Papillomacular Bundle and Inner Retinal Thicknesses Correlate With Visual Acuity in Nonarteritic Anterior Ischemic Optic Neuropathy

Abstract

Purpose.: To evaluate the ability of the papillomacular bundle (PMB) retinal nerve fiber layer and macular inner retinal layer thickness measurements with Spectralis optical coherence tomography (OCT) to differentiate eyes with nonarteritic anterior ischemic optic neuropathy (NAION) from uninvolved eyes and to evaluate whether their thicknesses correlate with visual acuity.

Methods.: An observational, cross-sectional study was performed, including 29 eyes with NAION and 29 uninvolved eyes from 29 patients. Eyes underwent scanning with Cirrus OCT (peripapillary and macular scanning) and Spectralis OCT (N-site axonal peripapillary scan and a new automated segmentation macular scan to measure individual retinal layers) in both eyes.

Results.: The NAION eyes showed significant thinning versus uninvolved eyes in the macular retinal nerve fiber (P < 0.05), ganglion cell layer (GCL; P < 0.001), and inner plexiform layer (IPL; P < 0.01) by Spectralis and in the GCL-IPL by Cirrus (P < 0.02). Average and sectors of peripapillary retinal nerve fiber layer (pRNFL) and total macular thickness (TMT) were significantly reduced in NAION eyes, with both Spectralis and Cirrus OCT (P < 0.05). Spectralis temporal (ρ_Spearman = −0.768; P < 0.001) and PMB pRNFL thicknesses (ρ_Spearman = −0.675; P < 0.001), as well as central macular IPL thickness (ρ_Spearman = −0.735; P < 0.001), correlated strongly with best corrected visual acuity (BCVA). Quadratic regression using outer nasal TMT by Cirrus OCT and temporal pRNFL thickness by Spectralis were the best models to predict BCVA.

Conclusions.: Macular segmentation by Spectralis and Cirrus OCT revealed inner retinal layer atrophy in NAION eyes. The temporal and PMB pRNFL thicknesses and central macular IPL thickness by Spectralis-OCT and outer nasal TMT by Cirrus were strongly correlated with BCVA in NAION eyes.

Introduction

Nonarteritic anterior ischemic optic neuropathy (NAION) is a well-known clinical entity characterized by the sudden onset of painless visual loss with optic disc edema that resolves in a few weeks and is usually followed by optic disc pallor.¹ Optical coherence tomography (OCT) in NAION is useful to diagnose and quantify optic disc edema and to monitor peripapillary retinal nerve fiber layer (pRNFL) loss.^2–7 Contreras et al.² have reported that OCT is most useful at onset and 6 months after NAION, when pRNFL loss has reached a plateau. 

Several groups have used time-domain OCT to investigate if there is a relationship between pRNFL and macular thickness with visual acuity loss in eyes that have had NAION. Their findings suggest that reduction of both temporal optic nerve head quadrant thickness and macular thickness are good clinical indicators of central visual damage.^2–10 It has also been suggested that visual acuity loss after an episode of NAION depends on the level of damage to the papillomacular bundle (PMB) because these are the fibers that serve the macula.⁸ 

A new generation of OCT devices, based on spectral-domain technology, have improved resolution and advanced software and are capable of performing more precise measurements of the pRNFL as well as assessing the deeper retinal layers. Spectralis-OCT (Heidelberg Engineering, GmbH, Heidelberg, Germany) incorporates the N-site axonal protocol that includes an analysis of the PMB. This protocol might be especially useful in disorders in which there is predominant damage of the optic nerve temporal quadrant, such as multiple sclerosis, optic neuritis, and other optic neuropathies.¹¹ Furthermore, Spectralis OCT has developed a specific software that provides automated differentiation and quantification of the 10 retinal layers. Other improvements in Cirrus OCT technology (Carl Zeiss Meditec AG, Jena, Germany) include a software that can measure the ganglion cell layer (GCL) and inner plexiform layer (IPL) thickness at the macula. Up to 40% of the macular thickness is occupied by the ganglion cell layer. Fernandez-Buenaga and colleagues⁹ have found a significant correlation between nasal total macular thickness (TMT) and visual acuity in eyes after NAION. It is reasonable to expect that measurements of the macular GCL might better reflect the damage produced by NAION. 

Our hypothesis is that both PMB and GCL measurements will correlate strongly with visual acuity in NAION, more so than conventional pRNFL and TMT analysis. The purpose of this study was to investigate the ability of the newer PMB analysis software (Spectralis OCT), Ganglion Cell Analysis (GCA, Cirrus OCT), and macular inner retina layer measurements (Spectralis OCT) to differentiate eyes with NAION from uninvolved eyes and to identify the relationship between these parameters and visual acuity. 

