Oral contraceptives and thromboembolism: a reassessment.

The relationship between oral contraceptive usage and thromboembolism is controversial. Since thromboembolism is often undiagnosed, both clinically and at routine autopsy, most epidemiologic analyses rest or a very uncertain factual base. There are increases in blood coagulation factors in oral contraceptive users similar to, but less than, those seen in pregnancy, which is not associated with increased thromboembolism. Hematologists emphasize that these changes do not define a "hypercoagulable" state, and they do not define or predict the occurrence of thrombosis. Intrinsic vascular wall changes, unrelated to drug use, may play a role in sporadic cases of thromboembolism. When the incidence of thromboembolism in very large Phase III trials of conventional oral contraceptives is compared to that in other populations (patients admitted to the hospital, women who visit a physician, pregnant women, or users of nonestrogenic oral contraceptives), no difference is seen. Epidemiologic studies by the "case-control" ("trohoc") method consistently show an increase "relative risk" associated with oral contraceptive use in subjects with "idiopathic" thromboembolism but no increased risk in thromboembolism patients as a whole or in those with predisposing factors. This retrospective epidemiologic technique, its particular applications, and the inferences drawn are open to serious criticism, as are studies claiming a relationship between estrogen dose and thromboembolism incidence. An Australian prospective survey found no increased risk among users, and a large British study which initially reported an increased risk is currently undergoing recalculation. The only controlled clinical experiment (with random assignment of subjects to vaginal versus high-estrogen contraceptives) showed no increased incidence in the drug-treated group. Statistical associations derived from "trohoc" studies do not establish causal relationships; moreover, their risk estimates are in conflict with the findings of large Phase III clinical surveys including subjects using estrogen-free contraceptives, with at least one prospective clinical survey, and with a randomized, controlled clinical trial. The data relating estrogen dosage to thromboembolism incidence are ambiguous, at best. Thus, the claim of a causal relationship between oral contraceptive steroids and thromboembolism does not appear to be firmly founded, and the belief that predisposing factors increase the risk to contraceptive users is equally insubstantial.