Assessment of POPs in foods from western China: Machine learning insights into risk and contamination drivers

Abstract

Persistent organic pollutants (POPs), including PCDD/Fs, PCBs, and PBDEs, are major environmental and food safety concerns due to their bioaccumulative and toxic properties. However, comprehensive research on the concentrations and influencing factors of POPs across different food types and regions, particularly in underdeveloped regions of western China, remains scarce. This study conducted a comprehensive assessment of POPs contamination in six food types (pig liver, pork, freshwater fish, marine fish, beef, and eggs) from western China by integrating environmental, geographical, socio-economic data, and food POP concentrations with machine learning and multivariate analyses to evaluate distribution patterns, key influencing factors, and associated health risks. The results showed distinct contamination patterns across food types and regions. Among all food, pig liver exhibited the highest levels of ∑PCDD/Fs, while marine fish showed elevated PBDEs and ndl-PCBs, highlighting the influence of organ-specific bioaccumulation and global oceanic pollution. Freshwater fish displayed higher ∑PCDD/Fs due to localized agricultural and industrial pollution. Regional differences were most pronounced in pork, with higher contamination in Yunnan and Sichuan, driven by industrial emissions, biomass burning, and geographical factors. Regression models, particularly Random Forest and SHAP analyses, identified food type, latitude, GDP, and climatic conditions as key predictors of POP variability. Risk assessments indicated that dietary exposure to POPs from high-consumption foods remained within safety thresholds, posing no significant health risks to the general population. This study highlights the utility of advanced analytical tools in understanding contamination dynamics and emphasizes the need for systematic monitoring, targeted interventions, and enhanced food safety regulations, particularly in western China.

Keywords: Food safety; Machine learning; Polybrominated diphenyl ethers (PBDEs); Polychlorinated biphenyls (PCBs); Polychlorodibenzo-p-dioxin/furans (PCDD/Fs); Risk assessment.