A multivariate model to predict cancer upgrade from atypical ductal hyperplasia by core needle biopsy.

Abstract

3 Background: Atypical ductal hyperplasia (ADH) is a high-risk breast lesion usually diagnosed with core needle biopsy. Although upgraded to cancer at surgical excision in ~15 to 25% of cases, routine excision is questioned due to cost and overtreatment. We evaluated clinical, imaging, and histologic features associated with cancer upgrade and developed a multivariate model to predict risk of upgrade. Methods: With IRB approval a single institution retrospective review was performed of patients who underwent surgical excision of ADH diagnosed by core biopsy from 06/2005 to 06/2013. Review was performed of electronic records, breast imagin,g and biopsy slides. Multiple imputations were used for missing data. Association of cancer upgrade with various features was assessed with logistic regression. Results: 409 biopsies with ADH on core biopsy, with later surgical excision, were included. The overall upgrade rate was (16.1%, 95% CI:12.9-20.0%); 10 patients had invasive cancer at excision and 56 DCIS only. Features on core biopsy most strongly associated with upgrade were imaging estimated percent of lesion removed (upgrade 9% for 90% removed, 14% for 50 to 75%, and 27% for < 50% removed), individual cell necrosis (upgrade 34% with necrosis vs. 9.5% without), and # foci of ADH (22% for >1 focus vs 8% for 1 focus). A multivariate predictive model (see Table) showed an average C-statistic of 0.77. Women with no necrosis and either 1 focus with ≥ 50% removal or >1 focus with 90% removal (36% of the sample) have low risk of upgrade (5.0%, 95% CI:1.3-8.7%). Conclusions: ADH on core biopsy with low risk of upgrade to cancer is defined by percent of imaging lesion removed, # of foci of ADH, and lack of individual cell necrosis. If findings are validated, women whose biopsies meet low-risk criteria might be considered for chemoprevention and surveillance rather than surgical excision.[Table: see text]