Abstract
BACKGROUND: This study aims to externally validate the performance of the Oncotype DX (ODX) breast cancer (BC) recurrence score nomogram in predicting adjuvant chemotherapy (ACT) for BC after surgery and subsequently develop a machine learning-based model to predict postoperative overall survival (OS) and guide ACT, demonstrating superior comprehensive performance. METHODS: This analysis leveraged data from the SEER database spanning 2010-2020, alongside a BC cohort from the Beijing Hospital (BJH). Machine learning methods were applied for predictor selection by wrapper methods and the development of the predictive model. The optimal model was determined using the concordance index (C-index), time-dependent calibration curves, time dependent receiver operating characteristic (ROC) curves, and decision curve analysis (DCA). The benefit analysis of ACT was primarily conducted using Kaplan-Meier survival analysis. RESULTS: The ODX nomogram performed poorly in predicting ACT benefit in both the SEER cohort and the BJH cohort. Subsequently, we employed ten machine learning methods to develop ten prognostic models. The Accelerated oblique random survival forest model (AORSFM), exhibiting the highest prediction performance, was selected. The C-index for AORSFM is 0.799 (95% CI 0.779-0.823) in the SEER cohort and 0.793 (95% CI 0.687-0.934) in the BJH cohort. Furthermore, time-dependent calibration curves, time-dependent ROC analysis, and DCA indicate that the AORSFM demonstrates good calibration, predictive accuracy, and clinical net benefit. A publicly accessible web tool was developed for the AORSFM. Notably, the new staging system based on AORSFM can provide guidance for postoperative ACT in such patients. CONCLUSIONS: The AORSF has the potential to identify postoperative OS and guide ACT in patients with BC. This can assist clinicians in assessing the severity of the disease, facilitating patient follow-up, and aiding in the formulation of adjuvant treatment strategies.