OBJECTIVE: To estimate by meta-analysis the risk of developing cancer of the breast, uterine cervix, endometrium, ovary, and liver from age 20-54 years in the United States in women using oral contraceptives (OCs) for 4, 8, or 12 years. DATA SOURCES: Using pairs of terms (such as oral contraception and breast neoplasms), I searched for English-language literature on OC use and cancer published since 1980 and cited through July 1994 in the MEDLINE data base. METHODS OF STUDY SELECTION: I analyzed all epidemiologic studies reporting estimates of relative risk (RR) by duration and recency of OC use (79 independent studies in total). DATA EXTRACTION AND SYNTHESIS: Reported duration-specific and recency-specific estimates of RR, with the corresponding numbers of cases and controls or person-time at risk for cohort studies, were abstracted from each article. Relative risk of cancer as a function of both duration and recency of OC use was then estimated by weighted regression and applied, using life-table methods, to United States population-based data on age-specific mortality and cancer incidence. CONCLUSIONS: For every 100,000 women in the United States who never use OCs, the number developing cancer from age 20-54 years is estimated to be 2782 (breast), 425 (cervix), 438 (endometrium), 369 (ovary), and 20 (liver). For women using OCs for 8 years, the estimated number of additional or fewer cases per 100,000 users is +151 (breast), +125 (cervix), -197 (endometrium), -193 (ovary), and +41 (liver). Therefore, from a population perspective, there are only small cancer-related risks and benefits associated with OC use and, on balance, the net effect is negligible.