Eye-Tracking Test May Aid in Early Autism Diagnosis

An investigational eye-tracking device has identified a potential biomarker for autism spectrum disorder (ASD), as reported in 2 studies published by the Journal of the American Medical Association.

Although the majority of parents of children with ASD have concerns for their child’s development by age 2 years, the median age of diagnosis in the United States (US) is 4 to 5 years, and the delay is even longer for minority, low-income, and rural families. These delays pose major barriers to providing support and assistance at a developmentally sensitive time period. To aid in accurate and efficient diagnoses, researchers used an eye-tracking device, authorized by the US Food and Drug Administration (FDA), for the development of an objective measurement of social visual engagement as a prospective biomarker of ASD.

Identifying a Performance-Based Biomarker of ASD Using Eye-Tracking Data

The researchers first used the eye-tracking device to develop the data collection tool and algorithms of social visual engagement in an initial observational discovery study¹ (N=719; mean age, 22.4 months; girls, 31.2%) and then conducted a replication study¹ to test performance in an independent sample (N=370; mean age, 25.4 months; girls, 32.4%). Participants were children 16 to 30 months (discovery trial) or 16 to 45 months (replication trial) for whom there were concerns about ASD. Exclusion criteria included acute illness, clinically diagnosed hearing or visual impairments, and previously diagnosed genetic conditions associated with autism-like symptoms. All enrolled children in both studies completed an eye-tracking assessment (index test) and received clinician best-estimate diagnosis (reference standard) during a single testing site visit. The index test collected eye-tracking data to develop measures of the children’s social visual engagement while they watched video scenes of social interactions.

Primary end-point analyses assessed the diagnostic accuracy of the index test compared to the reference standard. Index test performance had an area under the receiver operating characteristic curve (AUROC) of 0.90 (95% CI, 0.88-0.92) in the discovery study and 0.89 (65% CI, 0.86-0.93) in the replication study. The discovery study index test results were able to predict reference standard diagnosis with sensitivity equal to 81.9% (95% CI, 77.3%-85.7%) and specificity equal to 89.9% (95% CI, 86.4%-92.5%). The replication study produced similar results, with index test prediction sensitivity equal to 80.6% (95% CI, 74.1%-85.7%) and specificity equal to 82.3% (95% CI, 76.1%-87.2%).  These results indicated the eye-tracking test had high convergent validity with expert clinician diagnosis, providing an efficient, automated aid to standard ASD diagnostics.

Eye-Tracking Assessment Accurately Predicts Diagnosis of ASD

To further evaluate the potential of the eye-tracking assessment in early diagnosis of ASD, an additional prospective, double-blind study was conducted in 6 specialty centers across the US. ² A total of 475 participants aged 16 to 30 months (mean age, 24.1 months; White, 74.1%) with normal hearing, normal vision, and no acute illness were enrolled. The primary endpoint again compared the eye-tracking index test against results from the expert clinicians’ reference standard diagnosis of ASD. In this study, a reference standard diagnosis was made by the expert clinicians using standardized assessments such as the Autism Diagnostic Observation Schedule, Second Edition (ADOS-2); Mullen Scales of Early Learning (Mullen); the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5); and each child’s medical and developmental history. In addition, clinicians were required to prospectively rate their level of certainty in each diagnosis of autism. 

For all participants, the index test performance had an AUROC equal to 0.85 (95% CI, 0.82-0.89), sensitivity of 71.0% (95% CI, 64.7%-76.6%), and specificity of 80.7% (95% CI, 75.4%-85.1%). However, given that 140 (29.5%) diagnoses were rated as uncertain, the researchers then stratified the test performance by clinician certainty. In children whose diagnosis was judged certain (n=335), the index test performance had an AUROC equal to 0.90 (95% CI, 0.87-0.94), sensitivity of 78% (95% CI, 70.7%-83.9%), and specificity of 85.4% (95% CI, 79.5%-89.8%). On the other hand, index test performance for children whose diagnosis was judged uncertain (n=140) had an AUROC equal to 0.70 (95% CI, 0.62-0.79), sensitivity of 56.3% (95% CI, 44.8%-67.3%), and specificity of 68.1% (95% CI, 56.4%-77.9%). Therefore, the eye-tracking test performs significantly better for cases when the reference standard ASD diagnosis is certain (z=4.65, P <.001), relative to uncertain cases.

Furthermore, this study extended the findings of the discovery/replication stage by including an exploratory analysis on race. When examining clinician ASD diagnoses by racial category, children who were Black/African American and Hispanic/Latino were given significantly more ASD diagnoses by clinicians, compared against White (χ²=11.24; P <.048) or non-Hispanic/non-Latino (χ²=14.39; P <.001) children, respectively. Notably, this discrepancy in ASD diagnosis did not occur when using eye-tracking data for diagnosis for either Black/African American children (χ²=2.16; P =.827) or Hispanic/Latino children (χ²=3.12; P =.210).

The results from these multisite, prospective, double-blind studies demonstrate that the development of an eye-tracking test of social visual engagement predicted ASD diagnosis had similar validity to expert clinician diagnosis. The accuracy and efficiency of this diagnostic tool has the potential to aid healthcare providers in earlier detection of ASD, particularly in underserved and minority populations. Identification of this biomarker could be vital in delivering support to children with ASD at an earlier time point than previously possible.

Warren Jones, PhD, the lead author of these studies and Director of Research at Marcus Autism Center in Atlanta, Georgia, stated in a press release, “These results leverage 15 years of basic science to demonstrate that biomarker-based technologies can provide a solution to the public health need for accessible and efficient early diagnosis and assessment of autism^.”3  

Co-author Ami Klin, PhD, Division Chief of Autism and Developmental Disabilities at Emory University School of Medicine, added that, “This is the first step toward alleviating the endless waits experienced by parents who need a prompt diagnosis for their children to access early intervention and supports, which, in turn, have the power to optimize lifetime outcomes.”³

A limitation of the current studies includes their limited testing environments, as the device was only tested in specialty centers and not in general medical settings. Furthermore, this tool is intended to supplement informed and experienced clinical judgment and is not a stand-alone assessment.

Disclosure: Multiple study authors declared affiliations with biotech, pharmaceutical, and/or device companies. Please see the original reference for a full list of disclosures.

This article originally appeared on Psychiatry Advisor