Between-Subject Differences Account for More Dissociation Between Rim Area and Visual Sensitivity Than Within-Subject Fluctuations

Abstract

 

Purpose

 

The aim of this study was to understand better the dissociation between rim area and visual sensitivity. Under the assumption that the true relationship is linear, we tested the hypothesis that between-subject differences account for more dissociation than within-subject fluctuations.

 

Methods

 

We tested the hypothesis independently for the superior-temporal (ST) and the inferior-temporal (IT) sectors. Twelve rim area and visual field examinations were taken for each of 30 eyes of 30 patients with glaucoma followed up over a period of 2 months [Artes et al ARVO 2011, #4148]. We assumed there was no progression. Rim area was expressed as percent normal mm² and sensitivity as percent normal 1 / L. Mean normal values were obtained for 173 visits for 77 eyes of 77 healthy controls. A linear model was fitted that accounted for noise in both sensitivity and rim area. For each patient, we calculated the centroid of the cloud of the 12 recorded rim areas and sensitivities. We corrected for between-subject differences by projecting each patient cloud onto the fitted line (see Figure). For both ST and IT, R² values were obtained for: all 360 datapoints (all variability), the 30 centroids (between-subject differences), and the 360 datapoints after correcting for between-subject differences (within-subject fluctuations).

 

Results

 

For ST, the R² values (and 95% confidence interval) were 23% (17% to 30%) for all variability, 26% (5% to 50%) for between-subject differences, and 86% (83% to 89%) for within-subject fluctuations. For IT, R² values were, respectively, 40% (33% to 47%), 45% (19% to 67%), and 87% (84% to 90%). The R² values were significantly larger for within-subject fluctuations in both ST and IT (with α = 0.05). Adjusting rim area for sector area with quantile regression had no significant effect.

 

Conclusions

 

If the true relationship between rim area and visual sensitivity is linear, between-subject differences account for more dissociation than within-subject fluctuations. This result emphasizes the potential of individualized tests and structure-function maps.

 

 

Correcting for between-subject differences. Twelve datapoints (open black circles) and corresponding centroids (solid black circles) are shown for each of 4 patients chosen for illustration, as there was little or no overlap between their data clouds. The corresponding projected clouds are shown in gray.

 

Keywords: 550 imaging/image analysis: clinical • 642 perimetry