Poster #31 - Autism Observation Scale for Infants Scores Predict Development of Autism-Related Traits in Unaffected Children

• In Event: Poster Session 09

Abstract

There is substantial evidence that autism-related traits are present in relatives of individuals with autism spectrum disorder (ASD), but less is known about the emergence and characteristics of these traits in early development. Understanding developmental trajectories of autism-related traits in individuals that do and do not go on to develop ASD is necessary to fully understand the broader autism phenotype. Georgiades et al. (2013) first investigated the emergence of autism-related traits in a sample of high-likelihood (HL) infant siblings of autistic probands and controls with no family history of ASD (low-likelihood – LL). Georgiades et al. (2013) in part validated the clinical utility of the Autism Observation Scale for Infants (AOSI) – a direct observational measure of signs of ASD in infants – and found that the cluster with higher AOSI scores had a higher proportion of HL infants and a greater prevalence of autism-related traits at 36 months.

Here, we sought to replicate Georgiades et al. (2013) in a new sample of HL siblings of autistic probands and LL controls and explore more fine-grained clustering solutions, comparing the outcomes using two, three, and four clusters. This sample is drawn from the multisite Infant Brain Imaging Study (IBIS Network). Similar to Georgiades et al. (2013), the 12 month-old HL subjects (n=270) showed significantly higher AOSI scores than the LL subjects (n=147), and both showed significantly lower AOSI scores than infants later diagnosed with ASD (n=88; Figure 1). After excluding the HL infants who received a diagnosis of autism, in all cluster solutions, the cluster with the highest AOSI scores at 12 months exhibited significantly higher levels of autistic-like traits at 24 and 36 months. In the two-cluster solution, while HL siblings had a relative risk of 1.31 (95% CI, 0.93-1.83) for membership in the higher AOSI cluster, there was no significant difference in the proportion of HL and LL infants in the two clusters (χ21 = 2.50; p = .11). Three- and four-cluster solutions better explained the data, as evidenced by the ascending Calinski-Harbasz statistic (two-cluster: CH = 854; three-cluster: CH = 1067; four-cluster: CH = 1368). HL siblings had a relative risk of 1.92 (95% CI, 1.12-3.29) for membership in the cluster with the highest AOSI scores in the three-cluster solution and 2.63 (95% CI, 1.11-6.19) in the four-cluster solution (see Table 1). The proportion differences between HL and LL infants was significant in the three-cluster solution (χ21 = 8.10; p = .02) and marginal in the four-cluster solution (χ21 = 6.78; p = .08).

Overall, these data replicate the general findings from Georgiades et al. (2013) and demonstrate the utility of the AOSI as an early measure of later emerging autistic traits. These data, however, show the superiority of three- and four-cluster solutions, highlighting that a more specified range of AOSI scores may offer even greater predictive information regarding the severity, prevalence, or phenotypic profile of autism-related traits later in life.

Authors

• Caitlin Sisk, University of Minnesota - Twin Cities
  Presenting Author

• Elayne Vollman, Lake Forest College

• Catherine Burrows, University of Minnesota - Twin Cities

• Kelly Botteron, Washington University in St. Louis

• Annette Mercer Estes, Uniiversity of Washington

• Stephen Dager, University Of Washington

• Heather Cody Hazlett, University of North Carolina at Chapel Hill

• Robert Schultz, Children's Hospital of Philadelphia

• Lonnie Zwaigenbaum, University of Alberta

• Joseph Piven, University of North Carolina at Chapel Hill

• Jed T. Elison, University of Minnesota - Twin Cities