31 A new artificial intelligence model provides more accurate forecasts of early-stage breast cancer recurrence than the widely used 21-gene Recurrence Score, according to research published in the journal npj Breast Cancer. The study was conducted by the ECOG ACRIN Cancer Research Group and Caris Life Sciences. It focuses on hormone receptor-positive and HER2-negative disease, the most common breast cancer subtype, accounting for roughly half of all US cases. “Powered by artificial intelligence integrating clinical, molecular, and histopathology data, this new test provides more reliable prognostic information for breast cancer recurrence,” said lead author Joseph A. Sparano, MD, of the Icahn School of Medicine at Mount Sinai. The model, named IICM+, combines digitised pathology images of the tumour with patient age, tumour size and grade, and molecular data from an expanded 42-gene panel. The benchmark 21-gene Recurrence Score (Exact Sciences Corporation’s Oncotype DX test) guides chemotherapy use in hormone receptor-positive, HER2-negative, lymph node-negative early breast cancer. However, it primarily predicts recurrence within five years of diagnosis. More than half of distant recurrences occur after five years, extending to 10, 15, or over 20 years later. Researchers developed IICM+ using tumour specimens and clinical data from 2,808 TAILORx trial participants. They then independently validated the completed model in a separate group of 1,621 TAILORx participants. Both cohorts had a median follow-up of over 11 years. In the independent validation group, IICM+ significantly outperformed the 21-gene Recurrence Score in distinguishing patient risk levels, as measured by the C-index, with higher scores indicating better performance. For overall distant recurrence, IICM+ achieved a C-index of 0.735, compared with 0.578 for the gene test (P less than 0.001). When evaluating early distant recurrence within five years, the model scored 0.791, compared with 0.722 (P equals 0.046). The disparity was most pronounced for late distant recurrence occurring after five years, where IICM+ scored 0.710, compared with 0.514 (P less than 0.001). Among patients with a Recurrence Score of 0–25, indicating lower genomic risk, IICM+ identified a subset with a higher observed risk of distant recurrence. Conversely, among those scoring 26–100, indicating higher genomic risk, IICM+ identified a subset whose actual recurrence risk was lower than predicted by the standard gene test. “Using multi-modal data (clinical data, Next Generation Sequencing data, and data obtained from imaging), integrated through modern artificial intelligence techniques, allowed us to create a truly novel means of interrogating breast cancers for important prognostic information,” said George W. Sledge, Jr, MD, and Chief Medical Officer of Caris. As part of a public-private partnership with ECOG-ACRIN, Caris converted archived TAILORx pathology slides into digital images and applied deep learning to extract new insights from long-stored specimens. “TAILORx continues to yield important new insights years after its original findings—a testament to the enduring value of Cooperative Group research,” said Peter J. O’Dwyer, MD, ECOG-ACRIN Group Co-Chair. “Applying today’s technologies to this exceptional research resource creates opportunities to further individualise breast cancer care, and with NCI’s concurrence and funding from Caris, this public-private partnership is a model for accelerating progress.” The findings do not yet establish IICM+ as a tool for selecting or altering treatment. Further studies are in development to validate the model in other patient populations and determine whether its clinical use can directly improve patient outcomes.