Autism Spectrum Disorders: A Dimension or Sub-Categories?

Background: Our group has recently found that many cases of autism spectrum disorders (ASDs) represent a category, qualitatively distinct from typical (non-autism) behavior in clinically ascertained samples (Frazier et. al., in press). The next question concerns whether a broad autism category is composed of sub-categories or is best characterized as a dimension of symptom severity when only autism-affected youth are examined. A recent study addressing this question identified unique sub-categories based upon dysmorphology/head circumference, social communication, and verbal/non-verbal ability scores (Ingram et al., 2008). However, interpretation of these findings is complicated by the aggregation of two different samples, a design feature which may bias toward category identification.

Objectives: The present study examined whether autism symptoms would identify sub-categories or a dimension of symptom severity, consistent with the notion of an autism spectrum. This distinction is relevant to future DSM nosology, screening and diagnosis, genetic and neurobiological study design, and identification of differential treatment effects.

Methods: Data were obtained from the Interactive Autism Network (IAN) and Autism Genetic Resource Exchange (AGRE) samples and analyzed separately to determine whether results replicate across samples and indicator sets. IAN preferentially recruits families with at least one affected child who has been diagnosed with an ASD. In the IAN sample, caregivers reported autism symptoms using the Social Communication Questionnaire (SCQ) and the Social Responsiveness Scale (SRS). In the AGRE sample, parents were interviewed using the Autism Diagnostic Interview – Revised (ADI-R). Autism symptom indicator sets were derived from each measure in each sample. Taxometric and latent variable models evaluated whether 1-group (dimensional) or 2-group (categorical) models fit the data better across indicator sets, demographic sub-samples, and IAN/AGRE samples. These models are “blind” to diagnostic status and thus provide an empirical test of whether sub-categories or a dimension best describe the data.

Results: In the IAN sample, 6875 and 2575 autism-affected participants had SCQ and SRS data, respectively. In the AGRE sample, 889 autism-affected individuals had ADI-R data. Results indicated that dimensional models fit the data better than categorical models. This was true across all taxometric and latent variable procedures, indicator sets, demographic sub-samples, and IAN/AGRE samples.

Conclusions: Previously, our group found that ASDs are best conceptualized as a category distinct from typical behavior. The present findings suggest that this broad category includes a continuum of symptom severity. Together, these findings suggest that ASDs may be conceptualized as a single, discrete entity that is distinct from typical behavior but that shows large variation in symptom severity. This conclusion does not rule out the possibility of autism sub-groups. However, the results imply that indicators beyond autism symptoms, such as cognitive or biological indicators, will be needed to identify autism sub-groupings.