Clinical Correlates of Computationally Derived Visual Field Defect Archetypes in Patients From a Glaucoma Clinic

Abstract

Purpose: Glaucoma diagnosis and monitoring would benefit from a quantitative visual field (VF) classification system like VF archetypal analysis, where Elze et al. showed that any VF can be represented as a weighted sum of 17 archetype (AT) patterns. We assessed the clinical construct validity of this system by testing for clinical correlates of VF ATs. Methods: From 30,995 reliable Humphrey VFs (24-2), 243 patients were retrospectively selected, comprising the 10 to 20 patients whose VFs had the highest decomposition coefficients for each AT. Corresponding patient systemic and ocular characteristics were obtained and compared between each AT and all others using the two-tailed t-test or Fisher exact test. Results: Mean deviation and pattern standard deviation varied with VF loss severity and focality respectively. Cup-to-disc ratio was more often ≥0.7 for AT6 (central island; P=0.002), AT14 (superior paracentral defect; P=0.016), and AT16 (inferior paracentral defect; P=0.016). Other associations included: AT6 (central island): African ancestry (P<0.001) and younger age (P<0.0001); AT13 (diffuse inferior defect): African ancestry (P=0.006) and chronic angle closure glaucoma (P=0.005); and AT16 (inferior paracentral defect): female sex (P=0.017). Conclusions: Several expected clinical associations support computationally derived VF ATs’ clinical construct validity. New associations identified by exploratory analysis may merit further investigation.