Meta-analysis of randomized trials for health care interventions: one for all?

Abstract

Meta-analysis of randomized trials stands at the top of the epidemiological designs to test the effect of health care interventions. Ideally, it systematically and comprehensively collects the relevant randomized trials, and pools the trials to generate a single estimate of effects. However, the rationale for meta-analysis may not always be thoughtfully considered, and overlooking underlying principles may cause illusions. This article revisits the rationale for meta-analysis, discusses how meta-analysis can be misleading, and summarizes approaches to critically and judiciously interpret and use the results of meta-analysis.