Project AIM: Autism intervention meta-analysis for studies of young children

Abstract

In this comprehensive systematic review and meta-analysis of group design studies of nonpharmacological early interventions designed for young children with autism spectrum disorder (ASD), we report summary effects across 7 early intervention types (behavioral, developmental, naturalistic developmental behavioral intervention [NDBI], TEACCH, sensory-based, animal-assisted, and technology-based), and 15 outcome categories indexing core and related ASD symptoms. A total of 1,615 effect sizes were gathered from 130 independent participant samples. A total of 6,240 participants, who ranged in age from 0-8 years, are represented across the studies. We synthesized effects within intervention and outcome type using a robust variance estimation approach to account for the nesting of effect sizes within studies. We also tracked study quality indicators, and report an additional set of summary effect sizes that restrict included studies to those meeting prespecified quality indicators. Finally, we conducted moderator analyses to evaluate whether summary effects across intervention types were larger for proximal as compared with distal effects, and for context-bound as compared to generalized effects. We found that when study quality indicators were not taken into account, significant positive effects were found for behavioral, developmental, and NDBI intervention types. When effect size estimation was limited to studies with randomized controlled trial (RCT) designs, evidence of positive summary effects existed only for developmental and NDBI intervention types. This was also the case when outcomes measured by parent report were excluded. Finally, when effect estimation was limited to RCT designs and to outcomes for which there was no risk of detection bias, no intervention types showed significant effects on any outcome. (PsycINFO Database Record (c) 2019 APA, all rights reserved).

Figure 1.

PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) flow diagram.

Figure 2.

Decision tree (adapted from Yoder et al., 2013) used to code whether a study outcome was proximal or distal to treatment targets.

Figure 3.

Decision tree (adapted from Yoder et al., 2013) used to code whether a study outcome measured a potentially context-bound or more highly generalized characteristic.

Figure 4.

Summary of quality indicator ratings for studies of behavioral, developmental, and naturalistic developmental behavioral intervention (NDBI) types.

Figure 5.

Summary of quality indicator ratings for sensory-based, Treatment and Education of Autistic and related Communication-handicapped Children (TEACCH), and technology-based intervention types.

Figure 6.

Forest plot of robust variance estimation (RVE) summary estimates with small sample bias correction for each outcome by intervention type, when all outcomes from quasi-experimental and RCT group design studies are included. * denotes summary effect size estimates with confidence intervals that do not overlap with zero.

Figure 7.

Forest plot of robust variance estimation (RVE) summary estimates with small sample bias correction for each outcome by intervention type, when all outcomes from RCTs are included. * denotes summary effect size estimates with confidence intervals that do not overlap with zero.

Figure 8.

Forest plot of robust variance estimation (RVE) summary estimates with small sample bias correction for each outcome by intervention type, when only non-caregiver report outcomes from RCTs are included. * denotes summary effect size estimates with confidence intervals that do not overlap with zero.

Figure 9.

Forest plot of robust variance estimation (RVE) summary estimates with small sample bias correction for each outcome by intervention type, when only outcomes from RCTs that are not threatened by detection bias are included. † denotes summary effect size estimates that have p-values < 0.10.