Controversy of oral contraceptives and risk of rheumatoid arthritis: meta-analysis of conflicting studies and review of conflicting meta-analyses with special emphasis on analysis of heterogeneity.

ABSTRACT

The authors analyze the heterogeneity present in the combined results of past observational studies that investigated the association between oral contraceptive use and rheumatoid arthritis. The authors also evaluate discrepancies among meta-analyses that focus on the same relation. Of the 15 initially reviewed studies, 10 were selected for this meta-analysis, which also includes a qualitative summary of study characteristics and a critical appraisal of study quality. The authors used the direct method to combine the study results when there was no evidence of heterogeneity and the DerSimonian-Laird method when heterogeneity was present. Using a meta-regression to assess the sources of heterogeneity, the authors weighted summary estimates by sample size and undertook a sensitivity analysis. There was a strong indication of heterogeneity when combining all studies (x2 = 29.34, p = 0.00060) with the source of controls explaining most of the heterogeneity. The most important factor in explaining the differences among the overall summary estimates given by the meta-analyses is that different effect estimates had been selected for the same studies. There is no conclusive evidence of a protective effect of oral contraceptives on the risk of developing rheumatoid arthritis. Consensus is needed on how meta-analyses of observational studies should be conducted.

ABSTRACT

PIP Epidemiologists extracted data from 10 different studies to conduct a meta-analysis in order to compare the differences among the case-control studies and cohort studies that examine the association between oral contraceptives (OCs) and rheumatoid arthritis. They also analyzed the differences among four meta-analyses that study the same association. The epidemiologists used the direct method to combine the findings of the 10 studies when heterogeneity was absent and the DerSimonian-Laird method when heterogeneity was present. They applied a meta-regression to assess the sources of heterogeneity. Specifically, the epidemiologists weighted summary estimates by sample size and conducted a sensitivity analysis. When all the studies were combined, heterogeneity was strong (p = 0.0006). The source of controls explained most of the heterogeneity. As for the meta-analyses, the most important factor in accounting for the differences among the overall summary estimates is that the researchers had selected different effect estimates for the same studies. It is still not clear whether or not OCs have a protective effect against developing rheumatoid arthritis. Epidemiologists need to reach a consensus on how they should conduct meta-analyses of observational studies.