and perhaps toThis protocol should be initiated astively (grade 1C)volumetric indices (and even to microcir-soon as hypoperfusion is recognizedRationale. Early goal-directed resusci-culation changes) (30 33). Technologiesand should not be delayed pendingtation has been shown to improve sur-currently exist that allow measurementICU admission. During the rst 6 hrsvival for emergency department patientsof ow at the bedside (34, 35). Futureof resuscitation, the goals of initialpresenting with septic shock in a ran- goals should be making these technolo-resuscitation of sepsis-induced hypo-perfusion should include all of the fol-domized, controlled, single-center study gies more accessible during the criticallowing as one part of a treatment pro-(16). Resuscitation directed toward theearly resuscitation period and research totocol:previously mentioned goals for the initial validate utility. These technologies areCentral venous pressure 8 12 mm6-hr period of the resuscitation was ablealready available for early ICU resuscita-Hgto reduce 28-day mortality rate. The con-tion.Mean arterial pressure (MAP) 65 sensus panel judged use of central venous2. We suggest that during the rst 6 hrsmm Hg and mixed venous oxygen saturation tar- of resuscitation of severe sepsis or sep-300Crit Care Med 2008 Vol. 36, No. 1 6. Table 5. Other supportive therapy of severe sepsisStrength of recommendation and quality of evidence have been assessed using the GRADE criteria, presented in parentheses after each guideline Indicates a strong recommendation, or we recommendIndicates a weak recommendation, or we suggest Blood product administration Give red blood cells when hemoglobin decreases to 7.0 g/dL ( 70 g/L) to target a hemoglobin of 7.09.0 g/dL in adults (1B). A higherhemoglobin level may be required in special circumstances (e.g., myocardial ischaemia, severe hypoxemia, acute hemorrhage, cyanotic heartdisease, or lactic acidosis)Do not use erythropoietin to treat sepsis-related anemia. Erythropoietin may be used for other accepted reasons (1B)Do not use fresh frozen plasma to correct laboratory clotting abnormalities unless there is bleeding or planned invasive procedures (2D) Do not use antithrombin therapy (1B)Administer platelets when (2D)Counts are 5000/mm3 (5 109/L) regardless of bleedingCounts are 500030,000/mm3 (530 109/L) and there is signicant bleeding riskHigher platelet counts ( 50,000/mm3 [50 109/L]) are required for surgery or invasive procedures Mechanical ventilation of sepsis-induced ALI/ARDS Target a tidal volume of 6 mL/kg (predicted) body weight in patients with ALI/ARDS (1B) Target an initial upper limit plateau pressure 30 cm H2O. Consider chest wall compliance when assessing plateau pressure (1C) Allow PaCO2 to increase above normal, if needed, to minimize plateau pressures and tidal volumes (1C) Set PEEP to avoid extensive lung collapse at end-expiration (1C)Consider using the prone position for ARDS patients requiring potentially injurious levels of FIO2 or plateau pressure, provided they are not putat risk from positional changes (2C) Maintain mechanically ventilated patients in a semirecumbent position (head of the bed raised to 45) unless contraindicated (1B), between 30and 45 (2C)Noninvasive ventilation may be considered in the minority of ALI/ARDS patients with mild to moderate hypoxemic respiratory failure. Thepatients need to be hemodynamically stable, comfortable, easily arousable, able to protect/clear their airway, and expected to recover rapidly (2B) Use a weaning protocol and an SBT regularly to evaluate the potential for discontinuing mechanical ventilation (1A) SBT options include a low level of pressure support with continuous positive airway pressure 5 cm H2O or a T piece Before the SBT, patients should be arousable be hemodynamically stable without vasopressors have no new potentially serious conditions have low ventilatory and end-expiratory pressure requirement require FIO2 levels that can be safely delivered with a face mask or nasal cannula Do not use a pulmonary artery catheter for the routine monitoring of patients with ALI/ARDS (1A) Use a conservative uid strategy for patients with established ALI who do not have evidence of tissue hypoperfusion (1C) Sedation, analgesia, and neuromuscular blockade in sepsis Use sedation protocols with a sedation goal for critically ill mechanically ventilated patients (1B) Use either intermittent bolus sedation or continuous infusion sedation to predetermined end points (sedation scales), with dailyinterruption/lightening to produce awakening. Re-titrate if necessary (1B) Avoid neuromuscular blockers where possible. Monitor depth of block with train-of-four when using continuous infusions (1B) Glucose control Use intravenous insulin to control hyperglycemia in patients with severe sepsis following stabilization in the ICU (1B) Aim to keep blood glucose 150 mg/dL (8.3 mmol/L) using a validated protocol for insulin dose adjustment (2C) Provide a glucose calorie source and monitor blood glucose values every 12 hrs (4 hrs when stable) in patients receiving intravenous insulin (1C) Interpret with caution low glucose levels obtained with point of care testing, as these techniques may overestimate arterial blood or plasmaglucose values (1B) Renal replacementIntermittent hemodialysis and CVVH are considered equivalent (2B)CVVH offers easier management in hemodynamically unstable patients (2D) Bicarbonate therapy Do not use bicarbonate therapy for the purpose of improving hemodynamics or reducing vasopressor requirements when treating hypoperfusion-induced lactic acidemia with pH 7.15 (1B) Deep vein thrombosis prophylaxis Use either low-dose UFH or LMWH, unless contraindicated (1A) Use a mechanical prophylactic device, such as compression stockings or an intermittent compression device, when heparin is contraindicated (1A)Use a combination of pharmacologic and mechanical therapy for patients who are at very high risk for deep vein thrombosis (2C)In patients at very high risk, LMWH should be used rather than UFH (2C) Stress ulcer prophylaxis Provide stress ulcer prophylaxis using H2 blocker (1A) or proton pump inhibitor (1B). Benets of prevention of upper gastrointestinal bleed mustbe weighed against the potential for development of ventilator-acquired pneumonia Consideration for limitation of support Discuss advance care planning with patients and families. Describe likely outcomes and set realistic expectations (1D) GRADE, Grades of Recommendation, Assessment, Development and Evaluation; ALI, acute lung injury; ARDS, acute respiratory distress syndrome; PEEP, positive end-expiratory pressure; SBT, spontaneous breathing trial; ICU, intensive care unit; CVVH, continuous veno-venous hemoltration; UFH, unfractionated heparin; LMWH, low-molecular weight heparin. tic shock, if ScvO2 or SVO2 of 70% or hematocrit of 30% and/or adminis-Rationale. The protocol used in the65%, respectively, is not achieved with tration of a dobutamine infusion (upstudy cited previously targeted an increaseuid resuscitation to the central ve- to a maximum of 20 gkg 1min 1)in ScvO2 to 70% (16). This was achievednous pressure target, then transfusionbe used to achieve this goal (grade by sequential institution of initial uid re-of packed red blood cells to achieve a2C).suscitation, packed red blood cells, andCrit