Eye-tracking measures could help make the right autism diagnosis sooner

By Pooja Toshniwal PahariaReviewed by Danielle Ellis, B.Sc.Jun 5 2024

In a recent study published in JAMA Network Open, researchers investigated the reliability of eye-tracking biological markers in distinguishing autistic children from non-autistic ones during clinical evaluations in community-based settings. They also determined whether combining biological markers with primary care physician (PCP) diagnoses and diagnosis certainty improved diagnostic outcomes.

Background

Racial and ethnic minority children, as well as underprivileged communities, increase autism diagnosis gaps. Long wait periods for assessments are due to the large number of children who require evaluations, which exceeds the number of specialists. Diagnostic delays impede early, evidence-based therapies, lowering long-term care expenses.

To address the problem, new community-based care delivery models are under development that combine clinical and biobehavioral methods to increase early diagnostic accuracy and timeliness. Eye-tracking biological markers, which are non-invasive, low-cost, and practicable, show promise for detecting early autism diagnostic biomarkers.

About the study

In the present prospective diagnostic study, researchers investigated the reliability of eye-tracker biological markers used in primary care clinical evaluations to identify autistic children in community settings. They determined whether combining these biomarkers with PCP diagnoses would improve diagnostic accuracy.

Between 7 June 2019 and 23 September 2022, the Early Autism Evaluation (EAE) PCPs recommended a sequential pediatric sample for a blinded ophthalmologic monitoring index assessment and expert evaluation at follow-up. The study comprised 146 individuals aged between 14 and 48 months referred by seven EAE Hub centers. Of the 154 youngsters who participated, 146 produced adequate data for one or more eye-tracking metrics.

Child caregivers completed electronic surveys, and study members gathered EAE Hub PCP data. Within 16 weeks of the EAE Hub examination, the team performed a follow-up criterion-standard autism diagnostic assessment and eye-tracking biomarker battery test.

A professional clinical psychologist established the diagnosis based on the Autism Diagnostic Observation Schedule, Second Edition (ADOS-2), Vineland Adaptive Behaviour Scale, Third Edition (VABS-3), Mullen Scales of Early Learning (MSEL), and a caregiver interview. The EAE Hub PCPs provided categorical diagnoses and diagnostic certainty, with levels of confidence divided into certain and uncertain (somewhat, somewhat, or not at all certain).

The primary study outcomes were the specificity and sensitivity of the eye-tracking index test, the composite measure including significant indices for eye tracking compared to the standard reference autism diagnosis by clinical psychologists. Secondary outcomes included the specificity and sensitivity of the combined strategy that integrated the index test, PCP diagnoses, and diagnostic certainty.

The researchers used five eye-tracking biomarker batteries to assess non-social preference, attentional disengagement, pupillary light reflex (PLR) latency and amplitude, tonic pupil size, oculomotor metrics, and passive visual exploration. The battery consisted of five paradigms: GeoPreference, gap-overlap, PLR test, resting eye-tracking task, and passive visual exploration task. The researchers used binary logistic regressions, Pearson correlations, and a classification and regression tree (CART) analysis to identify the best predictors for reference standard autism diagnosis.

Results

Among study participants, the mean age was 2.6 years; 71% (n=104) were male; 14% (n=21) were Hispanic or Latino; and 66% (n=96) were non-Latino. Seventy percent (n=102) had a standard reference autism diagnosis, and 77% (n=113) showed autism outcomes consistent with the biomarker composite (index) and standard reference endpoints, with 78% sensitivity and 77% specificity. Integrating the index test composite biomarkers, primary care physician diagnoses, and certainty, 90% (114 out of 127) of individuals showed results consistent with the reference, with 87% specificity and 91% sensitivity.

Six eye-tracking indices showed associations with reference standard autism outcomes, including non-social preference, no-shift percentage, PLR latency and amplitude, and resting and exploratory fixation lengths. Correlational analyses of significant biomarkers and autism severity, developmental levels, and adaptive skills revealed that a higher non-social preference percentage was related to lower MSEL and VABS-3 scores for the autism reference group but not for the non-autism group.

Individual biomarkers were unrelated, except identical ideas tested in two tasks (fixation time). The mean area under the receiver operating characteristic curve (AUC) for the three training runs that selected our model was 0.93, and the mean cross-validation AUC was 0.90, which was slightly higher than the other two models identified, indicating superior performance for the selected model. The mean cross-validated AUC (0.90) was high, demonstrating excellent out-of-sample performance.

Conclusion

Based on the study findings, a multimethod approach to early autism diagnosis might enhance access to reliable diagnoses in places with few neurodevelopmental experts. Multiple eye-tracking indices may be sensitive to autism and provide information in addition to the outcome and certainty of the PCP diagnosis.

The composite eye-tracking biomarker was associated with the best-estimate clinical diagnosis of autism, and when combined with primary care physician diagnosis and certainty, it exhibited 87% specificity and 91% sensitivity.

The findings indicate that providing primary care physicians with a multimethod diagnostic strategy might considerably increase access to fast and accurate autism diagnoses.