Abstract PS3-09-09: Validation and Enhancement of the Tennessee Nomogram Incorporating Ki-67 for Predicting Oncotype DX Risk in Hormone Receptor-Positive, HER2-Negative Breast Cancer in a Korean Cohort

Abstract Oncotype DX (ODX) is widely used for predicting recurrence risk and guiding adjuvant therapy in hormone receptor-positive (HR+), HER2-negative breast cancer. However, access limitations have prompted the development of alternative tools, such as the Tennessee nomogram. Previous studies in a Korean population reported moderate accuracy but highlighted discrepancies, particularly among tumors with aggressive features. This study aimed to validate the predictive accuracy of the Tennessee nomogram in a Korean cohort and evaluate whether integrating Ki-67 could enhance its performance. Methods: We retrospectively analyzed data from 1,324 patients with invasive ductal or lobular carcinoma, HR-positive, HER2-negative, and node-negative breast cancer who underwent ODX testing. Based on an analysis of discordant cases from the original Tennessee model, we developed a new nomogram incorporating Ki-67 as an additional variable. We then compared the performance of this revised nomogram with the original Tennessee model using sensitivity, specificity, accuracy, PPV, NPV, and area under the curve (AUC). Additional analyses were performed by histologic grade. Results: The original Tennessee nomogram demonstrated an accuracy of 85.8%, with sensitivity of 12.8%, specificity of 99.1%, and an AUC of 0.784. Discordant predictions were often associated with aggressive tumor features such as high histologic grade, PR negativity, and elevated Ki-67, especially among false negatives whose ODX scores clustered between 25 and 30. Incorporating Ki-67 into the nomogram improved sensitivity to 29.2% and slightly increased overall accuracy to 86.7%, although specificity declined modestly to 97.3%. The number of false negatives decreased from 197 to 160, indicating better identification of high-risk patients. However, false positives rose from 11 to 34, reflecting a trade-off between higher sensitivity and somewhat reduced specificity. Despite this, the revised model achieved a higher AUC of 0.812, suggesting improved overall discrimination and better clinical utility when ODX testing is not readily available. In subgroup analyses by histologic grade, the revised model showed notable gains. For grade 2 tumors, sensitivity increased from 4.2% to 15.3% while maintaining high specificity (>99%), leading to higher overall accuracy and an improved AUC from 0.675 to 0.739. In grade 3 tumors, sensitivity nearly doubled from 35.9% to 68.8%; however, this was accompanied by a significant drop in specificity (87.5% to 48.2%) and only a modest improvement in AUC. Conclusions: The Tennessee nomogram remains a useful tool when ODX testing is unavailable, but its sensitivity is limited, particularly in tumors with aggressive features. Incorporating Ki-67 improves predictive performance overall and offers benefits in grade 2 tumors, where sensitivity improves without substantial loss of specificity. These results underscore the clinical value of Ki-67 as a biomarker in intermediate-grade breast cancer and support its inclusion in future predictive nomograms. Citation Format: S. Lee, H. Lee, S. Lee, J. Ahn, J. Kim, I. Lee, S. Park, N. Son. Validation and Enhancement of the Tennessee Nomogram Incorporating Ki-67 for Predicting Oncotype DX Risk in Hormone Receptor-Positive, HER2-Negative Breast Cancer in a Korean Cohort [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 PS3-09-09.