Multivariable Logistic Regression Model: A Novel Mathematical Model that Predicts Visual Field Sensitivity from Macular Ganglion Cell Complex Thickness in Glaucoma

Purpose: To design a mathematical model that can predict the relationship between the ganglion cell complex (GCC) thickness and visual field sensitivity (VFS) in glaucoma patients. Design: Retrospective cross-sectional case series. Method: Within 3 months from VFS measurements by the Humphrey field analyzer 10-2 program, 83 eyes underwent macular GCC thickness measurements by spectral-domain optical coherence tomography (SD-OCT). Data were used to construct a multiple logistic model that depicted the relationship between the explanatory variables (GCC thickness, age, sex, and spherical equivalent of refractive errors) determined by a regression analysis and the mean VFS corresponding to the SD-OCT scanned area. Analyses were performed in half or 8 segmented local areas as well as in whole scanned areas. A simple logistic model that included GCC thickness as the single explanatory variable was also constructed. The ability of the logistic models to depict the real GCC thickness/VFS in SAP distribution was analyzed by the x2 test of goodness-of-fit. The significance of the model effect was analyzed by analysis of variance (ANOVA). Results: Scatter plots between the GCC thickness and the mean VFS showed sigmoid curves. The x2 test of goodness-of-fit revealed that the multiple logistic models showed a good fit for the real GCC thickness/VFS distribution in all areas except the nasal-inferior-outer area. ANOVA revealed that all of the multiple logistic models significantly predicted the VFS base