Interpretation of data from multiple trials: a critical review.

Abstract

The objective was to review the current status of meta-analysis, the process of combining the results of similar investigations, which was described in 1976. It has evolved rapidly, and thousands have already been published. Meta-analysis may be performed in a narrative form or in the preferred statistical format, which is more rigorous and demanding. The criteria for designing and performing a meta-analysis of high quality are presented and discussed in detail, including selection criteria, quality, biases and heterogeneity, both of component studies and of results. Critical reviews of meta-analyses are presented. Examples are given from my own field--hepatology--and from my own research experience. Although meta-analysis is simple in concept, it is complex in execution. If performed with insufficient insight or attention to detail, it may give ambiguous or erroneous results.