Development and validation of a clinical-pathological tool for RS prediction: A practical alternative for resource-limited settings.

e13637 Background: International guidelines recommend the 21-gene Recurrence Score (RS) to guide adjuvant chemotherapy decisions in early-stage luminal breast cancer. However, high costs and limited availability create significant barriers to access in countries like Mexico, where only a minority of patients can afford genomic testing. We developed a multivariable linear regression equation based on standard clinico-pathological features to predict RS, aiming to provide a practical tool for treatment optimization in resource-constrained environments. Methods: We analyzed 90 Mexican patients with HR+/HER2- early breast cancer (Stages IA-IIB, pT1b-c, pN0-1) . Clinicopathological variables were correlated with RS results using SPSS v31 . RS was analyzed as a continuous and a dichotomous variable (Low-Intermediate: 0-25; High: > = 26). A multivariable regression model was constructed using an interaction term between SBR grade and Ki67 . The model's performance was compared against actual RS results and the Magee Score (Equation 1) using 2x2 contingency tables to calculate sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV). Results: Median age was 57 years; 32% were premenopausal, 37% were node-positive (1-3 nodes), and median Ki67 was 13% . Univariate analysis showed that Progesterone Receptor (PR) intensity, Nottingham (SBR) score/grade, and Ki67 significantly correlated with RS (p < 0.006), while age, tumor size, and nodal status did not . The multivariable model (R = 0.630, R^2 = 0.396, p < 0.001) yielded the equation: RS = 2.512 - 3.979(PR) + 3.009(SBR\Sum) - 9.094(G3) - 0.252(G1\times Ki67) + 0.150(G2\times Ki67) + 0.487(G3\times Ki67). When dichotomized, the model achieved a sensitivity of 46.7%, a specificity of 98.7%, a PPV of 87.5%, and an NPV of 90.2% (p< 0.001). Conclusions: The proposed Mexican regression equation is a highly specific tool for predicting RS risk groups. This model demonstrates strong alignment with established benchmarks such as the Magee Score, while offering particularly robust specificity (98.7%) and PPV (87.5%) within our specific population . These results suggest a reliable and cost-effective strategy to identify patients who are unlikely to require genomic testing, serving as a practical alternative to expand the reach of precision medicine and optimize resource allocation in oncology settings with limited access to commercial platforms . Performance and predictive accuracy. Actual Recurrence Score (Oncotype DX) Mexican Multivariate Regression Model High Risk (>=26) Low-Intermediate Risk (0-25) Total High Risk 7 (87.5% PPV) 1 (46.7% Sensitivity) 8 Low-Intermediate Risk 8 (90.2% NPV) 74 ( 98.7% Specificity) 82 Total 15 75 90