Glaucoma Detection Ability of Ganglion Cell-Inner Plexiform Layer Thickness by Spectral-Domain Optical Coherence Tomography in High Myopia

Abstract

Purpose.: To compare the glaucoma detection ability of macular ganglion cell-inner plexiform layer (GCIPL) thickness measured with Cirrus spectral-domain optical coherence tomography (SD-OCT) with that of peripapillary retinal nerve fiber layer (RNFL) thickness in high myopia.

Methods.: In 49 highly myopic and 54 nonhighly myopic glaucoma patients—along with 78 healthy myopic subjects—two scans, including one macular scan and one peripapillary RNFL scan, were obtained using Cirrus SD-OCT. For 44 randomly selected glaucoma patients, three macular scans were taken for reproducibility measurements. The glaucoma detection abilities of macular GCIPL and peripapillary RNFL thicknesses were compared between the highly myopic and nonhighly myopic groups. Diagnostic power was assessed by area under the receiver operating characteristic (AUROC) curves and sensitivity. Repeatability was assessed by intraclass correlation coefficient (ICC) and coefficient of variation (CV).

Results.: All of the macular GCIPL and peripapillary RNFL thickness measurements excepting the 3 o'clock peripapillary RNFL sector showed an AUROC over 0.5. The best parameters for discriminating normal from glaucomatous eyes were inferior RNFL (0.906) and inferotemporal GCIPL (0.852) thickness in the highly myopic group, and average RNFL (0.920) and minimum GCIPL (0.908) thickness in the nonhighly myopic group. The best peripapillary RNFL and macular GCIPL thickness parameters showed no statistically significant differences. All of the ICCs of the macular GCIPL ranged between 0.96 and 0.99, and the CV was <3%.

Conclusions.: In cases of high myopia, the glaucoma detection ability of macular GCIPL thickness was high and comparable with that of peripapillary RNFL thickness.

Introduction

Glaucoma is an optic neuropathy characterized by loss of retinal ganglion cells (RGCs) and their axons. Because it is chronic and progressive, glaucoma is a major cause of blindness worldwide. Therefore, timely detection and continuous monitoring is critically important for glaucoma patients. 

Traditionally, glaucoma has been diagnosed based on ophthalmoscopy, optic disc photography, and visual field (VF) tests. However, these methods do not reflect the exact extent of RGC loss as there is no proportional relationship between RGC loss and visual sensitivity. ¹ Furthermore, clinically detectable structural change can precede the onset of VF loss by up to 5 years. ^2,3 An accurate method of measuring RGC loss, then, would enable early detection of glaucoma and monitoring of its progression. The macula offers a theoretical advantage over other regions in this respect given that more than 50% of all RGCs are concentrated and multilayered there and that RGC bodies are 10 to 20 times the diameter of their axons. ^4,5  

Several recent studies have shown that measurement of macular ganglion cell complex (GCC) thickness using optical coherence tomography (OCT; RTVue; Optovue, Inc., Fremont, CA) offers an effective glaucoma detection ability comparable with that of the peripapillary RNFL. ^6–13 GCC is defined as the sum of the three innermost retinal layers: the RNFL, ganglion cell layer (GCL), and inner plexiform layer (IPL). 

We used a commercially available spectral-domain (SD) OCT device (Cirrus; Carl Zeiss Meditec, Dublin, CA) that measures the thickness of the macular ganglion cell-inner plexiform layer (GCIPL) using a ganglion cell analysis (GCA) algorithm. GCIPL thickness is sum of the GCL and IPL, and thus is less influenced by RNFL thickness variation than is GCC thickness. Previous studies have demonstrated that the GCA algorithm of Cirrus SD-OCT can successfully measure the macular GCIPL thickness with excellent intervisit reproducibility and furthermore, that it has a good glaucoma detection ability comparable to that of peripapillary RNFL thickness and optic nerve head (ONH) parameters. ^14–16  

