Childhood exposures to environmental chemicals and neurodevelopmental outcomes in congenital heart disease

Abstract

Background: Children with congenital heart defects have an increased risk of neurodevelopmental disability. The impact of environmental chemical exposures during daily life on neurodevelopmental outcomes in toddlers with congenital heart defects is unknown.

Methods: This prospective study investigated the impacts of early childhood exposure to mixtures of environmental chemicals on neurodevelopmental outcomes after cardiac surgery. Outcomes were assessed at 18 months of age using The Bayley Scales of Infant and Toddler Development-III. Urinary concentrations of exposure biomarkers of pesticides, phenols, parabens, and phthalates, and blood levels of lead, mercury, and nicotine were measured at the same time point. Bayesian profile regression and weighted quantile sum regression were utilized to assess associations between mixtures of biomarkers and neurodevelopmental scores.

Results: One-hundred and forty infants were enrolled, and 110 (79%) returned at 18 months of age. Six biomarker exposure clusters were identified from the Bayesian profile regression analysis; and the pattern was driven by 15 of the 30 biomarkers, most notably 13 phthalate biomarkers. Children in the highest exposure cluster had significantly lower adjusted language scores by -9.41 points (95%CI: -17.2, -1.7) and adjusted motor scores by -4.9 points (-9.5, -0.4) compared to the lowest exposure. Weighted quantile sum regression modeling for the overall exposure-response relationship showed a significantly lower adjusted motor score (β = -2.8 points [2.5th and 97.5th percentile: -6.0, -0.6]). The weighted quantile sum regression index weights for several phthalates, one paraben, and one phenol suggest their relevance for poorer neurodevelopmental outcomes.

Conclusions: Like other children, infants with congenital heart defects are exposed to complex mixtures of environmental chemicals in daily life. Higher exposure biomarker concentrations were associated with significantly worse performance for language and motor skills in this population.

Fig 1. Study enrollment and evaluation flowchart.

Fig 2

A) Exposure Biomarkers by Clusters: Heatmap to visualize the median (most typical) tertile of exposure that each chemical falls within for each cluster. Darker red colors indicate tertile 3 (highest exposure group), orange colors indicate tertile 2 (medium exposure group), and light orange indicate tertile 1 (lowest exposure group). B) Unadjusted Neurodevelopmental Scores by Cluster: Box and whisker plots of neurodevelopmental scores (unadjusted) for each cluster. Black lines within each boxplot represent the median value and the upper and lower ends of the boxes represent the 75th and 25th percentiles, respectively. The horizontal red line on the graph indicates 1 standard deviation (85) below population mean (100).

Fig 3. Assessment scores from linear regression models.

Crude and adjusted linear regression models for the three outcomes of interest and the clusters from the profile regression. Cognitive Score Model adjusted for genetic anomaly, gestational age, and raw Hollingshead SES score. Language Score Model adjusted for genetic anomaly, raw Hollingshead SES score, and sex. Motor Score Model adjusted for ethnicity, genetic anomaly, sex, and hospital length of stay at first operation. Cluster 2 is the lowest exposure cluster and is set as the reference group for each model.

Fig 4. Weighted quantile sum (WQS) regression effect estimate.

The WQS modeling results for selected neurodevelopmental outcomes displaying the mean and lower 2.5^th and upper 97.5^th percentiles for the WQS index coefficients obtained from the repeated holdout validation.

Fig 5. Weighted quantile sum regression.

Index weights for WQS index exposure-responses that were significantly associated with ND outcomes. (A) Language score index weights for girls; (B) Motor score index weights for entire cohort; (C) Motor score index weights for girls.