Statistics from Altmetric.com
Tertiary psychiatric hospital outpatient clinics and locally resident volunteers.
Adults, aged 18–68 years, consisting of 20 controls without autism, 20 subjects with autism and 19 subjects with attention-deficit hyperactivity disorder (ADHD). All subjects were male, IQ>75, right handed and without major psychiatric illness or medical condition affecting brain function.
Reference testing/standardisation: Participants underwent a psychiatric interview, physical examination and blood tests to exclude other disorders (eg, fragile X). Autism was diagnosed by ICD-10 (International Statistical Classification of Diseases and Related Health Problems 10th Revision) criteria and confirmed using the Autism Diagnostic Interview–Revised (17 cases) or the Autism Diagnostic Observation Schedule (3 cases). Patients with ADHD were diagnosed according to the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV criteria, http://allpsych.com/disorders/dsm.html).
All patients had an MRI scan of their brain, which was inspected initially by a radiologist for image quality and the existence of clinical abnormalities. Brain morphology parameters were inspected using software to assess if specific prespecified volumetric and geometric features of the brain differed between controls/patients with autism, and used to derive a discriminatory algorithm. The selected features were then tested by applying the rules to the 19 patients with ADHD.
The initially derived brain morphology rule showed good sensitivity and specificity in the training case/control setting using scans of the left hemisphere: sensitivity 90%, specificity 80%. When applied to the 19 patients with ADHD, the left hemisphere algorithm incorrectly identified 4 (21%) of individuals as having autism.
Right hemisphere values achieved lower accuracy in the training setting (sensitivity 60%, specificity 70%) and testing setting (47% incorrectly identified as autistic).
There may be subtle but definable differences in the volumetric and morphological parameters of brain images from adults with and without a clear diagnosis of autism, but the discriminatory algorithm requires further evaluation.
This widely reported study was the subject of numerous media reports and hailed as the ‘test that can diagnose autism in 15 min’.1 It follows previous papers that have described isolated differences in cerebral morphology in patients with autism. Here, a composite of five morphometric parameters was used simultaneously to classify control, ADHD and patients with autistic spectrum disorder (ASD).
The study focused on 20 adults with autism without comorbid conditions. This limits its usefulness in children. The key questions for us are as follows:
■ What happens if we try to replicate the findings in subjects whose brains are still maturing?
■ How does the test perform in children across the full spectrum of ASD severity and probability?
■ Can it differentiate between the ‘normal’, ‘the autistic’, ‘the hyperactive’ and the ‘very-large-caseload-of-children-referred-to-me-with-behavioural-problems-who-may-or-may-not-have-autism’. A short answer to this last question is, no: At least one-fifth of ADHD controls here were classified by the test as having ASD – an unacceptably high false-positive rate.
Additionally, a proper diagnostic study with, for example, adequate assessor blinding in an appropriate spectrum of patients is required. Until then, this technique is experimental.
There are other problems. With these initial estimates of test sensitivity and specificity, the test is not useful in situations where the odds of a diagnosis were low to begin with. For example, assuming a background prevalence for autism of 37.4% in a population with a statement of special educational needs2, a positive test result still only means 72% of those children had autism.
The combined problems of high patient referral rates, costs of neuroimaging, need for sedation for neuroimaging and potential side effects of general anaesthesia make this investigation even less feasible.
Current evidence does not support the use of this investigation in the paediatrician's armamentarium for the diagnosis of ASD. Further studies will need to focus on its diagnostic capability across the ASD spectrum and its ability to discriminate ASD from other neurodevelopmental disorders.
Sources of funding: Medical Research Council, National Institute for Health Research, Wellcome Trust.
Provenance and peer review Commissioned; internally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.