We conducted a nested case-control study to evaluate whether polymorphisms in two genes involved in estrogen metabolism, CYP17 and HSD17B1, were useful in developing a breast cancer risk model that could help discriminate women who are at higher risk of breast cancer. If polymorphisms in these genes affect the level of circulating estrogens, they may directly influence breast cancer risk. The base population for this study is a multiethnic cohort study that includes African-American, Non-Latina White, Japanese, Latina, and Native Hawaiian women. For this analysis, 1508 randomly selected controls and 850 incident breast cancer cases of the first four ethnic groups who agreed to provide a blood specimen were included (76 and 80% response rates, respectively). The CYP17 A2 allele and the HSD17B1 A allele were considered "high-risk" alleles. Subjects were then classified according to number of high-risk alleles. After adjusting for age, weight, and ethnicity, we found that carrying one or more high-risk alleles increases the risk of advanced breast cancer in a dose-response fashion. The risk among women carrying four high-risk alleles was 2.21 [95% confidence interval (CI), 0.98-5.00; P for trend = 0.03] compared with those who carried none. This risk was largely limited to women who were not taking hormone replacement therapy (relative risk, 2.60; 95% CI, 0.95-7.14) and was most pronounced among those weighing 170 pounds or less (RR, 3.05; 95% CI, 1.29-7.25). These findings suggest that breast cancer risk has a strong genetic component and supports the theory that the underlying mechanism of "complex traits" can be understood using a multigenic model of candidate genes.