Irving, Texas, USA, August 7, 2026
Caris Life Sciences has announced the publication of a new study demonstrating that a multimodal artificial intelligence (AI) model can accurately predict the risk of late distant recurrence in patients with hormone receptor-positive (HR+) early breast cancer, potentially supporting more personalized decisions regarding extended endocrine therapy. Published in Cancer Research Communications, the study was conducted through a collaboration involving Caris Life Sciences, ECOG-ACRIN Cancer Research Group, and NRG Oncology, bringing together expertise from both industry and leading cooperative clinical research organizations. The AI-driven model integrates digitized pathology images with clinicopathologic data to generate individualized recurrence risk assessments, offering a scalable and accessible alternative to existing genomic testing approaches. Since HR+ breast cancer accounts for nearly 70–80% of all breast cancer diagnoses, the ability to better identify patients at higher risk of recurrence beyond the initial five years of treatment could significantly improve long-term disease management while reducing unnecessary exposure to prolonged endocrine therapy and its associated side effects.
AI Model Predicts Late Breast Cancer Recurrence Risk
The published research focused on developing and validating a multimodal, multitask deep learning model capable of estimating late distant recurrence risk in patients with HR-positive early breast cancer. Unlike traditional risk assessment tools that rely primarily on genomic testing, the AI platform combines routinely collected hematoxylin and eosin (H&E) pathology images with standard clinical and pathological variables to produce personalized prognostic predictions. The model was initially developed using 2,271 archived tumor specimens obtained from patients enrolled in the NSABP B-42 clinical trial, followed by independent validation using 4,300 tumor samples from participants in the landmark TAILORx study through a public-private research collaboration. Results demonstrated that the AI model successfully distinguished patients with significantly different long-term outcomes, identifying an approximate 8% absolute difference in 10-year distant recurrence risk between high-risk and low-risk patient groups while maintaining prognostic value independent of conventional clinical risk factors and the widely used Oncotype DX Recurrence Score.
Study Supports Personalized Endocrine Therapy Decisions
One of the study’s most clinically important findings involved the potential use of artificial intelligence to guide decisions regarding extended endocrine therapy beyond the standard five-year treatment period. While prolonged endocrine therapy may reduce recurrence risk in selected patients, it is also associated with persistent adverse effects that can negatively affect long-term quality of life. Exploratory analyses suggested that patients identified by the AI model as high risk experienced greater absolute benefit from continued letrozole therapy than patients classified as low risk, indicating that AI-based risk stratification could help clinicians individualize treatment recommendations. Because the model relies on standard pathology slides and routinely collected clinical information, it may provide a more accessible and scalable alternative or complement to existing genomic assays, particularly in healthcare settings where advanced molecular testing remains limited by cost, availability, or turnaround time.
Advertisement
Caris Expands AI-Driven Precision Oncology Innovation
The publication further strengthens Caris Life Sciences’ position as a leader in AI-powered precision oncology by demonstrating how advanced computational methods can improve cancer prognosis using routinely available diagnostic information. The findings also complement the company’s recent launch of Caris MI Clarity™, a prognostic assay designed to assess both early and late distant recurrence risk in patients with HR-positive/HER2-negative, node-negative early-stage breast cancer while supporting treatment decisions related to chemotherapy and extended endocrine therapy. By combining comprehensive molecular profiling, artificial intelligence, machine learning, and extensive clinico-genomic databases, Caris continues to develop innovative diagnostic tools that enhance personalized cancer care. As precision oncology increasingly integrates digital pathology and AI-driven predictive analytics, technologies such as this multimodal deep learning model may help oncologists better identify patients who would benefit most from long-term therapy while minimizing unnecessary treatment for those at lower risk. The study represents another important advancement toward more individualized breast cancer management through data-driven clinical decision support.
Source: Caris Life Sciences press release
