Predicting progression-free survival in hormone-receptor positive (HR+/HER2–) metastatic breast cancer (MBC) treated with CDK4/6 inhibitors: A machine learning approach

Background In HR+/HER2– metastatic breast cancer (MBC), CDK4/6 inhibitors combined with endocrine therapy (ET) significantly improve progression-free survival (PFS). Machine learning (ML) approaches may improve individualized progression risk estimation. Methods We retrospectively analysed HR+/HER2– MBC patients treated with first-line CDK4/6i plus ET to develop CoxNet regression and Gradient Boosting Machine (GBM) models from baseline clinicopathological features. The primary endpoint was PFS prediction. The dataset was split into a 70/30 train/validation set. Performance was assessed by Harrell's C-index (1000 bootstrap replicates). Risk stratification was performed using Gaussian Mixture Modeling (GMM) to define high- and low-risk groups. Cox regression estimated the corresponding hazard ratios (HR). Early progression at 6 months (EP) prediction was evaluated using the area under the receiver operating characteristic curve (AUROC). Results 459 patients were included, with a median follow-up of 43.7 months (95% CI 39.6–48.3). Median PFS was 29.3 months (95% CI 24.0–33.7). Both ML models achieved strong predictive performance, with a Harrell's C-index of 0.74 (95% CI 0.67–0.80) in the validation set. The main predictors were liver metastases, Ki67 expression, and primary endocrine resistance. Stratification defined two risk groups with significantly different PFS in the validation set (HR 2.58, 95% CI 1.65–4.03, p = 3.3 × 10 −5 ). Median PFS was 34.8 (95%CI 24.0–52.4) in the low-risk and 10.6 (95%CI 7.7–14.6) in the high-risk group. For EP prediction, the model achieved an AUROC of 0.77 (95% CI 0.61–0.89). Conclusions This study supports the clinical applicability ML models using baseline clinicopathological variables for individualized risk stratification in HR+/HER2– MBC.