In highly myopic patients, there is variation of ONH size as well as structural variation such as peripapillary atrophy. Consequently, peripapillary RNFL thickness measurement is inaccurate and usually thinner in highly myopic compared with nonmyopic patients. ^17–22 What is more, subtle changes of optic disc cannot easily be detected by ophthalmoscopy or disc photography, due to the variability of ONH size and shape. Although Kim et al. ⁷ and Shoji et al. ^12,13 showed that macular GCC thickness offers a glaucoma detection ability comparable or superior to that of peripapillary RNFL thickness in highly myopic patients, little is known about macular GCIPL thickness or its diagnostic power in high myopia. In the present study, we utilized an SD-OCT device (Carl Zeiss Meditec) to compare the glaucoma detection ability of macular GCIPL thickness and peripapillary RNFL thickness in high myopia. 

Methods

Participants

This was a cross-sectional study conducted in conformity with all laws; institutional review board approval was obtained, and the research adhered to the tenets of the Declaration of Helsinki. 

The subjects were consecutively enrolled from the glaucoma clinic of Seoul National University Hospital from October 2011 to April 2012. Healthy control subjects who has been recruited as participants in routine eye examinations agreed to enroll in this study. All of the subjects underwent a complete ophthalmic examination, which included measurement of visual acuity (VA) and IOP (by Goldmann applanation tonometry); noncycloplegic refraction (Autorefractor KR-8900; Topcon Corporation, Tokyo, Japan); axial length (AL; Axis II PR; Quantel Medical, Inc., Bozeman, MT); slit lamp examination; gonioscopy; dilated fundus examination; color disc photography; red-free RNFL photography (Vx-10; Kowa Optimed, Tokyo, Japan); Humphrey Visual Field Analyzer (Carl Zeiss Meditec) using the Swedish interactive thresholding algorithm (SITA) 30-2; and an SD-OCT device (Carl Zeiss Meditec). The refraction data was converted to spherical equivalents (SE), and the subjects were divided into two groups: a highly myopic group (spherical equivalent ≤ −6.00 diopters [D] and > −20.00 D) and a nonhighly myopic group (spherical equivalent > −6.00 D and < −0.25 D). 

The inclusion criteria for patients with glaucoma were age ≥18 years, best-corrected visual acuity of at least 20/40 in the study eye, and open angle confirmed by gonioscopy. Subjects with a history or evidence of retinal pathology, diabetes, or nonglaucomatous optic nerve diseases were excluded, as were eyes that had undergone previous laser therapy or ocular surgery. For each of the subjects whose eyes both met the inclusion criteria, only one eye was randomly selected. 

Glaucomatous eyes were defined as glaucomatous damage to the optic disc as accompanied by two corresponding and reliable abnormal VF examinations with or without elevated IOP. Glaucomatous optic disc changes were defined on stereoscopic color disc photography as a large cupping (>0.7 vertical cup/disc ratio), cup/disc asymmetry between the glaucomatous and normal eyes greater than 0.2, neuroretinal rim thinning, notching, or excavation. In the case of the highly myopic group, two glaucoma specialists (CYJ, PKH) assessed stereoscopic disc photography in a masked fashion, without knowledge of the clinical data or OCT results. Only subjects classified by both observers as manifesting glaucomatous optic disc changes were included in the study. The criteria for glaucomatous VF defect were as follows: glaucoma hemifield test results outside normal limits; a pattern standard deviation (PSD) of P < 0.05; a cluster of three or more nonedge contiguous points in the pattern deviation plot in the same hemifield with P < 0.05, including one or more with P < 0.01. A VF was defined as reliable when fixation losses were <20%, and each of the false-positive and false-negative rates were <15%. 

Normal eyes were defined as those with no history of ocular symptoms, disease, or intraocular surgery. Normal eyes also had an IOP less than 22 mm Hg, an absence of glaucomatous optic disc appearance as determined by two masked observers, and no perimetric defects, irrespective of the spherical equivalents. 

OCT Measurements

Imaging was obtained using an SD-OCT device (Carl Zeiss Meditec). Two scans, including one macular scan (macular cube 200 × 200 protocol) and one peripapillary RNFL scan (optic disc cube 200 × 200 protocol), were acquired through a dilated pupil. For reproducibility measurements, three macular cube scans were consecutively taken for 44 randomly selected glaucoma patients by the same examiner on the same day. 

