Machine learning-enhanced meta-analysis unravels the global patterns of microplastic-heavy metal co-toxicity in...

The combined pollution of microplastics (MPs) and heavy metals (HMs) poses a significant threat to global food security and ecological health. However, the combined toxic effects of this co-contamination on terrestrial organisms and the underlying mechanisms remain unclear. This study synthesizes 1820 control-observation datasets from 113 toxicity experiments worldwide. By integrating meta-analysis with machine learning models, we systematically evaluated the synergistic toxic effects of MPs and HMs on terrestrial organisms. The results showed that the co exposure of MPs and HMs significantly inhibited plant growth (such as a 22.14 % decrease in plant height and a 17.80 % decrease in biomass) and microbial diversity (such as a 2.65 % decrease in Shannon index), and exacerbated animal toxicity (such as a 7.00 % decrease in survival rate and a 46.61 % increase in intestinal damage). The excellent predictive performance of the XGBoost model (R² = 0.71) indicates that it can be applied to the risk prediction of MPs and HMs combined pollution. Shapley additive explanations (SHAP) analysis and partial least squares path modeling revealed that HM concentration, exposure time, and MP particle size significantly modulated the magnitude of the combined toxicity, with nanoscale MPs exhibiting the most pronounced amplification effect on the combined pollution toxicity. This study elucidates the ecological risks associated with combined MP and HM pollution, providing a scientific basis for global soil pollution monitoring and remediation, and it underscores the importance of standardizing experimental protocols and promoting international collaboration in pollution management.