Corey Speers, Professor and Chair of Radiation Oncology at the University of Alabama at Birmingham, summarizes study data for an FDA-cleared AI system that works alongside standard breast cancer diagnostic tests, presented at ASCO 2026.
The following transcript has been edited for clarity.
Hello, I'm Corey Spears, Professor and Chair of Radiation Oncology at the University of Alabama at Birmingham. On behalf of SWOG and our collaborators at Artera, I am pleased to present a brief summary of our 2026 ASCO Annual Meeting abstract evaluating Artera AI Breast, an FDA-cleared, multimodal artificial intelligence model for prognosis and chemotherapy benefit prediction in postmenopausal women with node-positive, hormone receptor-positive breast cancer from the SWOG 8814 trial.
The purpose of this study was to determine whether a locked, previously validated MMAI model could provide prognostic information and predict chemotherapy benefit in this clinically important population. Artera AI combines routine H&E whole slide images with clinical variables, including age, tumor size, and nodal status to generate a patient-level risk score. This model has previously been validated for prognosis in the ABCSG-08 trial and for chemotherapy benefit prediction in the node-negative breast cancer patient population from the NSABP B-20 trial. This study extended those validation studies into a higher risk node-positive cohort.
SWOG 8814 was an ideal cohort for this question because it was a randomized phase III trial testing whether adjuvant chemotherapy improved outcomes when added to tamoxifen in post-menopausal women with hormone receptor-positive, node-positive breast cancer. In this analysis, the MMA model was applied in a locked fashion using the same previously validated cut points.
For clinical utility, the intermediate and high-risk groups were combined into a single non-low risk group, allowing for a binary assessment of chemotherapy benefit. The results were clinically meaningful. In our study, the MMAI was independently prognostic for both disease-free survival and overall survival after adjustment for age, tumor size, and nodal status.
At 10 years, the disease-free survival was 74 percent in the low-risk group, compared with 50 percent in the non-low risk group. More importantly, the model appeared to identify differential chemotherapy benefits.
In the overall cohort, low-risk patients had overlapping disease-free survival curves with tamoxifen alone versus chemo plus tamoxifen. In contrast, the non-low risk group had a great treatment effect, with a 23 percent relative reduction in 10-year disease-free recurrence risk with the addition of chemotherapy.
Perhaps the most clinically relevant finding was in patients with one to three nodes positive. In that subgroup, MMAI low-risk patients had minimal b– apparent benefit from chemotherapy, while the non-low risk patients had a 26 percent relative risk reduction in the 10-year disease recurrence risk.
Clinically, these findings suggest that multimodal AI might help us move beyond just nodal or clinical features in identifying patients who may benefit from chemotherapy. This approach may help identify which patients with hormone receptor-positive, node-positive breast cancer are most likely to benefit from chemotherapy and which patients may be reasonable candidates for de-escalation. The broader goal is more precise, scalable, and tissue-sparing decision support for both patients and clinicians in order to identify chemotherapy benefit, and we look forward to the continued validation of Artera AI MMAI in future studies.
Newsletters
Receive the latest pathologist news, personalities, education, and career development – weekly to your inbox.
