What Does High Variance of Optical Coherence Tomography (OCT) Retinal Nerve Fiber Layer (RNFL) Thickness Measurements Imply?

Abstract

Purpose : OCT glaucoma progression assessment relies on changes in longitudinal RNFL thickness, while its variance has been highly disregarded. We hypothesized that variance, like internal reflectivity, maybe indicative of active neural tissue damage prior to RNFL thinning. The purpose of this study was to investigate if longitudinal RNFL thickness variance is associated with glaucoma progression.

Methods : Longitudinal OCT data (Cirrus HD-OCT; Optic Disc 200x200 scan) on healthy and glaucomatous eyes (minimum 3 visits with 6 months interval) were exported. Healthy subjects were divided into normative data (ND) and individual testing (H) subgroups. Glaucoma subjects were divided into OCT progressers (OCTP, progressing on OCT glaucoma progression analysis (GPA)), visual field progressers (VFP, progression only on Humphrey VF GPA), and non-progressers (NP) subgroups. Longitudinal RNFL variance (normalized to mean RNFL) was calculated for each super-pixel (SP, 4x4 neighboring sampling points) outside of the 3.2 mm diameter circle centered at the optic disc. Two dimensional variance maps for every individual were generated using Euclidean distances of variance from ND at each SP (Fig A). Mean Euclidean distances of SP showing significant distances on the maps were calculated for each subject.

Results : 27 healthy (19 ND, 8 H) and 31 glaucoma subjects (13 OCTP, 8 VFP, 10 NP) were enrolled (Fig B). H and OCTP had lowest and highest variance, respectively. There was no significant difference detected between NP and VFP. But Pearson correlation coefficient between the variance and VF mean deviation (MD) was higher in VFP (0.72) than NP (0.26), seen as distinctive clusters (Fig C). When examining an additional glaucomatous eye showing two phases (NP then VFP), the mean variance was higher in NP than VFP phase (0.16 vs. 0.07).

Conclusions : Variance was lower in VFP than NP at the relatively healthy MD range, which is counter intuitive as active disease status may likely increase variance. We hypothesize that high variance without corresponding VF loss may indicate impending VF progression.

This is an abstract that was submitted for the 2016 ARVO Annual Meeting, held in Seattle, Wash., May 1-5, 2016.