Prognostic Risk Refinement using Artificial Intelligence in HR+/HER2- Early Breast Cancer: Implications for CDK4/6 Eligibility Criteria

Abstract

Abstract Patient selection and enrolment into phase III randomized clinical trials (RCTs) of adjuvant cyclin-dependent kinase 4 and 6 (CDK4/6) inhibitor therapies depend on accurate risk definition. However, standard clinicopathologic criteria incompletely capture recurrence risk, limiting their efficacy in treatment selection. To assess whether artificial intelligence (AI)-enhanced prognostication may enrich the clinical risk groups utilized in the adjuvant NATALEE trial, we evaluated Ataraxis Breast RISK (ATX), a multimodal AI test that integrates clinical data with morphological features from H&E-stained slides. ATX risk scores were generated for 2,228 patients with HR+/HER2- early breast cancer, of which 918 (41%) were classified as clinical high-risk and 1,310 (59%) were clinical low-risk. ATX was significantly associated with recurrence-free interval in both clinical risk groups and identified high-risk patients not captured by current clinical criteria, as well as individuals with limited benefit despite clinical high-risk classification. Consequently, integration of AI-enhanced risk assessment may improve selection of patients likely to benefit from adjuvant CDK4/6 inhibitors relative to current criteria.

Ähnliche Arbeiten