The Autism Biomarkers Consortium for Clinical Trials: Eye Tracking Interim Analyses

Oral Presentation
Saturday, May 12, 2018: 11:45 AM
Grote Zaal (de Doelen ICC Rotterdam)
F. Shic1, A. Naples2, E. Barney1, C. Sugar3, M. Murias4, J. Dziura5, C. Brandt5, R. Bernier6, K. Chawarska2, G. Dawson7, S. Faja8, S. Jeste3, C. A. Nelson8, S. J. Webb6 and J. McPartland2, (1)Center for Child Health, Behavior and Development, Seattle Children's Research Institute, Seattle, WA, (2)Child Study Center, Yale University School of Medicine, New Haven, CT, (3)University of California, Los Angeles, Los Angeles, CA, (4)Duke Center for Autism and Brain Development, Department of Psychiatry and Behavioral Sciences, Duke University, Durham, NC, (5)Yale University, New Haven, CT, (6)Psychiatry and Behavioral Sciences, University of Washington, Seattle, WA, (7)Department of Psychiatry and Behavioral Sciences, Duke Center for Autism and Brain Development, Durham, NC, (8)Boston Children's Hospital, Boston, MA

Eye tracking (ET) provides indices of cognitive processes via overt attention. ET measures robustly capture differences in social information processing in ASD, as well as associations with clinically-relevant phenotype. However, many prior studies have focused on theoretical mechanisms, relied on small samples, or have not been replicated.


To examine the suitability of several prominent ET paradigms for use as autism biomarkers.


ET data were acquired from 4-11 year-old (average: 7.2±2.2 years) children with ASD (n=24) and typical development (TD, n=26) with EyeLink 1000 Plus 500hz eye trackers during the following experiments: activity monitoring (AM: actresses engaged in shared activities), social interactive (SI: play interaction between children), biological motion (BM: side-by-side presentations of human motion point-light versus scrambled/rotating points), dynamic naturalistic scenes (DNS: social/emotional movies), visual search (VS: array of social/non-social objects), gap overlap (GO: orienting to a peripheral target from a central cue), pupillary light reflex (PLR: pupil size changes after bright flash), static scenes (SS: naturalistic social scenes), and spontaneous social orienting (SSO: actress looking and speaking to the camera). IQ was added as a covariate in group comparisons. Trial-level quality control (QC) requirements included: valid data acquired for more than 50% of stimulus presentation (%Valid), calibration uncertainty/error less than 2.5 degrees; experiment-level QC included 25% of trials being valid; subject-level QC included 1 or more valid experiments. Primary and secondary variables varied per experiment. Phenotypic associations were examined in the ASD group using Spearman's rank correlation.


Valid experiment acquisition was >90% in both groups for all experiments except GO (88%). 96% of the ASD group and 100% of TD provided at least one valid experiment. Lower %Valid was noted in ASD (p<.05) for DNS, PLR, SS, but also in sub-conditions of specific experiments (e.g. static images in AM). No between-group calibration differences emerged for any experiment. Primary variable differences (ASD<TD, *p<.05, ~p<.10) were noted in AM (%Heads*), SI (%Social~), VS (%Face*), SS (%Face*), and SSO (%Face*, Dyadic Bid and Joint Attention phases); secondary differences in DNS (%Eyes*), GO (overall reaction time*); no differences in BM and PLR. Positive associations in ASD were seen between %Valid and Overall IQ (AM~, SI*, VS*), Vineland Social standard scores and SI %Head~, ADOS overall severity and PLR latency~, and age and PLR constriction*; negative associations between IQ and calibration error (AM*,BM*,VS~), ADOS SA severity and %Valid (SS~,VS*), and SRS Social Awareness T and %Heads/%Face (AM*,SI~,SS*). The presentation will include updated data from the larger interim analysis sample.


ET paradigms generated valid data for nearly all participants. Paradigms with support for validation were mainly drawn from those with primary predictions of diminished face/head looking in ASD. Over multiple paradigms, decreased valid data collection in ASD was associated with lower IQ and increased symptom severity, suggesting QC variables may index pervasive characteristics of atypical attention in ASD. Complex patterns of phenotypic association point to nuances relating to selection of ET paradigms with ideal biomarker characteristics. In summary, these results highlight the promise of ET biomarkers for ASD.