Review ArticleArticie de revue

Total parenteral nutrition in the surgical patient;a meta analysis

Daren K, Heyland. MD, MSc;* Max Montalvo, MD;t Shaun MacDonald, MD;' Laurie Keefe, RD;tXiang Yao Su; John W. Drover, MDt

Objective: To examine the relationship bcwecn total pareiiteral nutrition(TPN) and complication anddcjth rites in siirgicil patients. Data sources: A computer search ot published research on MKDLINE,personal files and a review of rele\ant retcroncc lists. Study selection: A review ot'237 titles,, abstracts orpapers. Primar\' studies were included if they were randomized clinical trials of surgical patients thatevaluated the effect of TPN (compared to no TPN or standard care) on complication and death rates.Studies comparing TPN to enteral nutrition (EN) were excluded. Data extraction: Relevant data wereabstracted on the methodology and outcomes of primary' studies. Data were independently abstracted induplicate. Data synthesis: There were 27 randomized trials in surgical patients that compared the useot TPN to standard care (usual oral diet plus intravenous dextrose). When the results of these trials wereaggregated, there was no ctfecr on mortality' (risk ratio = 0.97, 95% contidence inten.-als, 0,76 to 1.24).There were fewer major complications in patients who received TPN, although there was signiticantheterogeneity in the overall estimate (risk ratio = 0.81,, 95% CI, 0.65 to 1,01). Because of this signiti-cant heterogeneity, several a priori hypotheses were examined. Studies that included only malnourishedpatients demonstrated a trend to a reduction in complication rates but no ditfl-rence in death rate whencompared with studies of patients who were not malnourished. Studies published in 1988 or earlier andstudies with a lower methods score were associated with a significant reduction in complication ratesand a trend to a reduction in death rate when compared with stLidies published after 1988 and studieswith a higher methods score. There was no ditference in studies that provided iipids as a component ofTPN when compared with studies that did not. Studies that iniriated TPN preoperatively demonstrateda trend to a reduction in complication rates but no difference in death rate when compared with studiesthat initiated TPN postoperatively. Conclusions: TPN does not intluence the death rate of surgicalpatients. It may reduce the complication rate, especially in malnourished patients, but study results areinfluenced by methodologic qualit\- and year of publication.

Objectif: Examiner le lien entre la nutrition parenterale totale (KPT) et les taux de complication et deniortalite chez les patients en chirurgie. Sources de donnees : Recherche inforniatique dans desrecherches publiees sur MEDLINE et dans des dossiers personnels, et examen dc listes de documents dereference pertinents. Selection d'etudes : Re\-ue de 237 titres, abreges ou communications. On a inclusdes etudes principales s'il s'agissait d'etudes cliniques randomisees portant sur des padents en chirurgieet qui ont cvalue Tetfet de la NPT (comparativemcnt a I'absence de NPT ou aux soins normaux) sur lestaux de complication et de mortaiite. On a exclu les etudes de comparaison de la NPT a Penteronutri-tion. Extraction des donnees : On a abrege les donnees peninentes sur la methodologie et les resultatsdes c'tLides principales. Les donnees ont ete abrcgees de fat on independante et en double. Synthese desdonnees : II y avait 27 etudes randomisees portant sur des patients en chirurgie au cours desquelles on acompare Tutilisation de ia NPT aux soins normaux (alimentation orale habituelle et dextrose par voieintraveineuse). L'agregation des resultats de ces etudes n'a revele aucun etict sur la mortaiite (risquerelatif = 0,97; inten-'alles de confiance a 95 %, 0,76 a 1,24). II y avait nioins de complications majeureschez les patients alimentes par NPT, mcme si I'on a constate une heterogeneite importante dans Testi-

From the 'Department of Medicine and tDepartment of Surger/. Queen's Universify, Kingston. Ont.. and f Nutritional Services.Kingston Generai Hospital. Kingston. Ont.

Dr. Heyiand is a Career Scientist of ttie Ontaria Ministry of Heaitt).

Accepted for publication Apr. 3. 2000.

Correspondence to: Dr. Daren K. Heyland. Angada 3, Kingsfon General Hospitai, 76 Stuart St.. Kingsfon ON K7L 2V7: fax 613 548-2577. dkh2@posf.queensu.ca

© 2001 Canadian Medical Associafion

102 Journal canadien de chirurgie. Vol. 44, N'=2, avril 2001

Parenterai nutrition in the surgicai patient

mation globale (RR = 0,81; IC a 95 %, 0,65 a 1,01). A cause de cette heterogeneite iniportantc, on aexamine piusieurs hypotheses a priori. Des etudes qui ont porte tiniquement sur des patients sous-alimcntes ont dcmontre une tendance a la reduction des taux de complication, mais aucuiie differenceai! nivcaii du taux de mortalitc comparativement aux etudes portant sur des patients qui nV'taicnt passous-alimcntc's. Par rapport aux etudes publiccs apres 1988 ct ayant obtenu uii re.sultat methodologiqiieplus clcvc, les etudes publices cn 1988 on antericurcmcnt et tes etudes ayant obtenu un resultatmethodologique moins elcve affichaicnt une reduction importantc des taux tic complication ainsiqu'unc tendance a une reduction du taux dc niortalite. (")n n'a constate aucune difference dans lesetudes qui comportaicnt Tadministration dc lipidcs par NPT comparativement aux cttidcs qui n'cn com-portaicnt pas. Les etudes ou on a cntrcpris la NPT avant rintcrvcntion ont revele une tendance a ia re-duction des taux de complication, mais aucunc difference au nivcau du taux dc mortalitc comparative-ment aux etudes oCi on a commence a administrer la NPT apres Tintcrvcntion. Conclusions : La NPTn'a pas d'cffct sur le taux de mortalite chez les patients en chirurgie. EUc pcut reduirc te tau.x dc compli-cation, particuliercmcnt chez les patients sous-alimcntcs. Les resultats des etudes varient ccpcndant cnfonction dc la qualite dc la methodologie ct dc I'anncc dc publication.

