Meta-analyses: what they can and cannot tell us about clinical research

The number of published clinical trials has increased exponentially over the past few decades. The majority of clinical research questions has been addressed by several small or moderately sized trials. Statisticians have developed guidelines for meta-analyses, which provide objective quantitative reviews to lighten the burden of clinicians who seek to keep themselves current in one or more areas of clinical investigation. The purpose of this review is to illustrate how to read critically a meta-analysis of nutrition support in critically ill patients.

Summary evidence based on a meta-analysis is observational, even though the individual elements of the meta-analysis are clinical trials. The level of evidence from a meta-analysis is thus similar to that derived from a case series or a cohort, with the stipulation that the 'subjects' are clinical trials instead of individuals. The effects of the combination of parenteral and enteral nutrition on mortality and infection rates in critically ill patients have been compared with enteral nutrition alone. Meta-analysis results show non-significant reductions in mortality and infectious complications favoring enteral nutrition. In patients with acute pancreatitis, the summary effects of parenteral versus enteral nutrition on the infection rate indicate a significant reduction in risk favoring enteral nutrition.

An interpretation of meta-analysis summary results is put in perspective with underlying medical assumptions. This critical review indicates what issues a meta-analysis can and cannot address.