Utility of the Autism Diagnostic Observation Schedule and the Brief Observation of Social and Communication Change for Measuring Outcomes for a Parent-Mediated Early Autism Intervention.

The ADOS-BOSCC total intervention analysis was preregistered, but all other analyses were exploratory and post-hoc, and the authors state that multiple testing makes them underpowered for robust interpretation. The Module 1 comparison of measures used 104 children with complete baseline and endpoint data, not the full 146 children retained in the trial. The supplied text does not describe how children were randomly allocated. ADOS assessments were coded blind to treatment group but not to timepoint, which may introduce bias. The trial used the same ADOS module at baseline and endpoint, so some endpoint assessments were not optimally aligned with the child's language level. Some ADOS-BOSCC items had poor reliability, especially eye contact and mannerisms, partly because the videos were 10-year-old low-definition recordings. The RRB subdomain had lower reliability and infrequent behaviours, so RRB findings should be interpreted cautiously. The significant ADOS effect in children with 'few to no words' was small-sample, not pre-specified, and not significantly different from the 'some words' group. Analyses that used change-score variance suggested a more modest ADOS effect.

The supplied text lists funding from the Efficacy and Mechanism Evaluation Programme, National Institute for Health Research, National Institutes of Health, Department of Health, South London and Maudsley NHS Foundation Trust, Biomedical Research Centre, King's College London, and the Medical Research Council. It does not report author conflicts of interest.

Do not treat the significant ADOS algorithm effect in the combined sample as the main result. The preregistered ADOS-BOSCC total effect was not significant, and the ADOS algorithm result is exploratory, post-hoc, and partly driven by Module 2 children.