Care Med 2008 Vol. 36, No. 1 301 7. then dobutamine. This protocol was asso- cular access device. Obtaining bloodBalancing risk and benet is therefore ciated with an improvement in survival.cultures peripherally and through a mandatory in those settings. Based on bedside clinical assessment and vascular access device is an important personal preference, a clinician may deemstrategy. If the same organism is recov-C. Antibiotic Therapy either blood transfusion (if hematocrit is ered from both cultures, the likelihood 30%) or dobutamine the best initialthat the organism is causing the severe1. We recommend that intravenous an- choice to increase oxygen delivery and sepsis is enhanced. In addition, if the tibiotic therapy be started as early as thereby elevate ScvO2, when uid resusci-culture drawn through the vascular ac-possible and within the rst hour of tation is believed to be already adequate. cess device is positive much earlier than recognition of septic shock (1B) and The design of the aforementioned trial did the peripheral blood culture (i.e., 2 hrs severe sepsis without septic shock not allow assessment of the relative contri- earlier), the data support the concept that (1D). Appropriate cultures should be bution of these two components (i.e., in-the vascular access device is the source of obtained before initiating antibiotic creasing oxygen content or increasing car- the infection (37). Quantitative cultures therapy but should not prevent diac output) of the protocol onof catheter and peripheral blood are also prompt administration of antimicro- achievement of improved outcome. useful for determining whether the cath-bial therapy (grade 1D).eter is the source of infection. Volume ofRationale. Establishing vascular ac- B. Diagnosis blood drawn with the culture tube shouldcess and initiating aggressive uid resus- 1. We recommend obtaining appropri-be 10 mL (38). Quantitative (or semi- citation are the rst priority when man-ate cultures before antimicrobial quantitative) cultures of respiratory tract aging patients with severe sepsis or septictherapy is initiated if such cultures secretions are recommended for the di-shock. However, prompt infusion of anti-do not cause signicant delay in an-agnosis of ventilator-associated pneumo-microbial agents should also be a prioritytibiotic administration. To optimizenia (39). Gram-negative stain can be use- and may require additional vascular ac-identication of causative organisms, ful, in particular for respiratory tractcess ports (42, 43). In the presence ofwe recommend at least two blood specimens, to help decide the micro-septic shock, each hour delay in achiev-cultures be obtained before antibiot- organisms to be targeted. The potential ing administration of effective antibioticsics with at least one drawn percuta-role of biomarkers for diagnosis of in- is associated with a measurable increaseneously and one drawn through eachfection in patients presenting with se- in mortality (42). If antimicrobial agentsvascular access device, unless the de-vere sepsis remains undened. The pro-cannot be mixed and delivered promptlyvice was recently ( 48 hrs) inserted. calcitonin level, although often useful,from the pharmacy, establishing a supplyCultures of other sites (preferably is problematic in patients with an acuteof premixed antibiotics for such urgentquantitative where appropriate), such inammatory pattern from othersituations is an appropriate strategy foras urine, cerebrospinal uid, wounds, causes (e.g., postoperative, shock) (40). ensuring prompt administration. Inrespiratory secretions, or other body In the near future, rapid diagnosticchoosing the antimicrobial regimen, cli-uids that may be the source of in- nicians should be aware that some anti-methods (polymerase chain reaction,fection should also be obtained be- microbial agents have the advantage ofmicro-arrays) might prove extremelyfore antibiotic therapy if not associ-bolus administration, while others re-helpful for a quicker identication ofated with signicant delay in quire a lengthy infusion. Thus, if vascularpathogens and major antimicrobial re-antibiotic administration (grade 1C). access is limited and many differentsistance determinants (41).agents must be infused, bolus drugs may Rationale. Although sampling should2. We recommend that imaging studiesoffer an advantage. not delay timely administration of antibi-be performed promptly in attempts otics in patients with severe sepsis (e.g.,2a. We recommend that initial empirical to conrm a potential source of in-anti-infective therapy include one or lumbar puncture in suspected meningi- fection. Sampling of potential tis), obtaining appropriate cultures be- more drugs that have activity against sources of infection should occur as all likely pathogens (bacterial and/or fore administration of antibiotics is es- they are identied; however, some sential to conrm infection and thefungal) and that penetrate in ade- patients may be too unstable to war- quate concentrations into the pre- responsible pathogens and to allow de-rant certain invasive procedures or escalation of antibiotic therapy after re- sumed source of sepsis (grade 1B). transport outside of the ICU. Bedside ceipt of the susceptibility prole. Samples studies, such as ultrasound, are use-Rationale. The choice of empirical an- can be refrigerated or frozen if processing ful in these circumstances (gradetibiotics depends on complex issues re- cannot be performed immediately. Im-1C). lated to the patients history, including mediate transport to a microbiological drug intolerances, underlying disease, lab is necessary. Because rapid steriliza- Rationale. Diagnostic studies may the clinical syndrome, and susceptibility tion of blood cultures can occur within aidentify a source of infection that re- patterns of pathogens in the community, few hours after the rst antibiotic dose,quires removal of a foreign body or drain-in the hospital, and that previously have obtaining those cultures before starting age to maximize the likelihood of a sat-been documented to colonize or infect therapy is essential if the causative organ- isfactory response to therapy. However, the patient. There is an especially wide ism is to be identied. Two or more bloodeven in the most organized and well-range of potential pathogens for neutro- cultures are recommended (36). In pa-staffed healthcare facilities, transport of penic patients. tients with indwelling catheters (for 48 patients can be dangerous, as can placing Recently used antibiotics should gener- hrs), at least one blood culture should be patients in outside-unit imaging devicesally be avoided. When choosing empirical drawn through each lumen of each vas-that are difcult to access and monitor.therapy, clinicians should be cognizant of302 Crit Care Med 2008 Vol. 36, No. 1 8. the virulence and growing prevalence ofopment of antimicrobial resistance or to sponse, undrainable foci of infection, oxacillin (methicillin)-resistant Staphylo-reduce cost is not an appropriate initialor immunologic deciencies, includ- coccus aureus (ORSA or MRSA) in some strategy in this patient population, onceing neutropenia (grade 1D). communities and healthcare settings (espe- the causative pathogen has been identi- 4. We recommend that if the presenting cially in the United States). If the preva-ed, it