The consequences of majorsurgerv' can lead to hypermetab-

olism and subsequent mainutridtin.'-The padent's previous nutridonal sta-tus, the concomitant or underlyingdisease, and the degree and duradonof other stresses can contribute to therisk of malnutrition.' Malnutritioncan lead to depletion of body mass,impaired tissue and organ funcdon,compromised immunit}' and poorwound healing. A strong associationexists between malnutrition and in-creased postoperadve morbidit\' andmortality' in surgical padents."*'

The administration of total par-enterai nutrition (TPN) can clearlyprevent the effects of stanation inpatients with a nontiinctioning gas-trointestinal tract. However, it is un-clear whether TPN can modulate thecatabolic response to surgical stressand reduce complications associatedwith hypercatabolism.^ Put ditler-ently, the perioperative administra-tion of TPN may result in signiticantimprovement in weight, nitrogenbalance, prealbumin levels and othernutridonal end points, but tlie effecton clinically important end points,such as mortalit)' and complications,is less certain. The purpose of thispaper is to systematically review, crit-ically appraise and statisdcally aggre-gate all studies evaluating the etfectof TPN on complication and deathrates in surgical padents.

A number of clinical trials' ** and ameta-analysis'' have suggested thatpreoperadve administration of TPNin severely malnourished patients

may be associated with a lower rateof postoperadve complications. Be-cause a number of trials have beenpublished subsequent to the meta-analysis, we decided to conduct an-other meta-analysis to summarize thecurrent literature.

Methods

Search strategy

We conducted a computerizedbibliographic search of MEDLINE(including pre-MEDLINE) from1980 to May 1999 to locate all rele-vant ardcles. The terms "randomizedcontrolled trial," "double blindmethod," "clinical trial," "placebo"and "comparadve study" were com-bined widi "parenterai ntitrition, to-tal." Citations were limited to Eng-lish studies reporting on adultpatients. We also searched referencelists of relevant review articles andpersonal files.

Selection criteria

Initially, 2 investigators screenedall citations and classitied them intoprimar\' studies, review articles orothers. We then retrie\'ed and inde-pendently reviewed all primar)' stud-ies. They were included in thisoverview if they met the followingcriteria;

• research design — randomizedclinical trial

• population — adult surgicalsubjects

• intervention — any form of TPN(protein, carbohydrates with orwithout lipids) compared to noTPN

• outcome — complicadons, lengthof hospital stay and mortalit}'.

We elected to include only ran-domized trials in this review becausestudies in which treatment is allocatedby any other method dian randomiza-tion tend to show larger (and tre-quently false-positive) treatment ef-fects than do randomized trials.'" Sincethe scope of our review was defined b\'our research question, we also ex-cluded stutlies diat compared TPN toenteral nutrition or other tbrms ofTPN. Finally, suidies that only evalu-ated the impact of TPN on nutritionaloutcomes (i.e., nitrogen balance,amino acid profile) were excluded. Al-though these end points may explainunderlying pathophysiologv', we con-sidered them as surrogate end points,"and we only included papers thatreported on clinically important out-comes (morbidity' and mortality').

Mettiodologic quality of primarystudies

We assessed the methodologiequalit\' of all selected ardcles indepen-dently in duplicate, according to thescoring system shown in Table 1.Even in randomized trials, failure toprevent foreknowledge of treatmentassignment can lead to an overesdma-don of treatment effect.'' Accordingly,we scored higher those studies thatreported that their randomization

Canadian Journal of Surgery, Vol, 44, No 2, April 2001 103

r- Heyiand et al

schema was concealed. Given the diffi-culties of blinding the administrationoF TPN, we only awarded points forstudies that blinded the adjudicationof study end points. We also evaluatedthe extent to which consecutive, eligi-ble padents were enrolled in tiie trial,whether groups were equal at baseline,if cointerventions were adequately de-scribed., whetlicr objecti\'e definitionsof infectious outcomes were employedand w hether all patients were properlyaccounted for in the analysis {inten-don-to-trcat analysis) (Table 1).

Data exiraction

Two investigators extracted data foranalysis and assessed the metJiodologicqualit\'; we resolved disagreement byconsensus. Not all studies reportedcomplication rates. Some reported to-tal complications per group not perpadent. When data were missing, un-clear or not reported on a per patientbasis., we attempted to contact the pri-mary' invesdgators to provide furtherinformation if the paper had beenpublished in the last 5 years.

Prior hypotheses regarding sourcesof heterogeneity

When conducting a meta-analysis,heterogeneity (major differences in

the apparent effect of the interven-tions across studies) is often found.When present, heterogeneity weakensany inferences that can be made fromthe results. The possible sources ofvariadon include the role of chance orditlcrenccs acrt)ss studies in popula-tion, intervention, outcome andmethods. A priori., we developed sev-eral hypotheses that might explainheterogcneit)' of study results.

• We considered that the premor-bid nutritional status of studypatients was a possible cause ofvariation in results. Wheneverpossible we grouped the resultsof studies that included only pa-tients who were malnourishedand compared them to the resultsof studies that included patientswho were not malnourished atthe dme of entr\' into the study.When possible, we used the defi-nition of malnourished pro\'idedin each individual study. If no de-finition was provided, we as-sumed patients who had greaterthan 10% weight loss to be mal-nourished .

• We hypothesized that studyresults may be related to themethodologic quality of thestudy. We planned a separateanalysis comparing the effect ofstudies with an overall method-

r" luuie 1 ~ ~ " • -

f Criteria Used to Assess Methodologic

11 Crilerion

1 Randomization

Blinding

Analysis

Patient selection

Comparability of group1 at baseiinei Extent of foilow-up

Treotment protocol

Cointerventions'

Outcomes*ThB axtent fo which antibiotics, equestkDfis 1 to 3 and 8 and 9, poss

0

Not blinded

OtherSelected patientsor cannot teiiNo Of not sure

<100%Poorly described

Not described

Not described

Quality

Score

1

Not conceaied ornot sure

Consecutive eligibiepatients

Yes

100%

ReproduciblydescribedDescribed but notequai or not sure

Partially described

j2

Concealed |

Adjudicotorsblinded jIntentton-to-treat ]

Well described andall equolObjectively detined 1

ntetal nuMHon. ventilation, oxvgen ond transfusions we:B oppted equollv ocioss goups. Forb<escore0.1 or2. Forquestlo*Ts4to7. posslblescoteOof 1. Totalposstoloscoio Is !4.

ologic qualit\' of 7 and greater tothose with a score of less than 7(median score = 7).

