We report a nested case-control study of serum biomarkers of 5 alpha-reductase activity and the incidence of prostate cancer. From a cohort of more than 125,000 members of the Kaiser Permanente Medical Care Program who underwent multiphasic health examinations during 1964-1971, we selected 106 incident prostate cancer cases. A control was pair matched to each case on age, date of serum sampling, and clinic location. Serum levels of total testosterone, free testosterone, androsterone glucuronide, and 5 alpha-androstane-3 alpha,17 beta androstanediol glucuronide (3 alpha-diol G) were measured on the stored samples and scored as quartiles. Potential confounders included alcohol, smoking, and body mass index. The adjusted odds ratios and 95% confidence intervals for a one quartile score increase were 1.00 (0.75-1.34) for total testosterone, 1.14 (0.86-1.50) for free testosterone, 1.13 (0.84-1.53) for androsterone glucuronide, and 1.16 (0.86-1.56) for 3 alpha-diol G. A limitation of this study is that there are two different 5 alpha-reductase isoenzymes, only one of which is expressed in high levels within the prostate, yet both of which may affect serum biomarkers. Since the two isoenzymes are encoded on different chromosomes, variation in one would act as an independent source of measurement error in any analysis of serum biomarker effects of the other. Consequently, the odds ratios may be underestimated and the study, although negative, cannot exclude the previously hypothesized possibility that a positive relationship between intraprostatic 5 alpha-reductase activity and prostate cancer may exist. A clinical trial to test this hypothesis is under way.