Normative variability in retinal nerve fiber layer thickness: Does it matter where the peaks are?

Abstract

Purpose : Retinal nerve fiber layer thickness (RNFLT) is an important biomarker for glaucoma, but it has a wide normative range affecting its sensitivity and specificity for abnormality detection. Many think that the inter-individual peak location variability causes this wide normative range. The purpose of this study was 1) to assess the effect of RNFLT peak normalization on normative variability and 2) to seek a structural representation of the RNFLT peaks, using retinal blood vessels (BVs) information.

Methods : 163 optical coherence tomography scans (Cirrus HD-OCT, 200x200 optic nerve head (ONH) scans, Zeiss, Dublin, CA) from 101 normal individuals were used. For each scan, a circumpapillary RNFLT profile at 1.7 mm radius circle around the ONH was obtained. Fovea-ONH axis (FOA) was calculated from corresponding scanning laser ophthalmoscope (SLO) images. Supra-temporal (ST) and infra-temporal (IT) RNFLT peaks from each profile were aligned to respective average peak locations. Normative ranges were calculated by averaging individual profiles before and after peak normalization. Mean BVs location (based on segmented vessels in ST and IT regions separately) and location of major BV arcades (manually annotated) were obtained at the 1.7 mm radius and correlated with the corresponding RNFLT peaks. All methods were also performed after alignment of FOA with the horizontal axis (HA).

Results : RNFLT peak normalization leads to localized reduction in variability around the peak locations, especially in the ST region (clock hour (CH) 11, p<0.05, Levene’s test) (Figure 1). Aligning FOA to HA prior to peak normalization further reduces variability in CHs 10, 11 and 12 (p<0.05). Of the two BV related measures, mean BVs locations in the ST and IT regions were better correlated (r = -0.51 and -0.61, respectively, p <0.05) with the respective RNFLT peak locations compared to the superior and inferior BV arcade locations (r = 0.03 and 0.15, respectively) (Figure2). Aligning FOA to HA prior to peak normalization did not improve correlations between BV measures and RNFLT peak locations.

Conclusions : RNFLT peak normalization has region specific effects on reducing variability in the normative range, with greater reduction in the superior compared to inferior hemifield. Mean BV location in each hemifield may be of utility in RNFLT peak normalization compared to the BV arcade locations.

This abstract was presented at the 2024 ARVO Annual Meeting, held in Seattle, WA, May 5-9, 2024.