Refining Stratification in Premenopausal Node-Positive HR⁺/HER2^- Breast Cancer With a Novel AI-Based Test

Genomic assays are considered the standard of care for stratification and chemotherapy benefit prediction in postmenopausal women with breast cancer. However, clinical trials indicate that these tools have limited utility for premenopausal women, especially when considering chemotherapy de-escalation in node-positive cases. Here, we evaluate Ataraxis Breast RISK (ATX), an artificial intelligence-based test integrating histopathology images and clinical data, for recurrence risk stratification, with an exploratory comparison to a clinically approved genomic assay.

ATX scores were generated for 222 HR + HER2^- premenopausal patients diagnosed with node-positive disease. Patients were classified into ATX high- and low-risk groups using a predetermined cut-off. Exploratory analyses were performed on a subset of 43 patients who had scores from ATX and a genomic assay. The primary endpoint used in this study was recurrence-free interval (RFI).

ATX had a pooled C-index of 0.75 and a 5-year time-dependent AUC of 0.63, demonstrating good discrimination across the 222-patient cohort. ATX-stratified patients into high- (n = 166, 75%) and low- (n = 56, 25%) risk groups, with significantly different Kaplan-Meier-estimated RFI (P = .04). High-risk patients had a more than twofold increase in 5-year recurrence rate compared to low-risk patients (17% vs. 7%). In contrast, genomic risk groups showed similar recurrence rates (17% vs. 14%).

ATX encodes meaningful prognostic information in premenopausal patients with node-positive breast cancer and may identify patients at elevated recurrence risk not captured by genomic assays.