may become apparent that none of clinical syndrome is determined to be lence is signicant, and in consideration of the empirical drugs offers optimal ther- due to a noninfectious cause, antimi- the virulence of this organism, empiricalapy; that is, there may be another drugcrobial therapy be stopped promptly therapy adequate for this pathogen would proven to produce superior clinical out- to minimize the likelihood that the be warranted. Clinicians should also con-come that should therefore replace em- patient will become infected with an sider whether candidemia is a likely patho-pirical agents.antibiotic-resistant pathogen or will gen when choosing initial therapy. WhenNarrowing the spectrum of antibiotic develop a drug-related adverse effect deemed warranted, the selection of empir-coverage and reducing the duration of(grade 1D). ical antifungal therapy (e.g., uconazole, antibiotic therapy will reduce the likelihoodRationale. Clinicians should be cogni- amphotericin B, or echinocandin) will be that the patient will develop superinfection zant that blood cultures will be negative tailored to the local pattern of the mostwith pathogenic or resistant organisms,in 50% of cases of severe sepsis or sep- prevalent Candida species and any priorsuch as Candida species, Clostridium dif- tic shock, yet many of these cases are very administration of azoles drugs (44). Riskcile, or vancomycin-resistant Enterococcus likely caused by bacteria or fungi. Thus, factors for candidemia should also be con- faecium. However, the desire to minimize the decisions to continue, narrow, or stop sidered when choosing initial therapy. superinfections and other complicationsantimicrobial therapy must be made on Because patients with severe sepsis or should not take precedence over the need the basis of clinician judgment and clin- septic shock have little margin for errorto give the patient an adequate course ofical information. in the choice of therapy, the initial selec- therapy to cure the infection that caused tion of antimicrobial therapy should bethe severe sepsis or septic shock. broad enough to cover all likely patho- D. Source Control gens. There is ample evidence that failure 2c. We suggest combination therapy for to initiate appropriate therapy (i.e., ther- patients with known or suspected 1a. We recommend that a specic ana- apy with activity against the pathogen Pseudomonas infections as a cause of tomical diagnosis of infection requir- that is subsequently identied as thesevere sepsis (grade 2D).ing consideration for emergent causative agent) correlates with increased 2d. We suggest combination empirical source control (e.g., necrotizing fas- morbidity and mortality (45 48).therapy for neutropenic patients withciitis, diffuse peritonitis, cholangitis, Patients with severe sepsis or septicsevere sepsis (grade 2D).intestinal infarction) be sought and shock warrant broad-spectrum therapy 2e. When used empirically in patientsdiagnosed or excluded as rapidly as until the causative organism and its an- with severe sepsis, we suggest thatpossible (grade 1C) and within the tibiotic susceptibilities are dened. Re-combination therapy should not berst 6 hrs following presentation striction of antibiotics as a strategy toadministered for 35 days. De- (grade 1D). reduce the development of antimicrobialescalation to the most appropriate 1b. We further recommend that all pa- resistance or to reduce cost is not an single therapy should be performed tients presenting with severe sepsis appropriate initial strategy in this patient as soon as the susceptibility prole isbe evaluated for the presence of a population.known (grade 2D).focus on infection amenable to All patients should receive a full load- Rationale. Although no study or meta-source control measures, specically ing dose of each antimicrobial. However, analysis has convincingly demonstrated the drainage of an abscess or local patients with sepsis or septic shock often that combination therapy produces a supe-focus on infection, the debridement have abnormal renal or hepatic functionrior clinical outcome for individual patho-of infected necrotic tissue, the re- and may have abnormal volumes of dis-gens in a particular patient group, combi- moval of a potentially infected device, tribution due to aggressive uid resusci-nation therapies do produce in vitro or the denitive control of a source tation. Drug serum concentration moni- synergy against pathogens in some models of ongoing microbial contamination toring can be useful in an ICU setting for(grade 1C). (Appendix A provides ex-(although such synergy is difcult to dene those drugs that can be measuredamples of potential sites needingand predict). In some clinical scenarios, promptly. An experienced physician or source control.)such as the two preceding, combination clinical pharmacist should be consulted2. We suggest that when infectedtherapies are biologically plausible and are to ensure that serum concentrations are peripancreatic necrosis is identied aslikely clinically useful even if evidence has attained that maximize efcacy and min- a potential source of infection, deni-not demonstrated improved clinical out- imize toxicity (49 52).tive intervention is best delayed untilcome (5356). Combination therapy for adequate demarcation of viable and 2b. We recommend that the antimicro- suspected known Pseudomonas pending nonviable tissues has occurred (grade bial regimen be reassessed daily tosensitivities increases the likelihood that at 2B). optimize activity, to prevent the de-least one drug is effective against that3. We recommend that when source con- velopment of resistance, to reduce strain and positively affects outcome (57).trol is required, the effective intervention toxicity, and to reduce costs (grade3. We recommend that the duration ofassociated with the least physiologic in- 1C). therapy typically be 710 days; longer sult be employed (e.g., percutaneous Rationale. Although restriction of an- courses may be appropriate in pa-rather than surgical drainage of an tibiotics as a strategy to reduce the devel- tients who have a slow clinical re-abscess (grade 1D).Crit Care Med 2008 Vol. 36, No. 1 303 9. 4. We recommend that when intravas- or crystalloids. There is no evidence-volume decit in patients with severe sepsis cular access devices are a possiblebased support for one type of uidvaries. With venodilation and ongoing cap- source of severe sepsis or septicover another (grade 1B).illary leak, most patients require continu- shock, they be promptly removed af-Rationale. The SAFE study indicated ing aggressive uid resuscitation during ter other vascular access has been that albumin administration was safe andthe rst 24 hrs of management. Input is established (grade 1C).typically much greater than output, andequally as effective as crystalloid (65). There Rationale. The principals of sourceinput/output ratio is of no utility to judgewas an insignicant decrease in mortality control in the management of sepsis in-uid resuscitation needs during this timerates with the use of colloid in a subset clude a rapid diagnosis of the specic siteperiod.analysis of septic patients (p .09). Previ- of infection and identication of a focusous meta-analyses of small studies of ICU on infection amenable to source controlpatients had demonstrated no difference