Machine learning prediction of glaucoma by heavy metal exposure: results from the National Health and Nutrition Examination Survey 2005 to 2008.

Using follow-up data from the National Health and Nutrition Examination Survey (NHANES) database, we have collected information on 2572 subjects and used generalized linear model to investigate the association between urinary heavy metal levels and glaucoma risk. In addition, we have developed an individualized risk prediction model using machine learning algorithms and further interpreted the model results through feature importance analysis, local cumulative analysis, and interaction effects. In this study, we found significant association between logarithmically calculated arsenic (As) metabolites, especially arsenochlorine (AC), and glaucoma after adjusting for a series of confounders, including urinary creatinine (β = 1.090, 95% CI: 0.313-1.835). The Shapley Additive Explanations (SHAP) analysis results and clinical risk scores also indicated that As metabolites promoted glaucoma more severely than other variables. This study applied machine learning for the first time to explore the relationship between heavy metals and glaucoma while analyzing the effects of multiple heavy metal exposures on the disease, improving the predictive power compared to conventional models. Our results provided important insights into the potential role of heavy metals in the pathogenesis of glaucoma, facilitated the discovery of new biomarkers for early diagnosis, risk assessment, and timely treatment of glaucoma, and guided public health measures to reduce heavy metal exposure.