The GCA algorithm, included in the software version of the SD-OCT device (Carl Zeiss Meditec) used, detects and measures macular GCIPL thickness within a 6 × 6 × 2 mm cube centered on the fovea. In this cube, the annulus has an inner vertical diameter of 1 mm and an outer diameter of 4 mm, and an inner horizontal diameter of 1.2 mm and an outer diameter of 4.8 mm. ¹⁴ The size of the inner ring was chosen so as to exclude the area wherein the macular GCIPL is very thin and difficult to detect accurately, whereas the dimensions of the outer ring were selected so as to include the area wherein the macular GCIPL is thickest in a normal eye. ^23,24 The GCA algorithm detects the outer boundary of the RNFL as well as the outer boundary of the IPL. The difference between these two outer boundaries yields the combined thickness of the RGC layer and the IPL. ¹⁴  

The optic disc cube 200 × 200 program obtains optic disc images through a 6 × 6 × 2 mm cube of data by means of 200 × 200 axial scans. From this data, the system extracts a B-scan in a 3.46-mm diameter circle, and the algorithm automatically detects the circle and positions it around the optic disc. ²⁵ After the anterior and posterior boundaries of the RNFL are detected, the system calculates the RNFL thickness at each point on the circle. 

The following macular GCIPL thickness measurements were analyzed: average, minimum, and sectoral (superonasal, superior, superotemporal, inferotemporal, inferior, and inferonasal). As for the peripapillary RNFL thickness measurements, the average thickness, superior, inferior, temporal, and nasal quadrant thicknesses, along with the 12-clock-hour thickness were included in the analysis. 

The software analyzes the values, compares them with the device's internal normative database, and generates a color-coded significance map. RNFL and GCIPL thickness readings in the normal range are shown in green; borderline values (at the 5% level) in yellow; and abnormal values (at the 1% level) in red. Images with a signal strength less than 6 on either the RNFL or macular scan, visible eye motion, blinking artifacts, or algorithm segmentation failure (i.e., on careful visual inspection of horizontal cross-sectional images on an output sheet, images with missing parts, misplacement of boundaries between retinal layers, or images showing seemingly distorted anatomy that resulted in readings of zero or otherwise abnormally low value), were excluded from the study. 

Statistics

The average values of the peripapillary RNFL and macular GCIPL thicknesses were compared between normal and glaucomatous eyes and between the highly myopic and nonhighly myopic groups using the Student's t-test and Mann-Whitney U test for independent variables. For multiple comparisons, we used Bonferroni correction to adjust for type I error. Abnormal peripapillary RNFL and macular GCIPL thickness, represented as yellow or red on the color scale, were used to calculate the sensitivity and specificity percentage; and sensitivity differences among all of the parameters were determined using the McNemar test. Also, a partial correlation coefficient was used to determine the relationship between the spherical equivalent and GCIPL thickness. 

To compare the diagnostic ability of peripapillary RNFL with that of macular GCIPL thickness in both the highly myopic and nonhighly myopic groups, area under the receiver operating characteristic (AUROC) curves were calculated. The resultant AUROC differences were tested using a statistical software package (MedCalc v. 12.0; MedCalc Statistical software, Marakierke, Belgium). Reproducibility was assessed with intraobserver means, test-retest standard deviation (TRTSD), the coefficient of variation (CV), and the intraclass correlation coefficient (ICC). 

All statistical analyses were performed using statistical analysis software (SPSS 18.0; SPSS, Inc., Chicago, IL): P values less than .05 were considered statistically significant. 

Results

Participants

During the enrollment period, a total 207 eyes were examined. Ten eyes were excluded from the study because of retinal disease, as well as four eyes due to optic nerve disease. Additionally, eleven eyes were excluded owing to unreliable VF, and one eye due to poor OCT signal strength. A total of 181 eyes were included in the analysis. The subjects, as above-stated, were divided into two, highly myopic (n = 71) and nonhighly myopic (n = 110) groups. There were 78 normal controls (22 and 56 in the highly myopic and nonhighly myopic groups, respectively) and 103 open-angle glaucoma patients. 

