Meta-analysis is the structured and systematic qualitative and quantitative integration of the results of several independent studies (Le. the epidemiology of results). As in any epidemiological study, a meta-analysis needs to start with clearly stated aims and objectives. Attention needs to be paid to selection bias in selecting the study population (all publications on the topic). An initial qualitative assessment (conducted blinded to results) categorises projects on the basis of their methods, as unacceptable (dropped from later analysis) and acceptable or good. Further analysis could be conducted by stratifying or weighting independent studies according to preset quality criteria. The quantitative assessment involves deriving a pooled measure of outcome (usually the relative or attributable risk). Tests for heterogeneity are required before pooling. By pooling the results from many settings using different methods, the ability to generalise them in terms of their public health relevance is increased.
S Afr Med J 1990; 78: 94·97.