Abstract
Background:
The visio-vestibular examination (VVE) of smooth pursuit, saccades, vestibulo-oculomotor reflex, near point of convergence (NPC), accommodative amplitude (AA), and complex tandem gait, has enhanced the physical examination of concussed individuals. The EyeBOX is a FDA-approved device for the diagnosis of concussion.
Hypothesis:
The goal of this investigation was to determine the diagnostic accuracy for concussion-related vision disorders of infrared eye-tracking via the EyeBOX.
Methods:
A cross-sectional study of 112 adolescents ages 11-18 years of age experiencing persistent post-concussion symptoms 4-12 weeks after concussion was conducted. All subjects had a clinical visio-vestibular examination (VVE) with a sports medicine physician, and a formal, comprehensive vision examination conducted by a pediatric developmental optometrist. A binocular eye tracking assessment was performed using the EyeBOX. Binocular pupil gaze positions are captured to ensure quality data acquisition and binocular movements in the X and Y axes are recorded at 500 Hz. Dynamic time series eye tracking data is acquired with 89 different metrics from eye movement in different planes recorded for each eye at 500 Hz captured over 220 seconds. From the time series data, a summary score, between 0 and 20, the BOX score, is generated.
Logistic regression was used to estimate the association between each clinical VVE parameter and concussion-related vision diagnosis. Univariate and multivariate analyses were performed to determine the predictive ability of all VVE parameters to predict a concussion-related vision diagnosis. The area under the receiver-operator (ROC) curve was estimated by treating the linear predictor from the logistic regression as a continuous diagnostic parameter. A p-value is reported for the Wald test of whether a logistic regression coefficient differs from zero.
Results:
The clinical VVE variables had an AUC = 0.64 for tandem gait errors (p=0.06), 0.64 for monocular AA (p=0.01 for each), and 0.66 for NPC (p<0.01). A multivariable model including all nine VVE variables had predictive ability to identify concussion-related vision disorders in adolescents with persistent post-concussion symptoms with an AUC = 0.73. The BOX score alone had an AUC = 0.67 (p<0.01).
Conclusions:
The addition of the BOX score, a composite measure of objectively acquired eye tracking data, to the multivariable model with all nine dichotomous VVE variables added significant additional diagnostic ability (p<0.01), yielding an AUC = 0.81. The combination of a clinical visio-vestibular examination and objective infrared eye tracking has relevant diagnostic value, facilitating identification and treatment of concussion-related vision disorders.