Visual Field Progression in Glaucoma: Comparison Between PoPLR and ANSWERS

Area of Science:

• Ophthalmology
• Medical Statistics
• Data Analysis

Background:

• Detecting visual field progression in glaucoma is crucial for timely intervention.
• Statistical modeling of visual field data, considering retest variability and spatial correlations, may enhance detection accuracy.
• Existing methods vary in complexity and computational demands.

Purpose of the Study:

• To compare the performance of a complex statistical model, Analysis with Non-Stationary Weibull Error Regression and Spatial Enhancement (ANSWERS), with a simpler method, Permutation of Pointwise Linear Regression (PoPLR), for detecting visual field progression in glaucoma patients.
• To evaluate the positive rates and false-positive rates of both methods across different datasets.

Main Methods:

• Three independent longitudinal glaucoma datasets were utilized.
• The study compared the positive rates of ANSWERS and PoPLR.
• False-positive rates were estimated by randomly re-ordering visual field series in time.
• Specificities of both models were equated to 0.05 for direct comparison of positive rates.

Main Results:

• For a statistical significance of 0.05, ANSWERS exhibited false-positive rates approximately three times greater than expected (15-17%), while PoPLR rates were lower (3-7%).
• After equating specificities, PoPLR showed positive rates that were, on average, 5% greater than ANSWERS (12-49% vs. 16-38%).

Conclusions:

• The simpler PoPLR method demonstrated comparable sensitivity to the more complex ANSWERS method in detecting visual field deterioration, particularly after equating specificities.
• PoPLR is less computationally demanding and offers an efficient alternative for analyzing visual field changes.
• Accurate control of false-positive rates is essential for reliable visual field analysis in clinical practice and trials.