Abstract PS16-07: Quantitative breast density measures and radiomic parenchymal phenotypes improve breast cancer risk prediction among Black and White women undergoing mammography screening

Abstract Breast parenchymal patterns identified on mammograms are associated with breast cancer risk. Breast density, which is clinically and visually assessed based on the BI-RADS lexicon, is a well-recognized risk factor for breast cancer that has modest reproducibility. Black women have, on average, lower BI-RADS categorized density than White women, despite having higher volumes of dense breast tissue when density is measured quantitatively. Beyond breast density, we recently identified and validated six parenchymal phenotypes, based on 390 radiomic texture features extracted from 2-D full-field digital mammograms performed from 2011 through 2014 from a racially diverse population of 30,000 women. The six phenotypes were also significantly associated with breast cancer risk after adjusting for BI-RADS density among Black and White women. The purpose of this study was to evaluate whether the combination of quantitative breast density measures (dense volume and volumetric percent density) with the parenchymal phenotypes improves breast cancer risk prediction, particularly among Black women. Eligible women were aged ≥35 with no prior history of breast cancer who underwent routine screening mammography at two large academic medical centers. Invasive breast cancer cases were identified and matched to controls in a 1:3 ratio based on age within 5 years, race, timing of images, and mammogram site. The six phenotypes were estimated on the 1055 invasive breast cancer cases and 2764 controls, of whom 144 cases and 390 controls were Black and 875 cases and 2298 controls were White. BI-RADS density was clinically assessed according to the 4th or 5th edition lexicon. Dense volume (DV) and volumetric percent density (dense volume / total breast area, VPD) were estimated using Volpara software. Conditional logistic regression was used to estimate the associations of BI-RADS density, quantitative density, and radiomic parenchymal phenotypes with breast cancer risk, adjusted for age and BMI. Area under the receiver operating curve (AUC) was also calculated for each model. Both dense volume and radiomic phenotypes were significantly associated with breast cancer risk among both Black and White women (p<0.05). Among White women, age and BMI adjusted models including BI-RADS density (AUC=0.600), volumetric percent density (AUC=0.596), and dense volume (AUC=0.601) yielded similar AUCs. For Black women, BI-RADS density had higher AUC (0.630) than VPD (AUC=0.599) or dense volume (AUC=0.607). The addition of parenchymal phenotypes to the models increased discrimination to a greater extent for Black women (BI-RADS + radiomics AUC=0.679, 95%CI=0.632-0.725, VPD + radiomics 0.666, 95%CI=0.619-0.713, and DV + radiomics 0.679, 95%CI=0.632-0.725) than for White women (BI-RADS + radiomics AUC=0.609, 95%CI=0.589-0.629, VPD + radiomics AUC=0.607, 95%CI=0.588-0.627), and DV + radiomics AUC=0.623, 95%CI=0.603-0.643). The model containing age, BMI, BI-RADS density, radiomic parenchymal phenotypes yielded significantly higher AUC among Black women (AUC=0.692, 95% CI 0.646-0.737) compared with White women (AUC= 0.625, 95%CI=0.606-0.645). These results suggest that incorporation of parenchymal phenotypes based on radiomic features improves breast cancer risk prediction and that the addition of quantitative breast density measures further improves risk prediction particularly among Black women. Citation Format: Anne Marie McCarthy, Stacey Winham, Christopher Scott, Aaron Norman, Walter C. Mankowski, Alex Nguyen, Sarah Ehsan, Matthew Jensen, Eric A. Cohen, Hannah Horng, Andrew D. A. Maidment, Kathleen Brandt, Emily F. Conant, Karla Kerlikowske, Despina Kontos, Celine M. Vachon. Quantitative breast density measures and radiomic parenchymal phenotypes improve breast cancer risk prediction among Black and White women undergoing mammography screening [abstract]. In: Proceedings of the San Antonio Breast Cancer Symposium 2024; 2024 Dec 10-13; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2025;31(12 Suppl):Abstract nr PS16-07.