Evaluating associations between early pregnancy trace elements mixture and 2nd trimester gestational glucose levels: A comparison of three statistical approaches

Abstract

Objective: Studies have shown that individual trace element levels might be associated with abnormal glycemic status, with implications for diabetes. Few studies have considered these trace elements as a mixture and their impact on gestational glucose levels. Comparing three statistical approaches, we assessed the associations between essential trace elements mixture and gestational glucose levels.

Methods: We used data from 1720 women enrolled in the Eunice Kennedy Shriver National Institute of Child Health and Human Development's Fetal Growth Study, for whom trace element concentrations (zinc, selenium, copper, molybdenum) were measured by inductively coupled plasma mass spectrometry (ICP-MS) using plasma collected during the 1st trimester. Non-fasting glucose levels were measured during the gestational diabetes mellitus (GDM) screening test in the 2nd trimester. We applied (1) Bayesian Kernel Machine Regression (BKMR); (2) adaptive Least Absolute Shrinkage and Selection Operator (LASSO) in a mutually adjusted linear regression model; and (3) generalized additive models (GAMs) to evaluate the joint associations between trace elements mixture and glucose levels adjusting for potential confounders.

Results: Using BKMR, we observed a mean 2.7 mg/dL higher glucose level for each interquartile increase of plasma copper (95% credible interval: 0.9, 4.5). The positive association between plasma copper and glucose levels was more pronounced at higher quartiles of zinc. Similar associations were detected using adaptive LASSO and GAM. In addition, results from adaptive LASSO and GAM suggested a super-additive interaction between molybdenum and selenium (both p-values = 0.04).

Conclusion: Employing different statistical methods, we found consistent evidence of higher gestational glucose levels associated with higher copper and potential synergism between zinc and copper on glucose levels.

Keywords: Bayesian kernel machine regression; Environmental mixtures; Generalized additive model; Gestational glucose levels; LASSO.