Comparison of within-eye and between-individual variability in macular structure-function relationships

Abstract

Purpose : Significant variability is observed when structure-function (SF) relationships are examined using cross-sectional data. However, respective contributions of within-eye vs. between-individual variability components are not well defined. We tested the hypothesis that within-eye variability in SF relationships between macular thickness and central visual field (VF) sensitivities are lower than between-individual variability.

Methods : Glaucoma patients with macular SD-OCT images (Spectralis’ Posterior Pole Algorithm) and central 10-2 VF on ≥3 visits were enrolled from a longitudinal study. 57 eyes (57 patients) in which macular and total deviation (TD) values varied through their entire range were selected. Macular images were segmented and ganglion cell/inner plexiform layer (GCIPL) thicknesses were exported as horizontal 8x8 arrays of 3° superpixels. VF locations were adjusted for ganglion cell displacement (Figure 1) and bivariate SF plots were drawn for GCIPL thickness in 24 central superpixels and corresponding TD values. Mixed-effects regression analyses were performed on data repeated over time where SF relationships were modeled with fixed effects (broken stick model or transformed variables) while within-eye and between-individual variability were estimated from the variance of residual errors and random intercepts, respectively, and their comparisons were expressed as intraclass correlation coefficients (ICC).

Results : Average (±SD) mean deviation of 10-2 VFs was –9.7 (±4.9) dB. A broken stick model best fit the data when TD values were expressed in dB whereas a linear fit modeled the data best with TD values in 1/L units. Regardless of model or change point chosen for the broken stick model (–8 dB based on Miraftabi et al. 2016 or –5.5 dB based on data), magnitude of within-eye variability was consistently larger than between-eye variability (σ²_BE =0.036 vs. σ²_WE =0.068 for the model with highest ICC using data at peak ganglion cell density or 5.4° eccentricity). ICCs ranged from 0.173 to 0.345.

Conclusions : Within-eye variability of macular/central field SF models were larger than between-individual variability. This finding seems to be at least partially related to measurement noise and the wide range of macular thickness before early functional loss can be detected. Eye-specific SF models may not reduce variability of models predicting functional from structural measurements.

This is an abstract that was submitted for the 2017 ARVO Annual Meeting, held in Baltimore, MD, May 7-11, 2017.