Measuring Interactive Developmental Pathways in ASD: A Dual-Domain Latent Growth Curve Model

Background:  Autism spectrum disorders (ASDs) are characterized by heterogeneity in both abilities and impairment across developmental domains.  How quickly individuals with ASDs develop skills in one important domain [e.g., social competence (SC)] may affect the rate of acquisition of multiple abilities [e.g., language (LANG)] in a cascade-like fashion. Uncovering whether and how such dynamic processes occur is crucial to understanding the pathophysiology of ASDs and to developing more effective interventions. This is possible through modeling the interactions between simultaneously measured growth curves of different developmental domains using a latent growth curve (LGC) analysis approach.  

Objectives:  SC and LANG are key predictors of outcomes and foci of intervention for children with ASD. We therefore aimed to: a) determine whether SC and LANG varied significantly in a cohort of preschoolers with ASDs; and b) model dynamic interactions between initial levels and rates of change over 1 year in SC and LANG trajectories.

Methods:  Data for 365 2- to 4-year-olds were obtained from a prospective longitudinal study of preschoolers recently diagnosed with an ASD in Canada. Children were assessed at time of diagnosis and then twice at 6-month intervals using the Vineland Adaptive Behavior Scale II Socialization domain as a measure of SC (relating to others, social play etc.) and the Preschool Language Scale 4 (Auditory Comprehension) as a measure of LANG. Latent variables represented the intercept (baseline level) and slope (rate of change) of SC and LANG.  Models were developed to measure the associations between intercepts and slopes of the two developmental domains.  These models were then compared in two groups: participants diagnosed before versus after the median age of diagnosis (39.5 months), and participants with cognitive abilities lower and higher than the sample median IQ of 55 as measured by the Merrill-Palmer-Revised Developmental Index. 

 

Results:  The LGC model demonstrated an excellent fit to the data (CFI=0.99; RMSEA=0.04).  Initial levels (intercepts) and rates of change (slopes) in LANG and SC varied significantly between individuals.  The intercept and slopes of SC were positively associated with the rate of change of LANG over the first year after diagnosis, with the rate of change in SC being more strongly predictive of LANG slope.  Initial levels of LANG did not predict rate of change in SC in the overall sample.    Children who were diagnosed before versus after 40 months of age differed only with respect to initial levels of LANG and SC.  By contrast, children with higher IQ demonstrated significantly higher initial SC and LANG,  and greater rates of change across both domains.  Furthermore, initial LANG  was a stronger predictor of subsequent growth in SC in the cognitively higher-functioning group compared to those with lower IQ. 

Conclusions:  Early social competence and its growth in the first year after diagnosis appear to be more important predictors of change in language ability than vice-versa. However, this growth appears to be moderated by IQ. These findings emphasize the importance of early ASD intervention focusing on social competence with additional interventions (e.g. language) tailored to children of differing cognitive abilities.