Abstract
Purpose :
Correctly classifying disease progression in moderate to advanced glaucoma is difficult. Clinical imaging metrics tend to reach a floor, hence visual field assessment is considered more useful as glaucoma advances. Point-wise visual field test-retest variability is very high for sensitivities below approximately 20dB, hence detecting progression requires many repeated visual fields. We hypothesised that a visual field testing approach that doesn’t attempt to threshold accurately in areas of moderate damage, but instead expends presentations testing more spatial locations, will improve the detection of progression.
Methods :
Our visual field procedure (a variant of the Bayesian procedure ZEST) applies the following approach: once a location has an estimated sensitivity of less than 17dB (a “defect”), it is simply checked that it is not an absolute defect (<0dB, “blind”). The saved presentations are used to test an extra location in the visual field. The location is selected near to the defect (between it and its most normal neighbour), to increase chance of progression detection. This permits visual field deterioration events to be either a) decreasing in the range of 40-17dB; b) decreasing from above 17dB to “defect”, or c) “defect” to blind. To test this approach we utilised a database of moderate-advanced 24-2 visual fields (162 patients) and “reverse engineered” progression via computer simulation so that at least one location per visit decreased by 2dB per visit, with series lengths between 10-18 visits. "Ground-truth" sensitivities between the 24-2 locations were calculated using natural neighbour interpolation.
Results :
With specificity and number of presentations approximately matched, reducing the range of the ZEST from 0:40 to {-1,defect,17:40}, and adding test locations, increases sensitivity to detecting a glaucomatous progression event (at least 2 points decreasing by 4 dB confirmed in 2 tests) in later stage glaucoma from 21% to 66%.
Conclusions :
Spending time trying to accurately measure visual field locations that have high variability is not productive. Our simulations indicate that giving up attempting to quantify size III white-on-white thresholds below 17dB and using the presentations saved to test locations not on the 24-2 grid should better detect progression in moderate to advance glaucoma without increasing test times.
This is an abstract that was submitted for the 2016 ARVO Annual Meeting, held in Seattle, Wash., May 1-5, 2016.