Sex-specific associations between prenatal exposure to phthalates and neurocognitive development in Taiwanese children: Application of environmental mixture methods to small study sample

Abstract

BACKGROUND AND AIM: Few studies have examined the sex-specific effect of phthalates as a mixture by applying multiple mixture methods to a small study sample. This study, therefore, examined sex-specific associations of prenatal exposure to urinary phthalates with neurocognitive development in Taiwanese children aged two to three years old. METHODS: The study included 127 children who were followed up at the ages two-three years in the Taiwanese Maternal and Infant Cohort Study. The Mental Development Index (MDI) and Psychomotor Development Index (MDI) were measured in children using the Bayley Scales of Infant Development-II (BSID-II). The monomethyl phthalate, monoethyl phthalate, mono-butyl phthalate (MBP), mono-benzyl phthalate, mono-2-ethylhexyl phthalate (MEHP), mono(2-ethyl-5-hydroxyhexyl) phthalate, mono (2-ethyl-5-oxohexyl) phthalate (MEOHP) were measured in urine samples collected from mothers during pregnancy. Multivariable linear regression (MLR), Weighted Quantile Sum (WQS) regression, quantile g-computation (qgcom), Bayesian Kernel Machine Regression (BKMR), the Super Learner with g-computation were applied and compared to examine the sex-specific effects of the phthalate mixture on MDI and PDI. RESULTS: The mediation MDI and PDI were 94.0 and 99.0, respectively. The spearman correlation ranged between -0.01 and 0.74. None of these methods could detect any sex-specific associations between phthalates and MDI. Results suggest that WQS and qgcomp were able to detect inverse associations between phthalates and PDI with a stronger effect in boys. BKMR and Super Learner with g-computation were unable to detect any precise associations of phthalates with PDI. However, mixture methods detected mixtures' negative directionality and linear dose-response effects on neurocognitive development. MBP, MEHP, and MEOHP dominated these associations with PDI. CONCLUSIONS: Mixture methods outperformed MLR. However, flexible approaches like BKMR and Super Learner with g-computation may perform better with larger samples. Therefore, the study recommends using multiple mixtures methods for getting comprehensive results from mixture studies. KEYWORDS: Mixture; Neurodevelopment; WQS; BKMR; G-computation