Patients and Methods

An observational, prospective, cross-sectional study was performed. Participants were recruited from the Neuro-Ophthalmology Unit at Ramón y Cajal University Hospital (Madrid, Spain). 

The study and data accumulation were performed with approval from the hospital institutional review board. The authors confirm that the study conformed to all country, federal, or state laws, and the study followed the principles of the Declaration of Helsinki. Informed consent for the research was obtained from all participants. 

Patients who had had unilateral NAION at least 6 months before the beginning of the study were considered for inclusion. Diagnosis of NAION was based on sudden loss of visual acuity; relative afferent pupillary defect; disc edema on fundus ophthalmoscopy at onset; visual field defects consistent with NAION; erythrocyte sedimentation rate and C reactive protein levels within normal values, with no signs or symptoms suggestive of giant cell arteritis; and resolution of disc edema in 2 months. Exclusion criteria were a refractive error greater than 5.0 diopters (D) of spherical equivalent or 3.0 D of astigmatism in either eye, media opacities that would preclude OCT scanning, glaucoma, coexistence of ophthalmic or neurologic disease, or other retinal pathologic processes, or previous ophthalmic surgery (other than uneventful cataract extraction). 

All patients underwent an extensive neuro-ophthalmologic evaluation including pupillary, anterior segment, and funduscopic examinations; assessment of best corrected visual acuity (BCVA); and automated visual field testing with a Humphrey Field Analyzer (Carl Zeiss Meditec AG) using the Swedish Interactive Threshold Algorithm Standard strategy (program 24-2). Visual fields were considered reliable when fixation losses were less than 20% and false-positive and false-negative errors were less than 15%. 

Both eyes of each subject were included in the study, with the uninvolved contralateral eye serving as the comparison group. 

Optical Coherence Tomography Measurements

All participants underwent macular and peripapillary OCT scanning with the Cirrus HD-OCT and Spectralis OCT. All scans were performed on the same day by a single well-trained operator (CS-S) in random order to prevent any fatigue bias. 

Cirrus Examination.

The pRNFL thickness was obtained by using the optic disc cube protocol on Cirrus software version 6.0. This protocol generates a cube of data through a 6-mm-square grid. A 3.46-mm-diameter circle is automatically centered on the optic disc. The analysis protocol provides average RNFL thickness and maps with four quadrants (superior, inferior, nasal, and temporal) and 12 clock-hours, including classification compared with an internal normative database. The operator checked that the OCT software had centered the measurement circle correctly and that it had correctly identified the RNFL limits. 

The macular cube 512 × 128 acquisition protocol was used to obtain both TMT and the GCL and IPL thicknesses. This scan protocol generates a cube of data through a 6-mm-square grid by acquiring a series of 128 horizontal scan lines comprising 512 A-scans. Total macular thickness was calculated by measuring the retinal thickness between the inner limiting membrane (ILM) and the retinal pigment epithelium (RPE). 

The analysis protocol provides thickness measurements for nine macular areas corresponding to the Early Treatment Diabetic Retinopathy Study map. The map is composed of three concentric circles with a diameter of 1 (central area), 3, and 6 mm. The area between the outer (6 mm) and middle (3 mm) circles forms the outer ring, while the area between the middle (3 mm) and inner circles (1 mm) forms the inner ring. Each ring is divided into superior, nasal, inferior, and temporal quadrants. 

The GC algorithm identifies the outer boundaries of the RNFL and IPL. The difference between the RNFL and the IPL outer boundary segmentations yields the combined thickness of the GCL and IPL (GCIPL). Cirrus OCT provides quantitative assessment of the GCIPL in six circular sectors centered in the fovea (superonasal, superior, inferonasal, inferotemporal, inferior, and superotemporal). It also gives information on the mean and minimum GCIPL thickness for each eye and compares these figures with a normative database. 

To be included all scans had to have a signal strength ≥6 and no movement artifacts. The operator checked all horizontal scans to confirm that the outer boundaries of the RNFL and IPL had been correctly identified and repeated the examination whenever necessary. If correct identification could not be achieved, the patient was excluded. 

Spectralis OCT.