F. Vasopressors measures (specically the drainage of an between crystalloid and colloid uid resus- abscess, debridement of infected necroticcitation (66 68). Although administration1. We recommend that mean arterial tissue, removal of a potentially infectedof hydroxyethyl starch may increase the pressure (MAP) be maintained 65 device, and denitive control of a sourcerisk of acute renal failure in patients withmm Hg (grade 1C). of ongoing microbial contamination)sepsis, variable ndings preclude denitive Rationale. Vasopressor therapy is re- (58). Foci of infection readily amenable torecommendations (69, 70). As the volume quired to sustain life and maintain perfu- source control measures include an in- of distribution is much larger for crystal- sion in the face of life-threatening hypo- tra-abdominal abscess or gastrointestinalloids than for colloids, resuscitation with tension, even when hypovolemia has not perforation, cholangitis or pyelonephri- crystalloids requires more uid to achieveyet been resolved. Below a certain mean tis, intestinal ischemia or necrotizing soft the same end points and results in more arterial pressure, autoregulation in vari- tissue infection, and other deep space in- edema. Crystalloids are less expensive. ous vascular beds can be lost, and perfu- fection, such as an empyema or septicsion can become linearly dependent on arthritis. Such infectious foci should be 2. We recommend that uid resuscita-pressure. Thus, some patients may re- controlled as soon as possible following tion initially target a central venousquire vasopressor therapy to achieve a successful initial resuscitation (59), ac- pressure of 8 mm Hg (12 mm Hg inminimal perfusion pressure and maintain complishing the source control objective mechanically ventilated patients).adequate ow (71, 72). The titration of with the least physiologic upset possibleFurther uid therapy is often re-norepinephrine to as low as MAP 65 mm (e.g., percutaneous rather than surgical quired (grade 1C).Hg has been shown to preserve tissue drainage of an abscess [60], endoscopic3a. We recommend that a uid challengeperfusion (72). In addition, preexisting rather than surgical drainage of biliary technique be applied wherein uidcomorbidities should be considered as to tree), and removing intravascular access administration is continued as longmost appropriate MAP target. For exam- devices that are potentially the source of as the hemodynamic improvementple, a MAP of 65 mm Hg might be too low severe sepsis or septic shock promptly (e.g., arterial pressure, heart rate, in a patient with severe uncontrolled hy- after establishing other vascular access urine output) continues (grade 1D). pertension, and in a young previously (61, 62). A randomized, controlled trial 3b. We recommend that uid challengenormotensive, a lower MAP might be comparing early vs. delayed surgical in- in patients with suspected hypovole-adequate. Supplementing end points, tervention for peripancreatic necrosis mia be started with 1000 mL ofsuch as blood pressure, with assess- showed better outcomes with a delayedcrystalloids or 300 500 mL of col- ment of regional and global perfusion, approach (63). However, areas of uncer-loids over 30 mins. More rapid ad-such as blood lactate concentrations tainty exist, such as denitive docu-ministration and greater amounts of and urine output, is important. Ade- mentation of infection and appropriate uid may be needed in patients with quate uid resuscitation is a fundamen- length of delay. The selection of optimalsepsis-induced tissue hypoperfusion tal aspect of the hemodynamic manage- source control methods must weigh(see Initial Resuscitation recommen-ment of patients with septic shock and benets and risks of the specic inter-dations) (grade 1D).should ideally be achieved before vaso- vention as well as risks of transfer (64). 3c. We recommend that the rate of uidpressors and inotropes are used, but Source control interventions may cause administration be reduced substan-using vasopressors early as an emer- further complications, such as bleed-tially when cardiac lling pressuresgency measure in patients with severe ing, stulas, or inadvertent organ in- (central venous pressure or pulmo-shock is frequently necessary. When jury. Surgical intervention should benary artery balloon-occluded pres-that occurs, great effort should be di- considered when lesser interventionalsure) increase without concurrent rected to weaning vasopressors with approaches are inadequate or when di-hemodynamic improvement (gradecontinuing uid resuscitation. agnostic uncertainty persists despite ra-1D). diologic evaluation. Specic clinical sit-2. We recommend either norepineph- Rationale. Fluid challenge must be rine or dopamine as the rst choice uations require consideration ofclearly separated from simple uid ad-vasopressor agent to correct hypo- available choices, patients preferences,ministration; it is a technique in whichtension in septic shock (administered and clinicians expertise.large amounts of uids are administered through a central catheter as soon asover a limited period of time under close one is available) (grade 1C). E. Fluid Therapymonitoring to evaluate the patients re-3a. We suggest that epinephrine, phenyl-1. We recommend uid resuscitationsponse and avoid the development of pul-ephrine, or vasopressin should not be with either natural/articial colloids monary edema. The degree of intravascular administered as the initial vasopres-304 Crit Care Med 2008 Vol. 36, No. 1 10. sor in septic shock (grade 2C). Vaso-septic shock, but with continued shockG. Inotropic Therapy pressin 0.03 units/min may be addedthe concentration decreases to normal to norepinephrine subsequently withrange in the majority of patients between 1. We recommend that a dobutamine in- anticipation of an effect equivalent to24 and 48 hrs (95). This has been called fusion be administered in the presence that of norepinephrine alone.relative vasopressin deciency because inof myocardial dysfunction as sug- 3b. We suggest that epinephrine be the the presence of hypotension, vasopressin gested by elevated cardiac lling pres- rst chosen alternative agent in sep-would be expected to be elevated. Thesures and low cardiac output (grade tic shock that is poorly responsive to signicance of this nding is unknown. 