Meta-analysis and bias in research reviews

This article provides an overview of meta-analysis, a statistical technique that has become common within the health sciences in recent years. In comparison with the traditional narrative review, meta-analysis provides a more effective means of tackling the difficulties that arise in the production of an integrative survey of research on a phenomenon. Even when a solution to a particular methodological problem has not yet been found, meta-analysis often serves a useful function by supplying a framework for characterizing and exploring the nature and extent of the problem. Publication bias is an enduring unsolved problem faced by all methods of research synthesis. Using the framework provided by meta-analysis, it is possible to determine the impact of publication bias on the frequency of spurious conclusions in a discipline. By ascertaining the strength of publication bias in a wide variety of circumstances, it is possible to distinguish those conclusions of a research review in which one can have confidence from those that are less well-founded.