We assessed the association of sex hormone levels with breast cancer risk in a case-control study nested within the cohort of 7054 New York University (NYU) Women's Health Study participants who were postmenopausal at entry. The study includes 297 cases diagnosed between 6 months and 12.7 years after enrollment and 563 controls. Multivariate odds ratios (ORs) (95% confidence interval (CI)) for breast cancer for the highest quintile of each hormone and sex-hormone binding globulin (SHBG) relative to the lowest were as follows: 2.49 (1.47-4.21), P(trend)=0.003 for oestradiol; 3.24 (1.87-5.58), P(trend)<0.001 for oestrone; 2.37 (1.39-4.04), P(trend)=0.002 for testosterone; 2.07 (1.28-3.33), P(trend)<0.001 for androstenedione; 1.74 (1.05-2.89), P(trend)<0.001 for dehydroepiandrosterone sulphate (DHEAS); and 0.51 (0.31-0.82), P(trend)<0.001 for SHBG. Analyses limited to the 191 cases who had donated blood five to 12.7 years prior to diagnosis showed results in the same direction as overall analyses, although the tests for trend did not reach statistical significance for DHEAS and SHBG. The rates of change per year in hormone and SHBG levels, calculated for 95 cases and their matched controls who had given a second blood donation within 5 years of diagnosis, were of small magnitude and overall not different in cases and controls. The association of androgens with risk did not persist after adjustment for oestrone (1.08, 95% CI=0.92-1.26 for testosterone; 1.15, 95% CI=0.95-1.39 for androstenedione and 1.06, 95% CI=0.90-1.26 for DHEAS), the oestrogen most strongly associated with risk in our study. Our results support the hypothesis that the associations of circulating oestrogens with breast cancer risk are more likely due to an effect of circulating hormones on the development of cancer than to elevations induced by the tumour. They also suggest that the contribution of androgens to risk is largely through their role as substrates for oestrogen production.