1C). norepinephrine or dopamine (gradeThe recent VASST trial, a randomized, 2. We recommend against the use of a 2B). controlled trial comparing norepineph- strategy to increase cardiac index torine alone to norepinephrine plus vaso-predetermined supranormal levels Rationale. There is no high-quality (grade 1B). primary evidence to recommend one cat- pressin at 0.03 units/min, showed no dif- echolamine over another. Much litera-ference in outcome in the intent to treat Rationale. Dobutamine is the rst- ture exists that contrasts the physiologic population. An a priori dened subgroup choice inotrope for patients with mea- effects of choice of vasopressor and com-analysis showed that the survival of pa-sured or suspected low cardiac output in bined inotrope/vasopressors in septictients receiving 15 g/min norepineph- the presence of adequate left ventricular shock (73 85). Human and animal stud- rine at the time of randomization was lling pressure (or clinical assessment of ies suggest some advantages of norepi- better with vasopressin. However, the adequate uid resuscitation) and ade- nephrine and dopamine over epinephrine pretrial rationale for this stratication quate mean arterial pressure. Septic pa- (the latter with the potential for tachy-was based on exploring potential benet tients who remain hypotensive after uid cardia as well as disadvantageous effectsin the 15 g norepinephrine require- resuscitation may have low, normal, or on splanchnic circulation and hyper- ment population. Higher doses of vaso-increased cardiac outputs. Therefore, lactemia) and phenylephrine (decrease in pressin have been associated with car-treatment with a combined inotrope/ stroke volume). There is, however, nodiac, digital, and splanchnic ischemia andvasopressor, such as norepinephrine or clinical evidence that epinephrine results should be reserved for situations where dopamine, is recommended if cardiac in worse outcomes, and it should be thealternative vasopressors have failed (96).output is not measured. When the capa- rst chosen alternative to dopamine or Cardiac output measurement to allow bility exists for monitoring cardiac out-maintenance of a normal or elevated ow put in addition to blood pressure, a vaso- norepinephrine. Phenylephrine is the ad-is desirable when these pure vasopressors pressor, such as norepinephrine, may be renergic agent least likely to produceare instituted. used separately to target specic levels of tachycardia but as a pure vasopressor 5. We recommend that low-dose dopa-mean arterial pressure and cardiac out- would be expected to decrease stroke vol-mine not be used for renal protection put. Two large prospective clinical trials ume. Dopamine increases mean arterialthat included critically ill ICU patients pressure and cardiac output, primarily (grade 1A).who had severe sepsis failed to demon- due to an increase in stroke volume and Rationale. A large randomized trialstrate benet from increasing oxygen de- heart rate. Norepinephrine increases and meta-analysis comparing low-doselivery to supranormal targets by use of mean arterial pressure due to its vasocon- dopamine to placebo found no difference dobutamine (99, 100). These studies did strictive effects, with little change in in either primary outcomes (peak serumnot specically target patients with se- heart rate and less increase in stroke vol-creatinine, need for renal replacement, vere sepsis and did not target the rst 6 ume compared with dopamine. Either urine output, time to recovery of normalhrs of resuscitation. The rst 6 hrs of may be used as a rst-line agent to correctrenal function) or secondary outcomes resuscitation of sepsis-induced hypoper- hypotension in sepsis. Norepinephrine is (survival to either ICU or hospital dis-fusion need to be treated separately from more potent than dopamine and may be charge, ICU stay, hospital stay, arrhyth- the later stages of severe sepsis (see Ini- more effective at reversing hypotension in mias) (97, 98). Thus, the available data do tial Resuscitation recommendations). patients with septic shock. Dopamine may not support administration of low doses be particularly useful in patients with com- of dopamine solely to maintain renal promised systolic function but causes more function. H. Corticosteroids tachycardia and may be more arrhythmo- genic (86). It may also inuence the endo-6. We recommend that all patients re-1. We suggest that intravenous hydro- crine response via the hypothalamic- quiring vasopressors have an arterialcortisone be given only to adult septic pituitary axis and have immunosuppressivecatheter placed as soon as practical ifshock patients after it has been con- effects. resources are available (grade 1D).rmed that their blood pressure is Vasopressin levels in septic shock have Rationale. In shock states, estimationpoorly responsive to uid resuscita- been reported to be lower than antici- of blood pressure using a cuff is com- tion and vasopressor therapy (grade pated for a shock state (87). Low doses of monly inaccurate; use of an arterial can-2C). vasopressin may be effective in raisingnula provides a more appropriate and re- Rationale. One French multicenter, blood pressure in patients refractory to producible measurement of arterialrandomized controlled trial (RCT) of pa- other vasopressors and may have otherpressure. These catheters also allow con- tients in vasopressor-unresponsive septic potential physiologic benets (88 93).tinuous analysis so that decisions regard-shock (hypotension despite uid resuscita- Terlipressin has similar effects but is long ing therapy can be based on immediate tion and vasopressors) showed a signicant lasting (94). Studies show that vasopres-and reproducible blood pressure informa-shock reversal and reduction of mortality sin concentrations are elevated in early tion. rate in patients with relative adrenal insuf-Crit Care Med 2008 Vol. 36, No. 1 305 11. ciency (dened as postadrenocortico-cortisol (protein-bound and free) while(110). It remains uncertain whether out- tropic hormone [ACTH] cortisol increasefree cortisol is the pertinent measure-come is affected by tapering of steroids.9 g/dL) (101). Two additional smaller ment. The relationship between free and6. We recommend that doses of cortico- RCTs also showed signicant effects on total cortisol varies with serum proteinsteroids comparable to 300 mg of shock reversal with steroid therapy (102,concentration. When compared with a hydrocortisone daily not be used in 103). However, a recent large, Europeanreference method (mass spectrometry), severe sepsis or septic shock for the multicenter trial (CORTICUS), which hascortisol immunoassays may over- or un-purpose of treating septic shock been presented in abstract form but not yetderestimate the actual cortisol level, af-(grade 1A). published, failed to show a mortality benet fecting the assignment of patients to re- with steroid therapy of septic shock (104). Rationale. Two randomized prospec-sponders or nonresponders (105). CORTICUS did show a faster resolution oftive clinical trials and a meta-analysesAlthough the clinical signicance is not concluded that for therapy of severe sep- septic shock in patients who received ste- clear, it is now recognized that etomi- roids. The use of the ACTH test (responders sis or septic shock, high-dose corticoste-date, when used for induction for intuba-roid therapy is ineffective or harmful and nonresponders) did not predict the tion, will suppress the hypothalamic- faster resolution of shock. Importantly, un-(111113). Reasons to maintain higherpituitary-adrenal axis (106).doses of corticosteroid for medical condi- like the French trial, which only enrolled shock patients with blood pressure unre- 3. We suggest that patients with septictions other than septic shock may exist. sponsive to vasopressor therapy, the COR- shock should not receive dexametha- 7. We recommend that corticosteroids TICUS study included patients with septic sone if hydrocortisone is availablenot be administered for the treatment shock, regardless of how the blood pressure (grade 2B).of sepsis in the absence of shock. responded to vasopressors. Although corti-Rationale. Although often proposed There is, however, no contraindication costeroids do appear to promote shock re-for use until an ACTH stimulation testto continuing maintenance steroid versal, the