AI Model Challenges Genomic Classifiers in Early Breast Cancer Risk Stratification: Nancy Lin, MD

A digital pathology-based multimodal artificial intelligence (AI) model has potential as an alternative to genomic classifiers for risk stratification in hormone receptor-positive, HER2-negative early breast cancer, according to data presented at the 2026 American Society of Clinical Oncology Annual Meeting (ASCO). Nancy Lin, MD, associate chief of the Division of Breast Oncology at Dana-Farber Cancer Institute and professor of medicine at Harvard Medical School, discussed findings in an interview with The American Journal of Managed Care^® (AJMC^®).

The study takes aim at one of the friction points in early breast cancer management: the cost, turnaround time, and tissue requirements of standard genomic assays. Lin explains how the AI model performed against the recurrence score in a real-world population, what it offers patients who fall into the ambiguous intermediate-risk zone, and why a lower-cost tool that performs comparably or better could meaningfully shift the access and cost landscape for early-stage breast cancer care.

This transcript has been lightly edited for clarity.

AJMC: Starting with the big picture, could you outline the goals of this study and what problem this study is trying to solve?

Lin: Genomic classifiers are widely used to assess the risk of distant metastasis (DM) and inform chemotherapy (CT) decisions in patients with HR+/HER2- early breast cancer (EBC). While clinically validated, these assays are associated with high cost, long turnaround time, and tissue consumption.

Artera has developed a digital pathology-based multimodal AI (MMAI) model. The locked MMAI algorithm was validated to estimate the risk of DM in the phase 3, NSABP B-14, NSABP B-39, and ABCSG-8 clinical trials and inform CT benefit. Further, MMAI was predictive for chemotherapy benefits in postmenopausal patients in NSABP B-20. MMAI offers the potential to reduce cost, reduce turnaround time, and avoid tissue consumption in comparison to genomic classifiers, so the primary question for this study was to directly compare risk stratification tools in a contemporary, real-world population and evaluate their concordance in a specific patient cohort.

AJMC: The intermediate Recurrence Score (RS) zone (scores of 11 to 25) has always been a difficult space for oncologists to counsel patients. What did the MMAI model do within that group, and does it actually help resolve the ambiguity that the recurrence score alone leaves?

Lin: In the RS Intermediate zone, MMAI was able to further refine these patients into distinct risk groups: 66% were classified as MMAI low risk and 34% as MMAI high risk. Importantly, there was a substantial difference in the 10-year risk of DM for these patients, most of whom did not receive chemotherapy—2.6% for MMAI low and 11.4% for MMAI high.

For the MMAI-low patients, interestingly, the outcomes were outstanding (10-year risk of DM, 0%) for those with concordant RS-low/MMAI-low risk scores. However, those with RS-low/MMAI-high scores experienced worse outcomes (10-year risk of DM, 20%), again suggesting that the MMRI algorithm is providing additional prognostic information beyond RS.

For those with MMAI high scores, it does warrant additional consideration that therapy intensification may be beneficial, whether that’s extended endocrine therapy or chemotherapy.

AJMC: For a managed care audience, what might the study results mean for the cost structure of early-stage breast cancer care going forward?

Lin: Genomic classifiers like the RS are well integrated in today’s clinical practice due to their ability to better personalize treatments and avoid both undertreatment and overtreatment. However, their high cost has been a long-standing issue in terms of market access, so the availability of a lower-cost tool that can perform similarly, if not better, should hopefully improve market access and reduce overall cost burden.