The impact of multipollutant exposure on hepatic steatosis: a machine learning-based investigation into multipollutant synergistic effects.

Introduction: This study examines the synergistic effects of multi-pollutant exposure on hepatic lipid accumulation in non-alcoholic fatty liver disease (NAFLD) through the application of an explainable machine learning framework. This approach addresses the limitations of traditional models in managing complex environmental interactions.

Methods: Using data from the National Health and Nutrition Examination Survey (NHANES) 2015-2016 (n = 494), we developed a stacked ensemble model that integrates LASSO, support vector machines (SVM), neural networks, and XGBoost to analyze urinary biomarkers of heavy metals, polycyclic aromatic hydrocarbons (PAHs), and volatile organic compounds (VOCs). The Environmental Pollution Exposure Index (EPEI) was constructed to quantify cumulative effects, with SHAP values employed to identify critical pollutants and thresholds. Subgroup analyses were conducted to assess heterogeneity across different Body Mass Index (BMI), diabetes, and hyperlipidemia statuses.

Results: 2-Hydroxynaphthalene was identified as the predominant pollutant (SHAP = 0.89), with cobalt and VOC metabolites (e.g., N-Acetyl-S-(2-carbamoylethyl)-L-cysteine) also contributing significantly. The EPEI demonstrated strong associations with obesity-related parameters (PLF: 7.02 vs. 3.41 in high/low-exposure groups,  < 0.0001) and hyperlipidemia (OR = 2.28 vs. 1.08,  = 2.7e-06). The model demonstrated an amplification of effects in subgroups with severe obesity (OR = 2.66, 95% CI: 2.08-3.24) and impaired fasting glucose.

Discussion: This study establishes a machine learning framework for assessing multi-pollutant risks in NAFLD, identifying 2-Hydroxynaphthalene as a significant hepatotoxicant and EPEI as a quantifiable metric of exposure. The findings highlight the metabolic vulnerabilities associated with obesity and early dysglycemia, thereby informing precision prevention strategies. Methodological advancements integrate exposomics with interpretable artificial intelligence, facilitating targeted interventions in environmental health.