Detection of Longitudinal Ganglion Cell/Inner Plexiform Layer Change: Comparison of Two Spectral-Domain Optical Coherence Tomography Devices

Abstract

Purpose: We compared rates of change of macular ganglion cell/inner plexiform (GCIPL) thickness and proportion of worsening and improving rates from 2 optical coherence tomography (OCT) devices in a cohort of eyes with glaucoma.

Design: Longitudinal cohort study.

Methods: In a tertiary glaucoma clinic we evaluated 68 glaucoma eyes with ≥2 years of follow-up and ≥4 OCT images. Macular volume scans from 2 OCT devices were exported, coregistered, and segmented. Global and sectoral GCIPL data from the central 4.8 × 4.0-mm region were extracted. GCIPL rates of change were estimated with linear regression. Permutation analyses were used to control specificity with the 2.5 percentile cutoff point used to define "true" worsening. Main outcome measures included differences in global/sectoral GCIPL rates of change between 2 OCT devices and the proportion of negative vs positive rates of change (P < .05).

Results: Average (standard deviation) 24-2 visual field mean deviation, median (interquartile range) follow-up time, and number of OCT images were -9.4 (6.1) dB, 3.8 (3.3-4.2) years, and 6 (5-8), respectively. GCIPL rates of thinning from Spectralis OCT were faster (more negative) compared with Cirrus OCT; differences were significant in superonasal (P = .03) and superotemporal (P = .04) sectors. A higher proportion of significant negative rates was observed with Spectralis OCT both globally and in inferotemporal/superotemporal sectors (P < .04). Permutation analyses confirmed the higher proportion of global and sectoral negative rates of change with Spectralis OCT (P < .001).

Conclusions: Changes in macular GCIPL were detected more frequently on Spectralis' longitudinal volume scans than those of Cirrus OCT. OCT devices are not interchangeable with regard to detection of macular structural progression.

FIGURE 1.

The region of interest from Cirrus and Spectralis optical coherence tomography macular volume scans consisted of a 4.8- × 4.0-mm ellipse centered on the fovea excluding a central foveal region 1.2 mm × 1.0 mm in size. The ganglion cell/inner plexiform layer thickness measurements within this ring-shaped region were divided into 6 pie-shaped sectors.

FIGURE 2.

Boxplot showing the distribution of global and sectoral ganglion cell/inner plexiform layer (GCIPL) rates of change for Spectralis and Cirrus optical coherence tomography (OCT) scans. The Spectralis OCT median rates of change were significantly lower than those of Cirrus OCT with all the pairwise differences being significant except for the inferonasal, inferotemporal, and superior sectors. G = global; I = inferior; IN = inferonasal; IT = inferotemporal; S = superior; SN = superonasal; ST = superotemporal.

FIGURE 3.

Comparison of the proportion of eyes with significant negative and positive ganglion cell/inner plexiform layer rates of change at the end of the follow-up period. Spectralis optical coherence tomography measurements detected a higher proportion of worsening in the inferotemporal (P = .001), superotemporal (P = .03), and superior (P = .03) sectors as well as globally (P = .01). The proportion of significant positive rates was small and varied between the 2 optical coherence tomography devices. G = global; I = inferior; IN = inferonasal; IT = inferotemporal; S = superior; SN = superonasal; ST = superotemporal.

FIGURE 4.

Bar graph shows the proportion of worsening macular ganglion cell/inner plexiform layer rates of change globally and for 6 sectors for Spectralis and Cirrus optical coherence tomography (OCT) scans based on permutation analyses. The 2.5 percentile cutoff point was used to define “true” worsening. Spectralis OCT measurements detected higher rates of worsening both globally and in all sectors compared with Cirrus OCT. G = global; I = inferior; IN = inferonasal; IT = inferotemporal; S = superior; SN = superonasal; ST = superotemporal.

FIGURE 5.

Bar graphs show root mean square error (RSME) measurements for regression models of global ganglion cell/inner plexiform layer thickness against time for Cirrus and Spectralis optical coherence tomography (OCT) devices. The RMSE measurements were lower for Spectralis OCT. The global RMSE was 1.59 (2.47) for Cirrus OCT vs 0.61 (0.46) for Spectralis OCT (P < .001).

FIGURE 6.

Scatterplots show global rates of ganglion cell/inner plexiform layer (GCIPL) change with Spectralis optical coherence tomography against global rates of change with Cirrus optical coherence tomography. The dashed line represents a spline fit.