Abstract
Abstract: :
Purpose: To generate an improved patient reliability index for threshold visual field (VF) tests. Introduction: Current VF reliability indices (false positives, negatives and fixation losses) are poor predictors of test–retest variability, accounting for only 5% and 17% of the variability in Full threshold and SITA algorithms respectively. We investigate the value of a spatial filter at deriving an index of reliability. Methods: A simulation program, in which patient response variability was altered (0–200% of normal values), generated VF data from 1) a normal eye and 2) 101 eyes contained within a normal database. The filter–effect variance (SD of the pointwise differences between raw and spatially filtered threshold data) was calculated for each generated result using a customized spatial filter developed by Gardiner and Crabb. The value of the filter at predicting patient variability was further tested on a population of 148 eyes with varying degrees of glaucomatous VF loss. In this population, test–retest variability was measured with a technique that compensated for the relationship between variability and threshold. Stepwise multiple linear regression analysis was used to analyze the relationship between variability (simulated and test–retest) and 3 exposure variables: filter–effect variance; mean deviation (MD) and pattern standard deviation (PSD). Results: Filter effect variance identified 78% (p<0.001) and 54% (p<0.001) of the variability in the simulated normal eye and 101 eyes within the normal database, respectively. However, filter effect variance, as a single parameter, only accounted for 10% (p<0.001) of the variability in the eyes with glaucomatous VF loss. The ability of the filter to identify variability declined with severity of VF loss, e.g. 32%, 28% and 15% for data truncated at MD greater than or equal to –2.5, –3.0 and –3.5, respectively. Multiple linear regression analysis, using all three exposure variables, demonstrated the importance of MD and PSD on variability with more advanced VF loss. Conclusions: Filter effect variance is a good predictor of patient variability in eyes with little VF loss and, in these cases, provides a better index of reliability than the traditional indices (false positives, negatives and fixation loss). Filter effect variance, unlike other indices, has no test time overhead.
Keywords: visual fields • perimetry