Unveiling the association between dietary metals and prostate health: Identifying benign prostatic hyperplasia using a SHAP-based interpretable machine learning model in US adults.

BACKGROUND

Benign prostatic hyperplasia affects roughly 50 % of men aged 51-60 and 70 % of those aged 61-70. While heavy metals have been linked to prostate cancer and inflammation, little is known about how the dietary intake of trace metals influences BPH risk. This study aimed to establish an interpretable, machine-learning-based framework using U.S. NHANES data to identify key dietary metals associated with BPH risk.

METHODS

We analyzed adult male participants from NHANES cycles 2005-2008, extracting daily intake estimates for six trace metals. This study compared six algorithms to elucidate the relationship between dietary metal intake and BPH risk, assessing their performance in associating BPH risk across seven evaluation metrics to identify the optimal model and enhance interpretability through analysis with SHapley Additive exPlanations (SHAP). Subgroup analyses explored nonlinear intake-risk relationships.

RESULTS

LightGBM demonstrated superior performance (AUC = 0.668). SHAP ranking identified zinc, iron, and selenium as the top protective nutrients. The most pronounced protective effect against BPH was observed within the moderate zinc and selenium intake strata. Iron demonstrated a consistent protective association, with risk reduction observed across its intake range. Copper intake demonstrated a shift from risk reduction for BPH at low to moderate levels to an increased risk at high intake levels.

CONCLUSIONS

These findings suggest that calibrated intake of these key micronutrients may attenuate BPH susceptibility, providing a data-driven foundation for targeted nutritional interventions and public-health strategies.