Grant Number:	1R01CA282725-01 Interpret this number
Primary Investigator:	Sprague, Brian
Organization:	University Of Vermont & St Agric College
Project Title:	Clinical Breast Cancer Risk Prediction Models for Women with a High-Risk Benign Breast Diagnosis
Fiscal Year:	2023

PROJECT SUMMARY/ABSTRACT Over 1.7 million core needle breast biopsies are performed in the U.S. every year among women with either a palpable mass or a suspicious finding on breast imaging.1 Benign breast disease is diagnosed in 75% of these biopsies2 and comprises a spectrum of diagnoses ranging from normal or non-proliferative changes to proliferative changes with or without atypical features. Early studies demonstrated that ~10-30% of women with specific benign breast disease diagnoses (referred to as “high-risk lesions”) had cancer identified on surgical excision of their lesion (referred to as “upgrade”).3,4 Thus, aggressive management with surgical excision is common for these high-risk lesions to rule out malignancy.1 However, there is widespread interest in avoiding surgery for high-risk lesions, the majority of which will not yield a malignant diagnosis on excision.5 Single- institution studies have identified various patient, pathologic, and radiologic variables associated with risk of upgrade on excision, but insufficient evidence is available to inform evidence-based guidelines as to which patients with high-risk lesions can safely avoid excision.5-7 There is also uncertainty about long-term breast cancer risk and the need for enhanced prevention and detection strategies among women who do not undergo excision, or who undergo excision without upgrade.8-10 Our goal is to provide definitive new evidence to guide clinical management recommendations for high-risk breast lesions. We propose a comprehensive study leveraging the rich infrastructure and longitudinal data of the Breast Cancer Surveillance Consortium, a large U.S. alliance of breast imaging and pathology registries. We will supplement the existing BCSC database with newly collected data from breast imaging reports, pathology reports, and other clinical medical records for over 7,000 high-risk lesions diagnosed at more than 100 healthcare facilities since 2010. Using an ensemble learning approach we will develop and validate prediction models for risk of upgrade on excision (Aim 1) and 5- year breast cancer incidence among women with high-risk lesions (Aim 2) based on patient, pathologic, and radiologic factors. In Aim 3 we will use a decision analysis framework paired with qualitative study of stakeholder perspectives to evaluate the potential population impact and acceptability of risk-based strategies for managing high-risk lesions. Our results will facilitate identification of women with acceptably low risk of upgrade on excision who can be safely managed by imaging surveillance, and identify women with elevated risk of a future cancer diagnosis who need enhanced long-term imaging surveillance and/or chemoprevention. The proposed study will be far larger and more representative of clinical practice in the U.S. than any study conducted to date, and will be the first to comprehensively evaluate patient, biopsy, pathologic, and radiologic factors in relation to high-risk lesion outcomes in a multicenter study. Evidence generated by this study will inform clinical recommendations for the management of high-risk lesions diagnosed with core needle biopsy, resulting in improved care and reduced harms for the >100,000 women diagnosed annually in the U.S.





None