As expected, the VF mean deviation (MD) and PSD were significantly worse in the glaucomatous eyes of both groups (Table 1). However, the other parameters did not significantly differ between the normal and glaucomatous eyes in either group. Further, there was no significant difference between the highly myopic and nonhighly myopic groups in visual acuity (0.04 ± 0.13, highly myopic; 0.06 ± 0.13, nonhighly myopic; P = 0.33) or IOP (13.30 ± 2.84, highly myopic; 13.58 ± 2.76, nonhighly myopic; P = 0.51); though they did differ, and significantly, with regard to age (45.79 ± 12.60, highly myopic; 51.52 ± 13.13, nonhighly myopic; P < 0.01); spherical equivalent (−8.81 ± 2.97, highly myopic; −2.46 ± 1.69, nonhighly myopic; P < 0.01); and axial length (26.63 ± 1.08, highly myopic; 24.49 ± 1.29, nonhighly myopic; P < 0.01). 

Table 1. 

 

Characteristics of Patients

	Highly Myopic Group, n = 71			Nonhighly Myopic Group, n = 110
	Normal, n = 22	Glaucoma, n = 49	P Value*	Normal, n = 56	Glaucoma, n = 54	P Value†	P Value‡
Age, y	44.05 (15.14)	46.57 (11.37)	0.50	49.27 (13.42)	53.85 (12.52)	0.07	<0.01
Female, n, %	13 (59.1)	17 (34.7)	0.05	29 (51.8)	25 (46.3)	0.57	0.37
VA, logMAR	0.07 (0.15)	0.03 (0.12)	0.30	0.04 (0.11)	0.08 (0.15)	0.16	0.33
IOP, mm Hg	13.55 (2.80)	13.19 (2.86)	0.63	13.60 (2.58)	13.56 (2.98)	0.95	0.51
SE, D	−9.07 (2.70)	−8.70 (3.11)	0.61	−2.38 (1.65)	−2.55 (1.74)	0.59	<0.01
AL, mm	26.16 (1.07)	26.83 (1.04)	0.06	24.23 (1.21)	24.73 (1.34)	0.13	<0.01
MD, dB	−0.94 (2.70)	−7.44 (4.85)	<0.01	−0.22 (1.67)	−7.31 (6.64)	<0.01	0.01
PSD, dB	2.45 (1.12)	8.90 (4.73)	<0.01	1.87 (0.64)	9.00 (4.36)	<0.01	0.01
Disc area, mm2	2.07 (0.54)	1.91 (0.59)	0.29	2.06 (0.39)	1.97 (0.48)	0.28	0.52

 The data are given as mean (SD).

*  Value for comparison of normal eyes and glaucoma in highly myopic group.

†  Value for comparison of normal eyes and glaucoma in nonhighly myopic group.

‡  Value for comparison of highly myopic and nonhighly myopic groups.

Reproducibility

For 44 randomly selected glaucoma patients, all of the ICCs of the macular GCIPL thicknesses ranged between 0.96 and 0.99, and the CV were <3%. The TRTSD was lowest for the average GCIPL (0.72 μm) and highest for the minimum GCIPL thickness (1.39 μm; Table 2). 

Table 2. 

 

Intraobserver Means, TRTSD, CV, and ICC of Macular GCIPL Thickness

Parameters	Mean	TRTSD, μm	CV, %	ICC, %
Superonasal GCIPL	76.6	1.03	1.3	0.99
Superior GCIPL	73.2	0.89	1.2	0.99
Superotemporal GCIPL	71.3	1.31	1.8	0.97
Inferotemporal GCIPL	66.8	0.98	1.5	0.99
Inferior GCIPL	67.0	1.28	1.9	0.97
Average GCIPL	71.6	0.72	1.0	0.98
Minimum GCIPL	62.8	1.39	2.2	0.96