• Since the practice of pro\'idingnutridonal support and the man-agement of surgical padents hasevolved over dme, we divided thestudies into groups comparingstudies published in 1988 or ear-lier with studies published since1989 (halt\\a)- point of this studyperiod).

• There are several randomized trialsof surgical patients that examinethe effect of amino acid infusionalone or in combination with acarbohydrate st)urce of calories(without the addition of lipids)on clinical outcomes. We hypothe-sized that there may be some ad-verse effects from the use oflipids.''''"' Accordingly, we sepa-rated trials into those that includedlipids and those that did not.

• We speculated that differences inthe timing of the interventionmay account for different results.To test this hypothesis weplanned a separate analysis com-paring studies that inidated TPNpreoperatively to studies thatstaitcd TPN postoperatively.

Analysis

The primar\' outcome was periop-eradve death (death within 30 daysof operation) or death in hospital.The secondary outct)me was majorcomplications. We defmed majorcomplications as pneumonia, intra-abdominal abscess, sepsis,catheter-related infecdon, myocardial infarc-don, pulmonar\' embolism, heart fail-ure, stroke, renal failure, li\er failureand anastomodc leak. Minor compli-cadons were deHned as wound infec-don, phlebids, urinar\' tract infectionand atelectasis. In 5 studies, the datawere not portrayed in a fashion thatallowed us to report major complica-tion rates so we reported total com-plicadons"'""' and total infectiouscomplications.'^"* There were somestudies in which their reporting

104 Joumal canadien de chirurgie. Vol. 44, N°2, avril 2001

Parenteral nutrition in the surgical patient

methods did not allow us to disag-gregate infectiotis from nonintec-tious complications. One study ran-domized patients to 3 groups(control versus statidard TPN versusTPN with branched chain aminoacids).''' We only included data fromthe control group and the standardTPN group. Two other studies ran-domi7xd patients to 3 groups (con-trol versus TPN without Iipids versusTPN with Iipids) and we includedboth experimental groups in theanalysis.''*'-' We also reported on theduration oF hospital stay,, althoughthese data were not aggregated ow-ing to infrequent and variable report-ing methods. Agreement betweenreviewers on the inclusion of articleswas measured by weighted kappa.

We combined data from all studiesto estimate the common relative riskof death and complications and asso-ciated 95% confidence inter\'als (CIs).We summarized the treatment eftectusing risk ratios (RRs). To avoid theproblem with bias and instabilit\- as-sociated with RR estimation in sparsedata,, we added one-half to each cell."In the meta-analysis, we used maxi-mum likelihood methods of combin-ing RR5 across all trials and examineddie data tor evidence ot heterogeneitywithin groups.^' The Mantel-Haenzel method was used to test thesignificance of treatment etiect.'^ Weused a random efFects model to esti-mate the overall RR.'"^" For the testof heterogeneity across subgroups,the f test for the difference betweenthe 2 subgroups was used. We con-sidered a p value of less tlian 0.05 tobe statistically signiticant.

Results

study identification and selection

In all 187 citations were identifiedfi-om the MEDLINE databases. Ourpersonal files and review of reterencelists yielded 57 additional articles torconsideration. Initial eligibilityscreening resulted in 47 articles se-lected for further evaluation. Of

these potentially eligible papers, 27met the inclusion criteria." **'" -' "^*

There was 100% agreement onthe inclusion of articles for this over-view. Reasons tor excluding relevantrandomized studies included studiesevaluating dit^erent kinds of TPN',^^"pseudorandomized studies,"'""" dupli-cate publications,'^'•'•' studies not re-porting clinically important out-comes,'"^ " a study available inabstract tbrm only^" and a study thatalso randomized patients to anabolicsteroids.'"

Impact of total parenteral nutritionon death and compiication rates

The 27 randomized trials, involv-ing 2907 patients, compared the useof TPN to standard care (usual oraldiet plus dextrose given intra-venously) in patients who unden\entsurgery."""''" '' The details of eachstudy are described in Table2_^«,i. -2i,r . Wa-ien the results of thesetrials were aggregated, there was noeffect on moitalit\' (RR = 0.97, 95%CI, 0.76 to 1.24) (Fig. 1). The testfor heterogeneit)' was not significantalthough a visual inspection suggeststhat the treatment effect of some ofthe studies was significantly ditfcrentfrom other studies. Twenty-twostudies reported major complica-tions. When these results were aggre-gated, TPN was associated with a re-duction in complication rates (RR =0.81, 95% CI, 0.65 to 1.01, p =0.06) (Fig. 2). The test for hetero-geneit\' was signiticant {p= 0.01),

Given that we found significantheterogeneity and in an attempt tobetter explain our findings, we exam-ined our a priori hypotheses. Wecompared those trials that includedonly malnourished patients withother trials. TPN was not associatedwith any difference in mortality(Fig. 3) in studies t)f malnourishedpatients (RR - 1.13,95% CI, 0.75 to1.71) or in studies of normaih' nour-ished patients (RR = 0.90, 95% CI,0.66 to 1.21, ^ = 0.38 tbr ditlerencesbetween subgroups). In studies of

malnourished patients, TPN was as-sociated witli a significant reductionin complication rates (RR = 0.52,95% CI, 0.30 to 0.91). The RR ofmajor complications in studies ofpatients who were not malnourishedwas 0.95 (95% CI, 0.75 to 1.21).When we compared the complica-tion rates associated with TPN instudies of patients who were notmalnourished with the rate in studiesof malnourished patients, the differ-ences were just short of statisticalsignificance (p = 0.066),

