Estimating the Joint Effect of Endocrine Disrupting Chemical Mixtures on Fetal Growth Using Bayesian Hierarchical Linear Regression Modeling: The HOME Study

Background: Pregnant women are exposed to endocrine disrupting chemicals (EDCs) through prolonged contact with personal care products, food containers and plastics. Gestational exposure to EDCs may be associated with changes in birth weight and fetal growth, placing children at risk for negative outcomes later in life. Objective: To examine the association between gestational exposure to four EDC classes (phthalates, polychlorinated biphenyls (PCBs), polyfluorinated compounds (PFCs) and polybrominated diphenyl ethers (PBDEs)) and infant birth weight. Methods: Using data from the Health Outcomes and Measures of Environment (HOME) study, 2003-2006 (n=387), we used Bayesian Hierarchical Linear Regression Modeling (BHLM) to examine the associations between four EDC classes and birth weight. Results: Colinearity among the chemical concentrations was severe with correlations as large as 0.98. In BHLM analyses that shrunk the beta coefficients towards the class-specific mean. We found that no individual metabolites from any of the four EDC classes were associated with birth weight. In the BHLM, the parameter μβ was the average beta coefficient within a chemical class (e.g. phthalates) in relation to birth weight. For a 10-fold increase in chemical concentration, the μβ values were respectively, 0.89g for phthalates, -1.48g for PCBs, -15.8g for PFCs and -2.14 for PBDEs. In order to put these effect sizes in context, we examined the relative impact of chemical exposure on birth weight versus the effect of poverty on birth weight, based on a Bayesian analysis of four socioeconomic (SES) indicators: low income, low education, no private health insurance, and unemployment. For SES, the estimated value of μβ was -63.1g. Thus poverty had a larger association with lower birth weight than gestational chemical exposure as measured by a multiplicative factor of x121 for phthalates, x193 for PCBs, x36.1 for PFCs, and x157 for PBDEs. Conclusion: Gestational exposure to phthalates, PCBs, PFCs or PBDE had only small associations with differences in birth weight among the HOME study participants. In particular, the average association between gestational EDC exposure and birth weight is approximately 36.1 to 193 times smaller than the effect of poverty on birth weight. While no μβ values attained significance, three were negative. PFCs and phthalates have the greatest associations with birth weight (one negative, one positive). The phthalate and PFC EDC classes, taken within the context of SES, should be considered for closer study.