OCT Measurements

The peripapillary RNFL thickness significantly differed between the normal and glaucomatous eyes in both groups (all P < 0.05), except for the nasal quadrant in the highly myopic group (P = 1.000) and the temporal quadrant in both the highly myopic and nonhighly myopic groups (P = 0.060 and 0.235, respectively). The average, superior, and inferior quadrant peripapillary RNFL thicknesses were significantly thinner in the highly myopic group than in the nonhighly myopic group. In the highly myopic group, significant differences were found in the peripapillary clock hour RNFL thicknesses for the 10- to 1-o'clock and 5- to 7-o'clock sectors (corresponding to the superior and inferior quadrants). Within the nonhighly myopic group, all sectors except 2, 3, 8, and 9 o'clock (corresponding to the nasal and temporal quadrants) showed significant differences between the normal and glaucomatous eyes. However, the nasal and temporal peripapillary clock hour RNFL sectors except 6 and 12 o'clock did not significantly differ between the highly myopic and the nonhighly myopic groups (Table 3). 

Table 3. 

 

Peripapillary RNFL and macular GCIPL Thickness Obtained Using SD-OCT

	Highly Myopic Group, n = 71			Nonhighly Myopic Group, n = 110
	Normal, n = 22	Glaucoma, n = 49	P Value*	Normal, n = 56	Glaucoma, n = 54	P Value†	P Value‡
Main RNFL thickness parameters, μm
Average	89.18 (10.97)	69.82 (10.97)	<0.001	93.48 (9.12)	72.74 (11.59)	<0.001	0.021
Superior	109.41 (21.05)	81.27 (21.21)	<0.001	118.09 (16.54)	88.00 (16.70)	<0.001	0.008
Nasal	62.82 (9.34)	61.41 (12.03)	1	66.70 (8.63)	60.96 (9.04)	0.023	1
Inferior	107.36 (15.60)	74.24 (19.37)	<0.001	119.95 (16.00)	78.63 (26.37)	<0.001	0.006
Temporal	75.09 (17.42)	62.92 (13.89)	0.06	68.96 (12.53)	62.44 (13.33)	0.235	1
RNFL clock hours thickness parameters, μm
12 Sup	103.77 (25.33)	75.98 (24.05)	0.001	115.20 (25.63)	89.46 (21.11)	<0.001	<0.001
1	110.64 (30.06)	81.33 (25.61)	0.002	116.11 (25.88)	85.56 (23.70)	<0.001	0.45
2	74.95 (17.37)	67.63 (17.85)	1	78.32 (15.47)	70.50 (15.93)	0.255	1
3 Nasal	56.50 (15.43)	57.04 (12.62)	1	55.66 (11.46)	55.56 (10.06)	1	1
4	64.86 (17.71)	59.98 (10.83)	1	65.77 (14.56)	57.76 (7.62)	0.012	1
5	102.18 (35.22)	71.37 (15.18)	0.015	109.11 (28.62)	74.87 (22.58)	<0.001	0.25
6 Inf	109.27 (24.72)	75.90 (22.28)	<0.001	128.38 (24.38)	82.98 (32.59)	<0.001	0.001
7	110.55 (25.71)	73.61 (29.50)	<0.001	121.50 (27.57)	78.46 (34.12)	<0.001	0.15
8	74.68 (22.06)	61.86 (16.55)	0.209	67.64 (17.17)	63.54 (20.40)	1	1
9 Temp	62.36 (13.86)	56.80 (14.57)	1	56.16 (8.71)	56.37 (13.17)	1	1
10	87.00 (20.94)	69.20 (17.91)	0.012	83.00 (17.34)	69.91 (20.40)	0.011	1
11	113.80 (27.57)	81.27 (21.21)	0.018	122.38 (24.55)	88.59 (30.40)	<0.001	0.65
Macular GCIPL parameters, μm
SN	77.55 (10.45)	69.47 (12.56)	0.26	82.54 (7.13)	74.48 (11.38)	<0.001	0.004
S	76.45 (7.85)	67.71 (10.05)	0.014	82.20 (7.69)	70.20 (15.35)	<0.001	0.048
ST	77.86 (8.34)	66.71 (11.78)	0.013	80.31 (7.19)	68.00 (12.81)	<0.001	0.329
IT	76.59 (13.62)	61.49 (11.18)	<0.001	80.96 (6.73)	63.96 (11.28)	<0.001	0.049
I	72.82 (7.89)	61.92 (11.00)	0.002	79.88 (7.09)	65.83 (10.10)	<0.001	<0.001
IN	75.91 (9.39)	65.96 (14.05)	0.022	79.93 (6.27)	70.98 (11.26)	<0.001	0.007
Avg	72.86 (17.50)	65.61 (9.89)	0.733	80.96 (6.43)	68.85 (9.69)	<0.001	0.001
Min	70.36 (7.42)	55.88 (14.37)	<0.001	77.84 (8.78)	59.46 (14.44)	<0.001	0.006

