ETD

• breast cancer
• grade 2
• Oncotype
• risk stratification

Histological intermediate-grade (G2) estrogen receptor-positive (ER+), HER2-negative breast cancer represents a highly heterogeneous and borderline subgroup, for which optimal prognostic stratification remains a significant clinical challenge. Oncotype DX Recurrence Score (RS) plays a pivotal role in guiding adjuvant chemotherapy decisions; its clinical interpretation varies across age groups. Utilizing RS-defined categories as a validated standard for low- and high-risk disease, this retrospective multicenter study evaluated the predictive potential of routinely available clinicopathological variables to develop a decision-support tool for therapeutic management.
A retrospective cohort analysis was conducted on 666 patients with early-stage G2 ER+/HER2-negative breast cancer who underwent Oncotype DX testing (2018–2025). The population was stratified into two age cohorts: patients > 50y (n = 523) and ≤ 50y (n = 143). Statistical associations between genomic RS and clinicopathological features (age, tumor size, histological subtype, ER, PgR and Ki-67 expression, and HER2 status) were evaluated using univariate and multivariable logistic regression. Receiver Operating Characteristic (ROC) analysis determined the discriminative performance of individual and combined markers, with clinical thresholds derived via the Youden Index.
In both age cohorts, quantitative ER and PgR expression were significantly inversely related to RS, whereas Ki-67 correlated directly (all p ≤ 0.03). Individually, PgR expression displayed the strongest discriminative capacity in both age groups (AUC > 0.85), with optimal thresholds at 25% for patients > 50y and 60% for those ≤ 50y, yielding high specificity (> 0.83) and an exceptional Negative Predictive Value (NPV, up to 0.96). Ki-67 showed moderate discrimination (AUC > 0.65) with cut-offs of 20% (> 50y) and 25% (≤ 50y). Despite statistical significance, ER provided limited clinical utility due to lower discrimination (AUC ≤ 0.64) and hardly-applicable thresholds (95% and 90%, respectively). Crucially, multivariable models integrating PgR, ER, and Ki-67 outperformed single parameters, achieving superior accuracy across both cohorts (AUC 0.89, specificity > 0.94, accuracy 0.90). Other variables (tumor size, histological subtype, HER2 status) showed no significant association with RS.
Progesterone receptor expression emerges as the most powerful individual marker for high-risk disease, operating through distinct, age-specific thresholds. However, integrating ER, PgR, and Ki-67 expression into a unified multivariate model significantly optimizes risk stratification for intermediate-grade breast cancer. Backed by an outstanding NPV, this integrated tool safely identifies low-risk patients who can be spared chemotherapy, thereby providing a reliable asset to streamline routine clinical decision-making.