Epidemiologic studies have consistently reported that endogenous steroid hormone levels are associated with postmenopausal breast cancer risk, but little is known on the associations by tumor grade, hormone receptor status, or age at diagnosis. We performed a case-cohort study of naturally postmenopausal women within the Melbourne Collaborative Cohort Study that included a random sample of 857 women and 197 breast cancer cases diagnosed during a mean of 9.2 years of follow-up. Concentrations of total estradiol, estrone sulfate, testosterone, DHEA sulfate, androstenedione, and sex hormone binding globulin were measured in plasma collected at baseline before diagnosis; free estradiol plasma concentration was calculated. Cox regression was used to estimate associations adjusted for known and potential confounders. The HR for breast cancer comparing fourth and first quartiles was 1.44 [95% confidence interval (95% CI), 0.89-2.35] for total estradiol, 1.75 (95% CI, 1.06, 2.89) for free estradiol, 2.05 (95% CI, 1.24-3.37) for estrone sulfate, 1.25 (95% CI, 0.78-2.01) for testosterone, 1.41 (95% CI, 0.88-2.27) for DHEA sulfate, 1.49 (95% CI, 0.91-2.44) for androstenedione, and 0.33 (95% CI, 0.19-0.55) for sex hormone binding globulin. These associations did not differ by tumor grade and estrogen receptor/progesterone receptor status (all test for heterogeneity, P > 0.05). Risks associated with estrogen and androgen levels were stronger at older ages (test for interaction across age groups, P = 0.59 for total estradiol and P = 0.01 for testosterone). Our prospective study confirms earlier findings and suggests that the associations of endogenous hormones with postmenopausal breast cancer risk are independent of tumor grade, and hormone receptor status and might increase in strength with age.