Phenotypic age acceleration as a novel predictor of benign prostatic hyperplasia: a prospective cohort study

Abstract This study aims to investigate the predictive value of combined phenotypic age and phenotypic age acceleration (PhenoAgeAccel) for benign prostatic hyperplasia (BPH) and develop a machine learning-based risk prediction model to inform precision prevention and clinical management strategies. The study analyzed data from 784 male participants in the US National Health and Nutrition Examination Survey (NHANES, 2001–2008). Phenotypic age was derived from chronological age and nine serum biomarkers. PhenoAgeAccel, representing biological aging acceleration, was calculated as the residual from regressing phenotypic age on chronological age. Recursive Feature Elimination (RFE) identified 34 BPH-associated features, which were integrated into an XGBoost prediction model. Logistic regression evaluated PhenoAgeAccel-BPH associations, while SHapley Additive exPlanations (SHAP) quantified feature contributions to enhance model interpretability. The XGBoost model achieved an area under the curve (AUC) of 0.833 in the test set. Phenotypic age was strongly correlated with chronological age ( r = 0.833), and individuals with PhenoAgeAccel exhibited a significantly elevated risk of BPH ( p < 0.001). Adjusting the model with phenotypic age improved predictive performance (AUC = 0.853). SHAP analysis identified phenotypic age as the third most influential predictor (after trailing cancer history and lead exposure), highlighting its clinical relevance. Chronological age and serum biomarkers are critical predictors of BPH, while PhenoAgeAccel independently contributes to risk stratification. Integrating phenotypic age with machine learning provides a robust framework for the early detection of BPH and personalized risk assessment, aligning with advancements in aging biomarker research. This approach supports targeted interventions to mitigate BPH progression in aging populations.