Miss a day, miss a lot. Subscribe to The Defender's Top News of the Day. It's free.
A computer-based eye test correctly identified 71% of children with autism and recognized 81% of those without, leading researchers to conclude that automated eye-tracking “warrants further evaluation for early diagnosis and assessment of autism in young children referred to specialty clinics.”
According to a paper appearing Sept. 5 on the JAMA Network, the test worked somewhat better for children with confirmed autism spectrum disorder (ASD) based on current diagnostic standards, picking out 78% of affected subjects and 85% of those unaffected.
Strong correlations also emerged between the eye test’s measure of symptom severity and earlier evaluations for social, verbal and nonverbal signs of ASD.
This outcome confirmed earlier results, also published Sept. 5 on JAMA Network Open by the same research group.
How researchers conducted the study
The researchers, led by Warren Jones, Ph.D., at the Center for Translational Social Neuroscience, Emory University, were interested in how results from a product, EarliPoint Evaluation for Autism Spectrum Disorder, matched standard clinical diagnoses.
EarliPoint tracks and notes a child’s attention to specific aspects of on-screen social interaction scenes. It then compares those visual responses to those of a reference set of children of children who do not exhibit ASD symptoms.
Investigators recruited 475 children, ages 16-30 months, who had undergone assessment for ASD at treatment centers between April 2018 and May 2019.
Inclusion criteria included normal vision (with or without glasses), general good health and a parent or guardian who could understand informed consent information in English.
Researchers also collected information on race, ethnicity and maternal education level.
The children had previously received a diagnosis of ASD or non-ASD based on one or more standardized evaluations: the Autism Diagnostic Observation Schedule, Second Edition, the Mullen Scales of Early Learning, medical and developmental histories or DSM-5 (Diagnostic and Statistical Manual of Mental Disorders) criteria.
Since this was a double-blind study, test administrators were unaware of a child’s previous diagnosis.
Physicians who evaluated children for this study were required to provide an estimate of the reliability of their diagnosis.
According to the authors, while established diagnostic approaches represent “current best practice for diagnosing autism,” they are “based on the expert clinician’s subjective judgment.”
Investigators were interested in how differences in test scores between subjects and unaffected “reference” children correlated to a pre-enrollment diagnosis of ASD or non-ASD.
Sensitivity — how well EarliPoint identified pre-diagnosed, confirmed cases of autism — and specificity — how well the test identifies children without ASD — were calculated using standard statistical methods.
As a secondary measure of test effectiveness, researchers examined the relationship between deficiencies quantified by EarliPoint and the severity of clinician-identified symptoms for social disability and both verbal and nonverbal ability.
They included these additional measures because, even within a diagnosis of autism, children show a wide range of verbal and nonverbal responses and abilities, from seriously delayed to advanced.
Further details of this study are available at clinicaltrials.gov.
How the EarliPoint test works
The visual assessment tool used in this study, the EarliPoint Evaluation for Autism Spectrum Disorder, is a commercial product developed by EarliTec Diagnostics Inc. The second-generation product received FDA 510(k) clearance in July 2023 for use in children 16-30 months old.
Children undergoing EarliPoint Evaluation testing watch curated scenes of social interactions on a tablet. Tracking technology follows the child’s eyes as they survey the scenes, taking more than 120 data points per second.
Then, proprietary analysis software compares these results to age-appropriate reference values to note differences and deficits.
According to developers, this gives clinicians “a timely, objective and accurate reading of the presence of autism, the severity of the child’s social disability, and the child’s level of verbal ability and non-verbal learning.”
Young children tend to focus on the same area of a scene as it unfolds, a response known as entrainment. EarliPoint compares the responses of test subjects to those visual cues with the responses of unaffected children. Clinicians apply a statistical threshold to determine how the values diverge.
Testing took one day per child — a very long time for very young children — but 95% completed the assessment.
EarliTech posted a video showing how EarliPoint Evaluation works.
Why a ‘hard’ test is needed
A 2019 study reported about 1 in 6 U.S. children ages 3-17 years live with developmental disorders. These include physical, emotional and intellectual deficits.
According to the Centers for Disease Control and Prevention, 1 in 36 U.S. children have ASD, up from 1 in 150 in 2000. ASD occurs in all racial, ethnic and socioeconomic groups and is nearly 4 times more common among boys than girls.
Most ASD diagnoses do not involve a validated biochemical or quantitative, instrument-based assessment and are therefore open to interpretation.
According to the authors of the JAMA Network paper, an objective measure of ASD-related deficits could help by identifying affected children early and quantifying the extent of their illness.
A year typically passes before children showing signs of ASD see a specialist and a confirmed diagnosis takes even longer. This delay represents lost opportunities to intervene before maladaptive behaviors become established.
‘Not intended to replace expert clinicians’
The eye-tracking study included autistic and normatively developing children with widely divergent social and intellectual capabilities.
On that point, it succeeded in confirming its hypothesis that visual tests can predict or confirm an ASD diagnosis. Its prospective, double-blind design was another plus.
Prospective studies, which watch for outcomes during the trial after recruiting and categorizing study subjects, generally involve fewer sources of bias and confounding than retrospective studies.
Additionally, EarliPoint provides an objective, validated, predictive biomarker for ASD with the potential to reduce the time between the first signs of autism and intervention.
It could also potentially provide material for genetic and neurobiology studies, assist in monitoring treatments and help categorize subjects for future clinical trials.
On the negative side, investigators considered the 30% level of uncertainty in expert diagnoses of ASD or non-ASD as a “challenge,” despite this figure being 25% lower than a 2018 report.
Another limitation involved the study population. Since all subjects had been recently evaluated for ASD at the time of enrollment, they do not represent children in the general population, where the prevalence of ASD would be much lower.
On that basis, Jones and co-authors advised:
“This study is not intended to replace expert clinicians, but to supplement informed and experienced clinical judgment. Autism is a consequential diagnosis, and families of children with autism benefit from clinicians’ support to translate assessment results into plans of care and support as needed.”
