Integration of Environmental Variables into Autism Genetic Repository Data

Background:   Genetic repositories for research on autism spectrum disorders (ASD) often contain data on well-phenotyped individuals and have biospecimen resources, yet little information on environmental risk factors. Having detailed data in all these domains would allow the investigation of main effects of environmental exposures or gene-environment interactions. 

Objectives:  We assessed the ability to determine place of residence as an approach to characterizing environmental exposures, specifically evaluating two methods for gathering residential addresses for participants in the Autism Genetics Resource Exchange (AGRE): 1) Re-contact using a self-administered web-based questionnaire, and 2) Use of a publicly-available computerized address search engine. In this study, we focused on prenatal and early life exposures to air pollutants for children in AGRE.

Methods:  AGRE families with at least one child born 1994-2007 were eligible for our study.  Since AGRE is composed of children with ASD and their siblings, parents/guardians were contacted via email or postal letter and asked to complete a residential history online.  Separately, study contact (parent/guardian) information was used to conduct a computerized search of up to 10 historical addresses using LexisNexis.  All collected addresses were filtered based on dates of residences to identify those corresponding to the nine months before birth or two years after birth for eligible children.  Addresses were geo-coded and air quality data was assigned using the EPA’s AIRNOW monitoring network and roadway distance assigned using roadway geometry.   

Results:  A total of 1,373 families (4,508 children) from AGRE were eligible for our study.  All families were included in both address-finding methods.  The online residential history form was completed for 716 children from 305 families, reflecting only a 22% family participation rate.  LexisNexis search results were returned for 96% of families, providing at least one historical address for 3,119 children (69%).  We combined these two ascertainment methods to compile the most complete resource possible, and filtered for addresses pertaining to the exposure time window of interest.  This resulted in assignment of at least one residential address for 2,550 children (57%) during the prenatal and two-years postnatal period.  Of these children, roadway distance was assigned for 2,343 (92%).  Air quality assignments based on EPA monitor data were made for 43-83% of residential locations, resulting in sample sizes up to 1,948 children (in the case of Particulate Matter) with both air pollutant and roadway distance data for future analyses. 

Conclusions:   The study of geographically determined environmental exposures is possible for biorepository participants using ancillary resources, but requires access to personally identifiable data to allow the most complete address tracing.  The rich genetic, phenotypic, and family information, with new environmental data on hundreds of children, comprise a valuable resource for future studies, although careful analysis of subject non-participation or non-inclusion is needed.