Prenatal PBDE Exposure and Early ASD-Related Phenotype in a Risk-Enriched Pregnancy Cohort

Background:  Polybrominated diphenyl ethers (PBDE) are persistent organic pollutants (used as flame retardants in electronics, textiles, and furniture) where prenatal exposure has been linked to adverse neurodevelopmental outcomes in toxicologic studies.  One previous epidemiologic study found no association of ASD diagnosis with PBDE concentrations measured in children after the time of diagnosis; the other found a non-significant trend towards greater autistic behaviors assessed by the SRS (Social Responsiveness Scale) in 4-5 year olds with higher prenatal serum levels of BDE-28, but no association with other PBDE congeners.  

Objectives:  To investigate the association between prenatal PBDE exposure and early ASD-related phenotype in an ASD high-risk pregnancy cohort (the EARLI cohort – comprised of mothers of a child with ASD followed from the start of a subsequent pregnancy).  Specifically, we assessed whether elevated prenatal levels of PBDE congeners -28, -47, -99, -100, and -153 were associated with elevated scores on the Autism Observation Scale for Infants (AOSI) at 12 months.  (In preliminary analyses from a cohort subsample, 12-month AOSI was found to be predictive of 36-month best estimate clinical diagnosis – ROC AUC=87%, AOSI ≥7 Sn=70% Sp=90%.)

Methods:  PBDE congener concentrations were measured in the earliest available prenatal plasma samples for 172 mothers.   The association between each PBDE congener and AOSI score, alternatively parameterized as continuous (ln(total AOSI+1)) and dichotomous (total AOSI ≥7) outcomes, was estimated using regression approaches with adjustment for potential confounders.  PBDE exposure was categorized into quartiles (the exception being BDE-28, where 48% of the samples had no detectable level and were assigned to the lowest exposure category with the remaining subjects categorized in ascending tertiles).  A single summary of all five congener concentrations was also created and categorized into quartiles.  Confounders considered included: maternal age, race, ethnicity, pre-pregnancy BMI, parity, gestational age at birth, and geographic location (California vs. east coast study sites).

Results:  Median total AOSI score was 4 (IQR 2,8).  There were 54 (31%) children with total AOSI score ≥7. Direction of rank correlation between AOSI and PBDE congeners were positive for all except BDE-153, but magnitudes were consistently small and none were statistically significant (-0.04 ≤ r ≤0.13 with p-values 0.09 ≤ r ≤0.57).  The figure shows unadjusted ln(AOSI+1) regression results contrasting higher exposure groups to the lowest exposure category. Trend is seen for some congeners but the only unadjusted contrast that approached statistical significance was for the highest BDE-28 exposure category (β= 0.28; 95% CI -0.02, 0.58; p=0.07).  Adjustment for covariates decreased the magnitude of associations and increased p-values (e.g., for the highest BDE-28 category adjusted β= 0.21; 95% CI -0.13, 0.55; p=0.22 ). Findings from models of AOSI ≥7 yielded similar results.

Conclusions:  Findings from this analysis do not support an association between prenatal PBDE exposure and early ASD-related phenotype.  Correlations were strong among PBDE congeners, the exception being BDE-153, making it difficult to consider independent effects of individual congeners.