All participants were examined with two acquisition protocols with the Spectralis OCT: the fast macular cube and the RNFL-N axonal protocol, which differs from the standard RNFL scan because it starts and finishes in the nasal quadrant of the optic nerve. The pRNFL thickness was measured around the disc with 16 averaged consecutive circular B-scans (3.5-mm diameter, 768 A-scans). The RNFL Spectralis protocol generates a map showing the average thickness, maps with four quadrants (superior, inferior, nasal, and temporal), maps with six sector thicknesses (superonasal, nasal, inferonasal, inferotemporal, temporal, and superotemporal), thickness of the PMB, and the nasal-to-temporal (N/T) ratio. Inaccurate images due to RNFL segmentation algorithm errors (failure to detect the appropriate edges of the RNFL) were excluded from the analysis. 

The new prototype for retinal segmentation using the Spectralis OCT (software version 6.0) was used to identify and measure the thickness of each retinal layer. Images were acquired by using the image alignment eye-tracking software (TruTrack; Heidelberg Engineering GmbH) to obtain perifoveal volumetric retinal scans comprising 19 single axial scans (scanning area: 8.5 × 4.3 mm²) centered at the fovea. Segmentation of the retinal layers was performed automatically by the new segmentation application for the Spectralis OCT (Segmentation Technology; Heidelberg Engineering GmbH), in order to automatically identify the following 10 layers: ILM, RNFL, GCL, IPL, inner nuclear layer, outer plexiform layer, outer nuclear layer, outer limiting membrane, photoreceptors (rods and cones), and RPE. 

From the retinal layer thickness map, data are grouped in nine macular sectors within three concentric circles as defined by the Early Treatment Diabetic Retinopathy Study. The location of the center of the fovea is first identified by the minimum thickness on the map of layer 2 (GCL and IPL). This location defined the center of the concentric circles. The central circle has a diameter of 1.2 mm and represented the central foveal area. The second circle has a diameter of 3.1 mm and is subdivided into superior, nasal, inferior, and temporal parafoveal inner retinal sectors. The third circle has a diameter of 6 mm and is subdivided into superior, nasal, inferior, and temporal outer perifoveal retinal areas. Central, inner, and outer macular sectors contain 15, 38, and 74 data points, respectively. Measurements of the nine sectors of each individual layer were registered. 

The quality of the scans was assessed before the analysis and scans with misalignment, segmentation failure, decentration of the measurement circle, poor illumination, or those that were out of focus were excluded from the analysis. 

The Spectralis OCT uses a blue quality bar in the image to indicate signal strength. The quality score ranges from 0 (poor quality) to 40 dB (excellent quality). Only images that scored higher than 25 dB were included. 

Statistics

IBM SPSS Statistics version 20 (International Business Machines Corp., Armonk, NY, USA) was used for the statistical analysis of the data. Continuous variables were summarized as mean ± standard deviation. Binary variables were summarized as count (percentage). Kolmogorov-Smirnov test was used to assess normal distribution in continuous variables. Mean layer thicknesses of eyes with NAION and contralateral normal uninvolved controls were compared by using Student's paired t-tests. P values <0.05 were considered statistically significant. 

Spearman's ρ coefficient (ρ_Spearman) was used to assess the relationship between the different mean layer thicknesses and BCVA. A logistic regression analysis was performed to identify retinal layer thicknesses that predicted visual loss in NAION eyes. 

Receiver operating characteristic (ROC) curves were used to describe the ability of OCT parameters to discriminate NAION eyes from uninvolved contralateral eyes. The area under the ROC curves (AUROCs) was calculated to evaluate the sensitivity and specificity of each scanning region in distinguishing between normal and NAION eyes. 

Multiple linear and nonlinear regression models were used to find the best model fit for predicting BCVA from the OCT measures. Linear, logarithmic, inverse, quadratic, cubic, power, compound, logistic, growth, and exponential models were fitted, and the model with the minimum Aikake information criterion (AIC) was selected. 

Results

Twenty-nine patients met the inclusion and exclusion criteria. Average age of study participants was 69.6 ± 10.6 years (range, 50–83 years); 11 (39.3%) were female and 17 (60.7%) were male. Average time from the acute NAION attack was 18.7 months (range, 6–60 months). 

The NAION eyes had significantly worse logMAR visual acuity (0.413 ± 0.346 vs. 0.069 ± 0.104; P < 0.001), mean deviation (−14.69 ± 8.6 dB vs. −3.7 ± 3.1 dB; P < 0.001), and visual field index (58.4 ± 29 vs. 93.2 ± 6.8; P < 0.001) than the uninvolved control eyes. 

Tables 1, 2, and 3 show pRNFL, TMT, and GCIPL/GCL and individual inner retinal layer measurements for eyes with NAION and contralateral uninvolved eyes, with both devices. 

