Predicting risk of estrogen receptor positive breast cancers in postmenopausal women

1507 Background: Chemoprevention strategies for estrogen receptor positive (ER+) breast cancers are emerging, especially for postmenopausal women, but require methods of targeting appropriate populations. Our objective was to improve the Breast Cancer Risk Assessment Tool [Gail Model (GM)] for estimating ER+ breast cancer risk. Methods: A prospective cohort involving 161,809 postmenopausal women aged 50–79 years, (93,676 in the observational study (OS) and 68,132 in clinical trials (CT)) at Women’s Health Initiative (WHI) Clinical Centers had comprehensive assessment of lifestyle, medication use and breast cancer risk factors. Breast cancer risk from the GM and other models incorporating additional or fewer risk factors and five year incidence of ER + and ER negative (ER-) invasive breast cancers were determined. Main outcome measures were concordance statistics for models predicting breast cancer risk. Results: Of 148,266 women meeting eligibility criteria, (no prior breast cancer and/or mastectomy), 3,236 developed breast cancer. Chronological age and age at menopause, both GM components, were significantly associated with only ER+ but not ER- breast cancer risk (p<0.05 for heterogeneity test). The GM predicted population-based ER+ cancer risk with reasonable accuracy (concordance statistic 0.60, 95% confidence interval (CI) 0.58 to 0.62) but for ER- cancers, the results were equivalent to chance allocation (concordance statistic 0.49, 95% CI 0.45 to 0.54). For ER+ cancers, no additional risk factors improved the GM prediction. However, a simpler model, developed in the OS and tested in the CT population, including only age, family history, and benign breast biopsy was comparable to GM in ER+ breast cancer prediction (concordance statistics 0.58, 95% CI 0.56 to 0.60). Using this model, all women ≥ 55 years old (or ≥ 60 year old if African American) with either a prior breast biopsy or first degree breast cancer family history had five year breast cancer risk of ≥ 1.8%. Conclusions: In postmenopausal women with comprehensive mammography use, the GM identifies populations at increased risk for ER+ breast cancer but not for ER- cancer. A model with fewer variables provides a simpler alternative for identifying populations appropriate for breast cancer chemoprevention interventions. No significant financial relationships to disclose.