Title: Risk of cancer versus risk of cancer diagnosis? Accounting for diagnostic bias in predictions of breast cancer risk by race/ethnicity and breast density Project Leaders: Charlotte Gard, NMSU; Ruth Etzioni, Fred Hutch [The content of this pilot proposal is identical in the NMSU and Fred Hutch proposals] PROJECT SUMMARY/ABSTRACT Risk prediction modeling is critical for tailoring prevention and screening efforts and targeting cancer treatment. In breast cancer, both race/ethnicity and breast density are included as risk factors in a widely used risk prediction model from the Breast Cancer Surveillance Consortium (BCSC), an authoritative source of information on breast cancer screening and outcomes across the U.S. Since the BCSC is a screened population, variations in diagnostic intensity by race/ethnicity and breast density may have affected the relative risks associated with these factors. The broad objective of this pilot project is to develop an analytic approach to de-bias risk prediction models developed in settings where screening and/or biopsy frequencies vary across key risk factors. This will require decoupling the patterns of screening/biopsy from the underlying disease process. We propose to do this via a combination of statistical and simulation modeling using data from the BCSC. The statistical model (Aim 1) will establish the variations in screening and biopsy frequencies by race/ethnicity and breast density in the BCSC population. The simulation model (Aim 2) will estimate the natural history (onset and preclinical duration) of breast cancer by race/ethnicity and breast density. We will combine these models to determine (Aim 3) whether the dependence of breast cancer risk on race/ethnicity and breast density would change if screening and biopsy frequencies were similar across groups defined by these factors. This pilot project will bring together New Mexico University investigators with deep expertise in risk prediction using the BCSC and Fred Hutchinson investigators who are established leaders in cancer modeling. Its focus on racial and ethnic disparities fits squarely within the mission of this U54 partnership. The proposed work has the potential to greatly clarify our understanding of how breast cancer risk is modified by race/ethnicity and breast density. Further, if results show that natural history modeling is helpful in addressing bias in risk prediction models due to variations in screening practices, this could lead to a paradigm shift in risk prediction modeling for cancer.