 The data are given as mean (SD). Avg, average; I, inferior; IN, inferonasal; Inf, inferior; IT, inferotemporal; Min, minimum; S, superior; SN, superonasal; ST, superotemporal; Sup, superior; Temp, temporal.

*  Value for multiple comparison of normal eyes and glaucoma in highly myopic group.

†  Value for multiple comparison of normal eyes and glaucoma in nonhighly myopic group.

‡  Value for multiple comparison of highly myopic and nonhighly myopic groups.

The macular GCIPL thickness, comparing the normal and glaucomatous eyes in both the highly myopic and nonhighly myopic groups, was significantly different (all P < 0.05), except for the superonasal (P = 0.260) and average thickness (P = 0.733) within the highly myopic group. All of the GCIPL thicknesses except superotemporal sector (P = 0.329) were significantly thicker in the nonhighly myopic group than in the highly myopic group (Table 3). 

According to a partial correlation analysis adjusted for age and peripapillary RNFL thickness performed on 78 normal eyes, there were statistically significant relationships between the spherical equivalent and superonasal GCIPL (r = 0.296, P = 0.01); inferior GCIPL (r = 0.345, P < 0.01); and inferonasal GCIPL (r = 0.314, P < 0.01) thicknesses. However, there were no such relationships between the spherical equivalent and any of the other GCIPL parameters. 

Diagnostic Ability of Peripapillary RNFL and GCIPL Macular Thicknesses

The sensitivities and specificities of the SD-OCT device (Carl Zeiss Meditec) parameters are shown in Table 4. The superior and inferior peripapillary RNFL thicknesses showed better sensitivity than the nasal and temporal peripapillary RNFL thicknesses in both groups (P < 0.001, all parameters). The superonasal and inferonasal GCIPL showed worse sensitivity than the other GCIPL sectors (P < 0.05, all parameters), except between inferonasal and superior GCIPL sectors of both groups (P = 0.34, P = 0.63, respectively). When we used the most specific criteria (only “red” color scale as abnormal), all of the peripapillary RNFL and GCIPL thicknesses in both groups showed decreased sensitivity and increased specificity, as expected. 

Table 4. 

 

Discriminating Ability of Peripapillary RNFL and Macular GCIPL Thickness for Glaucoma Detection