Table 1

 

Mean Values of OCT pRNFL Parameters (in μm) With AUROC

	Involved Eyes		Uninvolved Eyes		P(Paired t-Test)	P(Independent Samples t-Test)	AUROC
	Mean	SD	Mean	SD
pRNFL Cirrus
Superior	62.8	22.5	102.6	19.6	<0.001	<0.001	0.90
Inferior	82.5	25.8	113.2	19.9	<0.001	<0.001	0.81
Temporal	49.5	8.7	61.6	8.7	<0.001	<0.001	0.84
Nasal	56.9	14.8	68.6	13.3	0.002	0.005	0.69
Average	62.9	12.9	86.5	10.7	<0.001	<0.001	0.91
pRNFL Spectralis*
Superonasal	54.5	24.4	96.9	25.7	<0.001	<0.001	0.88
Superotemporal	68.2	31.0	129.1	23.5	<0.001	<0.001	0.90
Inferonasal	87.9	32.5	115.0	32.4	<0.001	0.003	0.75
Inferotemporal	100.6	35.1	140.0	22.4	<0.001	<0.001	0.82
Average	62.8	14.8	95.9	12.7	<0.001	<0.001	0.95
PMB	34.0	13.7	55.4	7.9	<0.001	<0.001	0.90
N/T	1.3	0.6	1.1	0.3	0.008	0.024	0.34
Superior	61.3	24.7	112.9	21.8	<0.001	<0.001	0.91
Inferior	94.2	30.1	127.4	23.2	<0.001	<0.001	0.81
Nasal	53.2	22.1	73.1	17.9	<0.001	0.001	0.75
Temporal	42.4	14.3	70.3	9.2	<0.001	<0.001	0.94

 SD, standard deviation.

*  N-site axonal analysis.

Table 2

 

Mean Values of OCT TMT Parameters (in μm) With AUROC

	Involved Eyes		Uninvolved Eyes		P	AUROC
	Mean	SD	Mean	SD
TMT Cirrus
Central	242.8	31.8	262.3	24.5	0.002	0.74
Superior inner	283.1	20.5	319.8	18.9	<0.001	0.92
Superior outer	252.4	15.9	275.2	16.3	<0.001	0.87
Inferior inner	294.3	24.2	319.9	17.4	<0.001	0.77
Inferior outer	251.6	22.1	266.6	19.0	0.018	0.61
Temporal inner	284.0	17.4	311.8	14.3	<0.001	0.94
Temporal outer	251.8	12.9	254.1	38.4	0.802	0.65
Nasal inner	286.7	25.1	322.2	19.7	<0.001	0.89
Nasal outer	260.0	23.2	289.9	16.7	<0.001	0.86
TMT Spectralis
Central	255.08	22.56	265.96	26.20	0.122	0.51
Superior inner	296.16	19.16	326.20	18.93	<0.001	0.76
Superior outer	259.60	16.05	279.12	16.89	<0.001	0.78
Inferior inner	300.84	20.75	325.32	16.11	<0.001	0.73
Inferior outer	257.44	21.25	273.08	18.50	0.008	0.66
Temporal inner	291.40	12.78	316.80	13.77	<0.001	0.90
Temporal outer	259.64	15.62	278.36	16.84	0.002	0.78
Nasal inner	301.48	22.01	328.76	17.76	<0.001	0.71
Nasal outer	278.50	21.34	301.23	17.32	0.001	0.78

Table 3

 

Mean Values of OCT GCIPL and Individual Inner Retinal Layer Parameters (μm) With AUROC

