Peaks of Ability Combined with Speech Onset Identify Subgroups within Autism Spectrum

Background: Despite clinical recognition of contrasted phenotypes in autism spectrum, DSM-IV subtypes (e.g. autism vs. Asperger syndrome) are often considered as arbitrary segments, and might disappear in DSM-V. This decision stems from low reliability of the autism vs. Asperger syndrome distinction based on DSM-IV criteria and inconsistent findings about cognitive profiles. We propose that using cognitive profiles in conjunction with targeted behavioral characteristics to stratify the autism spectrum might result in more meaningful subgroups than using DSM-IV criteria.

Objectives: To investigate the association between different cognitive strengths and behavioral characteristics in order to identify subgroups within the autism spectrum.

Methods: All participants aged 6 to 40 years with a positive ADI-R score and Wechsler IQ above 70 were extracted from the Rivière-des-Prairies Hospital database. This procedure yielded a cohort of 146 participants (mean age 16.9 years, IQ 96.6). Performance on each Wechsler subtest was compared to the mean of all Wechsler subtests, using paired samples t tests (Bonferroni corrected), to reveal significant cognitive strengths and weaknesses.  The behavioral characteristics associated with these significant strengths were then investigated based on information collected with the ADI-R.  Data from 47 ADI-R items were used as predictors of performance on the Wechsler subtests in linear regression analyses. Two-steps cluster analyses allowed identifying subgroups among participants based on the cognitive strengths and behavioral characteristics identified.

Results: The Wechsler profile of the autism spectrum sample was consistent with previous descriptions, with significant strength in Block Design and Similarities, and weakness in Comprehension and Coding. The presence of several autism characteristics predicted higher performance in Block Design, among which delay in age of first single words or phrases, reduced social smiling, poor response to approaches of other children, pronominal reversal, repetitive use of objects/interest in parts of objects, and unexpected computational, drawing or visuospatial abilities. On the contrary, the absence of several autism characteristics (e.g. behaviours similar to typically developing children) predicted higher performance in Similarities :  no delay in first single words or phrases, interest in other children, spontaneous offering to share with others, and typical use of objects. Cluster analyses (with Block Design performance, Similarities performance and age of first phrases) identified a subgroup of participants (49% of sample) who had a strength in Block Design and developed 2 word phrases later (mean 47 months), and a subgroup (51% of sample) with higher Similarities, no strength in Block Design and who developed 2 word phrases earlier (mean 24 months). Going back to the clinical files of these participants, 97% of the participants in the first subgroup had a best estimate clinical diagnosis of autism, and 64% of the participants in the second subgroup had a diagnosis of Asperger syndrome.

Conclusions: Specific behaviors and cognitive strengths (e.g. visuospatial vs. verbal) aggregate within some individuals on the autism spectrum. It allows identifying meaningful subgroups within the spectrum and offers a more operational, objective definition of autism spectrum subgroups. Using cognitive strengths and specific behaviors could strengthen both the diagnosis and the development of targeted education methods.