161x Filetype PDF File size 0.09 MB Source: www.nlc-bnc.ca
Review Article Return to April 2001 Table of Contents Article de revue Total parenteral nutrition in the surgical patient: a meta-analysis Daren K. Heyland, MD, MSc;* Max Montalvo, MD;† Shaun MacDonald, MD;* Laurie Keefe, RD;‡ Xiang Yao Su; John W. Drover, MD‡ Objective: To examine the relationship between total parenteral nutrition(TPN) and complication and death rates in surgical patients. Data sources: A computer search of published research on MEDLINE, personal files and a review of relevant reference lists. Study selection: A review of 237 titles, abstracts or papers. Primary studies were included if they were randomized clinical trials of surgical patients that evaluated 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 were abstracted on the methodology and outcomes of primary studies. Data were independently abstracted in duplicate. Data synthesis: There were 27 randomized trials in surgical patients that compared the use of TPN to standard care (usual oral diet plus intravenous dextrose). When the results of these trials were aggregated, there was no effect on mortality (risk ratio = 0.97, 95% confidence intervals, 0.76 to 1.24). There were fewer major complications in patients who received TPN, although there was significant heterogeneity in the overall estimate (risk ratio = 0.81, 95% CI, 0.65 to 1.01). Because of this signifi- cant heterogeneity, several a priori hypotheses were examined. Studies that included only malnourished patients demonstrated a trend to a reduction in complication rates but no difference in death rate when compared with studies of patients who were not malnourished. Studies published in 1988 or earlier and studies with a lower methods score were associated with a significant reduction in complication rates and a trend to a reduction in death rate when compared with studies published after 1988 and studies with a higher methods score. There was no difference in studies that provided lipids as a component of TPN when compared with studies that did not. Studies that initiated TPN preoperatively demonstrated a trend to a reduction in complication rates but no difference in death rate when compared with studies that initiated TPN postoperatively. Conclusions: TPN does not influence the death rate of surgical patients. It may reduce the complication rate, especially in malnourished patients, but study results are influenced by methodologic quality and year of publication. Objectif : Examiner le lien entre la nutrition parentérale totale (NPT) et les taux de complication et de mortalité chez les patients en chirurgie. Sources de données : Recherche informatique dans des recherches publiées sur MEDLINE et dans des dossiers personnels, et examen de listes de documents de référence pertinents. Sélection d’études : Revue de 237 titres, abrégés ou communications. On a inclus des études principales s’il s’agissait d’études cliniques randomisées portant sur des patients en chirurgie et qui ont évalué l’effet de la NPT (comparativement à l’absence de NPT ou aux soins normaux) sur les taux de complication et de mortalité. On a exclu les études de comparaison de la NPT à l’entéronutri- tion. Extraction des données : On a abrégé les données pertinentes sur la méthodologie et les résultats des études principales. Les données ont été abrégées de façon indépendante et en double. Synthèse des données : Il y avait 27 études randomisées portant sur des patients en chirurgie au cours desquelles on a comparé l’utilisation de la NPT aux soins normaux (alimentation orale habituelle et dextrose par voie intraveineuse). L’agrégation des résultats de ces études n’a révélé aucun effet sur la mortalité (risque relatif = 0,97; intervalles de confiance à 95 %, 0,76 à 1,24). Il y avait moins de complications majeures chez les patients alimentés par NPT, même si l’on a constaté une hétérogénéité importante dans l’esti- From the *Department of Medicine and †Department of Surgery, Queen’s University, Kingston, Ont., and ‡Nutritional Services, Kingston General Hospital, Kingston, Ont. Dr. Heyland is a Career Scientist of the Ontario Ministry of Health. Accepted for publication Apr. 3, 2000. Correspondence to: Dr. Daren K. Heyland, Angada 3, Kingston General Hospital, 76 Stuart St., Kingston ON K7L 2V7; fax 613 548- 2577, dkh2@post.queensu.ca © 2001 Canadian Medical Association 102 Journal canadien de chirurgie, Vol. 44, No2, avril 2001 Parenteral nutrition in the surgical patient mation globale (RR = 0,81; IC à 95 %, 0,65 à 1,01). À cause de cette hétérogénéité importante, on a examiné plusieurs hypothèses a priori. Des études qui ont porté uniquement sur des patients sous- alimentés ont démontré une tendance à la réduction des taux de complication, mais aucune différence au niveau du taux de mortalité comparativement aux études portant sur des patients qui n’étaient pas sous-alimentés. Par rapport aux études publiées après 1988 et ayant obtenu un résultat méthodologique plus élevé, les études publiées en 1988 ou antérieurement et les études ayant obtenu un résultat méthodologique moins élevé affichaient une réduction importante des taux de