Dual-Level Pattern Tree for Visual Field Improves Glaucoma Progression and Polygenic Risk Prediction

Area of Science:

• Ophthalmology
• Genetics
• Data Science

Background:

• Glaucoma is a leading cause of irreversible blindness worldwide.
• Accurate characterization of visual field (VF) loss is crucial for predicting glaucoma progression.
• Existing methods may not fully capture the complexity of VF loss patterns.

Purpose of the Study:

• To develop a dual-level pattern tree for classifying VF loss subtypes.
• To enhance the prediction of glaucoma progression using these subtypes.
• To correlate VF patterns with glaucoma polygenic risk scores (PRSs).

Main Methods:

• Archetypal analysis was used to cluster 24-2 VFs into trunk and branch patterns.
• Cox regression models forecasted VF progression using various slope metrics.
• Multivariable regression analyses linked VF patterns with glaucoma PRSs.

Main Results:

• Identified 17 trunk patterns and 169 branch patterns, forming trunk-branch (T-B) classifications.
• T-B patterns significantly outperformed trunk-only patterns in predicting 5-year progression across multiple metrics.
• T-B patterns demonstrated superior predictive capability for glaucoma PRSs.

Conclusions:

• Dual-level trunk-branch VF classifiers are more effective than trunk-only classifications for predicting functional progression and glaucoma PRS.
• This approach offers a more nuanced understanding of glaucoma subtypes.
• Improved VF characterization can aid in personalized glaucoma management and research.