Combined effects of persistent organic pollutants and endocrine-disrupting chemicals on thyroid disease risk: A WQS-XGBoost analysis of US population data.

To assess the combined effects of mixed persistent organic pollutants and endocrine-disrupting chemicals on thyroid disease risk, we analyzed data from 1847 participants in the National Health and Nutrition Examination Survey 2011-2014. Environmental exposures included per- and polyfluoroalkyl substances, heavy metals, and phthalate metabolites. Thyroid disease was defined by self-reported physician diagnosis or laboratory criteria. We applied weighted quantile sum regression to assess mixture effects and identify individual chemical contributions. Machine learning analysis used extreme gradient boosting (XGBoost) with Bayesian optimization for prediction, with model interpretability assessed through Shapley additive explanations (SHAP) framework. The environmental chemical mixture shows a positive association with increased thyroid disease risk in the crude model (OR = 2.74, 95 % CI: 1.16-6.46, p < 0.05), which became marginally significant after covariate adjustment (adjusted OR = 2.30, 95 % CI: 0.97-5.47, p = 0.059). Five chemicals account for 68 % of the mixture effect: 2-(N-methyl-PFOSA) acetate (weight = 0.28), perfluorohexanesulfonic acid (0.14), blood cadmium (0.10), blood lead (0.08), and mono(2-ethyl-5-oxo) phthalate (0.08). The XGBoost model achieves excellent predictive performance (AUC = 0.978, precision = 0.986, recall = 0.900, and F1-score = 0.940). SHAP analysis reveals significant age-chemical mixture interactions, with individuals over 63 years showing dramatically increased susceptibility to environmental exposures (80 % higher predicted risk at high exposure levels). These results suggest that combined exposure to environmental chemical mixtures marginally increases thyroid disease risk in US adults, with older populations representing a particularly vulnerable subgroup. The cumulative effects of mixed exposures highlight the importance of mixture-based risk assessment approaches.