Comparison of three different eye-tracking tasks for distinguishing autistic from typically developing children and autistic symptom severity

Altered patterns of visual social attention preference detected using eye-tracking and a variety of different paradigms are increasingly proposed as sensitive biomarkers for autism spectrum disorder. However, few eye-tracking studies have compared the relative efficacy of different paradigms to discriminate between autistic compared with typically developing children and their sensitivity to specific symptoms. To target this issue, the current study used three common eye-tracking protocols contrasting social versus nonsocial stimuli in young (2-7 years old) Chinese autistic (n = 35) and typically developing (n = 34) children matched for age and gender. Protocols included dancing people versus dynamic geometrical images, biological motion (dynamic light point walking human or cat) versus nonbiological motion (scrambled controls), and child playing with toy versus toy alone. Although all three paradigms differentiated autistic and typically developing children, the dancing people versus dynamic geometry pattern paradigm was the most effective, with autistic children showing marked reductions in visual preference for dancing people and correspondingly increased one for geometric patterns. Furthermore, this altered visual preference in autistic children was correlated with the Autism Diagnostic Observation Schedule social affect score and had the highest discrimination accuracy. Our results therefore indicate that decreased visual preference for dynamic social stimuli may be the most effective visual attention-based paradigm for use as a biomarker for autism in Chinese children. Clinical trial ID: NCT03286621 (clinicaltrials.gov); Clinical trial name: Development of Eye-tracking Based Markers for Autism in Young Children. Autism Res 2019, 12: 1529-1540. © 2019 International Society for Autism Research, Wiley Periodicals, Inc. LAY SUMMARY: Eye-tracking measures may be useful in aiding diagnosis and treatment of autism, although it is unclear which specific tasks are optimal. Here we compare the ability of three different social eye-gaze tasks to discriminate between autistic and typically developing young Chinese children and their sensitivity to specific autistic symptoms. Our results show that a dynamic task comparing visual preference for social (individuals dancing) versus geometric patterns is the most effective both for diagnosing autism and sensitivity to its social affect symptoms.