The purpose of this study was to examine whether a high serum concentration of phytoestrogens reduces the risk of prostate cancer in a case-control study nested in a community-based cohort in Japan (Japan Collaborative Cohort (JACC) Study). Information on lifestyles and sera of the subjects were collected in 1988-90, and they were followed up to 1999. Incident and dead cases of prostate cancer and controls were matched for study area and age. Phytoestrogens and sex hormones in sera stored at - 80 degrees C were measured in 2002. Of 14,105 male subjects of the cohort who donated their sera, 52 cases and 151 controls were identified. Three datasets were analyzed; 1) all subjects, 2) 40 cases and 101 controls after excluding subjects with low testosterone levels who were suspected of having had medical intervention, and 3) 28 cases and 69 controls with prostate specific antigen level of </= 10.0 ng/ml. The odds ratio (OR) for the highest level to the lowest was 0.38 (95% confidence interval (CI); 0.13, 1.13) for genistein, 0.41 (0.15, 1.11) for daidzein, and 0.34 (0.11, 1.10) for equol for the second dataset. Genistein and daidzein showed similar findings in the third one. Equol and equol/daidzein ratio showed consistent findings in all three datasets (OR = 0.39, 95% CI; 0.13, 0.89, trend P = 0.02 for the first dataset). Their effects seemed to be independent of serum sex hormones. In conclusion, serum genistein, daidzein, and equol seemed to dose-dependently reduce prostate cancer risk.