Development and validation of machine learning prognostic models for overall survival in non-surgical prostate cancer patients with bone metastases - PubMed
5 hours ago
- #Machine Learning
- #Prostate Cancer
- #Survival Analysis
- Objective: Develop machine learning models to predict overall survival in non-surgical prostate cancer patients with bone metastases (PCBM).
- Methods: Used data from 3,378 SEER database patients to develop survival models, with the best model interpreted using SHAP.
- Results: Extra Survival Trees (EST) model performed best (validation AUC = 0.694, C-index = 0.643). Gleason score was the most critical survival factor, surpassing clinical T stage. Visceral metastasis and advanced age also increased mortality risk.
- Conclusion: EST model effectively assesses overall survival in non-surgical PCBM. Gleason score has greater prognostic value than local anatomical staging, suggesting early aggressive treatment for high-Gleason, high-burden patients.