[Meta-analysis: principles and pitfalls].

In a meta-analysis results of different studies are summarized quantitatively. This gives a much more precise effect estimate than is possible in individual studies. Prior to the analysis, all relevant studies should be traced, including studies with less favourable results. Exclusion of 'negative' studies may lead to overestimation of the effect. Because the results of a meta-analysis depend on the quality of the included studies, the quality and risk of bias of all studies should be systematically assessed. Quantitative summary of results is only practicable if the studies are homogeneous, i.e. sufficiently similar with respect to the study populations, interventions or exposure, outcomes and point of measurement. In cases of clinical heterogeneity, meta-analysis could be conducted in clinically homogeneous subgroups. Statistical heterogeneity can be discounted by using a random-effects model. Meta-analyses can be applied to research questions regarding aetiology, diagnosis, treatment or prognosis.