Predicting benign prostatic hyperplasia risks: model development and external validation based on three cohorts.

Area of Science:

• Urology
• Predictive Modeling
• Public Health

Background:

• Benign prostatic hyperplasia (BPH) is a common condition in aging males.
• Accurate risk prediction for BPH is crucial for timely intervention.
• Current methods for BPH risk assessment may lack simplicity or accuracy.

Purpose of the Study:

• To develop and validate a robust prediction model for identifying individuals at high risk of developing BPH.
• To utilize readily available medical characteristics for BPH risk assessment.
• To create a clinically applicable tool for BPH risk stratification.

Main Methods:

• Utilized UK Biobank data (n=210,408) for model development and CHARLS (n=5394) and Fengshen (n=294) for external validation.
• Employed six machine learning methods, including LightGBM, to construct prediction models.
• Assessed model performance using DeLong tests for area under the curve (AUC) comparisons and Cox regression for predictor significance.

Main Results:

• The LightGBM model demonstrated superior discriminative capability (AUC=0.688±0.004) with 17 predictors.
• Age was identified as the most significant predictor (HR=1.091).
• A simplified 5-predictor model (age, hypertension time, blood glucose, urate, serum creatinine) was developed and validated, with a user-friendly web tool created.

Conclusions:

• A simplified model using five accessible predictors offers acceptable predictive ability for incident BPH.
• This model can effectively identify individuals at high risk of BPH in the general population.
• The developed tool enhances clinical utility for BPH risk assessment and management.