To examine whether endogenous androgens influence the occurrence of prostate cancer, we conducted a nested case-control study among participants enrolled in the Carotene and Retinol Efficacy Trial. We analyzed serum samples of 300 cases diagnosed between 1987 and 1998, and 300 matched controls. Higher concentrations of testosterone (T) were not associated with increased prostate cancer risk. Relative to men with levels in the lowest fourth of the distribution, men in the upper fourth of total T had a risk of 0.82 [95% confidence interval (CI), 0.52-1.29]. The corresponding relative risks for free T (0.72; 95% CI, 0.45-1.14), percentage of free T (0.74; 95% CI, 0.46-1.19), and total T:sex hormone binding globulin ratio (0.52; 95% CI, 0.32-0.83) similarly were not elevated. Higher concentrations of androstenedione, dehydroepiandrosterone sulfate, and 3 alpha-androstanediol glucuronide were weakly associated with risk. Relative risks associated with being in the highest fourth for androstenedione, dehydroepiandrosterone sulfate, and 3 alpha-androstanediol glucuronide were 1.20 (95% CI, 0.76-1.89), 1.38 (95% CI, 0.86-2.21), and 1.27 (95% CI, 0.80-2.00), respectively. Men in the upper fourth of total estradiol (E2), free E2 and percentage of free E2 had relative risks of 0.71 (95% CI, 0.42-1.13), 0.52 (95% CI, 0.33-0.82), and 0.65 (95% CI, 0.40-1.05), respectively. The inverse association between E2 and prostate cancer risk was largely restricted to men with blood collection within 3 years of diagnosis. Our results add to the evidence that serum testosterone is unrelated to prostate cancer incidence. The suggestions that intraprostatic androgen activity may increase risk and that serum estrogens may decrease risk, warrant additional study.