While meta-analytic techniques are routine in the synthesis of data from randomized controlled trials, there are no clear guidelines on how best to summarize frequency data such as incidence and prevalence estimates. Based on data from two recent systematic reviews of the incidence and prevalence of schizophrenia, this paper explores some of the conceptual and methodological issues related to the meta-analyses of frequency estimates in epidemiology. Because variations in the incidence and prevalence of disorders such as schizophrenia can be informative, there is a case against collapsing data into one pooled estimate. Variations in frequency estimates can be displayed graphically, or summarized with quantiles around measures of central tendency. If pooled estimated are of interest, then researchers need to be aware that studies based on large samples will leverage greater weight on the pooled value. Based on systematic reviews of the incidence and prevalence of schizophrenia, we explore if these and related issues are of practical concern. When used with appropriate caution, meta-analysis can complement the synthesis of frequency data in epidemiology; however, researchers interested in variation should not rely on meta-analysis alone.
Copyright (c) 2008 John Wiley & Sons, Ltd.