XGBoost model for predicting erectile dysfunction risk after radical prostatectomy: development and validation using...

Area of Science:

• Urology
• Oncology
• Data Science

Background:

• Erectile dysfunction (ED) is a common, quality-of-life impacting complication after radical prostatectomy.
• Traditional methods for predicting ED risk are limited in capturing complex, nonlinear factors.

Purpose of the Study:

• To develop and validate a machine learning (ML) model for improved ED risk stratification post-radical prostatectomy.
• To enhance personalized patient management strategies for ED.

Main Methods:

• Analysis of 1,147 prostate cancer patients, with 285 (24.85%) developing ED.
• Utilized ML models (XGBoost, Random Forest, SVM, k-NN) trained on identified risk factors.
• Key predictors included age, smoking, Gleason score, prostate volume, T-stage, surgical approach, operative time, and PCT levels.

Main Results:

• XGBoost demonstrated superior predictive accuracy (AUC 0.960 in validation) and clinical utility.
• Calibration and decision curve analyses confirmed model performance and benefit.
• SHAP analysis identified key risk contributors for individualized assessment.

Conclusions:

• The XGBoost model offers robust predictive performance and clinical applicability for ED risk assessment.
• Provides a valuable tool for personalized postoperative management of ED after radical prostatectomy.