The presence and effect of bias in trials of early enteral nutrition in critical care

Abstract

Background: Randomized trials suggest that early enteral nutrition is beneficial in critically ill adults. However, methodologic bias can overestimate benefit.

Objective: To assess the potential effect of methodologic bias on these trials.

Study design: Systematic review and meta-analysis.

Data source: Randomized trials identified in electronic searches of PUBMED, EMBASE, and the Cochrane Library, and in various handsearches.

Methods: The primary (mortality, morbidity) and secondary (time on ventilator or in intensive care unit/hospital, cost) outcomes were abstracted from each identified trial comparing early enteral nutrition to no/delayed enteral nutrition. Each trial was assessed for six domains of methodologic bias (sequence generation, allocation concealment, blinding, intention-to-treat, selective outcome reporting, other). No low risk of bias trial (adequate in all six domains) was identified, so such trials could not be compared to the others. Instead, meta-analyses of trials with more or fewer risks were compared in the following ways: adequate methodology to deal with ≥3 or ≤2 domains; Jadad scores ≥3 or ≤2; adequate versus not adequate for each domain.

Data synthesis: In the 15 identified trials, early enteral nutrition appeared to improve mortality and infectious morbidity. Mortality benefit was observed only in trials with more risks of bias; infectious morbidity benefit was observed in some analyses of trials with fewer bias risks.

Limitations: Small numbers of trials and missing information.

Conclusions: The benefits attributed to early enteral nutrition were either seen only in trials with high risks of bias or may result from residual risks of bias.

Keywords: Critical care; Enteral nutrition; Randomized trials; Risk of bias; Systematic review.