Comparing ganglion cell-inner plexiform layer thickness with focal and global responses on multifocal electroretinogram in glaucoma.

BACKGROUND: The aim of this study was to evaluate responses on multifocal electroretinogram (mfERG) with ganglion cell-inner plexiform layer (GCIPL) thickness on cirrus spectral-domain optical coherence tomography (SD-OCT) in glaucoma.
METHODS: All diagnosed glaucoma patients attending glaucoma services at our institute from November 2012 to April 2013 were screened for this observational hospital-based study. Controls included patients attending our outpatient services for general eye checkup. Structural parameters on SD-OCT including GCIPL and retinal nerve fiber layer (RNFL) thickness were compared with functional parameters on mfERG in early (mean deviation <-6 dB), moderate (-6 to - 12 dB), and controls.
RESULTS: A total of 54 cases and 33 controls fulfilling inclusion criteria were recruited for the study. The average and minimum GCIPL thickness did not vary significantly between early and control eyes while moderate glaucoma eyes had marginally lower GCIPL thickness than early glaucoma eyes. The GCIPL minimum thickness on univariate regression was found to be influenced by N2 amplitudes (β = -0.5, P = 0.012) and global N2P1amplitudes (β =0.6, P = 0.01) in moderate glaucoma. In early glaucoma, these were influenced only by RNFL parameters with no association with functional mfERG responses. Multivariate logistic regression identified global N2P1 amplitude to be significantly influencing GCIPL average and minimum thickness (P = 0.01 and 0.02, R2 = 47.8% and 52.3%, respectively) in moderate glaucoma. Maximum area under the curve was found for GCIPL minimum (95% confidence interval [CI] 0.53-0.81) and N2P1 amplitude (95% CI 0.55-0.80).
CONCLUSIONS: The second order responses N2P1 and N2 amplitude on mfERG predict function that correlated with structural GCIPL thickness in moderate glaucoma. Early glaucoma may be best predicted by RNFL thickness rather than on mfERG responses.

Introduction

The visual fields (VFs) remain the gold standard for assessing function in glaucoma though it is known that it falls short of reflecting true function in very early or severe glaucoma.[123] These have urged clinicians to resort to structural imaging analyzing retinal nerve fiber layer (RNFL) thickness for assessing early structural loss rather than functional damage.[124] Several imaging technologies have ushered in different structural imaging parameters for detecting early structural loss including the spectral-domain optical coherence tomography (SD-OCT).[567] Studies have now focused on the macular ganglion cell-inner plexiform layer (GCIPL) to reflect changes in the area with maximum ganglion cell concentration affected early in the disease.[8] It is unclear if differential rate of loss of ganglion cells at different stages changes the structure-function relationship which is known to be different in early, moderate, or severe glaucoma. While studies have demonstrated a nonlinear relation between structure and function, topographical relationship in different stages of the disease, and utility of multifocal electroretinogram (mfERG) in revealing the structure-function relationship is unclear.[910]

mfERG can provide functions in different retinal locations with region-specific focal responses which allows a direct assessment of structure-function relation.[1112] Several studies have examined the relationships between local mfERG functional responses with structure in specific locations and have reported its utility in screening for glaucoma.[1113] This study attempts to evaluate the change in structure-function relationship between GCIPL thickness with functional mfERG responses in different stages of glaucoma.

Methods

All diagnosed glaucoma patients attending glaucoma services at our institute from November 2012 to April 2013 were screened for this observational hospital-based study which was approved by the Institutional Review Board of LV Prasad Eye Institute, Bhubaneswar, India. Controls included patients attending our outpatient services for general eye checkup. A detailed written informed consent was taken from all patients included in the study which followed the tenets of the Declaration of Helsinki. Comprehensive assessment for all patients included best-corrected visual acuity, refractive error, slit-lamp examination, Goldmann applanation tonometry, 4 mirror gonioscopy, +90 D fundus biomicroscopy, VFs, and nonstereoscopic fundus image acquisition by a single optometrist (SM).

Standard-automated perimetry was done using white-on-white stimulus and 24-2 SITA standard program in all patients by a single optometrist (DP) with appropriate refractive correction (Visual field analyzer Carl Zeiss Meditec, Dublin, CA, USA). Glaucoma was defined in the presence of VF defects confirmed by at least two reliable examinations with corresponding changes on optic disc as diffuse or focal neural rim thinning and/or RNFL defects. A glaucomatous VF defect was defined as a cluster of three or more nonedge points with a probability <5% on the pattern deviation map in at least one hemifield with at least one point with a probability <1%, pattern standard deviation (PSD) with a probability <5%, and glaucoma hemifield test “outside normal limits.” All included patients were divided into early (mean deviation [MD]>−6 dB) and moderate (−6 to −12 dB) glaucoma based on the MD. VF indices evaluated included MD, PSD, and VF index (VFI).

