Models of glaucomatous visual field loss

Area of Science:

• Ophthalmology
• Medical Statistics
• Glaucoma Research

Background:

• Glaucoma is a progressive optic neuropathy characterized by visual field (VF) loss.
• Accurate modeling of longitudinal VF data is crucial for predicting disease progression and managing glaucoma patients.
• Various mathematical functions are used to model VF changes, but their predictive accuracy varies.

Purpose of the Study:

• To compare the efficacy of linear, exponential, and logistic functions in modeling longitudinal visual field data in glaucoma.
• To evaluate the predictive performance of these models for future visual field loss.
• To assess the utility of a multiple-model approach versus single models for glaucoma progression prediction.

Main Methods:

• Longitudinal VF series from 782 eyes of glaucoma patients with over 6 years of follow-up were analyzed.
• Threshold sensitivities from the first 5 years were regressed using linear, exponential, and logistic functions.
• Model parameters were estimated, and root mean square error (RMSE) was used to assess fit and prediction accuracy at 1, 2, 3, and 5 years.

Main Results:

• The logistic function demonstrated the lowest average RMSE for model fit.
• The exponential model consistently yielded the lowest average prediction RMSE across all time intervals.
• A multiple-model approach did not outperform the best-performing single model (exponential) and showed higher prediction error.

Conclusions:

• While the logistic model best fits long-term glaucomatous visual field behavior, the exponential model offers superior predictive accuracy for future VF loss.
• Goodness of fit does not correlate with predictive ability in glaucomatous VF measurements.
• The exponential model is recommended for predicting visual field progression in glaucoma.