Development and validation of an associative model for the detection of glaucoma using pupillography

Abstract

Purpose: To develop and validate an associative model using pupillography that best discriminates those with and without glaucoma.

Design: A prospective case-control study.

Methods: We enrolled 148 patients with glaucoma (mean age 67 ± 11) and 71 controls (mean age 60 ± 10) in a clinical setting. This prototype pupillometer is designed to record and analyze pupillary responses at multiple, controlled stimulus intensities while using varied stimulus patterns and colors. We evaluated three approaches: (1) comparing the responses between the two eyes; (2) comparing responses to stimuli between the superonasal and inferonasal fields within each eye; and (3) calculating the absolute pupil response of each individual eye. Associative models were developed using stepwise regression or forward selection with Akaike information criterion and validated by fivefold cross-validation. We assessed the associative model using sensitivity, specificity and the area-under-the-receiver operating characteristic curve.

Results: Persons with glaucoma had more asymmetric pupil responses in the two eyes (P < 0.001); between superonasal and inferonasal visual field within the same eye (P = 0.014); and smaller amplitudes, slower velocities and longer latencies of pupil responses compared to controls (all P < 0.001). A model including age and these three components resulted in an area-under-the-receiver operating characteristic curve of 0.87 (95% CI 0.83 to 0.92) with 80% sensitivity and specificity in detecting glaucoma. This result remained robust after cross-validation.

Conclusions: Using pupillography, we were able to discriminate among persons with glaucoma and those with normal eye examinations. With refinement, pupil testing may provide a simple approach for glaucoma screening.