Comparison of Methods for Visual Field Progression in Eyes With Central Visual Field Defects.

Purpose

To investigate the agreement of various criteria for visual field (VF) progression in eyes with central VF defects, and to evaluate their performance in simulation datasets with and without age-related and glaucomatous change.

Methods

A total of 282 eyes of 197 primary open-angle glaucoma patients with 10-2 central VF defect at baseline with two or more years' follow-up and five or more visits for both 10-2 and 24-2 VF were included. Various progression detection methods were used: 10-2 clustered pointwise linear regression (PLR), 10-2 VF mean deviation (MD), 24-2 central VF mean total deviation (MTD), 24-2 VF MD, 24-2 PLR, guided progression analysis, Advanced Glaucoma Intervention Study, and Collaborative Initial Glaucoma Treatment Study scores. Progression was defined as a binary outcome at the final visit: ≤-0.7 dB/year for 10-2 VF MD and ≤-0.5 or ≤-1.0 dB/year for 24-2 central VF MTD. Pairwise agreements were evaluated using Cohen's kappa. To further assess the detection performance under controlled conditions, two simulation datasets were constructed: one incorporating realistic progression and another with no true change. The t-statistics from ordinary least squares regression were used to compute receiver operating characteristic curves and normalized partial area under the curves.

Results

Central progression was more frequently detected with 10-2 VF: 10-2 VF MD (35.1%) and 10-2 clustered PLR (20.6%) versus 24-2 central VF MTD at ≤-0.5 dB/y (17.7%) and ≤-1.0 dB/y (3.2%). Global progression was observed in 17.7% to 30.5%. The agreement among methods ranged from 67.0% to 85.1%, with kappas values of 0.11-0.25 between 10-2 and 24-2 MTD methods and 0.22-0.54 between 10-2 and 24-2 methods. Simulation analyses confirmed that 10-2 VF MD had the highest partial AUC across specificity levels.

Conclusions

Agreement among methods for central VF progression monitoring is low to moderate. Concordance between 24-2 and 10-2 VF methods is variable, with 10-2 detecting a higher proportion of central progression.

Translational relevance: Incorporating 10-2 VF testing alongside 24-2 is essential, as relying solely on 24-2 VF may underestimate central VF progression.