Abstract 2252: Predicting second breast cancers among women diagnosed with primary breast cancer using patient-level data and machine learning algorithms

Abstract

Abstract In 2020 2.3 million women were diagnosed with breast cancer. About 7.4% of women who have been diagnosed with primary breast cancer will have a second primary breast cancer within 10 years. This study builds a prediction model for second breast cancer for women who have had primary breast cancer. Readily available cancer registry data with machine learning methods for classification are employed. The best-performing model is selected based on the area under the receiver operator curve, and the key characteristics contributing to a high risk for second breast cancer are identified based on the prediction model. Using extreme gradient boosting (XGBoost) with limited patient features we find an area under the curve of 0.65-0.70 for the testing set. Among the most important features are days from incidence to treatment, size of primary tumor based on the pathology report, and oestrogen receptor status.This research is a step towards the development of a tool that will help doctors identify women very likely to develop second breast cancer, which will prioritize their follow-up or inform their course of treatment depending on their characteristics. Citation Format: Maria Eleni Syleouni, Nena Karavasiloglou, Laura Manduchi, Miriam Wanner, Dimitri Korol, Sabine Rohrmann. Predicting second breast cancers among women diagnosed with primary breast cancer using patient-level data and machine learning algorithms [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 2252.