We compared trials with a meth-ods score of less than 7 to trials witha score of 7 or better (Eig. 3). Trialswith the higher methods scoredemonstrated no effect of TPN onmortalit)' (RR = 1.08, 95% CI, 0.81to 1.43), whereas trials with a scoreof 7 or less suggested a trend towarda reduction in mortality associatedwith the use of TPN (RR = 0.75,95% CI, 0,47 to 1,19). The test forheterogeneity across subgroups wasnot significant {p = 0.21). With re-spect to complication rates, in studieswith a higher methods score therewas no effect of TPN on major com-plications (RR = 1.07, 95% C:i, 0.86to 1,32). In studies with a lowermethods sctjre, there was a significantreduction in complication rates (RR= 0.50, 95% CI, 0.32 to 0.76). Thetest for heterogeneity across sub-groups was significant {p = 0.005).

We next compared trials pub-lished in 1988 or earlier with snidiespublished since 1989 (see Eig. 3).Trials published in 1988 or earlierwere associated with a trend towarda decrease in death rates associatedwith the use of TPN (RR = 0.68,95% CI, 0.43 to 1.10). Trials pub-lished since 1989 were consistentwith no treatment eff ect associatedwith TPN ( R R - 1.11, 95% CI, 0.83to 1.48). The test for heterogeneityacross subgroups was short of con-ventional levels of significance ip =0.10). With respect to ctjmplicationrates, in studies published in 1988 orearlier there was a significant reduc-tion in major complications associ-

Canadian Journal of Surgery, Vol, 44. No 2, April 2001 105

Heyland et al

ou5

z

2I

TJo

OTJcO

8S

lipid

ocdoa;n

pids

—d0

a

(M — — — — — , — i n —

o ;z

CM — —

O 21- "

• ^ o

'^ c

CM

o2

c cCO •o§ ^ S

-o

CO

O

o

b-m

o

993"

CO

** "6C <D

"S D> Q.

<D

u -

Q)Q)

c0

eiio

n'

m

o

c0LJ.

o%n

guen

Eo

3nds

i

IDO)Co5I

D = ' i=i

o

c£1O)OX

(DO

: 5• a

ED

OQ.

o

106 Journai canadien de chirurgie. Vol. 44, N 2, avril 2001

Parenterat nutrition in ttie surgical patient -^

coo

oO

•'i^ s

p a= Ui•o

D ti' O

I D£

ho-'~o

00

<

E

rte

oX"•cJ I

1

ao0)Q."D

?>

CL

"Di o ;, « E ffi ;^

o "2 "5 3

2 EO ,0

5

,

eta

i.

13

'o

I M

eg

n an

d

o

0o

yenf

el

0

; V

on

00)Q"O0

* Y

am

— ~ -o

Q OE

<D CCO O

5 E

ated with TPN (RR = 0.42, 95% CI,0.26 to 0.68). The aggregated re-sults of studies published since 1989were consistent with no treatment ef-fect (RR - 1.09, 95% CI, 0.91 to1.31). The test fbr heterogeneityacross subgroups was statistically sig-nificant ( /= 0.002).

We then compared studies thatprovided intravenous Iipids as a com-ponent of TPN administration tothose studies that did not includeIipids. There was no difference inmortalit}' associated with TPN withIipids (RR - 1.05, 95% CI, 0.79 to1.40) or without Iipids (RR = 0.80,95% CI, 0.50 to 1.28). The test tbrheterogeneity across subgroups wasnot statistically signiticant (p= 0.35).There was a trend toward a reduc-tion in complications associated withTPN without Iipids (RR = 0.80, 95%CI, 0.63 to 1.02), but this did notdiffer significantly from the effect ofTPN with Iipids (RR = 0.86, 95%Cl, 0.63 to 1.19). The test for het-erogeneity' across subgroups was notstatistically significant {p - 0.72).

Einally, we compared studies thatinitiated TPN preoperatively to stud-ies that initiated TPN postoperatively.With respect to mortality', TPN didnot seem to have any treatinent effectwhen administered preoperatively(RR - 0.85, 95% CI, 0.61 to 1,20)or postoperatively (RR - 1.08, 95%CI, 0.73 to 1.58). The test for het-erogeneity' across subgroups was notstatistically significant {p = 0.39). Instudies that initiated TPN preopera-tively, there was a significant reduc-tion ill complication rates associatedwith the use of TPN (RR - 0.70,95% CI, 0.52 to 0.95), but in suidiesthat initiated TPN postoperativelythere was no such effect (RR = 1.01,95% CT, 0.70 to 1.46). The differ-ence between tlie complication ratesof these 2 groups was not statisticallysignificant (p= 0,15),

Only 13 studies reported the im-pact of TPN on duration of hospitalstay, 5 reporting median stay and 8reporting mean stay. In 7 studies, theduration of stay in hospital was

Canadian Journal of Surgery, Vol, 44, No 2, April 2001 107

r Heyland et al

shorter in the control group. Due tothe variability in duration of stay andvariability of reporting methods, wedid not attempt to aggregate theseresults, but they are displayed inTable 2.

Discussion

Although meta analyses have re-cently come under scrutiny,''" theyare potentially useful tools to helpevaluate the effieaey of medieal inter-ventions. In the last 2 decades, therehave been 27 randomized trials ex-amining the effect of TPN on themorbidity and mortality' of surgicalpatients. These studies ranged in sizefrom 18 to 678 patients, with themajorit}' of studies including fewerthan 100 patients. The death rate inthese studies ranged from 0% to 30%,with an overall a\erage death rate of8.6%. Individually, the majority ofthese studies were underpowered todemonstrate a significant treatment

effect witli TPN. The advantage of ameta-analysis is that it provides amethod of aggregating similar stud-ies to determine the best estimate ofoverall treatment effect.

