Abstract PS3-03-04: Predictive Factors of High Risk Were Indirectly Compared in Node-negative Stage I-II Breast Cancer Patients Examined Oncotype Dx or EndoPredict: The BRAIN Study

Abstract Purpose There have been reported different characteristics of multigene assays (MGAs) to discriminate high and low risk groups in estrogen receptor (ER)-positive/HER2-negative breast cancer patients. In real world data, however, direct comparison between MGAs is not easy because clinical guidelines recommended against using two or more MGAs in a patient simultaneously. In this study, clinicopathological characteristics and results of Oncotype Dx (ODx) and EndoPredict (EP) were indirectly compared to investigate characteristics of each assay. Methods A total of 1,782 node-negative stage I/II patients with ER-positive/HER2-negative breast cancer were retrospectively selected between 2013 and 2025. Of them, 1,537 (86.3%) patients performed ODx and 245 (13.7%) examined EP. High risk of ODx was defined considering clinical risk and recurrence score. EPclin was used to determine high risk of EP. Clinicopathological characteristics and assay results were compared using logistic regression models. After 1:1 propensity score matching (PSM) adjusting clinicopathological characteristics, predictive factors for high risk assay result were re-analyzed. Results High risk of ODx and EP was 25.4% and 24.5%, respectively. EP patients showed older age, more obese, and higher mixed/lobular/special histology. ODx patients demonstrated larger tumor size, higher grade II/III disease, and higher Ki-67 levels. Allred score of ER and progesterone receptor (PR) and HER2 immunohistochemical score were not different. Correlation analysis showed ODx recurrence score was significantly associated with ER (Spearman’s rho, -0.079) and PR (Spearman’s rho, -0.425) Allred score negatively and with Ki-67 levels (Spearman’s rho, 0.251) positively. EP 12-gene molecular score was negatively associated with PR Allred score (Spearman’s rho, -0.290) and positively related to Ki-67 values (Spearman’s rho, 0.516). Multivariate analysis adjusting clinicopathological characteristics showed that EP had a higher probability of high risk Hazard ratio, 1.783 (95% confidence interval, 1.330-2.390); p0.001. Subset analysis by each assay showed younger age, lower ER and PR Allred score, and HER2 2+ immunohistochemical score were major determinants of high risk in ODx patients. But larger T2-3 stage, histologic grade II/III, and Ki-67 ≥20% were significant factors of high risk in EP cases. After 1:1 PSM, assay method was not significant Hazard ratio, 1.286 (95% confidence interval, 0.761-2.174); p=0.348 but T2-3 stage, grade II/III, low PR, and higher Ki-67 were independently significant predictors of high risk result. Conclusion Each MGA showed different features of major determinants for high risk group. Although definition of high risk considering clinical risk factors and gene score might be partly related, ODx was strongly dependent on ER-related signatures and EP was significantly determined by proliferative features of histologic grade and Ki-67 levels. Primary physician and patient could select one among available MGAs based on understanding of its distinct characteristics. Citation Format: J. Ahn. Predictive Factors of High Risk Were Indirectly Compared in Node-negative Stage I-II Breast Cancer Patients Examined Oncotype Dx or EndoPredict: The BRAIN Study abstract. In: Proceedings of the San Antonio Breast Cancer Symposium 2025; 2025 Dec 9-12; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2026;32(4 Suppl):Abstract nr PS3-03-04.