Purpose
We recently developed and validated a model to objectively predict rates of visual field (VF) progression in patients with treated glaucoma. Despite its good performance in predicting final mean deviation (MD), the model does not take into account the localized nature of glaucomatous disease and progression, and hence is unable to predict sectorial VF deterioration. We aimed to enhance our prediction model by adding VF topographical information and validate it in a subset of glaucoma patients.
Methods
Data from 367 eyes with treated open-angle glaucoma (reference population) was used to develop the topographic model. All patients had at least 8 reliable VF SITA-Standard tests. Pointwise linear regression analysis was used to calculate the global and localized rates of VF sensitivity change. For VF topographic analysis, we used 6 VF sectors from the Garway-Heath (GH) structure-function map. Multivariable analysis, adjusted for follow-up time, was performed including known risk factors for glaucomatous VF progression. For each VF sector, we obtained one equation to predict sectorial rates of VF progression (decibels [dB]/year). To validate the model, each equation was tested on an independent validation population (50 eyes, 50 glaucoma patients) with similar characteristics as the reference population. Our outcome measures were the difference and correlation between predicted and observed rates of progression for each GH VF sector.
Results
There was a moderate correlation between predicted and observed topographic rates of progression (Table 1). The mean (95%CI) difference between observed and predicted values were -0.379 (-0.3092 to 0.158), 0.2247 (0.026 to 0.4234), 0.00494 (-0.1681 to 0.1780), 0.2851 (0.1271 to 0.4431), 0.1996 (0.04488 to 0.3544) and 0.05057 (-0.08542 to 0.1866) in the superior, superior-nasal, central, temporal, inferior-nasal and inferior sectors, respectively (all r ≥0.40; P <0.05). The parameter estimates of each variable differed based on the VF sector (Table 2).
Conclusions
Our enhanced prediction model is accurate in estimating glaucomatous progression in different VF sectors. This model may be helpful for clinicians in assessing areas of greater risk of progression and estimating the rate of progression in each VF sector as opposed to global indices.
Keywords: 758 visual fields •
642 perimetry •
473 computational modeling