Abstract
Purpose: :
To examine the ability of Spatial Frequency Doubling (FD) Illusion stimuli to predict progression of visual field loss and optic disc abnormalities in 218 subjects.
Methods: :
Our original 1991–92 study of 330 subjects determined contrast thresholds for several FD stimuli in up to 7 quarterly visits [Maddess et al. (1999) Vision Res., 39, 4258]. For this study we followed subjects for 12 yr for changes in: visual fields as measured by the Humphrey Field Analyser (HFA), and optic disc appearance. We then compared the diagnostic power of 4 FD hemifield thresholds, and 4 HFA measures, to predict progression of HFA visual fields and vertical cup/disc ratios (CDRs). The 4 HFA measures were the Mean Defects (MD) and Pattern Standard Deviations (PSD), which are derived from 52 thresholds/eye. Diagnostic ability was determined from Receiver Operator Characteristic (ROC) plots, which were constructed from linear discriminant classifier models. Bootstrap estimates of means and SD were obtained for all ROC plots and for ROC parameters such as d–prime, ROC area (ROCA), and the so called accuracy (the mean of the simultaneously highest sensitivity and specificity). The bootstrap also provided a degree of cross–validation. Visual field and optic disc progression was examined in weak, moderate and strong, glaucoma suspects, patients, and normals. Classification was based upon either the 4 FD thresholds or the 4 HFA measures per subject.
Results: :
Accuracies were less problematic than d–prime or ROCA. For visual field progression accuracies of 90 to 100% were achieved for the first 3 years but then accuracies fell to 70% or better for FD stimuli after 12 years in the moderate and strong suspect groups. MD and PSD performed relatively poorly. Both FD and HFA measures predicted CDR progression at about 80% accuracy in the moderate suspects, the FD measures performed better than HFA for glaucoma patients. Both FD and HFA discriminated HFA visual field progressors from non–progressors at 60% to 80% even after 12 years, the rate depending on patient group.
Conclusions: :
For the first 3 years the predictive power of the FD thresholds was excellent. The longer term prediction was still good enough to assist in decisions about which patients might require more aggressive treatment. The drop–off in predictive power over time observed for both FD and HFA measures suggests that glaucoma is not a deterministic process, perhaps being driven by environmental or patient variables that are not predictable from initial conditions.
Keywords: perimetry • optic disc • visual fields