complication ainsi qu’une tendance à une réduction du taux de mortalité. On n’a constaté aucune différence dans les études qui comportaient l’administration de lipides par NPT comparativement aux études qui n’en com- portaient pas. Les études où on a entrepris la NPT avant l’intervention ont révélé une tendance à la ré- duction des taux de complication, mais aucune différence au niveau du taux de mortalité comparative- ment aux études où on a commencé à administrer la NPT après l’intervention. Conclusions : La NPT n’a pas d’effet sur le taux de mortalité chez les patients en chirurgie. Elle peut réduire le taux de compli- cation, particulièrement chez les patients sous-alimentés. Les résultats des études varient cependant en fonction de la qualité de la méthodologie et de l’année de publication. he consequences of major may be associated with a lower rate • intervention — any form of TPN Tsurgery can lead to hypermetab- of postoperative complications. Be- (protein, carbohydrates with or olism and subsequent malnutrition.1,2 cause a number of trials have been without lipids) compared to no The patient’s previous nutritional sta- published subsequent to the meta- TPN tus, the concomitant or underlying analysis, we decided to conduct an- • outcome — complications, length disease, and the degree and duration other meta-analysis to summarize the of hospital stay and mortality. of other stresses can contribute to the current literature. We elected to include only ran- risk of malnutrition.3 Malnutrition domized trials in this review because can lead to depletion of body mass, Methods studies in which treatment is allocated impaired tissue and organ function, by any other method than randomiza- compromised immunity and poor Search strategy tion tend to show larger (and fre- wound healing. A strong association quently false-positive) treatment ef- 10 exists between malnutrition and in- We conducted a computerized fects than do randomized trials. Since creased postoperative morbidity and bibliographic search of MEDLINE the scope of our review was defined by mortality in surgical patients.4,5 (including pre-MEDLINE) from our research question, we also ex- The administration of total par- 1980 to May 1999 to locate all rele- cluded studies that compared TPN to enteral nutrition (TPN) can clearly vant articles. The terms “randomized enteral nutrition or other forms of prevent the effects of starvation in controlled trial,” “double blind TPN. Finally, studies that only evalu- patients with a nonfunctioning gas- method,” “clinical trial,” “placebo” ated the impact of TPN on nutritional trointestinal tract. However, it is un- and “comparative study” were com- outcomes (i.e., nitrogen balance, clear whether TPN can modulate the bined with “parenteral nutrition, to- amino acid profile) were excluded. Al- catabolic response to surgical stress tal.” Citations were limited to Eng- though these end points may explain and reduce complications associated lish studies reporting on adult underlying pathophysiology, we con- 5 11 with hypercatabolism. Put differ- patients. We also searched reference sidered them as surrogate end points, ently, the perioperative administra- lists of relevant review articles and and we only included papers that tion of TPN may result in significant personal files. reported on clinically important out- improvement in weight, nitrogen comes (morbidity and mortality). balance, prealbumin levels and other Selection criteria nutritional end points, but the effect Methodologic quality of primary on clinically important end points, Initially, 2 investigators screened studies such as mortality and complications, all citations and classified them into is less certain. The purpose of this primary studies, review articles or We assessed the methodologic paper is to systematically review, crit- others. We then retrieved and inde- quality of all selected articles indepen- ically appraise and statistically aggre- pendently reviewed all primary stud- dently in duplicate, according to the gate all studies evaluating the effect ies. They were included in this scoring system shown in Table 1. of TPN on complication and death overview if they met the following Even in randomized trials, failure to rates in surgical patients. criteria: prevent foreknowledge of treatment A number of clinical trials6–8 and a • research design — randomized assignment can lead to an overestima- 9 12 meta-analysis have suggested that clinical trial tion of treatment effect. Accordingly, preoperative administration of TPN • population — adult surgical we scored higher those studies that in severely malnourished patients subjects reported that their randomization Canadian Journal of SurgeryCanadian Journal of Surgery, Vol. 44, No, Vol. 44, No.. 2, April 20012, April 2001 103103 Heyland et al schema was concealed. Given the diffi- the apparent effect of the interven- ologic quality of 7 and greater to culties of blinding the administration tions across studies) is often found. those with a score of less than 7 of TPN, we only awarded points for When present, heterogeneity weakens (median score = 7). studies that blinded the adjudication