Association of exposure to multiple heavy metals during pregnancy with the risk of gestational diabetes mellitus and insulin secretion phase after glucose stimulation

Epidemiological evidence for the association between heavy metals exposure during pregnancy and gestational diabetes mellitus (GDM) is still inconsistent. Additionally, that is poorly understood about the potential cause behind the association, for instance, whether heavy metal exposure is related to the change of insulin secretion phase is unknown. We aimed to explore the relationships of blood levels of arsenic (As), lead (Pb), thallium (Tl), nickel (Ni), cadmium (Cd), cobalt (Co), barium (Ba), chromium (Cr), mercury (Hg) and copper (Cu) during early pregnancy with the odds of GDM, either as an individual or a mixture, as well as the association of the metals with insulin secretion phase after glucose stimulation. We performed a nested case−control study consisting of 302 pregnant women with GDM and 302 controls at the First Affiliated Hospital of Anhui Medical University in Hefei, China. Around the 12th week of pregnancy, blood samples of pregnant women were collected and levels of As, Pb, Tl, Ni, Cd, Co, Ba, Cr, Hg and Cu in blood were measured. An oral glucose tolerance test (OGTT) was done in each pregnant woman during the 24−28th week of pregnancy to diagnose GDM and C−peptide (CP) levels during OGTT were measured simultaneously. The four metals (As, Pb, Tl and Ni) with the highest effect on odds of GDM were selected for the subsequent analyses via the random forest model. Conditional logistic regression models were performed to analyze the relationships of blood As, Pb, Tl and Ni levels with the odds of GDM. The weighted quantile sum (WQS) regression and bayesian kernel machine regression (BKMR) were used to assess the joint effects of levels of As, Pb, Tl and Ni on the odds of GDM as well as to evaluate which metal level contributed most to the association. Latent profile analysis (LPA) was conducted to identify profiles of glycemic and C−peptide levels at different time points. Multiple linear regression models were employed to explore the relationships of metals with glycaemia−related indices (fasting blood glucose (FBG), 1−hour blood glucose (1h BG), 2−hour blood glucose (2h BG), fasting C−peptide (FCP), 1−hour C−peptide (1h CP), 2−hour C-peptide (2h CP), FCP/FBG, 1h CP/1h BG, 2h CP/2h BG, area under the curve of C−peptide (AUCP), area under the curve of glucose (AUCG)), AUCP/AUCG and profiles of BGs and CPs, respectively. Mixed−effects models with repeated measures data were used to explore the relationship between As (the ultimately selected metal) level and glucose−stimulated insulin secretion phase. The mediation effects of AUCP and AUCG on the association of As exposure with odds of GDM were investigated using mediation models. The odds of GDM in pregnant women increased with every ln unit increase in blood As concentration (OR = 1.46, 95% CI = 1.04–2.05). The joint effects of As, Pb, Tl and Ni levels on the odds of GDM was statistically significant when blood levels of four metals were exceeded their 50th percentile, with As level being a major contributor. Blood As level was positively associated with AUCG and the category of glucose latent profile, the values of AUCG were much higher in GDM group than those in non−GDM group, which suggested that As exposure associated with the odds of GDM may be due to that As exposure was related to the impairment of glucose tolerance among pregnant women. The significant and positive relationships of As level with AUCP, CP latent profile category, 2h CP and 2h CP/2h BG were observed, respectively; and the values of 1h CP/1h BG and AUCP/AUCG were much lower in GDM group than those in non−GDM group, which suggested that As exposure may not relate to the impairment of insulin secretion (pancreatic β−cell function) among pregnant women. The relationships between As level and 2h CP as well as 2h CP/2h BG were positive and significant; additionally, the values of 2h CP/2h BG in GDM group were comparable with those in non−GDM group; the peak value of CP occurred at 2h in GDM group, as well as the values of 2h CP/2h BG in high As exposure group were much higher than those in low As exposure group, which suggested that As exposure associated with the increased odds of GDM may be due to that As exposure was related to the change of insulin secretion phase (delayment of the peak of insulin secretion) among pregnant women. In addition, AUCP mediated 11% (p < 0.05) and AUCG mediated 43% (p < 0.05) of the association between As exposure and the odds of GDM. Our results suggested that joint exposure to As, Pb, Tl and Ni during early pregnancy was positively associated with the odds of GDM, As was a major contributor; and the association of environmental As exposure with the increased odds of GDM may be due to that As exposure was related to the impairment of glucose tolerance and change of insulin secretion phase after glucose stimulation (delayment of the peak of insulin secretion) among pregnant women.