New AI models integrating imaging, clinical, and molecular data from the TAILORx tissue biorepository show stronger prognostic performance than current methods to predict recurrence risk in early-stage breast cancer and guide long-term treatment decisions
The San Antonio Breast Cancer Symposium (SABCS), researchers presented the initial findings from a major multi-year collaboration between the ECOG-ACRIN Cancer Research Group (ECOG-ACRIN) and Caris Life Sciences® (Caris) focused on transforming recurrence risk assessment in early-stage breast cancer through artificial intelligence (AI). The public-private partnership pairs ECOG-ACRIN’s extensive clinical trial expertise and biorepository resources with Caris’ comprehensive MI Cancer Seek® whole exome and whole transcriptome profiling, whole slide imaging, and advanced machine learning platforms.
The research teams developed multimodal models to more precisely stratify recurrence risk in early-stage breast cancer. The models integrate histopathologic imaging, clinical, and molecular data generated from TAILORx, one of the world’s largest and most rigorously annotated breast cancer research repositories. This level of multimodal integration is unprecedented at this scale in early breast cancer prognostication.
Early-stage breast cancer represents a large and heterogeneous patient population in which treatment decisions frequently hinge on uncertain recurrence risk. Of the approximately 310,720 new cases diagnosed in the United States each year, an estimated 60% are early-stage (American Cancer Society), underscoring the broad application and clinical relevance of more accurate and individualized risk assessment.
Also Read: Quotient Sciences and Intrepid Labs partner to speed AI formulations
“Realized through collaboration between ECOG-ACRIN, NCI, and Caris Life Sciences, this public-private partnership represents a methodological, logistical, and collaborative integration of datasets from the historically impactful TAILORx trial to further extend the benefits for breast cancer patients,” said ECOG-ACRIN Group Co-Chair Peter J. O’Dwyer, MD. “The advance in personalized medicine afforded in this work, in turn, helps to advance the potential of AI to refine treatment and improve outcomes.”
Across analytic evaluations, the multimodal AI models demonstrated enhanced prognostic performance compared to existing recurrence-risk assessment methods, highlighting their potential to support more personalized treatment decision-making in early-stage breast cancer.
“By integrating imaging, clinical data, and molecular profiling, we are advancing beyond single-dimension diagnostics to deliver a more precise and comprehensive understanding of recurrence risk in breast cancer,” said Caris EVP and Chief Medical Officer George W. Sledge, Jr., MD. “The development of these models underscores the transformative power of multimodal AI and machine learning in precision oncology.”
Both AI models-including development approaches, integrated biomarker features, and demonstrated prognostic improvements- were presented in today’s SABCS sessions.
- Multimodal Artificial Intelligence (AI) Models Integrating Image, Clinical, and Molecular Data for Predicting Early and Late Breast Cancer Recurrence in TAILORx, presented by Joseph A. Sparano, MD (Mount Sinai Tisch Cancer Center). Late-Breaking Abstract GS1-08 was presented during SABCS General Session 1.
In this project, researchers developed and prospectively validated a multimodal model integrating pathomic imaging (I), clinical (C), and expanded molecular (M+) data from 4,462 TAILORx tumor specimens. The expanded M+ gene expression panel includes 42 tumor genes associated with breast cancer recurrence derived from five commercially available gene assays, including the Oncotype DX (ODX) 21-gene recurrence score and a set of highly variable genes. Based on the results of the TAILORx trial, ODX is widely used in clinical practice for its prognostic information on recurrence and predictive information on chemotherapy benefit; however, its ability to forecast recurrence beyond the 5-year mark is limited.
The findings from this study will ultimately provide crucial support for the development of a new diagnostic test for women with HR-positive, HER2-negative, node-negative breast cancer that more accurately estimates recurrence risk, especially late recurrence 5 or more years after diagnosis.
“Although the TAILORx trial was the first randomized trial to establish the role of the 21-gene recurrence score to guide chemotherapy use in early breast cancer, our goal was to take one step further in personalizing cancer therapy by developing a new diagnostic test using tumor specimens derived from the trial,” said Dr. Sparano.
Dr. Sparano noted that the team developed an AI model that evaluates not only tumor gene expression but also uses deep learning of digitized H&E slides used for routine pathologic assessment to provide better prognostic information about cancer recurrence risk.
“We found that the expanded gene panel was a strong predictor of early recurrence within 5 years after diagnosis, the pathomic imaging was a strong predictor of late recurrence after 5 years, and when combined, a test which added both features to the prognostic information provided by clinicopathologic factors was the strongest predictor of distant recurrence out to 15 years,” he said. - A Multimodal-Multitask Deep Learning Model Trained in NSABP B-42 and Validated in TAILORx for Late Distant Recurrence Risk in HR+ Early Breast Cancer, presented by Eleftherios (Terry) Mamounas, MD, MPH (NSABP Foundation, Inc. and AdventHealth Cancer Institute). Abstract RF3-07 was presented during SABCS Rapid Fire Session 3.
Patients with early-stage, hormone receptor–positive (HR+) breast cancer are at risk for distant recurrence several years after diagnosis and initial treatment, making long-term risk assessment critical. Assessment of clinical factors alone (tumor size, grade, node status) is insufficient for precise risk stratification. Furthermore, there is a lack of personalized tools to guide decisions about the use of extended endocrine therapy (EET) beyond the standard 5 years.
Dr. Mamounas presented a multimodal–multitask deep learning algorithm designed to estimate late distant recurrence (DR) risk and help identify patients most likely to benefit from EET. Originally developed and validated in the NSABP B-42 randomized phase 3 trial, the model demonstrated strong prognostic performance, identifying those with minimal recurrence risk after a standard 5-year course of adjuvant endocrine therapy who could be spared additional treatment.
The ECOG-ACRIN/Caris research team conducted a new external validation study of the model in 4,300 patients from the TAILORx trial. In TAILORx, the model demonstrated robust late distant recurrence prognostication independent of other known prognostic factors, supporting its potential clinical utility as a scalable, cost-effective alternative to genomic assays using routine H&E and clinical data.
Source: PRNewswire