Testing the Boundaries of Autism Spectrum Disorder in DSM-5

Background: Diagnostic criteria for autism will be radically revised in the fifth edition of the DSM, published in 2013. Changes include the creation of a single unitary category of autism spectrum disorder; the replacement of the autism triad with an autism dyad; the removal of the Asperger’s disorder and pervasive developmental disorder – not otherwise specified (PDD-NOS) diagnostic sub-categories; the inclusion of sensory abnormalities as a core ASD symptom; and the removal of language delay from diagnostic criteria. Empirical evidence already exists that these modifications will improve the reliability and validity of ASD. However, concerns have arisen that the threshold for receiving a diagnosis on the autism spectrum in DSM-5 is set too high; several studies have suggesting that DSM-5 ASD criteria will exclude around 40% of those currently meeting DSM-IV-TR criteria; and that sensitivity will be especially poor for high-functioning individuals. Proponents of DSM-5 have countered these claims, arguing that such research is flawed, as it used data that were collected to capture DSM-IV and DSM-III defined symptoms, and so failed to properly implement DSM-5 diagnostic criteria.

Objectives: Using detailed and comprehensive phenotyping, we aimed to accurately implement DSM-5 ASD criteria to test whether they exclude from the autism spectrum individuals with a DSM-IV Asperger’s, autism or PDD-NOS diagnosis.

Methods: Participants were 726 verbally able (mean verbal IQ=91) children and young people (mean age=9.5 years) who had been referred to a clinic for the assessment of social communication difficulties. All either met DSM-IV-TR criteria for PDD, or had significant, sub-diagnostic autistic traits characteristic of the broad autism phenotype. Symptoms were assessed by parent report using the Dimensional, Diagnostic and Developmental Interview (3Di).This interview contains over 200 questions capturing the full range of autistic phenomena, well beyond those described in DSM-IV-TR criteria. 140 of these questions were combined to create a DSM-5 algorithm, which mapped the full range of symptoms described in DSM-5 ASD criteria.

Results: Of the children with a DSM-IV-TR PDD diagnosis, 86% met criteria for DSM-5 ASD. Of the 231 children without a DSM-IV-TR PDD diagnosis, 61 (26%) met criteria for DSM-5 ASD. Most cases of DSM-IV-TR autism (175 of 185; 95%) and Asperger’s disorder (137 of 143; 96%) met DSM-5 criteria for ASD. Of the PDD-NOS cases 69% (115 of 167) met DSM-5 criteria for ASD. Of the 52 people with PDD-NOS who did not reach thresholds for ASD, 37 had insufficient repetitive behaviours for a DSM-5 diagnosis. Neither IQ nor age was associated with the risk of not meeting criteria for ASD.

Conclusions: Our findings contradict studies that have suggested severely reduced sensitivity of DSM-5 ASD criteria in higher-functioning individuals. We attribute this to our ability to use detailed phenotyping to accurately implement the full range of ASD symptoms described in DSM-5. Nevertheless, we identified a group of individuals who may miss an ASD diagnosis, due to insufficient repetitive and stereotyped behaviour. We anticipate that their profile of difficulties will be captured by ‘Social Communication Disorder’, a new category that will appear in DSM-5.