The Utility of Traditional Epidemiologic Designs in ASD Research

Background: Autism epidemiology can be informed through many design choices.  For example, registry-based data have very large samples sizes and increased precision. However, they often lack detailed information about presentation, subtypes, exposures, and biosamples.  More traditional case-control and cohort designs sacrifice precision given the typically much smaller sample sizes, but can capture deep phenotype and biomarker information.

Objectives: To describe the advantages and disadvantages of traditional case-control and cohort designs that include direct enrollment and participation, and to show the utility of such designs in ASD research.

Methods: We work primarily with three epidemiology studies of autism: a national population-based case-control study, an enriched risk pregnancy cohort, and a hospital system-based birth cohort.  The Study to Explore Early Development (SEED) is a 6-site case-control study that recruits ASD cases, non-ASD developmentally disabled children (DD), and typical children (POP) aged 3-5 years via medical and education service providers (ASD and DD), and vital records (POP). The Early Autism Longitudinal Risk Longitudinal Investigation (EARLI) recruits pregnant women from 4 sites who already have a child with ASD and follow them through the pregnancy and the baby’s first 3 years.  The Boston Birth Cohort (BBC) recruits women 1-3 days post-delivery and follows the babies through childhood via direct contacts at routine clinic visits and via electronic medical records.

Results: SEED includes extensive interview-based information on peri-natal risk factors, maternal and child medical records, extensive clinical assessment, and biosamples of children and parents, enabling examination of environmental, genetic, epigenetic, and gene-environment hypotheses. While the case-control design allows relatively large sample sizes, it precludes biosample and exposure assessment in the perinatal risk period and is subject to recall bias.  The EARLI study collects home environmental samples, parental, ASD case, and new sibling biosamples longitudinally beginning in pregnancy, extensive prospective questionnaire/interview data, and developmental phenotypes from 6 – 36 months.  This allows very detailed analysis of exposures, genetics, and epigenetics at relevant time windows in different tissue types, as well as examination of ASD and related quantitative and longitudinal phenotypes.  It is necessarily enriched risk to enable enough accumulated cases for study, but nevertheless is limited in total sample size. Finally, the BBC collects birth biosamples on mothers and babies, as well as later childhood biosamples, in addition to access to complete medical record information combined with prospective questionnaire data and neurodevelopmental screening.  It has the advantage of large sample size and nearly exhaustive medical record access.

Conclusions: Studies that involve direct contact offer advantages to registry and HMO-based research that passively use data collected for other purposes.  These include deep phenotyping data, biomarker availability, and a diversity of biosamples from relevant time windows.  However, such studies are limited by design challenges and sample size. All design choices inform ASD epidemiology, services, and policy research in complementary ways and cross-collaboration between them can help to maximize that complement.