Sources of heterogeneity in the meta-analysis of observational studies: the example of SIDS and sleeping position.

The study objective was to provide an example of how risk estimates might vary across studies of observational design, even when a causal association is present and to explore the possible sources of such variation. A meta-analysis of studies on the association between prone sleeping position and sudden infant death syndrome (SIDS) is used to illustrate how risk estimates might vary across studies. Data used were reported case-control studies of the association between sleeping position and SIDS that were published between 1970 and 1994. If the pooled odds ratio had been relied on to assess the association between sleeping position and SIDS without an accompanying examination of the reasons for heterogeneity, important insights into the causal significance of the relationship may have been lost. In meta-analyses of observational studies it is important to investigate the reasons for heterogeneity across studies.