Abstract
Purpose :
To investigate whether customizing visual field test locations for individuals at the start of a trial enables detection of a larger number of progressing visual fields, compared to a standardized grid.
Methods :
Using a novel data set of high-density visual fields, we simulated a 3 year trial with two baseline then 3 monthly visits. Two test grids were compared: “Regular” a 6x6° grid; and “Custom”. Custom tested the Regular grid at baseline then replaced points with sensitivity <17dB (unlikely to contribute to progression detection due to high variability or already “blind”) with new locations on scotoma edges. For each eye, from visit 3, the individualized custom grid remained the same. To assess progression detection, we created a synthetic series of fields seeded from data collected from 24 eyes with a range of visual field archetypes. Empirical data had 100 points on a 2° spaced square placed on scotoma bordersusing fundus-stabilised perimetry (Compass, iCare, Finland) controlled by the Open Perimetry Interface (Fig 1). We back-calculated progressing field series by adding 0.5, 1 and 2 dB/year to any damaged location, leaving locations with a Total Deviation of zero unchanged (creating series of 24x3 = 72 eyes). We then took the final and penultimate 13 fields of each series (72x2 = 144 eyes). These progressing series were then used as “ground-truth” for simulation of measured fields in the trial. We simulated visual field testing during the trial with two response profiles: reliable (3% FP, 3%FN); unreliable (15% FP/3%FN). Progression was called using the PoPLRmethod (specificity: 95% or 99%). Grids were compared as follows: 1) time to progression using a two-sided paired t-test on the visit number when classified as progressed; 2) number of eyes classified as progressed using a two-sided binomial test on proportions.
Results :
Table 1 shows that the Custom grid consistently found more progressors for all four conditions (column Custom > Grid; p < 0.04 for two conditions). For the eyes that progressed, time to progression was similar for both grids (all p-values > 0.4).
Conclusions :
Selecting an individualized test grid after baseline visual fields at the beginning of a clinical trial finds more progressing eyes, hence will likely increase the power of the trial.
This abstract was presented at the 2023 ARVO Annual Meeting, held in New Orleans, LA, April 23-27, 2023.