Hazardous Air Pollutants in Relation to Autism Diagnosis and Phenotype in a US Family-Based Study

Background: Determining modifiable risk factors for Autism Spectrum Disorder (ASD) is important given the well-supported theory that autism results from genetic liability combined with exogenous exposures during brain development.  Several studies have reported associations with perinatal exposure to hazardous air pollutants (HAPs), including some metals and volatile organic compounds.

Objectives: To expand the consideration to a larger set of HAPs across the US, and to explore whether prior findings hold in a family-based study design with information on autism phenotype and severity, we examined the association of perinatal HAP exposures with autism in a national study of multiplex families.

Methods:   We determined the geographical location of the residence at birth for 1,873 participants (1,435 with ASD and 438 unaffected siblings) from the Autism Genetics Resource Exchange (AGRE) born 1994-2007 across the US, using self-reported residential history or computerized search of historical addresses. We assigned ambient concentrations of HAPs from an annual census-tract emissions-based model, the National-scale Air Toxics Assessment (NATA) from 1996, 1999, 2002, and 2005.  We included 124 HAPs meeting our exposure variability criteria, linking to participants using census tract and closest birth year.  We included 3 measures of autism:  a broad ASD diagnosis based on report of autism, ASD, or PDD-NOS from the Autism Diagnostic Interview- Revised (ADI-R), a calibrated severity score created from individual items from the Autism Diagnostic Observation Schedule (ADOS) (range 1-10), and the total t-score of the Social Responsiveness Scale (SRS) (mean of 50) among all participants.  For each HAP, we modeled an association across the interquartile range, using each individual’s deviance from the family mean concentration, using hierarchical regression models (SAS hpmixed), including a family term to account for social and family factors and non-independence, family mean HAP concentration and birth year.

Results:   Only a few associations between HAPs and autism measures had 95% confidence intervals (CIs) that indicated statistically significant associations:  Protective associations with autism diagnosis included ethylene dibromide (odds ratio (OR)=0.5, CI: 0.3, 0.9) and ethylene dichloride (OR=0.4, CI: 0.2, 0.9). Elevated associations with higher SRS score included formaldehyde (+5.1 points, CI: 1.1, 9.0) and methylene chloride (+3.0 points, CI: 0.1, 5.8).  Autism severity scores were elevated among children with ASD for chlorobenzene: (+0.3, CI: 0.0, 0.6).  Other associations lacked statistical significance but showed consistent patterns across autism measures.  For example, 1,1,1-trichloroethane (methyl chloroform) associations were elevated for all 3 measures of autism.  Associations with benzene, carbon disulfide, methylene chloride, and propionaldehyde were elevated for the SRS and autism severity score.  Protective associations for ethylene dibromide were found for autism diagnoses and the SRS score.

Conclusions:   In these models, some HAPs showed associations with autism diagnosis and more severe autism phenotype.  Chlorinated solvents were disproportionately represented among the positive associations, consistent with previous studies. Additional statistical approaches to these data, including accounting for the correlation structure among pollutants, are planned.  Although limited in statistical power and by multiple comparisons, this study combining environmental data with genetic repository data adds to the evidence suggesting that some HAPs may be ASD risk factors.