Association between the co-exposure to endocrine disrupting chemicals and birth outcomes in Brazilian newborns: a cross-sectional study using elastic net and Bayesian kernel machine regression

Abstract

Exposure to environmental contaminants, such as endocrine disruptors (EDCs) and toxic metals, might have serious consequences for the fetus. Although the influence of these pollutants in causing adverse neonatal events is known, studies have focused on limited statistical techniques that limit the understanding of complex relationship between exposure and outcome. Thus, the application of advanced methodologies such as the elastic net and the Bayesian Kernel Machine Regression (BKMR) may be used. This study aims to investigate the effect of the co-exposure to six classes of EDCs (bisphenols, parabens, benzophenones, polycyclic aromatic hydrocarbons, polychlorinated phenyls and metals/metalloids) on neonatal outcomes using elastic net and BKMR. This proposal will be divided into two studies. In study I, biomonitoring of endocrine disruptors will be carried out in Brazilian pregnant women. For this, 300 healthy Brazilian pregnant women aged 18 to 44 years were recruited for a cross-sectional study. The determination of urinary levels of 15 EDCs will be done by liquid chromatography and sequential mass spectrometry. The concentrations of 16 metals/metalloids (lead, cadmium, arsenic, mercury, nickel, copper, zinc and manganese, cobalt, silver, cesium, iron, calcium, magnesium, sodium and potassium) will be determined by inductively coupled plasma mass spectrometer. Descriptive statistics for creatinine-adjusted urinary levels will be calculated for all contaminants and stratified by sociodemographic characteristics and lifestyle. In study II, the association between co-exposure to endocrine disruptors and neonatal outcomes in Brazilian newborns using elastic mesh and BKMR will be investigated. For the elastic network analysis, the hyper parameters (alpha and lambda) will be estimated using cross validation. BKMR will be used to investigate possible interactions and model the response function of outcomes in relation to co-exposure to contaminants. The selection of variables will be based on the posterior inclusion probabilities. With the application of these sophisticated statistical models, it is expected to identify complex relationships between exposure to environmental contaminants and neonatal outcomes. In addition, it is expected that this work can guide future public policies and other investigations evaluating exposure to multipollutants.