Plasma metabolomics combined with machine learning for postoperative prognostic stratification in hormone receptor-positive/human epidermal growth factor receptor 2-negative early breast cancer.

BACKGROUND AND AIMS: Hormone receptor-positive/human epidermal growth factor receptor 2-negative (HR+/HER2-) breast cancer (BC) accounts for the majority of BC cases. Although early-stage patients generally have favorable outcomes, recurrence and metastasis substantially worsen prognosis. We aimed to evaluate whether plasma metabolomics combined with machine learning could predict postoperative outcomes in HR+/HER2- early BC. METHODS: A total of 178 patients with HR+/HER2- BC were prospectively enrolled. Preoperative plasma samples were analyzed using untargeted metabolomics based on liquid chromatography-mass spectrometry. Differential metabolites between patients with short (< 5 years) and long (≥ 5 years) disease-free survival (DFS) were identified. Using least absolute shrinkage and selection operator feature selection followed by random forest classification, a prognostic model was constructed in a discovery cohort (n = 125) and validated in an independent test cohort (n = 53). RESULTS: A total of 209 metabolites were annotated, of which 100 were significantly altered between groups. Key dysregulated metabolic pathways included linoleic acid and galactose metabolism. A 10-metabolite prognostic model (10-PM) achieved an area under the receiver operating characteristic (AUROC) curve of 0.907 (95% CI: 0.883-0.925) in the test cohort, outperforming conventional clinicopathological models. The 10-PM model effectively stratified patients into high- and low-risk groups with significantly different DFS and overall survival, and remained an independent predictor in multivariable Cox regression analysis. CONCLUSION: We developed and validated a 10-PM model based on plasma metabolomics that accurately predicts short DFS in HR+/HER2- BC. This model outperforms conventional clinicopathological indicators, offering a specific, blood-based tool to stratify high-risk patients and guide individualized postoperative management.