Abstract
Purpose: :
To compare the time to initial progression and spatial distribution of progressing test locations with Guided Progression Analysis (GPA) and pointwise linear regression analysis (PLR) on pattern deviation data in eyes progressing on both.
Methods: :
363 eyes (214 patients) from UCSD's Diagnostic Innovations in Glaucoma Study (DIGS) and 928 eyes (582 patients) from UCLA's clinical database were enrolled. Inclusion criteria included a diagnosis of glaucoma or glaucoma suspect, follow-up of 3 years or longer and a minimum of 7 reliable SITA-standard visual fields. EMGT criteria were used for defining progression according to GPA. Presence of 3 or more points with a slope ≤-1.0 dB/year with p ≤0.01 on at least two consecutive regression analyses was required before progression was confirmed with PLR. Time to initial progression and topographic distribution of the worsening test locations were compared in eyes that progressed by both methods.
Results: :
: An average (± SD) of 9.2 (± 2.3) visual fields per eye was available. Mean (± SD) follow-up time was 6.1 (± 1.5) years. Twenty-three eyes (21 with abnormal baseline fields) progressed with both GPA and PLR. The mean (± SD) time to progression was 3.9 (± 1.7) years and 5.2 (± 1.7) years for GPA and PLR, respectively (p <0.001, paired t test). The median (range) number of initially progressing test locations was 3 (3-17) and 4 (3-13) for GPA and PLR (p =0.44, Wilcoxon's signed ranks test). All but one initially progressing test locations detected with GPA (100 of 101 test locations) and 95% of worsening test locations according to PLR (90 out of 95 locations) were either within the baseline defective area (locations with p<5% on pattern deviation plot at baseline) or in its immediate vicinity. About two-thirds of the progressing test locations by either method were detected by the other technique. In the DIGS cohort, GPA and PLR detected progression in 46 (12.7%) vs. 5 (1.4%) out of 363 eyes (p <0.001, McNemar's test). In this subset of patients, all progressing eyes detected by PLR also progressed by GPA.
Conclusions: :
In eyes progressing according to both techniques, GPA detected progression earlier than PLR. A majority of progressing test locations by both techniques either belonged to or immediately surrounded the area of initial field loss. In the DIGS cohort, GPA detected significantly more progression than PLR on pattern deviation data.