Evaluation of novel tomosynthesis density measures in breast cancer risk prediction

Breast screening has rapidly transitioned in the US to digital breast tomosynthesis (DBT), an x-ray technology in which 3-D images are reconstructed from a limited number of low-dose x-ray source projections. DBT offers superior tissue visualization allowing for the direct measurement of the actual volume of dense tissue, rather

than an estimated percent or volume (from a 2-D mammogram). Since breast density is a strong predictor of masking and risk, DBT volumetric density measures, including our recently developed and first of its kind, fully automated 3-D measure, have the potential to improve individualized breast cancer (BC) risk prediction. No

studies to date have evaluated DBT volumetric density measures in large, diverse cohorts or subpopulations to understand the impact of these measures to improve prediction of masking and risk in order to tailor prevention and screening approaches. Our goal is to comprehensively examine DBT volumetric density measures

as risk factors for invasive, interval and advanced BC, and evaluate their impact on clinically relevant BC risk models and artificial intelligence (AI) algorithms across multiple racial groups. We propose this research in three large breast screening cohorts that perform routine DBT, each with comprehensive clinical

risk factors, multiple DBT per woman, follow-up and BC outcomes. Specifically, we will establish a nested case-control study of over 3,000 invasive BC cases and 9,000 controls matched on facility, age, race, ethnicity, date of enrollment DBT and follow-up time and estimate novel research and commercial DBT volumetric

measures as well as ascertain clinical BI-RADS density from DBT screening exams from enrollment up to 6 months prior to diagnosis (or corresponding follow-up for controls). In Aim 1, we will evaluate DBT volumetric density measures and their combinations as risk factors for invasive, interval and advanced BC, at enrollment

DBT exam, as well as DBT exams within five years of the cancer (or follow-up for controls), using state of the art commercial and research algorithms. We will also assess differences in associations by time of DBT exam, age, race, menopausal status and body mass index. In Aim 2, we will evaluate the contribution of DBT

volumetric density measures to clinical BC risk models, including the BCSC 5-year risk model, the novel BCSC 6-year cumulative risk model for advanced cancer, and secondarily, the Tyrer-Cuzick model for both 5 and 10- year risk. Using these results, we will determine the impact of DBT density measures on high-risk thresholds

for preventative therapy and tailored imaging. Finally, in Aim 3, we will evaluate the contribution of DBT volumetric density measures to three novel AI algorithms developed for BC risk and detection, with risk of invasive, advanced and interval BC in the short- and longer-term. Our innovative study will be the largest to

inform how novel DBT volumetric density measures can augment BC risk-stratification and prediction across multiple races and build a diverse resource to evaluate new DBT measures and risk models as they evolve. These findings will build an evidence base to inform personalized prevention approaches.