24421
Exploration of Prognostic and Predictive Factors of Outcome in Participants in the Preschool Autism Communication Trial (PACT)
Objectives: To explore this issue we exploited data from the large Pre-school Autism Communication Trial (PACT) (Green et al., 2010) which had treatment sites in three UK locations. Families were randomised to receive PACT or Treatment as Usual (TAU). We wanted to identify which baseline child and family factors might differentiate between children who improved on a measure of autism symptoms (ADOS-G) and those who did not, over the period of a year. A further aim was to attempt to distinguish between factors that influence outcome for all children (i.e. prognostic factors) from those due to the intervention (predictive factors).
Methods: Data from 152 families who were randomised to receive PACT or TAU at three UK sites were analysed. ADOS scores (Lord et al 2000) from baseline to end of study (146 children) were used to assess changes in autism symptomatology over time. Jacobson and Truax’s (1991) Reliable Change Index (RCI) was used to identify children whose improvement or deterioration could be considered psychometrically reliable. The sample was divided into those children (n=43) who had made reliable improvement in ADOS scores over the trial period and those (n= 41) who had not.
Results: Baseline non-verbal ability was a significant prognostic indicator, being greater among Improvers than Non-Improvers, irrespective of randomisation with moderate effect size (d=.46). No other child factors were prognostic or predictive of treatment outcome. Contingency analysis indicated that treatment group x outcome status varied by trial site (see table). TAU families in London site were less likely to have child Improvers than Non-Improvers, compared with London families receiving PACT or families at the other trial sites, irrespective of randomisation. Exploration of parent variables suggested that lower levels of parent synchrony during parent-child interaction was associated with a poorer outcome on ADOS score for children in the TAU group, especially in London, but this was not the case for the children receiving the PACT treatment.
Conclusions: Our data indicate the importance of separating prognostic from predictive factors when analysing intervention effects. Although non-verbal ability was related to prognosis more generally, it was not specifically associated with treatment outcome. No other significant factors were identified although analysis of family factors suggested that, for parents with more limited synchronous communication, PACT therapy – with its focus on supporting parent sensitive interaction with the child – may mitigate the odds of children with autism maintaining or increasing their levels of symptom severity. Continuing research in this area and the need to distinguish carefully between prognostic and predictive factors in intervention research is important if clinicians are to be able to make reasoned decisions regarding for whom a specific therapy might be more or less beneficial.