	Involved Eyes		Normal Eyes		P	AUROC
	Mean	SD	Mean	SD
GCIPL Cirrus
Superior	56.4	14.7	74.8	14.5	<0.001	0.80
Superotemporal	54.9	14.6	76.2	11.1	<0.001	0.89
Superonasal	56.5	14.7	74.2	14.0	<0.001	0.81
Inferior	63.4	17.6	74.6	12.9	0.014	0.69
Inferotemporal	64.7	17.6	77.7	11.4	0.003	0.72
Inferonasal	59.1	14.5	73.9	13.7	0.001	0.77
Average	59.2	13.5	75.3	11.3	<0.001	0.83
Minimun	46.6	15.5	66.4	18.9	<0.001	0.76
RNFL macula Spectralis
Central	11.9	3.4	15.4	3.1	0.001	0.642
Superior inner	25.0	4.9	30.0	5.8	0.003	0.698
Superior outer	27.4	8.9	41.1	8.1	<0.001	0.843
Temporal inner	21.8	2.1	21.7	2.1	0.835	0.475
Temporal outer	23.2	2.1	23.7	2.4	0.572	0.556
Inferior inner	24.6	4.6	29.3	3.3	<0.001	0.772
Inferior outer	32.7	9.2	43.9	7.2	<0.001	0.682
Nasal inner	22.7	4.7	25.3	2.9	0.029	0.645
Nasal outer	36.1	13.4	46.4	6.8	0.006	0.762
GCL macula Spectralis
Central	8.6	7.0	17.2	6.6	<0.001	0.70
Superior inner	25.1	15.0	48.3	10.9	<0.001	0.84
Superior outer	18.5	5.8	27.5	3.8	<0.001	0.88
Temporal inner	22.8	12.6	46.4	7.5	<0.001	0.95
Temporal outer	20.3	7.3	31.4	5.4	<0.001	0.91
Inferior inner	28.9	15.8	53.0	11.8	<0.001	0.78
Inferior outer	20.9	6.8	27.5	4.6	<0.001	0.64
Nasal inner	25.5	18.6	49.9	11.1	<0.001	0.75
Nasal outer	24.8	8.7	35.5	5.5	<0.001	0.82
Inner plexiform macula Spectralis
Central	18.3	3.7	22.2	4.0	0.001	0.600
Superior inner	29.1	6.3	38.7	5.2	<0.001	0.781
Superior outer	22.9	2.2	25.5	2.4	<0.001	0.722
Temporal inner	28.6	6.1	39.2	4.1	<0.001	0.920
Temporal outer	26.6	3.7	32.8	3.7	<0.001	0.824
Inferior inner	29.9	7.3	38.9	4.8	<0.001	0.673
Inferior outer	23.1	2.9	25.5	3.2	0.009	0.577
Nasal inner	28.4	8.9	38.6	4.9	<0.001	0.614
Nasal outer	25.8	3.9	29.4	3.9	0.009	0.627

All average, quadrant, and sector pRNFL thickness measurements were significantly reduced in the eyes with NAION compared with their unaffected counterparts by Cirrus and Spectralis (Table 1). 

The N/T ratio was significantly higher in NAION eyes than healthy eyes (Table 1). 

The TMT measurements provided by both OCT devices were reduced in eyes with NAION compared with healthy contralateral eyes. The groups differed significantly in all parameters except at the temporal-outer sector by Cirrus and at the central sector by Spectralis (Table 2). 

All measurements of GCIPL by Cirrus were significantly reduced in affected eyes compared with uninvolved eyes. The inner retinal layer thicknesses measured by Spectralis (nerve fiber layer except for the temporal sectors, ganglion cell, and inner plexiform) were significantly reduced in eyes with NAION compared with control eyes (Figs. 1, 2A–C, Table 3). 

Figure 1

 

Comparison of retinal layer analysis determined by the new segmentation application of the Spectralis OCT between an eye with NAION (LE) and the healthy contralateral eye (RE). The software automatically delineated the following layers: ILM, RNFL, GCL, IPL, inner nuclear layer (INL), outer plexiform layer (OPL), outer nuclear layer (ONL), outer limiting membrane (OLM), photoreceptors (PR1, PR2), and RPE. The INLs were thinner in LE than in normal contralateral RE. Eye with NAION showed a decrease in the thickness of INLs (RNFL, GCL, and IPL).

Figure 2

 

A comparison of the mean average thickness in the central sector and the eight sectors between LE (NAION eye) and RE (contralateral healthy eye) in the macular inner retinal layers corresponding to the same patient as Figure 1. (A) Retinal nerve fiber layer. (B) Ganglion cell layer. (C) Inner plexiform layer.

No significant differences were detected between NAION and uninvolved eyes in outer and inner nuclear layer, outer plexiform, RPE, or photoreceptor (rods and cones) layer thicknesses. 

Areas under the ROC curves were similar for pRNFL (0.81–0.91) thickness with Cirrus and Spectralis (0.75–0.95). Areas under the ROC curves were also similar for TMT with Spectralis (0.51–0.90) and Cirrus (0.61–0.94). Areas under the ROC curves ranged from 0.72 to 0.89 for GCIPL measurements with Cirrus and from 0.64 to 0.95 for GCL values with Spectralis. The largest AUROC was for the average and temporal quadrant pRNFL thickness by Spectralis (0.95 and 0.94, respectively). 

The NAION eyes were analyzed for Spearman correlations between pRNFL, TMT, GCIPL, and individual macular layer thicknesses and BCVA (Table 4). Overall, the highest correlation was observed for the temporal quadrant of pRNFL with Spectralis (−0.768; P < 0.001) followed by the central sector of the IPL (−0.735; P < 0.001). For Cirrus OCT, the highest correlation was found for the outer nasal sector of TMT (−0.690, P < 0.001). 