	Highly Myopic Group, n = 71		Nonhighly Myopic Group, n = 110
	Sensitivity (%)	Specificity (%)	Sensitivity (%)	Specificity (%)
Main RNFL thickness parameters, μm
Average	75.5 (62.3–87.7)	86.4 (70.8–100.0)	68.5 (55.7–81.3)	98.2 (94.5–100.0)
Superior	75.5 (62.3–87.7)	86.4 (70.8–100.0)	59.3 (45.7–72.8)	98.2 (94.5–100.0)
Nasal	18.4 (7.3–30.2)	81.8 (64.3–99.3)	13.0 (3.7–22.2)	98.2 (94.5–100.0)
Inferior	77.6 (67.2–91.1)	72.7 (52.5–92.9)	74.1 (62.0–86.1)	94.6 (88.3–100.0)
Temporal	16.3 (5.7–27.6)	77.3 (58.3–96.3)	9.4 (2.5–19.8)	98.2 (94.5–100.0)
RNFL clock hours thickness parameters, μm
12 Sup	61.0 (39.5–68.8)	86.4 (70.8–100.0)	31.5 (18.7–44.3)	91.1 (83.1–98.8)
1	36.7 (21.4–49.5)	86.4 (70.8–100.0)	44.4 (30.8–58.1)	91.1 (83.1–98.8)
2	28.6 (15.8–42.5)	77.3 (58.3–96.3)	16.7 (6.4–26.9)	94.6 (88.3–100.0)
3 Nasal	6.1 (0.0–13.4)	100	3.7 (0.0–8.9)	96.4 (91.3–100.0)
4	10.2 (1.5–19.4)	95.5 (86.0–100.0)	7.4 (0.2–14.6)	98.2 (94.5–100.0)
5	46.9 (33.3–62.6)	77.3 (58.3–96.3)	37.0 (23.7–50.3)	94.6 (88.3–100.0)
6 Inf	65.3 (52.8–80.5)	72.7 (52.5–92.9)	64.8 (51.7–80.0)	91.1 (83.1–98.8)
7	73.5 (62.3–87.7)	95.5 (86.0–100.0)	72.2 (59.9–84.6)	100
8	18.4 (7.3–30.2)	86.4 (70.8–100.0)	9.3 (1.3–17.2)	98.2 (94.5–100.0)
9 Temp	6.1 (0.0–13.4)	81.8 (64.3–99.3)	5.6 (0.0–11.9)	96.4 (91.3–100.0)
10	20.4 (8.9–32.8)	90.9 (77.9–100.0)	22.2 (10.8–33.7)	98.2 (94.5–100.0)
11	65.3 (50.0–78.6)	81.8 (64.3–99.3)	53.7 (40.0–67.4)	98.2 (94.5–100.0)
Macular GCIPL parameters, μm
SN	49.0 (33.3–62.6)	54.5 (31.9–77.1)	33.3 (20.3–46.3)	92.9 (85.6–100.0)
S	61.2 (46.0–74.8)	50.0 (27.3–72.7)	46.3 (32.6–60.0)	94.6 (88.3–100.0)
ST	69.4 (55.1–82.4)	63.6 (41.8–85.5)	57.4 (43.8–71.0)	85.7 (75.8–95.1)
IT	79.6 (67.2–91.1)	68.2 (47.0–89.3)	75.9 (64.1–87.7)	91.1 (83.1–98.8)
I	67.3 (52.8–80.5)	68.2 (47.0–89.3)	63.0 (49.7–76.7)	94.6 (88.3–100.0)
IN	53.1 (37.4–66.7)	54.5 (31.9–77.1)	44.6 (29.0–56.2)	82.1 (71.3–92.3)
Avg	71.4 (57.5–84.2)	54.5 (31.9–77.1)	61.1 (47.7–74.5)	92.9 (85.6–100.0)
Min	81.6 (70.0–92.7)	40.9 (18.6–63.2)	77.8 (66.3–89.2)	91.1 (83.1–98.8)

For both groups, all of the peripapillary RNFL thicknesses excepting the 3 o'clock peripapillary RNFL sector in the highly myopic group showed AUROC curves above 0.5, and all of the macular GCIPL thicknesses showed AUROC curves above 0.5. None of the macular GCIPL or peripapillary RNFL thickness AUROC curves showed any significant differences between the highly myopic and nonhighly myopic groups (Table 5). The best parameters for discriminating normal from glaucomatous eyes were, in the highly myopic group, inferior peripapillary RNFL (0.906) and inferotemporal GCIPL thickness (0.852); and in the nonhighly myopic group, the best parameters were average RNFL (0.920) and minimum GCIPL thickness (0.908). However, there were no statistically significant differences between the best GCIPL thickness and the best quadrant peripapillary RNFL thickness in either group (Figs. 1, 2). 

Figure 1

 

Receiver operating characteristic (ROC) curves of best parameters of SD-OCT device in highly myopic group. There were no statistical significances between the best GCIPL thickness and the best peripapillary RNFL thickness.

