Evaluation of the t test as a method of detecting visual field changes

Automated perimetry is used to monitor changes associated with many ocular conditions such as glaucoma, neuro-ophthalmologic disorders, and retinal diseases. Comparisons among sequential visual fields are crucial to the determination of progression or regression of ocular anomalies, and to evaluate the efficacy of various therapeutic regimens. The Delta program for the Octopus perimeter uses a t test to provide an objective comparison of visual fields. To evaluate the utility of the t test for detecting visual field changes, the authors used computer simulation to generate pairs of visual fields with known alterations, and made comparisons with the t test procedure used by the Delta program. An initial visual field based on normal data was produced and stored. A second visual field, based on the same normal data but with a scotoma of known size and depth, was also produced and stored. A t test (as used by the Delta program) was then used to compare the two visual fields for statistically significant differences. Our results indicate that the t test was very sensitive to small changes of the entire visual field, such as those produced by both pathologic and nonpathologic (day-to-day sensitivity variations, changes in pupil size, etc.) factors. However, the t test was unable to reliably detect small to moderate scotomata (less than or equal to 18 degrees in diameter), even when the sensitivity loss was greater than 35 dB. These findings suggest that the t test has limited clinical utility for objective detection of pathologic visual field loss, and that additional procedures may be necessary to provide reliable detection of visual field changes.