We re-evaluated, using dimensional measures, the conventional ICD-10 triadic model of autistic symptomatology in a large clinically heterogeneous population of children with and without autism spectrum disorder (ASD). Conventionally defined elements of the autistic symptom triad were measured in 816 individuals (462 with ASD, 59 with DSM-5 defined Social Communication Disorder and 295 normal or miscellaneous clinical controls). Data output emulated those generated using the ADI-R algorithm. Measured dimensionally, social reciprocity and communication deficits were highly correlated and these symptom scores were therefore combined to create a new variable, namely social-communication impairment which was only moderately correlated with the restricted interests/stereotypy dimension. Multinomial logistic regression revealed a bidimensional model of symptom severity provided the most parsimonious fit to the data and predicted group membership as well as one based on traditional triadic symptomatology. A unidimensional model of autistic symptomatology was comparatively unsatisfactory, failing to predict any cases of Asperger syndrome or pragmatic language disorder. Our findings support the DSM-5 conceptualisation of ASD. Nevertheless, we caution that the boundaries of ASD remain ill-defined and potentially arbitrary.
Keywords: autism spectrum disorder, autism phenotype, bidimensional model