Abstract
Purpose:
Evaluation of a nonlinear logistic function as a model to represent the process of glaucomatous damage over the entire visual field (VF) range compared to linear and exponential functions.
Methods:
Reliable patient VF data, defined as <30% fixation loss, <30% false positive, and <30% false negative rates, from the Advanced Glaucoma Intervention Study (AGIS) and the Stein Eye Institute’s glaucoma division were used. All test were performed with a Humphrey Field Analyzer with a 24-2 test pattern, size III white stimulus and full threshold strategy or Swedish Interactive Threshold Algorithm (SITA) Standard and SITA Fast. The following functions were used to assess the pattern of threshold sensitivity deterioration at each VF location:<br /> Linear: y = a + bx; Exponential: ln(y) = a + bx; Logistic: y = g / (1 + exp( a + bx). VF locations of interest include those with an average of the initial two sensitivities greater than 30 dB, 26 dB, and 22 dB and with an average of the final two sensitivities less than 10 dB. Root mean squared error (RMSE) values were used to evaluate the goodness-of-fit for each regression model. The error was defined as a difference between the sensitivities predicted by the function and the observed sensitivities.
Results:
798 eyes from 583 patients were included. Average (±SD) follow-up time was 8.7 (± 2.2) years, and each eye had an average of 15.2 (±4.9) VF tests. For the VF locations with an initial sensitivity greater than 22 dB and final sensitivity less than 10 dB (938 locations), the logistic function had the lowest RMSE in 73.1% of the locations, the exponential function in 18.2%, and the linear function in 8.7%. This pattern held true for the subset of points with an average initial sensitivity greater than 26 dB and greater than 30 dB with a final sensitivity less than 10 dB.
Conclusions:
A pointwise logistic regression had the best ability to fit perimetric progression in a subset of locations that traverse the entire range of perimetric measurements from normal to perimetric blindness compared to linear and exponential functions. Our results show that perimetric measurements of glaucomatous visual field loss from early to advanced stages of glaucoma follows a pattern best represented by a logistic function. As such, the behavior of glaucomatous VF loss is nonlinear, and the rate of deterioration changes with the course of the disease.