Abstract Body

Background: Meta-analysis is a powerful method for quantitatively synthesizing multiple clinical trials of similar interventions. By combining multiple sources of information a sharper signal about the interventions can be calculated, both overall and in subgroups. While meta-analysis is a summarization of finished studies, pre-specification of the essential research question, study inclusion criteria, and analytic strategy increase rigor of the analysis. Conclusions: Two main statistical calculations for summarizing studies are discussed. The fixed effects approach is best when the intervention effects are relatively homogeneous and provides conclusions about the studies at hand. The random effects approach is more suited for more heterogeneous intervention effects and generalizes beyond the studies at hand. Techniques to explore and address heterogeneity will be discussed as well as more subtle issues such as the bias introduced by non-publication of null studies, the temptation to dismiss “aberrant” studies, methods to accommodate differential quality of studies, and a perspective on the strength of evidence that can be achieved by meta-analysis. Examples involving vaccines and drugs will illustrate the use of meta-analysis to explore safety signals as well as variation in efficacy.