Comparison of summary measures for analyzing longitudinal visual field data

Purpose: To devise a visual field (VF) summary measure for use in analyzing longitudinal VF data. Four possible measures were evaluated for their prediction by demographic and clinical variables. Methods: Humphrey VF results from an ongoing 2-yr, multi-center, open-label, single arm safety study of dorzolamide HCl were used to generate four summary measures: average (AVE) of the VF threshold values, the coefficient of variation of VF (CV= standard deviation/AVE), the Corrected Pattern Standard Deviation (CPSD) score, and the Advanced Glaucoma Intervention Study (AGIS) score. Multiple linear regression techniques were used to determine which of the following variables were significant predictors of each summary measure: patient age, investigator, study visit, sex, race, iris color, lens status,disease status (ocular hypertension vs glaucoma), IOP (baseline and concurrent), study eye (right eyes were always tested first), VF program (24-2 vs 30-2). Results: VF data was obtained from 302 of 303 study patients; 77% of patients continued in the study for at least 12 months. Iris color, VF program, and study visit were not significant predictors. Age, race, disease status, investigator and study eye were significant (p<0.05) for all 4 measures. Baseline IOP was significant for all measures but CPSD and concurrent IOP was significant only for CV. Sex was significant only for CPSD. Lens status was significant only for AVE. Conclusions: There was good agreement between measures. Increasing age, Afro-American race, and increasing baseline IOP, three accepted risk factors for VF damage, were significant predictors of poorer AVE, CV, AGIS scores. Of these 3 summary measures, lens status, a potential confounder, was not significant for CV and AGIS. These four summary measures are currently being used to evaluate the VF results from a randomized, comparative clinical trial.