Optical coherence tomography acquisition protocol

Spectral-domain OCT was performed with the cirrus HD-OCT (software version 4.0, Carl Zeiss Meditec, Inc.) in a dilated pupil as described elsewhere.[67] Only images with a signal strength >6 with no movement artifact were included for the study. The protocol used for RNFL assessment was the optic disc cube. The peripapillary RNFL thickness parameters automatically calculated by the Cirrus software and evaluated in this study included average/full circle (FC) thickness, RNFL-FC (360° measure), superior (temporal and nasal, [ST], and [SN], respectively) quadrant thickness, and inferior (nasal [IN] and temporal [IT]) quadrant thickness.

For ganglion cell analysis, macular parameters were acquired using the macular cube scan. This scan automatically segments the RNFL and the GCIPL layer (from the outer boundary of the RNFL to the outer boundary of the IPL). The average, minimum (lowest GCIPL thickness calculated over a single meridian), and sectoral (ST, superior, SN, IN, inferior, IT, nasal and temporal) thicknesses of the GCIPL are measured in an automated fashion, which were used for this study [Figure 1]. The RNFL and GCIPL parameters were used to compare structure-function in different quadrants in each eye with changes on mfERG responses in the same quadrant.

Figure 1

Top left panel shows the fundus photograph of a patient with moderate glaucoma in the left eye showing inferior notch and retinal nerve fiber layer defect and superior rim loss with associated quadrant-wise comparison of multifocal electroretinogram responses (each quadrant therefore containing 22 responses each as shown in top right panel) compared with macular cube showing ganglion cell-inner plexiform layer thickness (bottom left panel showing inferior thinning) and quadrant-wise retinal nerve fiber layer thickness (bottom right panel)

Multifocal electroretinogram protocol

All mfERG recordings were done after full refractive correction using a commercial mfERG system (Veris Science 5.1.10X; EDI, Redwood City, CA, USA) which conformed to ISCEV guidelines.[14] Stimuli are projected to the central 50° of the retina using the 103-scaled hexagon with each individual recording session consisting of 8 or 16 segments separated by a resting time of 8–16 min. Patient's fixation is actively monitored throughout the session using an infrared fundus camera.

The first negative trough, first positive peak, and second negative trough denote N1, P1, and N2, respectively. Figure 1 shows the custom-based design for analyzing summed responses from 22 focal responses for each quadrant (ST; SN; IT; IN) (veris science; EDI). The quantitative amplitudes and implicit times along with waveforms of the first and second order responses were obtained for global (all 4 quadrants) and quadrant-wise analysis using the automated custom quadrant analysis software.

The quantitative global and quadrant-wise implicit times and amplitude of N1, P1, and N2, N1-P1 and N2-P1 were compared with average and quadrant-wise GCIPL and RNFL thickness. Qualitative analysis was done to compare structure-function relation by comparing quadrant-specific reduction in amplitudes of mfERG responses with VF defects or changes on SD-OCT images [Figure 1].

Inclusion criteria for glaucoma included those with open angles on gonioscopy, best-corrected visual acuity better than 20/40, and at least 2 reliable VFs with false positive errors <15%, false negative errors <15%, and fixation loss <20%.

Age-matched controls were selected in the absence of any anterior or posterior segment pathology with normal optic nerve and VF with other criteria of refractive status and gonioscopy remaining similar to glaucoma patients. If both eyes of patients were eligible for the study, one eye was selected randomly for inclusion. Patients with neovascular glaucoma, anterior or posterior segment pathology precluding fundus examination or OCT acquisition, refractive errors >±4DS, and uncooperative or unconsenting patients were excluded from the study.

Statistical analysis

Statistical analysis was done using Stata (Stata Corp, Version 12, CA, USA). One-way ANOVA with post hoc Bonferroni (or Kruskal–Wallis for nonparametric data) was used to compare the response changes in each quadrant with GCIPL and RNFL thickness and VF parameters in early, moderate, and controls. Multivariate linear regression was done to assess the association between VF indices and structural OCT parameters and functional mfERG responses in each quadrant with statistical significance set at P < 0.05. The area under the curve (AUC) and receiver operating characteristic curves were plotted for various sensitivities and specificities of GCIPL and mfERG responses to classify the parameter distinguishing controls from glaucoma.

