Randomized Trial of Early Intervention for Spoken Communication in Autism Spectrum Disorder

Background: Applied behavior analysis interventions, including at least some discrete trial teaching (DTT), are often considered the standard of care for preschoolers with autism spectrum disorder (ASD) and limited language. DTT is a highly structured, adult-led teaching format that focuses on directly teaching sounds and words. A second category of approaches apply naturalistic, developmental behavioral interventions (NDBIs). These approaches focus on increasing social engagement by providing learning opportunities in the context of back-and-forth interpersonal interaction, often involving child-led play.

Objectives: To compare DTT to an empirically supported NDBI, the Interpersonal Developmental Approach (IDA) consisting primarily of JASPER.

Methods: Participants were 161 children with ASD, age 33-54 months at baseline, with limited language (<30 initiated spoken words during behavior observations). They were randomized to DTT (n=82) or IDA (n=79), and received one-hour intervention sessions five days per week for six months. Outcomes were assessed at baseline, end-of-treatment, and six-month follow-up. Outcome measures included the Reynell Developmental Language Scales and the Mullen Scales of Early Learning (MSEL), both assessing expressive and receptive language, and the Early Social Communication Scales (ESCS), an observational measure of nonverbal communication behaviors, including frequency of initiations of joint attention (IJA). Generalized linear mixed models (GLMM) were fitted with log link function for count outcomes and identity link function for Gaussian outcomes. Hurdle models were used for Reynell expressive language and IJA; this model examined whether or not each participant displayed the behavior (binary process) and, if so, how many (zero-truncated Poisson process).

Results:

Children in both DTT and IDA significantly improved from baseline to post-intervention and follow-up.

Reynell receptive language yielded non-significant post-intervention and follow-up differences between groups, F(1,308)=1.06, p=0.305, and F(1,308)=1.62, p=0.204, respectively. For expressive language, there were no significant post-intervention or follow-up difference between groups in either the binary, F(1,162)=0.06, p=0.799, or the truncated Poisson model, F(1,162)=0.06, p=0.799. On the MSEL, children in both IDA and DTT demonstrated significant gains in both expressive and receptive language from baseline to post-intervention, F(1,304)=185.97, p<0.001, and F(1,304)=156.36, p<0.001, respectively, and follow-up, F(1,304)=339.37, p<0.001, and F(1,304)=353.0, p<0.001, respectively. There were no significant treatment differences between IDA and DTT in the rate of improvement in both expressive and receptive language from baseline to post-treatment, F(1,304)=0.23, p=0.635, and F(1,304)=0.86, p=0.354 respectively.

For IJA, there was no significant treatment or maintenance in the binary model, F(1,159)=0.07, p=0.793; however, IDA was superior to DTT in the truncated Poisson model, F(1,159)=4.44, p=0.037. Children who displayed IJA at baseline made greater improvement in IJA if they received IDA than if they received DTT.

Conclusions: Participants in both DTT and IDA improved, and the groups did not significantly differ on most outcome measures. IDA did produce larger gains in IJA for participants who already demonstrated some joint attention skills when they began intervention. The similar improvements observed in approaches that differ markedly in intervention content and method suggest that it may be advantageous to pinpoint efficacious components of each approach and identify ways to match components to the needs of individual children with ASD.