Table 4

 

Spearman's ρ Correlation Coefficients Between pRNFL, TMT, GCIPL, and Inner Retinal Layer Thicknesses (in μm) and logMAR BCVA

	Spearman's ρ	P
pRNFL Cirrus
Superior	−0.052	0.801
Inferior	−0.324	0.107
Temporal	−0.380	0.055
Nasal	−0.151	0.460
Average	−0.464	0.017
pRNFL Spectralis N-Site
Superonasal	−0.057	0.772
Superotemporal	−0.236	0.227
Inferonasal	−0.157	0.424
Inferotemporal	−0.200	0.307
Average	−0.397	0.037
PMB	−0.675	<0.001
N/T	0.424	0.025
Superior	−0.200	0.307
Inferior	−0.276	0.155
Nasal	−0.068	0.731
Temporal	−0.768	<0.001
TMT Cirrus
Superior inner	−0.590	0.002
Superior outer	−0.482	0.023
Inferior inner	−0.507	0.012
Inferior outer	−0.455	0.038
Temporal inner	−0.623	0.001
Temporal outer	−0.526	0.014
Nasal inner	−0.579	0.003
Nasal outer	−0.690	<0.001
Central	−0.282	0.182
Spectralis TMT
Superior inner	−0.599	0.002
Superior outer	−0.229	0.282
Inferior inner	−0.482	0.017
Inferior outer	−0.259	0.222
Temporal inner	−0.473	0.020
Temporal outer	−0.266	0.381
Nasal inner	−0.513	0.010
Nasal outer	−0.575	0.013
Central	−0.377	0.069
GCIPL Cirrus
Superior	−0.372	0.067
Superotemporal	−0.455	0.022
Superonasal	−0.411	0.041
Inferior	−0.440	0.028
Inferotemporal	−0.491	0.013
Inferonasal	−0.458	0.021
Average	−0.536	0.006
Minimum	−0.288	0.162
Macular RFL Spectralis
Central	−0.412	0.051
Superior inner	−0.330	0.125
Superior outer	−0.272	0.209
Temporal inner	−0.290	0.179
Temporal outer	−0.216	0.478
Inferior inner	−0.377	0.076
Inferior outer	−0.377	0.084
Nasal inner	−0.438	0.037
Nasal outer	−0.444	0.074
GCL Spectralis
Central	−0.567	0.003
Superior inner	−0.663	<0.001
Superior outer	−0.386	0.057
Temporal inner	−0.703	<0.001
Temporal outer	−0.657	0.008
Inferior inner	−0.610	0.001
Inferior outer	−0.492	0.013
Nasal inner	−0.618	0.001
Nasal outer	−0.630	0.005
Inner plexiform Spectralis
Central	−0.735	<0.001
Superior inner	−0.611	0.002
Super outer	−0.164	0.456
Temporal inner	−0.661	0.001
Temporal outer	−0.463	0.111
Inferior inner	−0.487	0.018
Inferior outer	−0.219	0.327
Nasal inner	−0.616	0.002
Nasal outer	−0.602	0.011

For pRNFL parameters only average thickness, as measured with Cirrus, was moderately correlated with BCVA (−0.464; P = 0.017). However, both temporal and PMB thickness with the N-site axonal protocol of the Spectralis were strongly correlated with BCVA (−0.768; P < 0.001 and −0.675; P < 0.001, respectively). As opposed to the findings with pRNFL measures, a significant correlation with BCVA was found for most parameters of TMT and GCIPL with Cirrus. For TMT, the best correlation with BCVA was found for the outer nasal sector with Cirrus OCT (−0.690; P < 0.001) and inner superior sector with Spectralis (−0.599; P = 0.001). 

Multiple curvilinear and linear regression analysis models were built for BCVA prediction according to central IPL thickness (Fig. 3), temporal pRNFL thickness (Fig. 4), and Cirrus outer nasal (Fig. 5) TMT. The best fit models were found by using AIC. Quadratic regression using outer nasal TMT by Cirrus OCT was found to be the model with the better fit (AIC = 63.1), followed by the temporal pRNFL thickness by Spectralis OCT (AIC = 73.40). 

Figure 3

 

Dot plot representing the BCVA (logMAR) of the eyes from 29 patients with NAION versus their IPL thickness in the central circle, as measured by Spectralis OCT. The best fit model was an inverse regression model (R² = 0.375; P < 0.001).

Figure 4

 

Dot plot representing the BCVA (logMAR) of the eyes from 29 patients with NAION versus their temporal pRNFL thickness, as measured by Spectralis OCT. The best fit model was a quadratic regression model (R² = 0.657; P < 0.001).