lack of a clear improvement in can be administered, we no longer sug-therapy or to using stress-dose ste- mortality coupled with known side ef- gest an ACTH test in this clinical situa- roids if the patients endocrine or cor- fects of steroids, such as increased risk of tion (see the preceding point 3). Further-ticosteroid administration history infection and myopathy generally tem- more, dexamethasone can lead to warrants (grade 1D). pered enthusiasm for their broad use. Thus,immediate and prolonged suppression of Rationale. No studies exist that specif- there was broad agreement that the recom-the hypothalamic-pituitary-adrenal axisically target severe sepsis in the absence of mendation should be downgraded from the shock and offer support for use of stressafter administration (107). previous guidelines (Appendix B). There doses of steroids in this patient population. was considerable discussion and consider-4. We suggest the daily addition of oral Steroids may be indicated in the presence ation by the committee on the option of udrocortisone (50 g) if hydrocorti- of a history of steroid therapy or adrenal encouraging use in those patients whose sone is not available and the steroid dysfunction. A recent preliminary study of blood pressure was unresponsive to uidsthat is substituted has no signicant stress-dose level steroids in community- and vasopressors, while strongly discourag- mineralocorticoid activity. Fludrocor-acquired pneumonia is encouraging but ing use in subjects whose shock responded tisone is considered optional if hydro- needs conrmation (114). well to uids and pressors. However, this cortisone is used (grade 2C). more complex set of recommendations was Rationale. One study added 50 g of rejected in favor of the preceding single I. Recombinant Humanudrocortisone orally (101). Since hydro-Activated Protein C (rhAPC) recommendation (Appendix B).cortisone has intrinsic mineralocorticoid 2. We suggest that the ACTH stimulationactivity, there is controversy as to 1. We suggest that adult patients with sep-test not be used to identify the subset whether udrocortisone should be added. sis-induced organ dysfunction associ-of adults with septic shock who shouldated with a clinical assessment of high5. We suggest that clinicians wean thereceive hydrocortisone (grade 2B).risk of death, most of whom will have patient from steroid therapy when va- Rationale. Although one study sug- Acute Physiology and Chronic Health sopressors are no longer required gested those who did not respond toEvaluation (APACHE) II 25 or multi- (grade 2D). ACTH with a brisk surge in cortisol (fail- ple organ failure, receive rhAPC if there ure to achieve or 9 g/dL increase inRationale. There has been no compar- are no contraindications (grade 2B ex- cortisol 30 60 mins after ACTH admin-ative study between a xed-duration and cept for patients within 30 days of sur- istration) were more likely to benetclinically guided regimen or between ta-gery, for whom it is grade 2C). Relative from steroids than those who did re- pering and abrupt cessation of steroids.contraindications should also be consid- spond, the overall trial population ap-Three RCTs used a xed-duration proto-ered in decision making. peared to benet regardless of ACTH re-col for treatment (101, 103, 104), and in2. We recommend that adult patients sult, and the observation of a potential two RCTs, therapy was decreased after with severe sepsis and low risk of interaction between steroid use andshock resolution (102, 108). In four RCTs death, most of whom will have ACTH test was not statistically signicant steroids were tapered over several days APACHE II 20 or one organ failure, (101). Furthermore, there was no evi-(102104, 108), and in two RCTs (101, do not receive rhAPC (grade 1A). dence of this distinction between re-109) steroids were withdrawn abruptly.Rationale. The evidence concerning sponders and nonresponders in a recent One crossover study showed hemody- use of rhAPC in adults is primarily based on multicenter trial (104). Commonly used namic and immunologic rebound effectstwo RCTs: PROWESS (1,690 adult patients, cortisol immunoassays measure totalafter abrupt cessation of corticosteroidsstopped early for efcacy) (115) and AD-306 Crit Care Med 2008 Vol. 36, No. 1 12. DRESS (stopped early for futility) (116).rhAPC and prescribing information for emia, severe hypoxemia, acute hemor- Additional safety information comes from relative contraindications.)rhage, cyanotic heart disease, or lactic an open-label observational study, EN- Intracranial hemorrhage (ICH) oc- acidosis (see recommendations for ini- HANCE (117). The ENHANCE trial alsocurred in the PROWESS trial in 0.1% tial resuscitation), we recommend that suggested that early administration of (placebo) and 0.2% (APC) (not signi- red blood cell transfusion occur when rhAPC was associated with better out-cant) (106); in the ADDRESS trial 0.4%hemoglobin decreases to 7.0 g/dL comes. (placebo) vs. 0.5 % (APC) (not signicant)( 70 g/L) to target a hemoglobin of PROWESS involved 1,690 patients and(116); and in ENHANCE 1.5% (108). Reg-7.0 9.0 g/dL (70 90 g/L) in adults documented 6.1% in absolute total mor- istry studies of rhAPC report higher(grade 1B). tality reduction with a relative risk reduc- bleeding rates than randomized con- Rationale. Although the optimum he- tion of 19.4%, 95% condence intervaltrolled trials, suggesting that the risk ofmoglobin for patients with severe sepsis 6.6 30.5%, and number needed to treat bleeding in actual practice may be greater has not been specically investigated, the 16 (115). Controversy associated with thethan reported in PROWESS and AD- Transfusion Requirements in Critical results focused on a number of subgroupDRESS (120, 121).Care trial suggested that a hemoglobin of analyses. Subgroup analyses have the po- The two RCTs in adult patients were79 g/dL (70 90 g/L) when compared tential to mislead due to the absence of methodologically strong and precise andwith 10 12 g/dL (100 200 g/L) was not an intent to treat, sampling bias, and provided direct evidence regarding death associated with increased mortality in selection error (118). The analyses sug- rates. The conclusions are limited, how- adults (123). Red blood cell transfusion in gested increasing absolute and relativeever, by inconsistency that is not ade-septic patients increases oxygen delivery risk reduction with greater risk of deathquately resolved by subgroup analysesbut does not usually increase oxygen con- using both higher APACHE II scores and (thus the designation of moderate-qualitysumption (124 126). This transfusion greater number of organ failures (119).evidence). Results, however, consistentlythreshold of 7 g/dL (70 g/L) contrasts This led to drug approval for patients fail to show benet for the subgroup ofwith the early goal-directed resuscitation with high risk of death (such as APACHEpatients at lower risk of death and con- protocol that uses a target hematocrit of II 25) and more than one organ failure sistently show increases in serious bleed- 30% in patients with low ScvO2 (mea- in Europe. ing. The RCT in pediatric severe sepsis sured in superior vena cava) during the The ADDRESS trial involved 2,613 pa- failed to show benet and has no impor- rst 6 hrs of resuscitation of septic shock. tients judged to have a low