25646
Trajectories of Cognitive Development in Toddlers at-Risk for Autism Due to Language Delays

Saturday, May 13, 2017: 12:00 PM-1:40 PM
Golden Gate Ballroom (Marriott Marquis Hotel)
L. Henry1, C. Farmer1, L. B. Swineford2, S. S. Manwaring3 and A. Thurm1, (1)National Institute of Mental Health, Bethesda, MD, (2)Washington State University, Spokane, WA, (3)University of Utah, Salt Lake City, UT
Background: Toddlers with early language delays (LD) demonstrate variable outcomes, with some improving enough to “catch up,” and others developing a variety of disorders, including autism spectrum disorder (ASD; Miniscalco et al., 2006). Cognitive growth in other at-risk infants (siblings of children with ASD) is variable, but is often delayed (Brian, et al., 2014; Landa, & Garrett-Mayer, 2006; Landa et al., 2012; Estes et al., 2015). Less is known about cognitive development in toddlers at-risk for ASD due to early LD.

Objectives:  The aim of the present study was to examine the trajectories of both nonverbal and verbal cognitive development in toddlers with significant LD compared to TD, and investigate patterns of cognitive development in relation to outcome classification.

Methods: Data were combined from two longitudinal studies of language delay conducted at the National Institute of Mental Health and the University of Utah (n=91). Cognitive assessment included administration of the Mullen Scales of Early Learning (MSEL) at approximately 18 months of age. Dependent on MSEL language scores and history of delay, toddlers were categorized in to two groups: At-risk for ASD due to significant LD (N=30), or TD (N=61). Follow-up visits occurred at approximately 24 and 36 months. Growth mixture models (GMM) explored heterogeneity in cognitive development. In the full sample, a series of increasingly complex GMM were fitted and up to five classes were enumerated. Based on several fit indices and interpretability, the best solution was selected, and class membership was evaluated as a predictor of outcome grouping: No delays, non-spectrum delay, or ASD.

Results: The best-fitting models for nonverbal mental age (NVMA) and verbal mental age (VMA) are shown in Figure 1. The three-class solution was selected for NVMA (Age Appropriate, 78%, Delayed, 20%, and Significantly Delayed, 2%). Despite its small size, the Significantly Delayed trajectory was empirically supported and clinically meaningful. Unsurprisingly, NVMA class assignment was related to outcome (Figure 2; Fisher’s Exact Test, p<.001); there was no significant difference in assignment for toddlers with non-spectrum delay compared to ASD, though the sample size reduced power. The best-fitting model for VMA consisted of four classes: Age Appropriate (67%), Delay Catch-Up (9%), Delayed (17%), and Significantly Delayed (7%). VMA class assignment was also related to outcome (Figure 2; Fisher’s Exact Test, p<.001); toddlers with no delay outcomes comprised the Age Appropriate and Delay Catch-Up classes, while toddlers with non-spectrum delay and ASD predominantly formed the Delayed and Significantly Delayed classes, with no significant difference in assignment in toddlers with non-spectrum delay compared to ASD.

Conclusions: Results demonstrate significant heterogeneity in the cognitive development of toddlers with LD. Given that study recruitment was based on LD, the increased heterogeneity in trajectory observed for VMA compared to NVMA is unsurprising. Further, the four VMA classes were predictive of outcome classification, revealing variability in the development of toddlers with no delays, non-spectrum delay, and ASD outcomes, and demonstrating the link between impaired cognition and ASD in toddlers. We plan to extend these analyses to explore individual subscales of the MSEL.