Figure 5

 

Dot plot representing the BCVA (logMAR) of the eyes from 29 patients with NAION versus their outer nasal sector TMT, as measured by Cirrus OCT. The best fit model was a quadratic regression model (R² = 0.500; P < 0.001).

Discussion

Optical coherence tomography technology has been used to provide information about changes that occur in the optic nerve and the macula after NAION, using Stratus OCT^2–10 and SD-OCT.^12,13 

The purpose of the current study was to evaluate the newer SD-OCT algorithms in NAION eyes, to establish which device and/or protocol has the best correlation with visual acuity. These new protocols, including PMB thickness, N/T ratio, and the 10 paramacular individual retinal layer thicknesses, have not been previously evaluated in NAION eyes. 

Our results corroborate the findings of previous studies on pRNFL thickness. The average, quadrant, and sector pRNFL thicknesses yielded by two SD-OCT devices, Cirrus and Spectralis, were significantly lower in NAION eyes than in the uninvolved eyes. 

It has been suggested that temporal quadrant pRNFL loss might be a good clinical indicator of central visual damage in NAION eyes and that visual acuity loss depends on the severity of the damage sustained by the PMB.^2,4,7,9,10 

In the current study, the largest AUROC was for the central and temporal quadrant pRNFL thickness provided by the N-site axonal protocol by Spectralis OCT (0.95 and 0.94, respectively). 

Furthermore, we found that the mean thickness of the PMB was significantly reduced and the N/T ratio was higher in NAION eyes than uninvolved eyes, indicating that the RNFL damage was more severe in the temporal than in the nasal quadrant. 

Of all measurements provided by pRNFL analysis with Cirrus OCT, only average RNFL thickness was moderately correlated with BCVA (ρ_Spearman = −0.464). However we found a significant strong correlation between BCVA and both the temporal quadrant and the PMB pRNFL thickness measured with Spectralis (ρ_Spearman = −0.768 and −0.675, respectively). 

The percentage of temporal pRNFL thickness loss in NAION eyes compared to contralateral eyes was higher with Spectralis (35%) than with Cirrus (19%). This finding is not surprising, since the N-site axonal protocol provided by Spectralis has been designed to focus on the temporal quadrant and PMB. It differs from the standard pRNFL scan, because it starts and ends in the nasal side of optic nerve. Hence, this protocol would provide more information in optic neuropathies in which damage typically predominates in the temporal quadrant.¹⁴ 

Overall, the temporal pRNFL thickness by Spectralis had the strongest correlation with BCVA (ρ_Spearman = −0.768; P < 0.001), followed by the central IPL macular thickness (ρ_Spearman = −0.735; P < 0.001) and the nasal outer quadrant TMT measured by Cirrus (ρ_Spearman = −0.690; P < 0.001). Furthermore, quadratic regression using outer nasal TMT by Cirrus OCT and temporal pRNFL thickness by Spectralis OCT were the best models to predict BCVA. 

Although in NAION the damage occurs at the optic nerve head, a significant proportion of the axons affected have their cell bodies in the macula (representing 30%–40% of total macular thickness together with the RNFL), arranged in layers of two to six cells, as opposed to the single-layer organization in other areas of the retina. Hence, their loss could lead to significant macular thinning. Total macular thickness and volume have been previously investigated in NAION.^7,9,15 In a previous cross-sectional study performed with Stratus OCT, we reported that only reduction of both the outer and inner nasal macular thickness after NAION is significantly correlated with BCVA.⁹ Using regression analysis, it was found that for every 11 μm lost in mean nasal outer macular thickness there was a one-line drop in BCVA. Therefore, we hypothesized that visual acuity loss after an episode of NAION depends on the severity of the damage to the PMB. Using Stratus-OCT, Papchenko et al.⁷ have reported that total thickness and volume of the macula correlate strongly with visual field damage. 

Our study sought to expand the investigation of the relationship between visual damage and macular structure, by evaluating TMT and each individual layer with the improved technology and resolution of SD-OCT. The TMT measurements provided by both OCT devices were reduced in eyes with NAION compared with uninvolved eyes. The groups differed significantly in all parameters except in the temporal-outer sector with Cirrus and in the central sector with Spectralis. The AUROCs of all TMT parameters for NAION obtained with both devices were similar. The central IPL macular thickness by Spectralis and the nasal outer sector TMT measurement with Cirrus OCT had the highest correlation with BCVA. 

Taking into account that PMB travels from the outer nasal sector to the PMB pRNFL sector, it is not surprising that both the outer nasal TMT provided by Cirrus and the PMB thickness by Spectralis had a good correlation with BCVA (ρ_Spearman = −0.690 and −0.675, respectively), thus supporting the hypothesis that visual acuity is correlated with damage to the PMB. 

