Performance of glaucoma progression analysis software in a glaucoma population

Abstract

To compare visual field progression in glaucoma patients assessed by the glaucoma progression analysis (GPA) used in the Humphrey Field Analyzer perimeter and by objective clinical criteria. Retrospective cross-sectional study of 93 eyes of 93 consecutive glaucoma patients with at least five reliable visual fields. Progression of the visual field damage was analyzed by a masked observer using both GPA and defined clinical criteria. Prevalence of progressive visual field damage was determined by clinical criteria and GPA analysis. Agreement between both methods was quantified by kappa analysis. GPA performance was also calculated using clinical criteria analysis as the reference standard. The prevalence of progressive visual field damage was 30% and 29% with GPA evaluation and clinical criteria analysis respectively. Where two consecutive visual fields showing progressive damage were needed to confirm progression, the kappa index of agreement between these two approaches was 0.87 +/- 0.06 (mean +/- standard error of the mean). Where three consecutive fields showing progressive damage were required, the kappa index of agreement was 0.64 +/- 0.1. The GPA performance showed a sensitivity and specificity of 93% (95% CI, 83-100%) and 95% (95% CI, 90-100%) respectively, and a positive likelihood ratio of 20 if the two consecutive visual fields criterion was used. The performance was worse if three consecutive progressing visual fields were required to confirm progression. There is a strong correlation between GPA identification of glaucomatous progression and a thorough objective clinical assessment of the visual fields. GPA could be a useful test to aid clinicians in the detection of glaucomatous progression, with high specificity, strong positive likelihood ratio, and good sensitivity and negative likelihood ratio.