any inferences that can be made from • Since the practice of providing of study end points. We also evaluated the results. The possible sources of nutritional support and the man- the extent to which consecutive, eligi- variation include the role of chance or agement of surgical patients has ble patients were enrolled in the trial, differences across studies in popula- evolved over time, we divided the whether groups were equal at baseline, tion, intervention, outcome and studies into groups comparing if cointerventions were adequately de- methods. A priori, we developed sev- studies published in 1988 or ear- scribed, whether objective definitions eral hypotheses that might explain lier with studies published since of infectious outcomes were employed heterogeneity of study results. 1989 (halfway point of this study and whether all patients were properly • We considered that the premor- period). accounted for in the analysis (inten- bid nutritional status of study • There are several randomized trials tion-to-treat analysis) (Table 1). patients was a possible cause of of surgical patients that examine variation in results. Whenever the effect of amino acid infusion Data extraction possible we grouped the results alone or in combination with a of studies that included only pa- carbohydrate source of calories Two investigators extracted data for tients who were malnourished (without the addition of lipids) analysis and assessed the methodologic and compared them to the results on clinical outcomes. We hypothe- quality; we resolved disagreement by of studies that included patients sized that there may be some ad- consensus. Not all studies reported who were not malnourished at verse effects from the use of complication rates. Some reported to- the time of entry into the study. lipids.13,14 Accordingly, we sepa- tal complications per group not per When possible, we used the defi- rated trials into those that included patient. When data were missing, un- nition of malnourished provided lipids and those that did not. clear or not reported on a per patient in each individual study. If no de- • We speculated that differences in basis, we attempted to contact the pri- finition was provided, we as- the timing of the intervention mary investigators to provide further sumed patients who had greater may account for different results. information if the paper had been than 10% weight loss to be mal- To test this hypothesis we published in the last 5 years. nourished. planned a separate analysis com- • We hypothesized that study paring studies that initiated TPN Prior hypotheses regarding sources results may be related to the preoperatively to studies that of heterogeneity methodologic quality of the started TPN postoperatively. study. We planned a separate When conducting a meta-analysis, analysis comparing the effect of Analysis heterogeneity (major differences in studies with an overall method- The primary outcome was periop- Table 1 erative death (death within 30 days of operation) or death in hospital. Criteria Used to Assess Methodologic Quality The secondary outcome was major Score complications. We defined major Criterion 0 1 2 complications as pneumonia, intra- Randomization Not concealed or Concealed abdominal abscess, sepsis,catheter- not sure related infection, myocardial infarc- Blinding Not blinded Adjudicators tion, pulmonary embolism, heart fail- blinded ure, stroke, renal failure, liver failure Analysis Other Intention-to-treat and anastomotic leak. Minor compli- Patient selection Selected patients Consecutive eligible or cannot tell patients cations were defined as wound infec- Comparability of groups No or not sure Yes tion, phlebitis, urinary tract infection at baseline and atelectasis. In 5 studies, the data Extent of follow-up <100% 100% were not portrayed in a fashion that Treatment protocol Poorly described Reproducibly described allowed us to report major complica- Cointerventions* Not described Described but not Well described and tion rates so we reported total com- equal or not sure all equal plications8,15,16 and total infectious Outcomes Not described Partially described Objectively defined *The extent to which antibiotics, enteral nutrition, ventilation, oxygen and transfusions were applied equally across groups. For complications.17,18 There were some questions 1 to 3 and 8 and 9, possible score 0, 1 or 2. For questions 4 to 7, possible score 0 or 1. Total possible score is 14. studies in which their reporting 104 Journal canadien de chirurgie, Vol. 44, No2, avril 2001 Parenteral nutrition in the surgical patient methods did not allow us to disag- these potentially eligible papers, 27 malnourished patients, TPN was as- gregate infectious from noninfec- met the inclusion criteria.6-8,15–21,27–43 sociated with a significant reduction tious complications. One study ran- There was 100% agreement on in complication rates (RR = 0.52, domized patients to 3 groups the inclusion of articles for this over- 95% CI, 0.30 to 0.91). The RR of (control versus standard TPN versus view. Reasons for excluding relevant major complications in studies of TPN with branched-chain amino randomized studies