The relation between the risk of breast cancer before 45 years of age and oral contraceptive use was examined in a case-control study conducted in New York, Philadelphia, Baltimore, and Boston from 1983 to 1986 of 407 patients with breast cancer and 424 controls. With allowance for confounding, for ever use, the multivariate relative risk estimate was 2.0 (95% confidence interval (CI), 1.4-2.9). For less than 10 years of use, the estimate approximated 2.0 in all categories of duration, including less than three months; for 10 or more years of use it was 4.1 (95% CI, 1.8-9.3). The association was apparent in virtually all subgroups examined, including younger and older women, and women at low and high underlying risk of breast cancer. Contrary to some previous reports, the association was not stronger for use before a first term pregnancy or at an early age. The results suggest that oral contraceptive users, particularly those with very long durations of use, may be at increased risk of breast cancer. However, information bias, particularly for short-term use, could not be ruled out. There may also have been selection bias if oral contraceptive users were under more intensive medical surveillance. It has not been possible to reconcile the findings of the various studies to date, including the authors' earlier results showing no association. The latter results were derived from data collected using methods almost identical to those used in the present study.