Endogenous androgens have been suggested as determinants of risk of prostatic cancer. To examine this possibility, baseline sex hormone levels were measured in 1008 men ages 40-79 years who had been followed for 14 years. There were 31 incident cases of prostatic cancer and 26 identified from death certificates with unknown dates of diagnosis. In this study, total testosterone, estrone, estradiol, and sex hormone-binding globulin were not related to prostate cancer, but plasma androstenedione showed a positive dose-response gradient. Age-adjusted relative risks of prostatic cancer for low (0-2.2 nM), middle (2.3-3.1 nM), and high (3.2+ nM) tertiles of androstenedione were 1.00, 1.34, and 1.98, respectively (P trend less than 0.05). The linear gradient of risk persisted after adjustment for age and body mass index. If confirmed, these data suggest that androstenedione might increase the occurrence of clinically manifest prostatic cancer.