Oral contraceptive use and cardiovascular disease: Is the relationship real or due to study bias?

Epidemiologic studies link oral contraceptive (OC) use with several cardiovascular events, but the literature is difficult to summarize, and potential biases remain poorly addressed. This study uses meta-analysis to summarize study results and to analyze the influence of study characteristics, including susceptibility to bias, on study outcome. 47 case-control and cohort studies of OCs and 4 cardiovascular events were coded for relative risk (RR) and study characteristics, including adherence to 14 bias-control standards. Key RRs were pooled to summarize findings for each disease type. Univariate determinants of the magnitude of the RRs were identified, and partial correlation analysis was performed for each disease type. RRs were significantly greater then 1.0 for venous thromboembolism (RR=2.8, confidence interval ,CI]=2.4-3.2), stroke (RR=1.8, CI=16-2.0), and myocardial infarction (RR=1.6, CI=1.4-1.8, but not for death due to any cardiovascular cause (RR=1.0, CI=0.8-1.3). Study characteristics were diverse, and potential biases were frequently uncontrolled. For 3 of 10 study characteristics identified as independently influencing RR, methodologically stronger studies of venous thromboembolism tended to have higher RRs. In studies that were methodologically stronger where variables were identified as important, the RRs for stroke and myocardial infarction were stronger. In studies with cardiovascular death, bias-control standards identified as important were generally handled by the studies. OC use does not appear to increase overall cardiovascular mortality. The associations noted with stroke and myocardial infarction may be due to methodologic flaws within the studies, while the association with venous thromboembolism is more likely to be valid. author's modified