risk of deathtant limitations. Thus, for low-risk and at the time of enrollment. The 28-daypediatric patients, we rate the evidence as2. We recommend that erythropoietin mortality rate from all causes was 17% onhigh quality. not be used as a specic treatment of placebo vs. 18.5% on APC, relative riskFor adult use there is probable mor-anemia associated with severe sepsis 1.08, 95% condence interval 0.921.28 tality reduction in patients with clinicalbut may be used when septic patients (116). Again, debate focused on subgroup assessment of high risk of death, most of have other accepted reasons for ad- analyses; analyses were restricted towhom will have APACHE II 25 or mul- ministration of erythropoietin, such as small subgroups of patients with tiple organ failure. There is likely no ben-renal failure-induced compromise of APACHE II score 25 or more than oneet in patients with low risk of death, red blood cell production (grade 1B). organ failure, which failed to show bene-most of whom will have APACHE II 20 Rationale. No specic information re- t. However, these patient groups also or single organ dysfunction. The effects garding erythropoietin use in septic pa- had a lower mortality than in PROWESS. in patients with more than one organ tients is available, but clinical trials in Relative risk reduction of death was failure but APACHE II 25 are unclear,critically ill patients show some decrease numerically lower in the subgroup of pa- and in that circumstance one may use in red cell transfusion requirement with tients with recent surgery (n502) in clinical assessment of the risk of death no effect on clinical outcome (127, 128). the PROWESS trial (30.7% placebo vs. and number of organ failures to supportThe effect of erythropoietin in severe sep- 27.8% APC) (119) when compared withdecision. There is a certain increased sis and septic shock would not be ex- the overall study population (30.8% pla- risk of bleeding with administration ofpected to be more benecial than in other cebo vs. 24.7% APC) (115). In the AD-rhAPC, which may be higher in surgical critical conditions. Patients with severe DRESS trial, patients with recent surgerypatients and in the context of invasivesepsis and septic shock may have coexist- and single organ dysfunction who re- procedures. Decision on utilization de-ing conditions that do warrant use of ceived APC had signicantly higher 28- pends on assessing likelihood of mor-erythropoietin. day mortality rates (20.7% vs. 14.1%, ptality reduction vs. increases in bleed- .03, n 635) (116). ing and cost. (Appendix D provides the 3. We suggest that fresh frozen plasma Serious adverse events did not differ in nominal committee vote on recommen- not be used to correct laboratory clot- the studies (115117) with the exception dation for rhAPC.) A European regula- ting abnormalities in the absence of of serious bleeding, which occurred more tory mandated RCT of rhAPC vs. pla- bleeding or planned invasive proce- often in the patients treated with APC:cebo in patients with septic shock is dures (grade 2D). 2% vs. 3.5% (PROWESS; p .06) (115);now ongoing (122). Rationale. Although clinical studies 2.2% vs. 3.9% (ADDRESS; p .01) (116); have not assessed the impact of transfu- 6.5% (ENHANCE, open label) (117). TheJ. Blood Product Administrationsion of fresh frozen plasma on outcomes pediatric trial and implications are dis- in critically ill patients, professional or- cussed in the pediatric consideration sec- 1. Once tissue hypoperfusion has resolvedganizations have recommended fresh fro- tion of this article. (Appendix C providesand in the absence of extenuating cir-zen plasma for coagulopathy when there absolute contraindications to use ofcumstances, such as myocardial isch-is a documented deciency of coagulationCrit Care Med 2008 Vol. 36, No. 1 307 13. factors (increased prothrombin time, in- 2. We recommend that plateau pressuresvolumes to demonstrate a mortality ben- ternational normalized ratio, or partialbe measured in patients with ALI/et (135). thromboplastin time) and the presence ofARDS and that the initial upper limitHigh tidal volumes that are coupled active bleeding or before surgical or in- goal for plateau pressures in a pas- with high plateau pressures should be vasive procedures (129 131). In addition,sively inated patient be 30 cm H2O. avoided in ALI/ARDS. Clinicians should use transfusion of fresh frozen plasma in Chest wall compliance should be con- as a starting point the objective of reducing nonbleeding patients with mild abnor- sidered in the assessment of plateau tidal volume over 12 hrs from its initial malities of prothrombin time usually failspressure (grade 1C). value toward the goal of a low tidal vol- to correct the prothrombin time (132). ume ( 6 mL/kg PBW) achieved in con- There are no studies to suggest that cor-Rationale. Over the past 10 yrs, sev-junction with an end-inspiratory plateau rection of more severe coagulation ab- eral multicenter randomized trials havepressure 30 cm H2O. If plateau pressure normalities benets patients who are not been performed to evaluate the effects ofremains 30 after reduction of tidal vol- bleeding.limiting inspiratory pressure throughume to 6 mL/kg PBW, tidal volume shouldmoderation of tidal volume (135139). be reduced further to as low as 4 mL/kg 4. We recommend against antithrombin These studies showed differing resultsadministration for the treatment of se- PBW. (Appendix E provides ARDSNet ven-that may have been caused by differencestilator management and formulas to calcu-vere sepsis and septic shock (grade 1B).between airway pressures in the treat-late predicted body weight.) Rationale. A phase III clinical trial of ment and control groups (135, 140). The No single mode of ventilation (pres- high-dose antithrombin did not demon-largest trial of a volume- and pressure-sure control, volume control, airway strate any benecial effect on 28-day all- limited strategy showed a 9% decrease ofpressure release ventilation, high-fre- cause mortality in adults with severe sepsis all-cause mortality in patients with ALI or quency ventilation) has been consistently and septic shock. High-dose antithrombin ARDS ventilated with tidal volumes of 6 shown advantageous when compared was associated with an increased risk of mL/kg of predicted body weight (PBW), with any other that respects the same bleeding when administered with heparinas opposed to 12 mL/kg, and aiming for aprinciples of lung protection. (133). Although a post hoc subgroup anal-plateau pressure 30 cm H2O (135). The ysis of patients with severe sepsis and high 3. We recommend that hypercapnia (al-use of lung-protective strategies for pa- risk of death showed better survival in pa- lowing PaCO2 to increase above its pre-tients with ALI is supported by clinical tients receiving antithrombin, antithrom- morbid baseline, so-called permissivetrials and has been widely accepted, but bin cannot be recommended until further hypercapnia) be allowed in patientsthe precise choice of tidal volume for an clinical trials are performed (134).with ALI/ARDS if needed to minimizeindividual patient with ALI may require plateau pressures and tidal volumes 5. In patients with severe sepsis, we sug- adjustment for such factors as the plateau (grade 1C).gest that platelets be administered pressure achieved, the level of