Developmental Trajectories of Response to Treatment for Toddlers with Autism: Findings from the Early Social Interaction Project

Background: Early intervention for toddlers with ASD is critical, given increasing rates of early diagnosis and the potential for treatment to impact brain development. Early intervention programs have resulted in gains in cognitive, language, and/or social skills at the group level (Dawson et al, 2010; Kasari et al., 2008, 2010). However, these effects are often modest and variable across children, suggesting different patterns of response to treatments. A randomized controlled trial (RCT) of the Early Social Interaction (ESI) Project demonstrated group-level efficacy across measures of language, social communication, and autism symptoms (Wetherby et al., 2014), but variability in degree of response was also observed. Despite promising findings at the group level, existing studies have not yet determined which children respond best to intervention.

Objectives: To identify patterns of response to intervention in children with ASD enrolled in ESI, using latent class growth analyses (LCGA).

Methods: 82 children diagnosed with ASD were enrolled in the ESI Project at 16-20 months of age. Children were randomly assigned to receive nine months of 1) individual-ESI offered in 2-3 weekly sessions to teach parents to support their child’s social communication development or 2) group-ESI offered weekly. Children completed the ADOS and CSBS at baseline, bimonthly, and at the end of treatment, yielding 5-6 time points.

Results: Four LCGA models were conducted for CSBS Social, Speech, and Symbolic composites, and ADOS Social Affect severity scores. Three-trajectory-class solutions provided the best fit for the CSBS Social, Speech, and Symbolic models. For the Speech analyses, Class1 showed the highest scores at baseline and the highest rate of improvement, Class2 showed low baseline scores and significant growth during intervention, and Class3 also showed low baseline scores, but no change during treatment. For the CSBS Social and Symbolic analyses, Class1 showed the highest baseline scores with the highest rate of improvement, Class2 showed moderate baseline scores with significant but lower improvement than Class1, and Class3 had the lowest scores at baseline and no change during treatment. The two-trajectory-class solution provided the best fit for ADOS Social Affect scores: Class1 showed moderate social impairment and significant improvement while Class2 showed severe impairment and no change during treatment. Children in individual-ESI and group-ESI were relatively evenly distributed throughout trajectory classes. Significant predictors of treatment trajectory class membership validate these distinct patterns of response to treatment. 

Conclusions: Findings support that group-level analyses, while important for RCTs documenting efficacy, can obscure individual-level variability in response to treatment. Significant variability in baseline skills and rate of improvement were observed and significantly related, with trajectory classes with the highest baseline scores showing the highest rate of improvement. Children in individual-ESI and group-ESI were represented across all trajectory classes, demonstrating that substantial variability in response exists across treatment type/intensity. LCGAs may help researchers identify which children respond best to which treatments, as well as those children who are not showing optimal response to treatment. This may ultimately improve our ability to identify which treatments work for which children and tailor treatments based on early pattern/rate of change.