Eye-Tracking Tool May Aid in Autism Diagnoses

Eye-tracking-based measurement of social visual engagement — how kids look at and learn from their social environment — was predictive of autism diagnosis by clinical experts in young children, researchers said.

In a multisite prospective study, the eye-tracking tool that measures social visual engagement had a sensitivity of 78% and a specificity of 85.4% in a subgroup of 335 children whose autism diagnosis was certain, with an area under the curve (AUC) of 0.90 (95% CI 0.87-0.94), reported Warren Jones, PhD, of Marcus Autism Center at Children’s Healthcare of Atlanta, and colleagues in JAMA.

In contrast, for 140 children for whom the diagnosis was uncertain, the test’s sensitivity was 56.3% and specificity was 68.1%.

“The index test’s performance metrics were significantly better when the reference standard was certain relative to uncertain” (P<0.001), Jones and team noted.

Eye-tracking test results also correlated with expert clinical assessments of individual levels of social disability (r = -0.75, 95% CI -0.79 to -0.71), verbal ability (r=0.65, 95% CI 0.59-0.70), and nonverbal cognitive ability (r=0.65, 95% CI 0.59-0.70).

“Autism is currently diagnosed behaviorally, by highly trained and specialized clinicians,” Jones told MedPage Today in an email. “The gold-standard diagnostic instruments measure the presence of autistic social disability through both behavioral observation of the child and through parent interview. Best practice also calls for standardized assessments of each child’s cognitive and language skills.”

“Unfortunately, there are often very long waitlists to access expert clinicians using gold-standard instruments,” he said. “There aren’t enough experts to meet the community need, and the assessments themselves are quite lengthy.”

“In addition, beyond highly trained specialists, practitioners in the general community often don’t use gold-standard instruments,” he noted. “As a result, children who don’t reach specialty centers often receive diagnostic labels without comprehensive evaluations and without use of any standardized assessments to guide individual supports for each child.”

Hopefully, these “objective measurements can help speed that process,” he added.

In an accompanying editorial, Geraldine Dawson, PhD, of the Duke Center for Autism and Brain Development in Durham, North Carolina, noted that the study “represents a milestone in the development of autism diagnostic biomarkers,” but more work is needed before an eye-tracking test can be used in clinical practice.

“Demonstrating that an eye-tracking test improves diagnostic certainty would require following children whose diagnosis was uncertain longitudinally to determine whether the test improves prediction of a later definitive autism diagnosis,” she explained.

Furthermore, future studies will need to examine how “feasible, acceptable, reliable, and efficient” the eye-tracking tool is, how information from the tool should be “weighted and integrated with other sources of information,” and the influence of demographic factors and co-occurring conditions, she added.

“The risks and benefits of incorporating objective biomarkers into standard clinical care should be carefully evaluated based on input from clinicians, caregivers, autistic individuals, healthcare administrators, and other stakeholders to ensure that this new approach to improving autism diagnostic certainty leads to long-term benefit for autistic individuals and their families,” Dawson concluded.

The study builds on research simultaneously published in JAMA Network Open.

In two prospective, consecutively enrolled diagnostic studies, Jones and colleagues showed that objective eye-tracking-based measurements of social visual engagement predicted reference standard expert clinician diagnosis with a sensitivity of 81.9%, a specificity of 89.9%, and an area under the receiver operating characteristic curve of 0.90 (95% CI 0.88-0.92) in the discovery study, and a sensitivity of 80.6%, a specificity of 82.3%, and an area under the receiver operating characteristic curve of 0.89 (95% CI 0.86-0.93) in the replication study.

Eye-tracking test results correlated with expert clinical assessments of children’s individual levels of ability, explaining 68.6%, 63.4%, and 49.0% of variance in reference standard assessments of social disability, verbal ability, and nonverbal cognitive ability, respectively, the authors wrote.

“Across the two papers, these results present data from more than 1,500 children, across three studies (one discovery and then two replications), validating an objective, performance-based biomarker to aid in the early diagnosis and assessment of autism,” Jones told MedPage Today, adding that the diagnostic tool is now FDA-cleared.

“We hope that the clinical data in these papers, and the availability of this tool, can help answer profound challenges that currently exist in the healthcare system (in the U.S. and globally), in which children with concerns about autism typically face a 2-year-long clinical odyssey before receiving a diagnosis,” he said.

“For a child, the data collection procedure is as straightforward as watching TV, but approximately 10 minutes of watching by the child yields actionable clinical assessment data for the clinician, with results that are highly correlated with gold-standard assessments that otherwise require hours of behavioral testing by highly trained experts,” he noted.

However, Jones and colleagues pointed out that their tool was not intended for general population screening.

“Studies in general population settings, where the prevalence of autism is lower than in the specialty clinics in the current study, will be of interest,” Jones and colleagues wrote in JAMA. “Relatedly, the tool evaluated in this study is not intended to replace expert clinicians, but to supplement informed and experienced clinical judgment.”

Study Details

For the multisite prospective study, Jones and colleagues included 499 children ages 16 to 30 months — which is significantly younger than the median age at autism diagnosis in the U.S. (4 to 5 years) — across six U.S. specialty centers from April 2018 through May 2019. Of these children, eye-tracking measurement of social visual engagement was successful in 95.2% (mean age 24.9 months, 74.2% boys, 67.4% white).

By expert clinical diagnosis, 46.5% of children had autism. In all children, measurement of social visual engagement had a sensitivity of 71.0% and a specificity of 80.7%.

In the two diagnostic studies, Jones and colleagues included 1,089 children ages 16 to 30 months, 719 of whom were included in the discovery study (mean age 22.4 months, 68.8% boys) and 370 in the replication study (mean age 25.4 months, 67.6% boys).

The researchers developed an objective eye-tracking-based index test and compared its performance with a reference standard diagnosis of autism (discovery study), and then replicated the findings in an independent sample (replication study). Reference standard diagnoses were made using best-practice standardized protocols by specialists blind to eye-tracking results.

Based on reference standard expert clinical diagnosis, 53.7% of patients had non-autism diagnoses and 46.3% had autism diagnoses in discovery, and 49.7% had non-autism diagnoses and 50.3% had autism diagnoses in replication.

Please enable JavaScript to view the

comments powered by Disqus.