Results

A total of 54 cases and 33 controls fulfilling inclusion criteria were recruited for the study with no statistical difference in age between the glaucoma cases and controls [Table 1]. The VF and RNFL parameters were statistically significant between the glaucoma and controls with maximum difference between early or controls and moderate glaucoma, P < 0.001. Table 1 gives the other clinical and demographic variables in controls and glaucoma cases.

Table 1

Demographic and structural parameters on spectral-domain optical coherence tomography of patients with early or moderate glaucoma and controls in this study

The average and minimum GCIPL thickness did not vary significantly between early and control eyes with moderate glaucoma eyes having marginally lower GCIPL thickness than early glaucoma eyes [Table 1 and Figures 2, 3]. The GCIPL thickness was maximum in the superior followed by inferior, nasal, and temporal quadrants in that order in all stages of glaucoma [Table 1].

Figure 2

A case with early glaucoma in the right eye and normal left eye showing decreased ganglion cell-inner plexiform layer thickness (a) corresponding superior visual field defect (b) involving fixation, inferior notch, and retinal nerve fiber layer defect (d) and retinal nerve fibrer layer thickness (e) in inferotemporal quadrant. The corresponding global amplitudes and latencies of N1P1 and N2P1 are not very different between the right and left eye (c)

Figure 3

A case with early glaucoma in the right eye with inferior retinal nerve fiber layer defect and inferior and superior retinal nerve fiber layer defect and moderate glaucoma on fundus photography (a) and spectral-domain optical coherence tomography (b) in the left eye. The ganglion cell-inner plexiform layer thickness (minimum) is decreased in the left eye while the average ganglion cell-inner plexiform layer is not very different (c)

The RNFL-FC differed significantly between glaucoma eyes and controls and between early/moderate glaucoma. Similar changes were observed maximally for inferior and superior RNFL thickness [Table 1].

The GCIPL average thickness in all eyes correlated significantly with RNFL parameters in all quadrants except for the SN quadrant [Table 2]. Similar results were obtained separately for early and moderate glaucoma [Figures 2 and 3]. The GCIPL average and minimum thickness also correlated significantly with MD, PSD, and VFI, P < 0.001 each [Figure 2]. When analyzed separately in different stages of glaucoma, the correlation was sustained in moderate glaucoma and in controls though GCIPL average and minimum showed only weak correlation with MD and VFI in early glaucoma (r = 0.2 and 0.4, respectively, P = 0.05 each).

Table 2

First and second order responses on multifocal electroretinogram in early to moderate glaucoma and controls in this study

The mean N1P1 amplitudes or implicit time (P = 0.8 and 0.1), N1 or P1 amplitude (P = 0.8 and 0.9), or implicit time (P = 0.6 and P = 0.8) did not vary significantly between different stages of glaucoma [Table 2 and Figure 2].

The GCIPL minimum thickness on univariate regression was found to be influenced by N2 amplitudes (β = −0.5, P = 0.012) and global N2P1 amplitudes (β = 0.6, P = 0.01) in moderate glaucoma [Figures 4 and 5]. In early glaucoma, these were influenced only by RNFL parameters [Table 3] with no association with functional mfERG responses [Figure 2]. In control eyes, the GCIPL average and minimum were associated with the implicit time on mfERG (β = 0.3 and 0.3, P = 0.05 and 0.04, respectively).

Figure 4

A case with early glaucoma in the right eye with inferior retinal nerve fiber layer defect and inferior and superior retinal nerve fiber layer defect with moderate glaucoma on fundus photography (a) and spectral-domain optical coherence tomography (b) in the left eye with corresponding visual field defects (d). The ganglion cell-inner plexiform layer thickness minimum is reduced in the left eye (e) while multifocal electroretinogram shows decrease in global N2P1 amplitudes and similar implicit times (c)

Figure 5

A case of moderate glaucoma showing inferior notch (a) with diffuse retinal nerve fiber layer loss (b) and superior field defect (c). The multifocal electroretinogram responses show decreased inferotemporal N2P1 and N2 amplitude while implicit times of N2P1 and amplitudes/latencies of N1, P1 are similar (d)

Table 3

Correlation of structural parameters including ganglion cell-inner plexiform layer and retinal nerve fiber layer thickness on spectral-domain optical coherence tomography in this study

Multivariate logistic regression identified global N2P1 amplitude to be significantly influencing GCIPL average and minimum thickness (P = 0.01 and 0.02, R2 = 47.8% and 52.3%, respectively) in moderate glaucoma. Maximum AUC was found for GCIPL minimum and N2P1 amplitude [Table 4]. Quadrant-wise comparison of structure and function showed similar results in all quadrants except for SN quadrant with minimal or no association with mfERG parameters.

