Based on an extensive literature survey, this report discusses some central topics in meta-analysis, heterogeneity, publication bias and the assessment of quality. For handling heterogeneity it describes random effects methods and meta-regression. It is concluded that meta-regression is preferable when this is possible. A chapter describes statistical methods to diagnose publication bias and simulation studies that evaluate the properties of these methods. The Egger regression method is the preferable test for publication bias. The trim and fill method can be used to assess the influence of publication bias on the overall effect estimate. A chapter argues the case for quality assessment and the necessity of taking quality into account in meta-analyses. (Author/publisher)
Abstract