For this meta-analysis, we defineda specific research question, con-ducted a comprehensive literaturesearch and used explicit eriteria forstudy selection and methodologicquality assessment."' In the overallanalysis, there was no effect on mor-tality' and no significant reduction incomplication rates v\'hen comparingthe use of TPN to standard care(usual oral diet plus intra\'enous dex-trt)se). 'I'he degree of heterogeneityof the results weakens the inferenceswe can make from the overall resultsand makes the results of the pre-specified subgroups more com-pelling. Furthermore, it is the explo-ration of why this heterogeneityexists that sheds light on the poten-tial benefits or risks of IPN.

Our subgroup analysis, based on

our a priori hypotheses, showed thatthe significant reduction in complica-tion rates associated with the use ofTPN was found only in those studiesthat were published in 1988 or ear-lier or those with a methodologicqualit\' score less than 7. This treat-ment effect was systematically differ-ent from the effect observed in stud-ies since 1989 and in studies ofhigher methodoiogic quality. In-deed, if we accept the more reeent,high-qualit\' studies as providing thebest estimate of the current treat-ment effect, TPN is associated withno proven benefit in surgical pa-tients. Finally, we found no differ-ence in outcome in comparing stud-ies that used lipids versus studies thatdid not use lipids iii TPN administra-tion. There were no significant dif-ferenees in mortality' associated withTPN in any of the subgroups ex-plored in this meta-analysis.

When examining just studies ofmalnourished patients and studies

Abel, 1976Holler, 1977

Freund. 1979Thompson, 1981

Mullet. 1982Garden. 1983

Yamada, 1983Brennan, 1994

Askanazi, 1986Mullet, 1986

Bellantone. 1988Meguid, 1988

Stniih. 1988Fan, 1989

Figgetas, 1988Wooltson, 1989

Gys, 1990Reilly.1990

VATPN, 1991Von Meyenfeldl. 1992

Hwang, 1993aHwang, 1993b

Satidsltom, 1993Fan, 1994

Jimenez, 1995Doglietto, 1996

Hu, 1998

Overall

0.001 0.01

TPN betieficial

1 10 100

TPN hatmful

Risk ratio {log scale)

Holier. 1977

Freund, 1979Thompson, 1981

Muller, 1982Yamada, 1983Brennan, 1994

Askana^i, 1986Muller, 1986

Bellantone. 1988

Smith, 1988

Fan, 1989

Figueras, 1988Woolfson, 1989

Gys, 1990

VATPN, 1991

Von Meyenfeldt, 1992Hwang. 1993aHwang, 1993b

Fan, 1994Jimenez, 1995

Doglietto, 1996

Overall

0,001 0.01

TPN beneticial

0 1 10 100

TPN hatmful

Risk ratio (log scale)

FIG. 1. Risk ratios and associated 95% confidence intervals for mortal-ify reported in 27 randomized trials of total parenteral nutrition (TPN) insurgical patients. VATPN = Veterans Affairs Total Parenteral NutritionCooperative Study Group.

FIG. 2. Risk ratios and associated 95% confidence in-tervais for major complications reported in 21 ran-domized trials of total parenteral nutrition (TPN) in sur-gical patients. VATPN = Veterans Affairs TotalParenteral Nutrition Cooperative Study Group.

108 Journal canadien de chirurgie. Vol 44. avril 2001

Parenteral nutrition in the surgical patient —i

that initiated TPN preoperatively,t)ne finds a significant reduction incomplication rates associated withthe use of TPN. However., the test ofheterogeneity' across tliese subgroupswas not statistically significant, sug-gesting that the differences acrossthese subgroups (malnourished ver-sus normally nourished., preoperativeversus postoperative) may be due tochance alone. This weakens any in-ference that TPN is beneficial in mal-nourished patients or preoperatively.If nutrition support is indicated inmalnourished surgical patients, thereare preliminary' data that enteral sup-plementation may significantly re-duce complication rates." " ^

There are a number of strengthsto this meta-analysis. First, we in-cluded only randomized controlledtrials and did not include pseudoran-domized studies in the data analysis.Second, unlike previous reviews ofthe use of TPN in surgical patients,our study provided a thoroughanalysis of heterogeneity' of the data.By exploring the heterogeneity' of

the various studies, we have come toa better understanding about the ob-served treatment effects (or lackthereof). Third, our updated data setincluded many recent studies.

A previous meta-analysis on peri-operative parenteral nutrition pub-lished in 1987 aggregated die resultsof 18 trials of surgical patients.'' Simi-lar to our findings, no overall effectof TPN on morbidity' and mortalityand an inverse correlation betweenbetter qualit\' studies and treatmenteffect of TPN was shown. That is,the more methodologically rigorousstudies demonstrated less treatmenteffect with TPN. Following thismeta-analysis, the results of a largerandomized trial of preoperativeTPN in surgical patients were pub-lished.^ In this study of 395 patients,overall death and complication rateswere similar in the group receivingTPN preoperatively and in the con-trol group. There were, however,more infectious complications in dieTPN group {14.1% versus 6.4%, p =0.01). Again, consistent with our

ComplicationsMalnourished

Not malnourished

Quality score <7Quality score >7

Pubiistied before 1988Published after 1989

LipidsNo lipids

Preop.Postop.

Overall effect

Deaths MalnourishedNot malnourished

Quality score <7Quality score >7

Published before 1988Published after 1969

LipidsNo lipids

Preop.Postop.

Overall effect

Risk ratio (log scale)

FIG. 3. Results of subgroup analysis, examining the effect of total parenteral nutrition(TPN) on death and complication rates in surgical patients, p values represent testsof heterogeneity across subgroups.

fuidings, in a subgroup analysis,these investigators found that TPNwas associated with a trend to alower rate of major complications inpatients who were considered to beseverely malnourished (25.8% versus47.4%,V=0-12).

