The Search for Biomarkers and Clinically Meaningful Subtypes of ASD Based on Eye Tracking Data

Background: Our understanding of autism as a heterogeneous disorder and the pressing need to discover and understand potential subtypes has emerged as a high research priority.  Markers that precede diagnosis or indicate ASD subtypes are extremely valuable because of their potential implications about causes, mechanisms, prognosis, and treatments. However, discovering and understanding such subtypes requires large sample sizes. In 2011 we published an eye tracking study that included 110 toddlers (Sample 1) and in 2015 examined an additional sample of 334 children (Sample 2) and raised the possibility that markers based on eye tracking, if robust, could be considered biomarkers of ASD (Pierce et al., 2011, 2015).  In both studies we discovered that a percentage (~20%) of ASD toddlers showed unusually high fixation levels on geometric rather than social images, a profile not often seen in typically developing or non-ASD contrast groups (Specificity 98%, PPV 89%).  Since that time we have collected data on another independent sample (Sample 3) of 387 toddlers that has not yet been published. Presumably if experience dependent mechanisms play an important role in how a child develops, the degree to which abnormal visual attention persists during early development may have some power to predict later outcome.  Moreover, if a true “biological subtype” is discovered, persistent visual attention traits within an individual toddler might be in evidence across different eye tracking paradigms.

Objectives: As part of an education motivated panel, the primary objective is to discuss the ways in which such a large dataset could be examined.  Topics relating to disentangling heterogeneity and the value of eye tracking indices as diagnostic and prognostic biomarkers will be considered. Other topics include examination of state and trait markers and stability of responding across time and different paradigm types to discover a true biotype.

Methods: Combining Samples 1-3, a total of 837 toddlers participated and included 242 ASD and 595 contrast toddlers (LD, DD, Typical, Typical Sibling and “Other”) ranging in age from 12-48 months.  Using a Tobii wireless eye tracker, all toddlers participated in the “GeoPref Test”, a 1-minute movie containing both dynamic geometric and social images. Fixation duration within each area of interest and number of saccades were recorded. Toddlers were diagnostically and psychometrically evaluated at time points both proximal and distal to the eye tracking session. In order to examine stability of responding across paradigms, toddlers participated in several different eye tracking tests beyond the GeoPref Test, such as one containing complex social interactions.

Results: Data from the new unpublished sample (Sample 3), replicates the finding that overall ASD toddlers fixate more on geometric images than other diagnostic groups (e.g., ASD vs TD, t₂₁₂ = 8.37, p<.0001, CI of the difference = 17.6%-28.5%).  Other analyses will be performed to understand subtypes, examine visual attention across time and consider how a child’s visual fixation patterns at ages 1-2 years might predict his outcome at ages 3-4 years. 

Conclusions: Eye tracking is a powerful tool for discovering subtypes of ASD and for understanding its early course.