The latest OCT investigations involve segmentation of specific retinal layers, allowing quantification of both axonal damage (RNFL) and neuronal degeneration. In addition to RNFL loss, the acute ischemia of the optic nerve in NAION results in reduction of the function and in death of retinal ganglion cells. Gonul et al.¹⁵ have evaluated the macular ganglion cell complex (GCC) in NAION eyes. They found that GCC parameters, such as focal loss volume and global loss volume, show the best ability to detect ganglion cell loss in patients with NAION, suggesting that macular scans might provide a good alternative or a complementary practice to RNFL scans in the detection of damage in patients with NAION. Newer Spectralis-OCT software can delineate the inner retinal complex, composed of the axons, retinal ganglion cell bodies, and dendrites (RNFL, GCL, and IPL), from the rest of the retina (Fig. 1). 

Measuring the thickness of these specific layers rather than TMT may provide more useful information, as these are the layers that are thought to be predominantly damaged after NAION. 

This issue has been investigated in other optic neuropathies. Patients with relapsing-remitting multiple sclerosis (RRMS) exhibit a significant thinning of pRNFL and inner retinal layers. The OCT-macular GCIPL thickness is reduced before the pRNFL thickness and correlates better with visual acuity than RNFL thickness in RRMS patients.^16–20 

To the best of our knowledge, there are no previous reports investigating the paramacular individual retinal layers in NAION eyes. In the current study, the average, minimum, and all six sectors of GCIPL thickness values by Cirrus were significantly reduced in NAION eyes compared with the contralateral uninvolved eyes. The main weakness of the GCIPL analysis provided by Cirrus is that it does not differentiate the retinal ganglion cell and the inner plexiform layers, and therefore it measures the thickness of both layers together. With the new segmentation software of the Spectralis OCT used in the present study, all retinal layers are differentiated and measured separately. We found a statistically significant reduction in the thickness of the RNFL, GCL, and IPL in affected eyes compared with healthy eyes. The AUROC curves for GCIPL by Cirrus and for each inner individual retinal layer by Spectralis were similar. 

The relationship between BCVA and GCIPL thickness with Cirrus and BCVA and macular RNFL thickness with Spectralis was moderate. However, most sectors of individual macular GCL and macular IPL thickness measured with Spectralis were strongly related with BCVA. 

Nonarteritic anterior ischemic optic neuropathy seems to have no effect on the outer and inner nuclear layer, outer plexiform, photoreceptor, and RPE layers because we observed no significant changes in these layers. 

This study had some limitations. We only collected data from patients at least 6 months after the NAION episode; therefore, information concerning dynamic changes in the acute phase are lacking. Several reports have found that approximately 6 months after the ischemic episode a stable morphologic end point is reached and there appears to be no further RNFL loss.^2,8,11 Regarding macular segmentation, a direct comparison between parameters by two different SD-OCT devices, such as Cirrus and Spectralis, is not possible owing to differences in the segmentation software, in the divisions of the macula performed, and in the parameters provided. 

To the best of our knowledge, this is the first study in which N-site axonal analysis and automated retinal segmentation macular scanning were used to evaluate patients with NAION. Our data proved that in addition to pRNFL loss, NAION reduces the TMT, as well as the thickness of the GCIPL and of individual layers such as the RNFL, GCL, and IPL (Fig. 2). Several OCT parameters of both devices showed strong correlations with visual acuity. Overall, the correlations of pRNFL and both GCL and IPL thickness parameters provided by Spectralis with BCVA were stronger than those of the corresponding parameters provided by Cirrus. The only exception was the outer nasal TMT provided by Cirrus, which provided the best regression model for BCVA prediction. 

The significance of our findings is that these parameters may be an important surrogate for determining the extent of vision loss in NAION. 

Further longitudinal studies are required to describe the dynamic changes of OCT-derived variables in NAION. Identifying the location and timing of the changes could help establishing a therapeutic window for this condition. 

Acknowledgments

CSS and JJGL are PhD candidates at the Department of Surgery, Universidad de Alcalá School of Medicine (Madrid, Spain). JJGL has received study grants from Alcon, Novartis, and Merck Sharp & Dohme (MSD), and has provided unpaid consultancy to Bayer. 

Disclosure: G. Rebolleda, None; C. Sánchez-Sánchez, None; J.J. González-López, Bayer (C), Alcon (R), Novartis (R), MSD (R); I. Contreras, None; F.J. Muñoz-Negrete, None