Randomized Multisite Replication of Early Start Denver Model Outcomes

Background: Dawson et al (2010,2012) reported widespread improvements in outcomes of young children with ASD treated for two years for 15 hours per week with Early Start Denver Model, a manualized, comprehensive individualized intervention fusing developmental and behavior analytic methods. The ESDM group significantly surpassed a randomized comparison group in IQ gain, language, adaptive behavior, autism symptoms and diagnostic changes.  Furthermore, an EEG study revealed normalized patterns in the ESDM group involving stronger responses to social than nonsocial stimuli, a pattern opposite that of the comparison group.

Objectives: The objectives of this study were to replicate the original ESDM study in a large, randomized, multisite, controlled trial.

Methods:  98 children with ASD, without medical risk factors, between 15 and 28 months of age, were randomized into either 20 hours per week of ESDM for 24 months or to community treatment at three sites. There were no significant differences in participant characteristics across sites, and no significant differences between experimental and comparison participants on any dimension.  ESDM intervention was delivered in homes by paraprofessionals trained and supervised by professional staff and averaged 14.8 hours per week. Treatment fidelity measures demonstrated that all sites delivered the intervention at acceptable, and similar, levels.  All measures were gathered pretreatment, after 12 months of treatment, and after 24 months of treatment.   

Results: Significant differences in outcomes across sites required that we analyze each site separately, rather than pooling of data.  Two of the three sites replicated the ESDM advantage in IQ and language previously reported. In site 1, in which the comparison group averaged more intervention hours than did the ESDM group, the ESDM group mean IQ significantly increased from 69 to 90 while the comparison group increased from 67 to 85. In site 2, mean IQ increased from 73 to 91 compared to the comparison group from 68 to 67.  In contrast, in site 3, the comparison group made greater IQ gains (from 67 to 83) than did the ESDM group (64 to 72), though the increases in both groups were significant. There were significant drops in autism symptom severity in both assignment groups in all sites. Significant group differences in language occurred in 2 sites.

Conclusions: This randomized multisite study demonstrated significant developmental improvements related to ESDM assignment in two of three sites, with the mean IQ gains in those sites mirroring or surpassing the IQ gain reported in the original study.  It is interesting that in this study and in many others, treatment effects are much larger for developmental variables rather than for core autism symptoms. Developmental variables, rather than autism severity per se, are the best predictors of outcomes and therefore these improvements are critical for later functioning. Significant effects of ESDM compared to randomized comparison groups have now been demonstrated at three, independent sites. This demonstrates that, according to Nathan and Gorman’s criteria, ESDM is an efficacious, empirically validated 1:1 intervention model for young children with ASD.