Multicenter Normative Data for Mesopic Microperimetry.

Area of Science:

• Ophthalmology
• Visual Neuroscience
• Biostatistics

Background:

• Microperimetry is crucial for assessing visual function.
• Establishing normative data is essential for accurate interpretation of microperimetry results.
• Existing models for predicting visual sensitivity may require refinement.

Purpose of the Study:

• To establish a large, multi-center normative dataset for the Macular Integrity Assessment (MAIA) microperimeter.
• To compare the goodness-of-fit and prediction accuracy of various hill-of-vision models.
• To identify the most effective model for predicting normative visual sensitivity.

Main Methods:

• Included microperimetry data from 1137 tests on 531 healthy eyes across multiple study groups.
• Employed linear mixed models (LMMs) to develop interpretable hill-of-vision models.
• Utilized cross-validation with site-wise splits to compare regression models, evaluating Mean Absolute Error (MAE) and miscalibration area.

Main Results:

• A Bayesian mixed model showed the lowest MAE (2.13 dB) and miscalibration area (0.13).
• A parsimonious linear model demonstrated comparable performance with an MAE of 2.17 dB and miscalibration area of 0.14.
• Both models effectively predicted normative visual sensitivity based on age and eccentricity.

Conclusions:

• Normal variations in mesopic microperimetry visual sensitivity are well-explained by a linear model incorporating age and eccentricity.
• The developed dataset and code vignette facilitate the estimation of normative values for customized microperimeter testing patterns.
• The findings support the use of simplified models for understanding visual sensitivity across diverse retinal locations.