positivewhen counts are 5000/mm3 (5 end-expiratory pressure chosen, the com-Rationale. An acutely elevated PaCO2109/L) regardless of apparent bleeding. pliance of the thoracoabdominal com-may have physiologic consequences thatPlatelet transfusion may be consideredpartment, and the vigor of the patientsinclude vasodilation as well as an in-when counts are 5000 30,000/mm3breathing effort. Some clinicians believe creased heart rate, blood pressure, and(530 109/L) and there is a signi- it may be safe to ventilate with tidal vol- cardiac output. Allowing modest hyper-cant risk of bleeding. Higher plateletumes 6 mL/kg PBW as long as the pla-capnia in conjunction with limiting tidalcounts ( 50,000/mm3 [50 109/L]) teau pressure can be maintained 30 cm volume and minute ventilation has beenare typically required for surgery or H2O (141, 142). The validity of this ceil-demonstrated to be safe in small, nonran-invasive procedures (grade 2D). ing value will depend on breathing effort,domized series (145, 146). PatientsRationale. Guidelines for transfusion as those who are actively inspiring gen-treated in larger trials that have the goal of platelets are derived from consensuserate higher transalveolar pressures for aof limiting tidal volumes and airway pres- opinion and experience in patients under-given plateau pressure than those who sures have demonstrated improved out- going chemotherapy. Recommendationscomes, but permissive hypercapnia wasare passively inated. Conversely, pa- take into account the etiology of throm- not a primary treatment goal in these stud-tients with very stiff chest walls may re- bocytopenia, platelet dysfunction, risk of ies (135). The use of hypercapnia is limitedquire plateau pressures 30 cm H2O to bleeding, and presence of concomitantin patients with preexisting metabolic aci-meet vital clinical objectives. One retro- disorders (129, 131).dosis and is contraindicated in patientsspective study suggested that tidal vol-with increased intracranial pressure. So-umes should be lowered even with pla- II. SUPPORTIVE THERAPY OFdium bicarbonate or tromethamineteau pressures 30 cm H2O (143). An SEVERE SEPSIS(THAM) infusion may be considered in se-additional observational study suggestedlected patients to facilitate use of permis-that knowledge of the plateau pressures sive hypercarbia (147, 148). A. Mechanical Ventilation ofwas associated with lower plateau pres- Sepsis-Induced Acute Lung4. We recommend that positive end-sures; however, in that trial plateau pres- Injury (ALI)/Acute Respiratory sure was not independently associatedexpiratory pressure (PEEP) be set so Distress Syndrome (ARDS) with mortality rates across a wide range as to avoid extensive lung collapse at 1. We recommend that clinicians target of plateau pressures that bracketed 30 cmend-expiration (grade 1C).a tidal volume of 6 mL/kg (predicted) H2O (144). The largest clinical trial em-Rationale. Raising PEEP in ALI/ARDSbody weight in patients with ALI/ ploying a lung-protective strategy cou- keeps lung units open to participate inARDS (grade 1B).pled limited pressure with limited tidalgas exchange. This will increase PaO2308Crit Care Med 2008 Vol. 36, No. 1 14. when PEEP is applied through either antidal volumes, and in those who improved ratory failure (responsive to relatively endotracheal tube or a face mask (149 CO2 exchange as a result of proning (159). low levels of pressure support and 151). In animal experiments, avoidance of A second large trial of prone positioning, PEEP) with stable hemodynamics who end-expiratory alveolar collapse helpsconducted for an average of approximatelycan be made comfortable and are eas- minimize ventilator-induced lung injury 8 hrs/day for 4 days in adults with hypox- ily arousable; who are able to protect when relatively high plateau pressuresemic respiratory failure of low-moderate the airway and spontaneously clear are in use. One large multicenter trial ofacuity, conrmed improvement in oxygen-the airway of secretions; and who are the protocol-driven use of higher PEEPation but also failed to show a survival ad- anticipated to recover rapidly from the in conjunction with low tidal volumes did vantage (160). However, a randomized precipitating insult. A low threshold not show benet or harm when compared study that extended the length of time for for airway intubation should be main- with lower PEEP levels (152). Neither the proning each day to a mean of 17 hrs for a tained (grade 2B). control nor experimental group in thatmean of 10 days supported benet of pron-Rationale. Obviating the need for air- study, however, was clearly exposed toing, with randomization to supine positionway intubation confers multiple advan- hazardous plateau pressures. A recent an independent risk factor for mortality by tages: better communication, lower inci- multicenter Spanish trial compared amultivariate analysis (161). Prone position-dence of infection, reduced requirements high PEEP, low-moderate tidal volumeing may be associated with potentially life-for sedation. Two RCTs demonstrate im- approach to one that used conventionalthreatening complications, including acci-proved outcome with the use of NIV when tidal volumes and the least PEEP achiev-dental dislodgment of the endotrachealit can be employed successfully (162, ing adequate oxygenation. A marked sur- tube and central venous catheters, but vival advantage favored the former ap-165). Unfortunately, only a small percent- these complications can usually be avoidedage of patients with life-threatening hy- proach in high-acuity patients with ARDSwith proper precautions. (153). Two options are recommended forpoxemia can be managed in this way. PEEP titration. One option is to titrate6a. Unless contraindicated, we recom- 8. We recommend that a weaning proto- PEEP (and tidal volume) according tomend that mechanically ventilatedcol be in place and that mechanically bedside measurements of thoracopulmo- patients be maintained with the head ventilated patients with severe sepsis nary compliance with the objective of ob- of the bed elevated to limit aspirationundergo spontaneous breathing trials taining the best compliance, reecting arisk and to prevent the developmentregularly to evaluate the ability to dis- favorable balance of lung recruitment of ventilator-associated pneumonia continue mechanical ventilation when and overdistension (154). The second op-(grade 1B).they satisfy the following criteria: a) tion is to titrate PEEP based on severity 6b. We suggest that the head of bed be They are arousable; b) they are hemo- of oxygenation decit and guided by the elevated approximately 30 45dynamically stable (without vasopres- FIO2 required to maintain adequate oxy- (grade 2C).sor agents); c) they have no new po- genation (135) (Appendix D). WhicheverRationale. The semirecumbent posi- tentially serious conditions; d) they the indicator compliance or oxygen-tion has been demonstrated to decrease have low ventilatory and end-expira- ationrecruiting maneuvers are reason-the incidence of ventilator-associated tory pressure requirements; and e) able to employ in the process of PEEP pneumonia (VAP) (162). Enteral feeding their F IO 2 requirements could be selection. Blood pressure and oxygen- increased the risk of developing VAP;safely delivered with a face mask or ation should be monitored a