Early Vocalization Trajectories are Associated With ASD Diagnosis in High-Risk Infants

• In Event: Poster Session 08
  In Poster Session: PS 08 Topic - Developmental Disabilities

Abstract

Background. Recent research has identified patterns of early vocalization development that differentiate infants at high familial risk for autism from infants at low risk (Paul, Fuerst, Ramsay, Chawarska, & Klin, 2011; Northrup & Iverson, 2015), but little is known about which high-risk infants will develop ASD. Significant heterogeneity in language development in high-risk infants has made it difficult to identify specific linguistic markers of enhanced autism risk. Recently, there has been a push to use naturalistic language sampling to characterize language development in ASD, rather than (or in addition to) standardized assessments and parent-report questionnaires (Barokova & Tager-Flusberg, 2018). However, prospective naturalistic vocalization sampling has yet to be used to identify markers of elevated autism risk in samples of infants already at high familial risk for ASD.
Objectives. This study utilizes a data-driven, latent class approach to (1) identify trajectories of language development from 6-24 months in a large sample of infants at high familial risk of ASD, and (2) assess the relative likelihood of being diagnosed with ASD in each pathway.
Method. Infants at high familial risk of developing ASD (N=232) were assessed at 6 (AOSI; Bryson, Zwaigenbaum, McDermott, Rombough, & Brian, 2008), 12 (AOSI or CSBS; Wetherby & Prizant, 1993), and 24 (CSBS) months as a part of the multi-site Infant Brain Imaging Study, with diagnostic assessment at 24 months. Trained and reliable coders identified speech-like vocalizations (non-vegetative or affective) from videos. Vocalizations (count per 10 minutes) were entered into a general latent class mixed model fitted using maximum likelihood. The optimal number of classes was detected using goodness-of-fit statistics (BIC). Posterior classification probabilities and class membership by diagnosis are reported; diagnosis odds ratios were calculated using logistic regression.
Results. The optimal number of classes in this dataset is 2 (BIC=4599.64; Table). Class 1 (58.19%) showed steep growth in the production of speech-like vocalizations from 6-24 months (Figure). Class 2 (41.81%) demonstrated an attenuated slope, with large class differences by 24 months. Class was significantly associated with diagnostic group (b=-1.27, z=-3.96, p<0.001); high-risk infants in class 2 (slow growth) were more likely to be diagnosed with ASD at 24 months than high-risk infants in class 1 (OR=3.57).
Discussion. Wide variability in the phenotypic profiles of infants at high familial risk for ASD renders it challenging to identify early markers that confer likelihood of diagnosis. This study utilized data driven methods to identify two latent classes of speech-like vocalizing that were associated with differential likelihood of developing ASD. Infants with attenuated growth in speech-like vocalizations were 3.57 times more likely to be diagnosed with ASD than infants showing steeper growth. Infants in the steep growth class had a rate of diagnosis similar to rates seen in other populations of infant siblings (14.81%; Ozonoff et al., 2015) whereas the class with slower growth had a diagnosis rate of 38.14%. Class assignment did not identify 100% of infants later diagnosed with ASD, but could be utilized to identify infants who should be prioritized for diagnostic assessment and early intervention.

Group Authors

The Infant Brain Imaging Study (IBIS) Network

Authors

• Samantha Plate, University of Pittsburgh
  Presenting Author

• Victoria Petrulla, Florida International University
  Non-Presenting Author

• Lisa Yankowitz, University of Pennsylvania
  Non-Presenting Author

• Meredith Cola, Children’s Hospital of Philadelphia
  Non-Presenting Author

• Whitney Guthrie, Children's Hospital of Philadelphia
  Non-Presenting Author

• Birkan Tunç, Children's Hospital of Philadelphia
  Non-Presenting Author

• Juhi Pandey, Children's Hospital of Philadelphia
  Non-Presenting Author

• Meghan Swanson, University of Texas at Dallas
  Non-Presenting Author

• Shoba S Meera, National Institute of Mental Health and Neurosciences
  Non-Presenting Author

• Annette Mercer Estes, Uniiversity of Washington
  Non-Presenting Author

• Natasha Marrus, Washington University in St. Louis
  Non-Presenting Author

• Sarah Paterson, James S McDonnell Foundation
  Non-Presenting Author

• John R. Pruett, Washington University in St. Louis
  Non-Presenting Author

• Heather Cody Hazlett, University of North Carolina at Chapel Hill
  Non-Presenting Author

• Stephen Dager, University Of Washington
  Non-Presenting Author

• John Constantino, Washington University in St. Louis
  Non-Presenting Author

• Tanya St. John, University of Washington
  Non-Presenting Author

• Kelly Botteron, Washington University in St. Louis
  Non-Presenting Author

• Lonnie Zwaigenbaum, University of Alberta
  Non-Presenting Author

• Joseph Piven, University of North Carolina at Chapel Hill
  Non-Presenting Author

• Robert Schultz, Children's Hospital of Philadelphia
  Non-Presenting Author

• Julia Parish-Morris
  Non-Presenting Author