An artificial intelligence (AI) model created by integrating clinical, molecular, and histopathology data significantly improved recurrence risk stratification in hormone receptor (HR)-positive, HER2-negative breast cancer, according to results presented at the San Antonio Breast Cancer Symposium (SABCS), held December 9-12, 2025.
HR-positive, HER2-negative breast cancer is the most common subtype of breast cancer, and at least 50% of recurrences in this subtype occur more than five years after diagnosis, explains Joseph A. Sparano, M.D., chief of hematology-oncology at Mount Sinai Tissue Cancer Center. The Oncotype DX (ODX) 21-gene recurrence score, a unimodal molecular test that provides prognostic information for distant recurrence and predictive information for chemotherapy benefit, is widely used in clinical practice, but its ability to predict recurrence after 5 years is limited, Sparano noted.
Our goal was to develop a new diagnostic test that provides better prognostic estimates of recurrence risk, including late recurrence risk, by studying tumor specimens obtained in the TAILORx trial. We developed an AI model that evaluates both digitized slide images used for routine pathological evaluation and molecular and clinical features of breast cancer, providing better prognostic information about the risk of cancer recurrence up to 15 years out, including early recurrence within 5 years after diagnosis and late recurrence after 5 years. ”
Joseph A. Sparano, MD, Director of Hematology-Oncology, Mount Sinai Tisch Cancer Center
This included the development of a new molecular test with an expanded gene panel derived from five commercially available genetic assays, including ODX.
The research team used digitized tissue images and molecular RNA expression data from 4,462 tumor samples and corresponding clinical data from TAILORx study participants. These data were used to train and validate several risk models. The prognostic performance of the model was compared to the performance of the ODX results used in the trial to guide the use of chemotherapy and was evaluated using the concordance index (C-index). The C-index is a statistical test that measures a diagnostic test's ability to accurately rank recurrence risk. A C-index of 0.5 indicates that the test run is only by chance, while a C-index of 1 indicates a perfect prediction.
ICM+, a multimodal model that integrates pathological imaging (I), clinical (C), and extended molecular (M+) models, improved ODX for overall distant recurrence at 15 years (C-index 0.705 vs. 0.617) and late recurrence at 5 years (C-index 0.656 vs. 0.518) in a training/5-fold cross-validation set including 2,806 patients. showed significantly better performance. ICM+ also showed similar superior prognostic performance compared to ODX in a holdout validation set including 1,621 patients for total recurrence (C-index 0.733 vs. 0.631) and late distant recurrence (C-index 0.705 vs. 0.527).
The results of this study will ultimately lead to the availability of new diagnostic tests that more reliably estimate the risk of recurrence in women with HR-positive, HER2-negative, node-negative breast cancer, which accounts for about half of all breast cancers in the United States, Professor Sparano explained.
“This study shows the potential of using AI to develop better diagnostic tests that can more accurately estimate recurrence risk and individualize treatment decisions,” Sparano said. Currently available molecular assays, whether performed at central reference laboratories or CLIA-certified regional laboratories, require advanced equipment and technical expertise, he noted. “AI-based pathology tools that rely on the evaluation of tissue sample slides routinely generated in clinical settings can be captured by scanners and even widely available smartphones, uploaded electronically, and analyzed centrally at minimal cost,” he added.
A limitation of the study is that it was not designed to develop a test to predict the benefit of chemotherapy or the benefit of continuing adjuvant endocrine therapy beyond five years, Sparano said.
The study is a public-private partnership between the federally funded ECOG-ACRIN Cancer Research Group and Charis Life Sciences, with support from the Breast Cancer Research Foundation, the National Institutes of Health's National Cancer Institute, and the U.S. Postal Service Breast Cancer Research Stamp Fund. Sparano serves as a consultant for AstraZeneca, Delphi Diagnostics, Genentech, Genomic Health/Exact Sciences, Novartis, and Pfizer. He is a member of the Scientific Advisory Board of PreciseDX. Receives institutional research support from Olema Oncology.
sauce:
American Association for Cancer Research
