Abstract Prenatal exposure to heavy metals (HMs) has been the focus of international research. However, current studies tend to examine individual metals in isolation and rely on traditional linear regression models, which may not adequately reflect the complex, non‐linear and interactive effects of mixed metal exposure. The aim of this study was to investigate the relationship between maternal mixed urinary HM exposure levels during pregnancy and adverse birth outcomes such as preterm birth (PTB), low birth weight (LBW) and small for gestational age (SGA) infants using advanced machine learning methods. This study was conducted at a tertiary hospital in Guilin, from 2022 to 2023. A total of 489 pregnant women were enrolled. First‐trimester urine samples were collected to quantify HM concentrations using Inductively coupled plasma mass spectrometry. Demographic and clinical data were obtained through structured questionnaires. Bayesian Kernel Machine Regression analysis revealed a significant cumulative effect of mixed metal exposure on adverse pregnancy outcomes, with distinct dose‐response relationships. The risk of PTB and LBW increased monotonically with higher exposure levels; the adjusted odds ratios were elevated as metal mixture concentrations increased from the 25th to the 75th percentile. In contrast, the association with SGA exhibited a non‐monotonic pattern—higher risk at lower exposure levels and a marked decline in risk at higher concentrations. Inorganic arsenic was identified as the primary toxic component in univariate models. Multivariate response modeling demonstrated the joint influence of metal mixtures on adverse outcomes (AUC = 0.697), with no significant statistical interactions between individual metals, as indicated by parallel dose‐response curves ( p > 0.05).
Sun et al. (2025) studied this question.