AI Multimodal Model Outperforms Oncotype DX for Long-Term Breast Cancer Recurrence Risk

Results presented at the San Antonio Breast Cancer Symposium showed stronger risk ranking in HR-positive, HER2-negative, node-negative disease, including late recurrence beyond five years. In cross-validation on 2,806 cases, the ICM+ model posted a C-index of 0.705 for 15-year distant recurrence and 0.656 for late recurrence versus 0.617 and 0.518 with Oncotype DX. In a 1,621-patient holdout set, ICM+ achieved a C-index of 0.733 for overall recurrence and 0.705 for late distant recurrence compared with 0.631 and 0.527 for Oncotype DX. Investigators described potential deployment through digital pathology workflows, with slide scans or smartphone images uploaded for centralized analysis. The study did not evaluate prediction of chemotherapy benefit or extended endocrine therapy benefit, and the authors called for further validation, clinical-utility studies, and regulatory review before clinical use.