Abstract 5586: Breast cancer risk prediction: An update to the Rosner-Colditz breast cancer incidence model

Abstract

Abstract Purpose: To update and build on the Rosner-Colditz breast cancer incidence model by evaluating the contributions to breast cancer risk of childhood and adolescent somatotype, fruit and vegetable intake, breastfeeding, and physical activity. Secondarily, we evaluated the contributions of these variables to breast cancer incidence models defined by ER and PR status of the tumor. Methods: Using data from the Nurses’ Health Study (NHS) and Nurses’ Health Study II (NHSII), we added childhood somatotype (average of somatotype at ages 5 and 10; 9 unit scale), adolescent somatotype (average of somatotype at ages 10 and 20; 9 unit scale), fruit and vegetable intake (servings/day), breastfeeding (months), and physical activity (METs/week) to the Rosner-Colditz breast cancer incidence model to determine whether these variables improved the discriminatory power of the model. We evaluated the contribution of these variables to a model for all invasive breast cancer as well as to models for ER+/PR+, ER+/PR-, and ER-/PR- breast cancer. Results: In the NHS/NHSII, we accrued over 3000 cases of invasive breast cancer with complete data on the selected breast cancer risk factors. Compared to the original Rosner-Colditz model, the models that included childhood or adolescent somatotype were significantly improved (p<0.001). The relative risk (RR) of invasive breast cancer for a 4 unit change in somatotype was 0.69 for childhood somatotype and 0.59 for adolescent somatotype. Including childhood somatotype increased the age-adjusted concordance statistic from 0.613 to 0.616 whereas including adolescent somatotype increased the age-adjusted concordance statistic from 0.613 to 0.618. Childhood and adolescent somatotypes also significantly improved ER+/PR+ and ER-/PR- breast cancer incidence models (p<0.02). Breastfeeding and vegetable intake suggestively improved the ER-/PR- model. The RR for five servings of vegetables per day was 0.82, while the RR for 12 months of breastfeeding was 0.87. While these variables were significantly associated with ER-/PR- breast cancer risk in the updated models, the differences in concordance statistics comparing models with to models without these variables were not statistically significant. Conclusion: Childhood and adolescent somatotype significantly improved breast cancer risk classification in the NHS/NHSII. Citation Format: Megan Rice, Shelley Tworoger, Susan Hankinson, Rulla Tamimi, A Heather Eliassen, Graham Colditz, Bernard Rosner. Breast cancer risk prediction: An update to the Rosner-Colditz breast cancer incidence model. [abstract]. In: Proceedings of the 106th Annual Meeting of the American Association for Cancer Research; 2015 Apr 18-22; Philadelphia, PA. Philadelphia (PA): AACR; Cancer Res 2015;75(15 Suppl):Abstract nr 5586. doi:10.1158/1538-7445.AM2015-5586