Abstract 268: Absolute breast cancer risk according to three risk prediction models: Inverse associations with risk of death and poor prognostic features

Abstract

Abstract Predicted absolute breast cancer risk is an important factor for women to consider in making decisions about screening and prevention. If breast cancer risk models preferentially predict risk of indolent cancers, then using these tools to guide screening may reduce mortality less than expected. In contrast, if models perform well in determining risk of fatal breast cancers, these tools may help reduce mortality by guiding targeted interventions. Accordingly, we assessed relationships of women's predicted breast cancer risk, as estimated with the Breast Cancer Risk Assessment Tool (BCRAT; http://www.cancer.gov/bcrisktool/), Breast Cancer Surveillance Consortium (BCSC) risk calculator (https://tools.bcsc-scc.org/BC5yearRisk/) and Barlow model, with risk of death and tumor features. Using information collected in the BCSC from 37,939 women prior to the diagnosis of invasive breast cancer (1996-2007), we calculated predicted 5-year breast cancer risk (stratified as <1%; 1-2.5%; >2.5%) with the BCRAT, BCSC and Barlow models. Cox models were used to estimate the association between breast cancer risk strata and 10-year survival following a breast cancer diagnosis. Separate models were analyzed for women less than age 50 years and 50 years or older at diagnosis, adjusted for covariates. To elucidate relationships between predicted breast cancer risk and risk of death, we compared features of cancers by risk strata. Mean age at breast cancer diagnosis was 60 years. There were a total of 6,021 deaths of which 2,993 (49.7%) were ascribed to breast cancer. Estimated breast cancer risk varied by model; Barlow assigned 36.5% of women to the highest risk strata (>2.5% absolute risk at 5 years) versus 21.0% for BCSC and 5.5% for BCRAT. For each model, women at lowest breast cancer risk experienced the highest risk of breast cancer death. Women above age 50 years whose predicted breast cancer risk was in the two highest strata were at lower risk of death than those in the lowest risk stratum (HRs for breast cancer death ranged from 0.69-0.84 for the 3 models, with all but one estimate statistically significant). Compared with women at highest breast cancer risk, women at lowest risk developed a significantly higher percentage of breast cancers that were larger than 2 cm, grade 3, associated with positive nodes, higher than stage IIA and classified as interval cancers (detected following a negative screening mammogram). Analyses excluding women with missing risk factor data yielded similar interpretations. Higher breast cancer risk, as predicted with BCRAT, BCSC and Barlow models, is inversely associated with risk of breast cancer death. Application of existing breast cancer risk models to guide screening and prevention may yield less than expected reductions in mortality. Citation Format: Mark E. Sherman, Laura Ichikawa, Diana Miglioretti, Pamela Vacek, Jeffrey Tice, Ruth Pfeiffer, Karla Kerlikowske, Gretchen Gierach. Absolute breast cancer risk according to three risk prediction models: Inverse associations with risk of death and poor prognostic features. [abstract]. In: Proceedings of the 105th Annual Meeting of the American Association for Cancer Research; 2014 Apr 5-9; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2014;74(19 Suppl):Abstract nr 268. doi:10.1158/1538-7445.AM2014-268