included studies patients who were not malnourished 19 44–46 acids). We only included data from evaluating different kinds of TPN, was 0.95 (95% CI, 0.75 to 1.21). the control group and the standard pseudorandomized studies,47–52 dupli- When we compared the complica- TPN group. Two other studies ran- cate publications,53,54 studies not re- tion rates associated with TPN in domized patients to 3 groups (con- porting clinically important out- studies of patients who were not trol versus TPN without lipids versus comes,55–57 a study available in malnourished with the rate in studies 58 TPN with lipids) and we included abstract form only and a study that of malnourished patients, the differ- both experimental groups in the also randomized patients to anabolic ences were just short of statistical 7,20,21 59 analysis. We also reported on the steroids. significance (p = 0.066). duration of hospital stay, although We compared trials with a meth- these data were not aggregated ow- Impact of total parenteral nutrition ods score of less than 7 to trials with ing to infrequent and variable report- on death and complication rates a score of 7 or better (Fig. 3). Trials ing methods. Agreement between with the higher methods score reviewers on the inclusion of articles The 27 randomized trials, involv- demonstrated no effect of TPN on was measured by weighted kappa. ing 2907 patients, compared the use mortality (RR = 1.08, 95% CI, 0.81 We combined data from all studies of TPN to standard care (usual oral to 1.43), whereas trials with a score to estimate the common relative risk diet plus dextrose given intra- of 7 or less suggested a trend toward of death and complications and asso- venously) in patients who underwent a reduction in mortality associated ciated 95% confidence intervals (CIs). surgery.6–8,15–21 The details of each with the use of TPN (RR = 0.75, We summarized the treatment effect study are described in Table 95% CI, 0.47 to 1.19). The test for 6–8,15–21,27–43 using risk ratios (RRs). To avoid the 2. When the results of these heterogeneity across subgroups was problem with bias and instability as- trials were aggregated, there was no not significant (p = 0.21). With re- sociated with RR estimation in sparse effect on mortality (RR = 0.97, 95% spect to complication rates, in studies 22 data, we added one-half to each cell. CI, 0.76 to 1.24) (Fig. 1). The test with a higher methods score there In the meta-analysis, we used maxi- for heterogeneity was not significant was no effect of TPN on major com- mum likelihood methods of combin- although a visual inspection suggests plications (RR = 1.07, 95% CI, 0.86 ing RRs across all trials and examined that the treatment effect of some of to 1.32). In studies with a lower the data for evidence of heterogeneity the studies was significantly different methods score, there was a significant within groups.23 The Mantel– from other studies. Twenty-two reduction in complication rates (RR Haenzel method was used to test the studies reported major complica- = 0.50, 95% CI, 0.32 to 0.76). The 24 significance of treatment effect. We tions. When these results were aggre- test for heterogeneity across sub- used a random effects model to esti- gated, TPN was associated with a re- groups was significant (p = 0.005). mate the overall RR.25,26 For the test duction in complication rates (RR = We next compared trials pub- of heterogeneity across subgroups, 0.81, 95% CI, 0.65 to 1.01, p = lished in 1988 or earlier with studies the t-test for the difference between 0.06) (Fig. 2). The test for hetero- published since 1989 (see Fig. 3). the 2 subgroups was used. We con- geneity was significant (p = 0.01). Trials published in 1988 or earlier sidered a p value of less than 0.05 to Given that we found significant were associated with a trend toward be statistically significant. heterogeneity and in an attempt to a decrease in death rates associated better explain our findings, we exam- with the use of TPN (RR = 0.68, Results ined our a priori hypotheses. We 95% CI, 0.43 to 1.10). Trials pub- compared those trials that included lished since 1989 were consistent Study identification and selection only malnourished patients with with no treatment effect associated other trials. TPN was not associated with TPN (RR = 1.11, 95% CI, 0.83 In all 187 citations were identified with any difference in mortality to 1.48). The test for heterogeneity from the MEDLINE databases. Our (Fig. 3) in studies of malnourished across subgroups was short of con- personal files and review of reference patients (RR = 1.13, 95% CI, 0.75 to ventional levels of significance (p = lists yielded 57 additional articles for 1.71) or in studies of normally nour- 0.10). With respect to complication consideration. Initial eligibility ished patients (RR = 0.90, 95% CI, rates, in studies published in 1988 or screening resulted in 47 articles se- 0.66 to 1.21, p = 0.38 for differences earlier there was a significant reduc- lected for further evaluation. Of between subgroups). In studies of tion in major complications associ- Canadian Journal of Surgery, Vol. 44, No.2, April 2001 105
no reviews yet
Please Login to review.