Recently, another critical reviewof the medical literature appraisingthe use of nutritional support wasconducted by a panel of experts re-cruited from the National Institutesof Health, the American Societ\' forParenteral and Enteral Nutrition andthe American Society for ClinicalNutrition.''*' They synthesized datafrom 33 trials involving o\'cr 2500surgical patients. They did not de-scribe the search strategy they usedto find relevant articles, nor did theydo a methodologic qualit\' assess-ment of primary studies included inthe re\'iew process. Consistent withour findings, they found that thepreoperative use of TPN was associ-ated with a 10% risk reduction incomplication rates. Unlike our find-ings, they noted a 10% increase incomplications associated with the useof TPN postoperatively.

Both these review articles includednonrandomized studies, weakeningany inferences one can make fromtheir results. In addition, neither re-view assessed or explained the hetero-geneity across studies included intheir review. As previously stated, inour meta-analysis, the tests for het-erogeneit)' suggest tliat the treatmenteffect was only observed in studiespublished in 1988 or earlier and inthose with methodologic qualityscores less than 7. Such findings fiir-ther weaken any inferences that canbe made by previous review articles.

There are a number of limitationsto the methods used in our meta-analysis. First, our computerized liter-ature search was restricted to studiespublished in English. Second, wewere not able to assess whether thecomposition and amount of TPNinfluenced study outcomes since in-cluded studies did not consistentlyreport these data. We could not assess

Canadian Journal of Surgery, Vol. 44, No 2, April 2001 109

r- Heyland et al

the extent to whieh the adverse orneutral effects of TPN may he due tohv'perglyeemia or overfeeding in eer-tain studies.'''''^ In addition, we werenot able to assess whether the eticetof TPN varied across different surgi-cal populations. We were unable toclassiR' patients according to severityof illness aiid degree of comorbid ill-nesses present at baseline. With theexeeption of 3 studies,'"'-'^ all studiesincluded patients who underwentmajor gastrointestinal surger\'. Per-haps the treatment effect is differentin other subgroups of surgical pa-tients not well characterized or repre-sented in the current literature. Fi-nally., the definition of malnourishedwas not standard aeross studies. Wehad to rely on the investigators' deti-nition, whieh may vary across studies.

Conclusions

The results of this meta-analysissuggest that., o\era!l, there is no ad-vantage to using TPN peri ope rativelyin surgical patients. The apparent re-duction in complication rates withTPN is associated with a significantdegree of heterogeneity across stud-ies. Possibly the treatment effect isstrongest in malnourished patients.However, the beneficial effects ofTPN in surgical patients are onlyseen in those studies done prior to1988 and studies of low method-ological quality-. Given the potentialfor increased ctjsts and complicationsassociated with the use of TPN inthe surgical patients (with no appar-ent reduetion in mortalit\' or lengthof stay), further sttidies are needed toconfirm the benefits of TPN in thispatient population.

Acknowledgements: We tlunk Dr.Djxid Spain who reviewed earlier ver-sions ot this niaiiLiscript.

References

] . Clifron GL, Robertsuji CS, Choi SC:, As-sessment ot" nutritional requirements ofhead injured patients. / Neurosurjf 1986;64:895-901.

2. Monk DN, Plank LD, 1 ranch-Areas G,Finn Pf, Streat SJ, Hill GI.. Sequentialchanges in the mctaholic response in criti-eally injured patients during the first 25days afi:er blunt trauma. Ami Siirjt 1996;223(4 ):395-405.

3. Souba W\V. Nutritional support [review],NErijrlJMcd !997;336(1): 41-8.

4. Daley I, Kluiri SF, Henderson VV, Hur K,Gibbs IO, Barbour G, et al- Risk adjust-ment ot" the postoperative morbidity' ratetor the comparative assessment of thequjlit\' of surgical care: results of the NA-tional Veterans Aftairs Surgical Risk Study.JAm CollSurg 1997;185(4):328-40.

5. Dempscy DT, Mullen JL, Buzby GP. Thelink between nutritional status and clinicaloutcome: Can nutritional lnter\entionmollify it? ^»i/a(«N«n-1988;47:352-6.

6. Muller JM, Brenner U, Dienst C,Pichlmaicr H. Preoperative parenteralfeeding in patients with gastrointestinalcaneer. Lancet 1982;1:68-71.

7. The Veterans Atlain Total Parenteral Nutri-tion CxK)perative Snidy Group. Pcriopera-ti\e t()t.i! piirentera] nutrition in surgical pa-tients. NEnjjlJMcd 1991;325(8l:525-32.

8. Bellantone R, Doglietto GB, Bossola M,Paeelli F, Negro F, Sotb L, et al. Preopera-tive parenteral initritiun in the high risksurgical patient. JPEN J Paremer EiitemlNntr l9S8-ni2): 195-7.

9. Detsky AS, Baker IP, O'Rourke K, GoelV, Perioperative parenterai nutrition: ameta-analysis. Ann Intern Med 1987;I07:195-203.

10. Sacks HS, Chalmers TC, Smith H Ir, R.ui-domized versus historical assignment incontrolled trials. N En^l} Med 1983;309:1353-61,

11. Fleming TR, DeMets DL. Surrogate endpoints in clinical trials: Are we being mis-led? Atin Intern Med 1996; 125:605-13.

12. Chalmers TC, Celano P, Sacks HS, SmithH. Bias in treatment assignment in eon-trolled clinical trials. N Enjjl J Med 1988;309:1358-61.

13. Noriienstrom 1, Jarstrand C, Wiernik A.Decreased chemotactie and random mi-gration of leukocytes during Intralipid in-fusion. Am J Clin Nun 1979;32(12):2416-22.

14. Battistella FD, Widergren IT, AndersonIT, Siepler IK, Weher IC, MaeColl K. Aprospective, randomized trial of intra-venous fat emulsion administration intrauma victims requiring total parenteralnutrition. / Trauma 1997;43( 1 ):52-6().

15. Fan ST, Lau \V\', Wong KK, Chan YP.Preoperative parenteral nutrition in pa-tients witli oesophageal cancer: a prospec-tive randomized clinical trial. Clin Niitr1989;8:23-7.

