Associations between prenatal multiple metal exposure and preterm birth: Comparison of four statistical models

Exposure to some heavy metals has been demonstrated to be related to the risk of preterm birth (PTB). However, the effects of multi-metal mixture are seldom assessed. Thus, we aimed to investigate the associations of maternal exposure to metal mixture with PTB, and to identify the main contributors to PTB from the mixture.

The population in the nested case-control study was from a prospective cohort enrolled in Wuhan, China between 2012 and 2014. Eighteen metals were measured in maternal urine collected before delivery. Logistic regression, elastic net regularization (ENET), weighted quantile sum regression (WQSR), and Bayesian kernel machine regression (BKMR) were used to estimate the overall effect and identify important mixture components that drive the associations with PTB.

Logistic regression found naturally log-transformed concentrations of 13 metals were positively associated with PTB after adjusting for the covariates, and only V, Zn, and Cr remained the significantly positive associations when additionally adjusting for the 13 metals together. ENET identified 11 important metals for PTB, and V (β = 0.23) had the strongest association. WQSR determined the positive combined effect of metal mixture on PTB (OR: 1.44, 95%CI: 1.32, 1.57), and selected Cr and V (weighted 0.41 and 0.32, respectively) as the most weighted metals. BKMR analysis confirmed the overall mixture was positively associated with PTB, and the independent effect of V was the most significant. Besides, BKMR showed the non-linear relationships of V and Cu with PTB, and the potential interaction between Zn and Cu.

Applying different statistical models, the study found that exposure to the metal mixture was associated with a higher risk of PTB, and V was identified as the most important risk factor among co-exposed metals for PTB.