Study to Validate a Model of Surgical Deescalation in Atypicals Breast Lesions

Routine screening mammography has increased the non-palpable suspicious lesion detection rate, requiring histopathological evaluation. The discovery of atypical lesions on the breast biopsy is associated with two interrelated risks: 1. The risk of underestimating the severity of the lesion biopsied currently leads systematically to achieve a surgical resection for these patients while a breast cancer (BC) will finally identified at surgery in 10 to 25% of cases. Thus, unnecessary surgeries will be performed in 75 to 90% of cases (no cancer). 2. These breast lesions confer long-term increase in the subsequent risk of breast cancer (cumulative incidence of 15 to 25% at 25 years). Thus, women identified with atypical lesions are then followed using clinical and mammographic annual evaluation. The goal would be to get a model to assess the risk and no longer operate those patients at high risk of BC. Several groups have attempted to identify predictors of concurrent or secondary BC associated with the discovery of an unusual lesion at biopsy. However, they are often focused on a subtype of atypia (eg atypical metaplasia, atypical ductal hyperplasia, atypical lobular hyperplasia) and no prediction model has been validated in prospective multicenter. Based on a large retrospective study at Gustave Roussy, a prediction model (Nomat) has been developed, common to all atypical lesions, which can predict the presence of a BC at excision surgery (risk of concurrent BC). This model is based on the age of the patient, the disappearance of radiographic abnormalities (microcalcifications in general) after the biopsy and initial radiological lesion size. This model has good performance with an area under the curve (AUC) of 0.72. In previous series, with a BC high-risk threshold of 20%, negative predictive value was 90%, and this model would have prevented the surgery in 51% of patients (low risk patients by model). It is essential to validate this model by forward-looking and in