Table 4

Area under the curve for structural parameters on spectral-domain optical coherence tomography and functional parameters on multifocal electroretinogram predicting moderate glaucoma

Discussion

This study showed that GCIPL thickness is only marginally different between early/moderate glaucoma but predicts glaucoma more aptly than mfERG responses. While GCIPL correlated with implicit times in controls, it correlated well with global N2P1 and N2 amplitude in moderate glaucoma while in early glaucoma, this only correlated with RNFL parameters. This suggested that N2P1 and N2 amplitudes are measures of function only in moderate glaucoma while early glaucoma representing very minimal functional disturbance may be missed on mfERG as on VFs.

The superior performance of the GCIPL is attributed to anatomic variations such as peripapillary atrophy or abnormal size/shapes of optic disc which can create measurement errors in optic nerve head (ONH) scans.[815161718192021222324] Furthermore, it is worth noting that variability in the position of peripapillary vessels which are also accounted in the measurement of RNFL thickness may also contribute to RNFL thickness measurement in ONH scans. Therefore, the GCIPL thickness measurement is much more accurate than temporal or macular RNFL thickness for structural evaluation in glaucomatous eyes. Our study also confirmed earlier findings and also suggested a supportive role of mfERG amplitudes in moderate glaucoma with GCIPL.[15222324]

Structure-function disparity has been a concern in glaucoma diagnosis and monitoring despite the evolution of modern imaging technologies.[123] This has been attributed to a nonlinear relationship between structure and function which essentially means that rate of structural loss may not correlate with extent of functional loss in a similar way in all stages of glaucoma.[391023] The GCIPL algorithm measures the thickness of the ganglion cell layer and IPL and is reportedly a reliable measure of early glaucomatous damage more than RNFL thickness.[2122] The hemifield analysis of the ganglion cell complex (GCC) analysis allows topographic correlation of this parameter with sectoral functional parameters as evaluated in this study. Kim et al. reported that macular GCC thickness on Fourier domain-OCT (RTVue) had similar structure–function relationships and peripapillary RNFL with VF sensitivity.[15] Other studies including one by Na et al. have reported better structure-function relationship of GCC than RNFL parameters.[202223] The cirrus SD-OCT GCIPL algorithm calculates the combined thickness of the retinal ganglion cells (RGCs) layer and the IPL excluding the RNFL. Kim et al. also reported better structure-function relation of GCIPL with the macular mean visual sensitivity.[15] Mwanza et al. reported a comparable diagnostic performance of GCIPL and RNFl thickness with maximum AUC for GCIPL minimum and IT thickness.[22] We evaluated the structure-function correlation using mfERG where a strong correlation between global N2P1 amplitude and N2 amplitudes were found with GCIPL thickness only in moderate glaucoma. This may be partly because the GCIPL does not actually correlate with axons at the ONH while mfERG gives a more appropriate topographical representation.

In our study, we did not find significant relationship between GCIPL with mfERG responses in early glaucoma. One explanation is that the macula is overrepresented in the cortex and also the proportion of RGCs at macula is considerably more than the axons at the ONH. Hence, it may be possible that in early glaucoma, loss of RGC extent may not be so severe and may be compensated unlike in established disease in moderate glaucoma. This suggests that GCIPL may be a more appropriate measure of moderate or established glaucoma while RNFL would be better predictor of early disease. Correspondingly, mfERG and N2 amplitudes were found to be significantly altered in moderate glaucoma, reflecting that these represent established disease forms. It remains unclear from this study as to how much GCIPL thickness loss may reflect functional damage on mfERG in each stage.

Conclusions

GCIPL and global N2P1 amplitudes may predict moderate glaucoma while structural RNFL parameters seem to predict early glaucoma. Future studies are required to evaluate utility of other mfERG responses for monitoring early stages of glaucoma.

Financial support and sponsorship

Nil.

Conflicts of interest

There are no conflicts of interest.