Nutritional support for liver disease.Cochrane Database Syst Rev. 2012 May 16CD
Weight loss and muscle wasting are commonly found in patients with end-stage liver disease. Since there is an association between malnutrition and poor clinical outcome, such patients (or those at risk of becoming malnourished) are often given parenteral nutrition, enteral nutrition, or oral nutritional supplements. These interventions have costs and adverse effects, so it is important to prove that their use results in improved morbidity or mortality, or both.
To assess the beneficial and harmful effects of parenteral nutrition, enteral nutrition, and oral nutritional supplements on the mortality and morbidity of patients with underlying liver disease.
The following computerised databases were searched: the Cochrane Hepato-Biliary Group Controlled Trials Register, the Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library), MEDLINE, EMBASE, and Science Citation Index Expanded (January 2012). In addition, reference lists of identified trials and review articles and Clinicaltrials.gov were searched. Trials identified in a previous systematic handsearch of Index Medicus were also considered. Handsearches of a number of medical journals, including abstracts from annual meetings, were done. Experts in the field and manufacturers of nutrient formulations were contacted for potential references.
Randomised clinical trials (parallel or cross-over design) comparing groups of patients with any underlying liver disease who received, or did not receive, enteral or parenteral nutrition or oral nutritional supplements were identified without restriction on date, language, or publication status. Six categories of trials were separately considered: medical or surgical patients receiving parenteral nutrition, enteral nutrition, or supplements.
DATA COLLECTION AND ANALYSIS
The following data were sought in each report: date of publication; geographical location; inclusion and exclusion criteria; the type of nutritional support and constitution of the nutrient formulation; duration of treatment; any nutrition provided to the controls; other interventions provided to the patients; number, sex, age of the study participants; hospital or outpatient status; underlying liver disease; risks of bias (sequence generation, allocation concealment, blinding, incomplete outcome reporting, intention-to-treat analysis, selective outcome reporting, others (vested interests, baseline imbalance, early stopping)); mortality; hepatic morbidity (development or resolution of ascites or hepatic encephalopathy, occurrence of gastrointestinal bleeding); quality of life scores; adverse events; infections; lengths of stay in the hospital or intensive care unit; costs; serum bilirubin; postoperative complications (surgical trials only); and nutritional outcomes (nitrogen balance, anthropometric measurements, body weight). The primary outcomes of this review were mortality, hepatic morbidity, quality of life, and adverse events. Data were extracted in duplicate; differences were resolved by consensus.Data for each outcome were combined in a meta-analysis (RevMan 5.1). Estimates were reported using risk ratios or mean differences, along with the 95% confidence intervals (CI). Both fixed-effect and random-effects models were employed; fixed-effect models were reported unless one model, but not the other, found a significant difference (in which case both were reported). Heterogeneity was assessed by the Chi(2) test and I(2) statistic. Subgroup analyses were planned to assess specific liver diseases (alcoholic hepatitis, cirrhosis, hepatocellular carcinoma), acute or chronic liver diseases, and trials employing standard or branched-chain amino acid formulations (for the hepatic encephalopathy outcomes). Sensitivity analyses were planned to compare trials at low and high risk of bias and trials reported as full papers. The following exploratory analyses were undertaken: 1) medical and surgical trials were combined for each nutritional intervention; 2) intention-to-treat analyses in which missing dichotomous data were imputed as best- and worst-case scenarios; 3) all trials were combined to assess mortality; 4) effects were estimated by absolute risk reductions.
Thirty-seven trials were identified; only one was at low risk of bias. Most of the analyses failed to find any significant differences. The significant findings that were found were the following: 1) icteric medical patients receiving parenteral nutrition had a reduced serum bilirubin (mean difference (MD) -2.86 mg%, 95% CI -3.82 mg% to -1.89 mg%, 3 trials) and better nitrogen balance (MD 3.60 g/day, 95% CI 0.86 g/day to 6.34 g/day, 1 trial); 2) surgical patients receiving parenteral nutrition had a reduced incidence of postoperative ascites only in the fixed-effect model (RR 0.65, 95% CI 0.48 to 0.87, 2 trials, I(2) = 70%) and one trial demonstrated a reduction in postoperative complications, especially infections (pneumonia in particular); 3) enteral nutrition may have improved nitrogen balance in medical patients (although a combination of the three trials was not possible); 4) one surgical trial of enteral nutrition found a reduction in postoperative complications; and 5) oral nutritional supplements had several effects in medical patients (reduced occurrence of ascites (RR 0.57, 95% CI 0.37 to 0.88, 3 trials), possibly (significant differences only seen in the fixed-effect model) reduced rates of infection (RR 0.49, 95% CI 0.24 to 0.99, 3 trials, I(2) = 14%), and improved resolution of hepatic encephalopathy (RR 3.75, 95% CI 1.15 to 12.18, 2 trials, I(2) = 79%). While there was no overall effect of the supplements on mortality in medical patients, the one low risk of bias trial found an increased risk of death in the recipients of the supplements. Three trials of supplements in surgical patients failed to show any significant differences. No new information was derived from the various subgroup or sensitivity analyses. The exploratory analyses were also unrevealing except for a logical conundrum. There was no difference in mortality when all of the trials were combined, but the trials of parenteral nutrition found that those recipients had better survival (RR 0.53, 95% CI 0.29 to 0.98, 10 trials). Either the former observation represents a type II error or the latter one a type I error.