What Drives Microplastic Exposure in Human Blood and Feces? Machine Learning Reveals Potential Key Influencing...

Microplastics are pervasive environmental pollutants, making human exposure unavoidable. Although previous studies have detected microplastics in human blood and feces, these investigations were limited by small sample sizes and key contributors to microplastic biomonitoring remain underexplored. In this study, we analyzed 229 blood and 227 fecal samples using pyrolysis-gas chromatography-mass spectrometry to quantify microplastic exposure and identify key influencing factors through machine learning modeling. Seven polymer types were detected in both biological matrices, including polyethylene, poly(vinyl chloride) (PVC), polypropylene, polystyrene, polyamide 66, poly(ethylene terephthalate), and poly(methyl methacrylate), with polyethylene, PVC, and polystyrene being the most prevalent. A significant negative correlation was observed between blood and fecal PVC levels, while other polymers showed no significant intermatrix association. Demographic, lifestyle, socioeconomic, and dietary information were collected via questionnaires. Age, sex, geographic, and indoor environmental variations in microplastic levels were observed. Machine learning models were developed to predict microplastic levels from anthropometric measurements and questionnaire data. Explainable AI tools revealed that the drinking water source was the strongest predictor of blood PVC levels. Socioeconomic factors, including income and education, were also significant predictors, with lower-income individuals showing higher microplastic burdens. This approach provides a framework for understanding and mitigating microplastic exposure.