ABSTRACT
This chapter discusses two issues that must precede a statistical meta-analysis. First, a suitable “effect measure” or “metric” must be chosen; for example, the prevalence of a disease, a treatment effect, or a correlation between variables. Choice of the effect measure depends on the aims of the analysis and the type of data, but statistical issues are also relevant. Second, once a systematic review has identified suitable studies, data from each study must be extracted in a way that allows estimation of the chosen effect measure. We discuss some pitfalls and tricks in doing this, with particular focus on difficulties with time-to-event data and on settings where insufficient data may be reported; for example, cluster-randomized trials where the intra-class correlation may not be reported.
