Improving Breast Cancer Risk Prediction Models: The Addition Of A Genetic Risk Score, Mammographic Density, And Endogenous Hormones

Proceedings: AACR 107th Annual Meeting 2016; April 16-20, 2016; New Orleans, LABackground: Risk prediction models have important public health implications for targeted disease prevention by identifying women at higher risk who would most likely benefit from screening, chemoprevention, or other risk-reducing regimens. Among such models, the Gail and Rosner-Colditz models have been well validated and used to identify high-risk women. Each of these models generally includes only traditional breast cancer risk factors (e.g., parity and family history of breast cancer). No prior study has examined the joint contribution of a genetic risk score (GRS), mammographic density (MD), and postmenopausal endogenous hormone levels to these models.Method: We conducted a nested-case control study within the Nurses’ Health Study (NHS, follow-up 1990-2010) and NHS II (1997-2009). We created a breast cancer GRS based on 67 single-nucleotide polymorphism (SNPs). We measured percent MD as well as circulating testosterone, estrone sulfate, and prolactin levels in prediagnostic plasma samples. Using receiver-operating-characteristic curve analyses, we calculated the area under the curve (AUC), adjusting for age, as a measure of discrimination for the 5-year risk of invasive breast cancer and estrogen receptor (ER) and progesterone receptor (PR) positive disease (ER+PR+) by menopausal status, after adding GRS, MD, and/or circulating hormones to the modified Gail and Rosner-Colditz models.Results: We documented 11,880 women (4,006 cases/7,874 controls; age range: 34-80 yr) for Gail model and 8,210 women (2,665 cases/5,455 controls) for Rosner-Colditz model analyses. In both models, about 45% women were premenopausal, 25% were postmenopausal women not using postmenopausal hormone (HT), and 30% were postmenopausal women using HT at blood draw. The AUC was statistically significantly improved by 7.0 units in premenopausal women (Gail+GRS+MD), 11.8 units in postmenopausal women not using HT (Gail+GRS+MD+hormones), and 6.1 units in postmenopausal women using HT (Gail+GRS+MD+prolactin, all p-valueu003c0.001). For the Rosner-Colditz model, the corresponding AUC improved by 5.4 units, 7.9 units, and 5.4 units (all p-valueu003c0.001). For ER+PR+ tumors specifically, among postmenopausal women not using HT, the AUC was improved by 13.2 units for the Gail model and 9.9 units for the Rosner-Colditz model (all p-valuesu003c0.001).Conclusion: While each of genetic markers, percent MD, and hormones independently improved risk prediction for invasive breast cancer and ER+PR+ disease, incorporating all of these factors simultaneously improved models the most, especially among postmenopausal women not using HT. If validated in independent populations, our findings could help identify women at a higher risk, who would most benefit from chemoprevention, other risk-reducing regimens or screening.Citation Format: Xuehong Zhang, Megan Rice, Shelley S. Tworoger, Bernard A. Rosner, A. Heather Eliassen, Rulla M. Tamimi, Jing Qian, Graham A. Colditz, Walter C. Willett, Susan E. Hankinson. Improving breast cancer risk prediction models: the addition of a genetic risk score, mammographic density, and endogenous hormones. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr 2600.