Multimodal AI Model Strengthens Prognostic Accuracy for Late Distant Recurrence in HR-Positive, HER2-Negative Early Breast Cancer
Key Clinical Takeaways
- Multimodal artificial intelligence (AI) models integrating clinical, expanded molecular, and histopathology features markedly improved prognostic accuracy for distant recurrence compared with Oncotype DX, particularly for late recurrence where Oncotype DX had minimal predictive value.
- The ICM+ model delivered the highest overall prognostic performance, achieving a C-index of 0.713 for overall recurrence and demonstrating strong predictive value for both early and late events.
- Molecular features primarily informed early recurrence prediction, whereas histopathology features enhanced late recurrence prognostication, underscoring the complementary value of multimodal AI approaches.
Joseph Sparano, MD, Mount Sinai Medical Center, New York, New York, discusses new findings from the TAILORx multimodal artificial intelligence (AI) analysis, which integrated clinical, molecular, and histopathology features to improve prognostication of early, late, and overall distant recurrence risk among patients with HR-positive, HER2-negative early breast cancer.
Multimodal AI models, particularly the ICM+ model, significantly outperformed the traditional Oncotype DX recurrence score, especially for predicting late distant recurrence where current tools offer limited value. These results highlight the potential of AI-driven risk stratification to personalize long-term management strategies in early breast cancer.
These findings were presented at the 2025 San Antonio Breast Cancer Symposium in San Antonio, Texas.
Source:
Sparano J, Wang V, Gray RJ, et al. Multimodal artificial intelligence (AI) models integrating image, clinical, and molecular data for predicting early and late breast cancer recurrence in TAILORx. Presented at SABCS 2025. December 9 - 12, 2025. San Antonio, Texas. Abstract GS1-08