16. Figueras I, Puig P, R;ifecas A, Bianchi A,Hernandez F, Pi F, et al. Postoperativehypocaloric parenteral nutrition, A snidyin patients without neoplasm. Acta Chir

17. Sandstrom R, Drott C, Hyltander A., Ar-t%idsso]i B, Schersten T, VVickstroni I, etal. The effect of postoperative intravenousfeeding (TPN) on outcome following ma-jor surger\' evaluated in a randomizedstudy. ^nw.VHrfll993;217(2):185-95.

18. Hu SS, Fontaine F, Kelly B, Bradford DS,Nutritional depletion in staged spinal re-constnictive surgcr.'. The effect of total par-enteral nutrition. Spine 1998;23:I401-5.

19. Reiily J, Mehta R, Teperman L, Cemaj S,Tziikis A, Yjnaga K, et al. Nutritional sup-port attcr li\er transplantation: A random-ized prospective study |comnicnt|. JPE-NJ Paraiter Enteral Nutr 1990;14(4l:386-91. Ck)mment in: JPEN J Parniter En-teral Nutr 1991;15(5):583-4.

20. Hwang TL, Mou SC, Chen MF. The im-portance of a source of sutticicnt proteinin postoperative hypocaioric partial par-enteral nutritional support. /PEN J Par-enter Enteral Nutr 1993; 17:254-256.

21. MuUcr JM, Keller PiW, Brenner U, WalterM, Holzmuller W, Indications and etlcctsof preoperative parenteral nutrition [re-view]. World J Surji \986;\0{\):S^-6^.

22. Naylor AF. Small sample considerations incombining 2 x 2 tables. Biometries 1967;23:349-56.

23. Rothman IR. Modem epidemiology.Boston: Litde Bmwii; 1986. p. 177-237.

24. Rothman KI, Boice ID. Epidcmioiogicalanalysis with a programmable calculator.Washington: National Institutes ofHealth; 1979. paper 79-1649,

25. Berlin lA, Laird NM, Sacks HS, C:halmersTC. A comparison of statistical methodsfor combming event rates from clinical tri-als. StatMed 1989;8(2):141-51.

26. Whttehead A, Whitchead ), A generalparametric approach to the mcta-ana!ysisof randomized trials. Stat Med 1991;10:i665-77.

27. Garden OJ, Smith A, Harris NW, ShenkinA, Sim AI, Ciirter DC. The ertect of iso-tonic amino acid infijsions on scrum pro-teins and muscle breakdown followingsiirgcn. Br/Surj! 1983;70(2):79-82,

28. Hogbin BM, Smith ,VV1, Craven AH. Anevaluation of peripheral essential aminoacid infusion following major surgery.fPFN J Parenter Enteral Nntr 1984;8:511-4.

29. Doglietto GB, tallitclli I., Pacclli F, Bel-lantone R, Malerba M, Sgadari A, er al.Protein-sparing therapy after major ab-dominal surgery-: lack of clinical etlects.Protein-Sparing Therapy Study Cin)up.Ann Siirjj 1996;223 (4):357-62.

110 Journal canadien de chirurgie, Vol, 44, N"2, avri! 2001

Parenteral nutrition in the surgical patient

30. Jimenez FJ, Leyba CO, Jimenez I,M,V'aldeirasJis MS, Montero JG. Study ot"li jKKralorit peripheral parenteral nutritionin postoperative patients, (.'.lin Nutr1995;14:88-96.

31. Bremiiin Mf, Pisters PW, Posner M, Qiic-sada (), Shikc M. A prospective random-ized trial of total parentenl nutrition aftermajor pancreatic resection ft)r malignancy.Ami Surs 1994; 220(4):436-44.

32. Askanazi J, HL-nsle TW, Starker PM,Lockhan SH, LaSala PA, CXsson C, et al.Effect of immediate postoperative nutri-tional support on lenjith of hospitaliza-tion. Ann Surg i986;203 (3j:236-9.

33. Thompson BR, Julian TB, Scrumple JW.Perioperative total parenteral nutrition inpatients with gastrointestinal cancer. /Surfj Res ]98\;iOA97-^00.

34. Fan ST, U) CM, Lai EC, Chu KM, LiuCL, Wong J. Perioperativc nutritionalsupport in patients undergoing hepatec-Comy for hepatocclkilar carcinoma. N En^l/,Ufrfl994;33I(23): 1547-52.

35. Abel RM, Fischer JE, Buckley MJ, BarnettGO, Austen WG. Malnutrition in cardiacsurgical patients. Results of a prospective,randomized evaluation of eariy postojxra-tive parenteral nutrition. Arch Surcj 1976;

43. Freund H, Hoover HC!], Atamian S, Fis-cher JE. Intlision of the hranched chainamino acids in postojwrativc patients. Ann

elective abdominal surgety. hifusionthera-

36. Smith RC, Hartcmink R. Improvement ofnutritional measures during preoperativeparenteral nutrition in patients selected bythe prognostic nutritional index: A ran-domized controlled tnal. JPEN J ParenterEnteral Nutr 1988;12(6);587-91.

37. Holter AR, Fischer JE. The effects of pcri-operative hyperalimentadon on complica-tions in patients with careinonia andweight loss. JSurjf Res I977;23:31-4.

38. Meguid MM, Ciirtas MS. Mcgiiid V,C'ampos AC . EtTects of pie-operative TPNon surgical risk — prcliininaiy status re-port. Br} Clin Pract Suppl 1988;63:53-8.

.^9. Woolfson AM, Smitli JA. Elective nutri-tional support after major surgcr\': aprospective randomized trial. CHn Niitr1989;8:15-21.

40. Von Mcyenfcldt ME, Meijennk WJ, Rou-Hart MM, Builmaasen MT, Soeters PB.l\-rioperati\e nutrirional support: a ran-domized clinical trial. Clin Nutr 1992;11:180-6.

41. Vamada N, Koyama H, Hioki K, YamadaT, Yamamoto M. Effect of postoperativetotal parenteral nutrition (TPN) as an ad-junct to gastrcctomy for advanced gastriccarcinoma. BrJSiirji 1983;70(5):267-74.

