Meta-analysis of nonpharmacological interventions for young autistic children finds limited moderation by cognitive or language scores

BACKGROUND: Young autistic children have a range of language and cognitive abilities and, as a result, may differentially benefit from interventions supporting skills in these and related domains. Although studies have previously examined the extent to which participant characteristics interact with intervention effects, they have primarily restricted the analyses to a single intervention approach. METHOD: In the present study, we drew on data from a comprehensive meta-analysis of group design, nonpharmacological intervention studies for young autistic children to test these effects. Specifically, we conducted a secondary meta-regression analysis to examine whether cognitive and language standard scores and age equivalents at study entry significantly moderated intervention effects across intervention type on adaptive, cognitive, language, and social communication outcomes and separately across outcome type for behavioral, developmental, naturalistic developmental behavioral interventions and technology-based interventions. Cognitive and language ability was quantified using reported or estimated standard scores, and cognitive and language level was quantified using reported or estimated age-equivalent scores. Analyses within outcome type were conducted using a data set of 1,911 effect sizes from 202 independent samples, and analyses within intervention type were conducted using a data set of 2,137 effect sizes from 144 independent samples. RESULTS: Few studies reported standard scores and/or age equivalents for participant language. None of the putative moderators significantly predicted intervention effects by outcome domain (i.e., adaptive, cognitive, language, and social communication). Both cognitive standard and age-equivalent scores positively and significantly predicted effects of technology-based interventions exclusively, but we did not find robust evidence that language standard or age-equivalent scores significantly predicted effects by intervention type. CONCLUSIONS: These findings are exploratory and warrant cautious interpretation. Future intervention researchers should extensively characterize participant samples in terms of their language and cognitive ability to aid meta-analytic investigation. The field would benefit from additional high-quality randomized controlled trials testing whether intervention effects vary by participant characteristics, using preplanned moderator analyses, valid measures, and large representative samples. SUPPLEMENTAL MATERIAL: https://doi.org/10.23641/asha.31967844.