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

作者：Hao Zi, Yongbo Wang, Qiao Huang, Yuanyuan Zhang, Fazhi He, Limin Xing, Yao Yan, Binghui Li, Li-Sha Luo, Fei Li, Shi‐Di Tang, Xian‐Tao Zeng, Jiao Huang · 发表于：Global Health Research and Policy · 年份：2025 · DOI：10.1186/s41256-025-00456-4 · 被引用次数：3 · 研究领域：Urinary Bladder and Prostate Research、Prostate Cancer Diagnosis and Treatment、Sexual function and dysfunction studies

BACKGROUND: As benign prostatic hyperplasia (BPH) becomes increasingly prevalent, there is a growing need for simple and accurate methods to predict its risk. This study aimed to develop and validate a prediction model to identify males at high risk of developing BPH. METHODS: The model was developed using data from 210,408 participants in the UK Biobank and externally validated with 5394 participants from the China Health and Retirement Longitudinal Study (CHARLS) and 294 participants from the Fengshen study. Six methods were employed to construct prediction models utilizing readily available medical characteristics at baseline. The DeLong tests were used to assess the differences in the area under the curves (AUCs). Cox regression was adopted to examine the relationships between the predictors and BPH. RESULTS: During a median follow-up period of 13.2 years (interquartile range [IQR] 12.3-14.0), 7.0 years (IQR 6.8-7.0) and 4.0 years (IQR 2.2-5.0), 18,681 males in the UK Biobank, 309 males in the CHARLS, and 27 males in the Fengshen study developed BPH. The model developed using the LightGBM method exhibited the highest discriminative capability among the six methods. Following feature reduction based on importance ranking, a full model with 17 predictors was established for BPH prediction (AUC = 0.688 ± 0.004). Age was the most important feature that contributed to the model, with older males showing a higher hazard ratio (HR) of 1.091 (95% confidence interval [CI] 1.089-1....