42. Gys T, Peeters R, Hubens A. The value ofshort-term peripheral parenteral nutritionafter colorccta! surgcr>: a comparativestudy with conventional postoperative in-travenous fluid. Acta Cbir Bcljt 1990; 90:234-9.

44. Bower RH, Muggia-Sullam M, VallgrenS, Hurst )M, Kern KA, LaFrance R, et al.Branched ch<un amino acid-cnrichcd s(.>lu-tions in the septic patient. A randomized,prospective trial. Ann Sarjj 1986;203(l):13-20.

45. Vente JP, Soeters PB, von MeyenfeldtMF, Routlart MM, \\m der Linden CJ,Gouma DJ. Prospective randomized dou-ble-blind trial of branched chain aminoacid enriched versus standard parenteralnutrition solutions in traumatized and sep-tic patients. World J Sur^ii 1991;15(1):128-33.

46. Brow]i RC), Biionpane EA, Vehe KL,Hickerson Wl., Luther RW. C^omparisonof modified amino aeids and standardamino acids in parenteral nutrition sup-port of thermally injured patients. CritCan-Mfd \99{):, 18( I0):1096-101.

47. Preshaw RM, Attisha RP, HollingsworthWJ, Todd JD. Randomized sequentialtrial of parenteral nutrition in healing ofcolonic anastomoses in man. Can J Sitrj^1979; 22:437-9.

48. Heatley RV, Williams RH, Lewis MH.Pre-operative intravenous feeding — acontrolled trial. Postgrad Med j I979;55(646):54l-5.

49. Moghissi K, Horn.shaw ), Teasdale PR,Dawes EA. Parenteral nutrition in carci-noma of the esophagus treated by surgery:nitrogen balance and clinical studies. Br ]Sw)2rl977;64:125-8.

50. Collins JP, Oxby C:B, Hill GL. Intra-venous aminoacids and intravenous hyper-alimentation as protein-sparing therapy af-ter major surger\'. l-ancct l979;l:788-92.

51. Jensen S. Clinical effects of enteral andparenteral nutrition preceding cancersurger\'. Med Oncol Tumor Pharmacather1985;2:225-9.

52. Hensle TW. Protein-sparing in cystcctoniypatients. / Urol 1978;119: 355-8.

53. Bellantone R, Doglietto G, Bossola M,Pacelli F, Negro E, Sotb L, et al. Preopera-tive parenteral nutntion of malnourishedsurgical patients. Aeta Chir Scand1988;154(4):249-51.

54. Herndon DN, Stein ML>, Rutan TC, Ab-ston S, Linares H. Failure of TPN supple-mentation to improve liver ftmction., im-munity, and mortality' in thermally injuredpatients./Tra«wfl 1987;27{2): 195-204.

55. C'ulebras-Fernandez JM, de la Hoz RiescoM, Garcia CV, Fernandez-LlamazaresGFl, Villalba AA. Jmprovement of thenutritional condition with hypocaloricperipheral parenteral nutrition (HPPN) inthe immediate postoperative period of

56. Bonaii RA, Jeevanandam M, Daly JM.High-branched chain amino aeid solu-tions: relationship of composition to effi-cacy. JPEN J Panntcr Enteral Nutr 1984;8(6):622-7.

57. Neuvonen P, Salo M, Perttila J, Havia T.Lack of modulation of postoperati\'e im-munosuppression by isotonic amino acidinfusion. JPEN J Parenter Enteral Nutr1986;10(2): 160-5.

58. Simnis JM, Oliver K, Smith JA. A study oftotal parenteral nutrition (TPN) in majorgastric and esophageal resection fbr neo-plasia. JPEN J Parenter Enteral NutrI98O;4:422.

59. Hansel! DT, Davies JW, Shenkin A, Gar-den O| , Burns HJ, C:arter DC. "l*he effectsof an anabolic steriod and peripherally ad-ministered intravenous nutrition in theearly postoperative period. JPEN J Par-enter Enteral Nutr \9H9-\ 3(4):349-58.

60. Meta-analysis under scrutiny [editorial].

61. Oxman AD, Guyatt GH. Guidelines forreading literature reviews. CMAJ 1988;138:697-703.

62. Flynn MB, I^eigluty FF. PreojTcrative out-patient nutiitional support of patients withsquamous cancer of tlie upper aerodiges-tive tract. >1»J/ . .W^ 1987;154(4):359-62.

63. Shukla HS, Rao RR, Banu N, Gupta KM,Yadav RC. Enteral hyperalimentation inmalnoiLrished surgical patients. Indian }A/fr-rfK«1984;80: 339-46.

64. Beier-Holgersen R, Boesby S. Influence ofpostoperative entera! nutrition on postsur-gical infecti.)ns. Gut l996;39:833-5.

65. Hasse JM, Blue I.S, Licpa ClU, GoldsteinRM, Jennings LW, Mor E, et al. Early en-teral nutrition support in patients under-going liver transplantation. JPEN J Par-enter Enteral Nutr 1995;I9(6):437-43.

66. Klein S, Kinney J, Jeejeebhoy K, Alpers D,Hellerstein M, Murray M, et al. Nutritionsupport in clinical practice: review of pub-lished data and recommendations for Ri-turc research directions. National Insti-tutes of Health, American Societ\' forParenteral and Enteral Nutrition, andAmerican Society fbr Clinical Nutrition[review]. JPEN J Parenter Enteral Nutr1997;21(S):133-56.

67. Schloerb PR, Henning JF. Patterns andproblems of adult total parenteral nutri-tion use in US academic medical centers.Arch Surg ]998^33:7-'[2.

68. Pomposelli JJ, Baxter JK 3rd. BabineauTJ, Pomfret EA, Driscoll DF, Forse RA,et al. Eariy postoperative glucose controlpredicts nosoconiial infection rate in dia-betic patients. JPEN J Parenter Enteral

Canadian Journal of Surgery, Vol. 44. No 2, April 2001 111