Early Interventions for Autism: Mechanism and Developmental Science

Background: Suitably designed trials of early interventions for autism, combining the benefits of randomisation with repeated measures follow-up in a developmental context, can illuminate both the processes behind treatment effectiveness, and causal mechanisms within developmental science (Green and Dunn 2008).

Objectives: To identify the active ingredients of autism intervention and to use intervention trials to illuminate developmental science.

Methods: The presentation will draw from findings within two randomised controlled trials of parent-mediated, video-aided and developmentally-focused social communication intervention for autism.

1) The Preschool Autism Communication trial (PACT; Green et al 2010), a three site RCT (N=152) of the PACT social communication intervention against usual care in 2-5 year old children diagnosed with core autism; pre-designed to enable the study of treatment mechanism. The PACT mediation analysis (Pickles et al 2014) extended traditional techniques by including exploration of instrumental variables to address post-randomisation confounding, and exploited repeated measures to address measurement error in the mediator.

2) Intervention within the British Autism Study of Infant Siblings (iBASIS), a two site RCT (N=54), testing parent mediated intervention for infants at familial risk for autism, not selected for atypicality. Baseline at 9 months, treatment endpoint 14 months, follow up 24 months. The design tested intervention impact on pre-specified infancy risk markers for later autism using intention to treat analysis.

Results: PACT mediation analysis showed that a treatment effect on parental communicative synchrony with the child (effect size (ES) 1.22 (0.85,1.59) at 13 month endpoint, 1.44 at 6 months), strongly mediated (70%) an improvement in child communication initiation with parent (ES 0.41 (0.08, 0.74) at 13 months, 0.5 at 6 months). Other parental interactive behaviours did not significantly contribute to this mediation. In turn, the improvement in the child's communication initiation strongly mediated (73%) a more modest treatment effect on autism symptoms (ADOS; ES -0.24 (-0.59, 0.11) at endpoint).

New complimentary data will be presented from iBASIS on the intervention effect on parent-infant interaction (parent nondirectiveness, child attention), atypical infant behavior (AOSI), infant cognition (GAP attention disengagement) and brain function (ERP to speech sounds).

Conclusions: PACT mediation suggests a causal chain of effect from parent behavior with child to child behaviour with parent, to child behavior with researcher; identifying the active process of intervention and supporting the intervention theory. It also suggests the ADOS symptom change may be meaningful in direction and amenable to further improvement if the causal chain of effect can be enhanced; thus implying logical ways of strengthening treatment effectiveness. For developmental science, the causal direction of effect from parent behaviour to improved child social communication amplifies previous findings on the assocation between early parental synchrony and later child functioning (eg Siller et al 2008).

The comparative data from iBASIS will show how intervention can similarly effect parental communication behaviour in infancy; along with impact on infant's attention to caregiver, atypical behaviour, ERP response, attention disengagement, language and adaptive outcomes. Implication of these findings for the developmental science of the autism prodrome will be presented.