Figure 2

 

ROC curves of best parameters of SD-OCT device in nonhighly myopic group. There were no statistical significances between the best GCIPL thickness and the best peripapillary RNFL thickness.

Table 5. 

 

AUROC Curve Values between Normal and Glaucomatous Eyes

	Highly Myopic Group, n = 71	Nonhighly Myopic Group, n = 110	P Value
Main RNFL thickness parameters, μm
Average	0.899 (0.041)	0.920 (0.025)	0.66
Superior	0.816 (0.057)	0.884 (0.031)	0.29
Nasal	0.541 (0.071)	0.672 (0.052)	0.14
Inferior	0.906 (0.035)	0.891 (0.034)	0.76
Temporal	0.682 (0.066)	0.628 (0.053)	0.52
RNFL clock hours thickness parameters, μm
12 Sup	0.785 (0.062)	0.773 (0.044)	0.87
1	0.745 (0.062)	0.808 (0.041)	0.4
2	0.595 (0.073)	0.664 (0.052)	0.44
3 Nasal	0.419 (0.075)	0.504 (0.055)	—*
4	0.537 (0.075)	0.663 (0.051)	0.16
5	0.785 (0.068)	0.824 (0.038)	0.62
6 Inf	0.839 (0.049)	0.851 (0.038)	0.85
7	0.840 (0.046)	0.840 (0.041)	1
8	0.671 (0.066)	0.551 (0.055)	0.16
9 Temp	0.603 (0.073)	0.505 (0.056)	0.29
10	0.756 (0.064)	0.678 (0.052)	0.34
11	0.745 (0.062)	0.807 (0.042)	0.41
Macular GCIPL parameters, μm
SN	0.647 (0.067)	0.719 (0.049)	0.39
S	0.741 (0.060)	0.782 (0.044)	0.58
ST	0.756 (0.059)	0.832 (0.040)	0.29
IT	0.852 (0.044)	0.891 (0.033)	0.48
I	0.797 (0.053)	0.860 (0.036)	0.33
IN	0.694 (0.062)	0.737 (0.047)	0.58
Avg	0.754 (0.063)	0.854 (0.035)	0.17
Min	0.830 (0.048)	0.908 (0.029)	0.16

 The data are given as AUROC (SE).

*  The AUROC curve of the highly myopic group was not significantly better than 0.5, and thus they were omitted from the statistical comparison.

Discussion

The present study was designed for the main objective of comparing the glaucoma detection ability of macular GCIPL thickness with that of peripapillary RNFL thickness in highly myopic patients. We confirmed that the peripapillary RNFL and macular GCIPL have similar diagnostic ability for discriminating between healthy and glaucomatous eyes in both highly myopic and nonhighly myopic patients. 

Generally, peripapillary RNFL measurements are known to a good parameter for glaucoma discrimination, because nearly all of the axons arising from RGCs radiate toward the ONH, and are not affected by macular pathologies such as age-related macular degeneration or diabetic retinopathy. However, as the optic disc cube protocol of the SD-OCT device (Carl Zeiss Meditec) uses a 3.46-mm fixed-diameter circle, peripapillary RNFL measurements in highly myopic patients are not always reliable. The optic disc of highly myopic eyes is frequently associated with tilting, oval configuration, and peripapillary atrophy, ^17,26 and it can influence the disc margin definition algorithms. Furthermore, as RNFL thickness decreases with increasing distance from the disc margin, optic disc size also affects peripapillary RNFL thickness. ^{20,22,27–29} Although some study reported no relationship between peripapillary RNFL thickness and spherical equivalent, ^29,30 highly myopic patients usually tend to have thinner RNFL than the normal population. ^{18,19,21,22,31–34} Also, Shoji et al. ¹³ recently showed that the diagnostic accuracy of peripapillary RNFL measurement in highly myopic eyes decreased significantly. 

Macular GCIPL thickness is a new parameter that reflects the thickness of RGC bodies directly affected by glaucoma with their axons. Because it is still unknown whether axonal pathology precedes or follows RGC loss in glaucoma, ²³ assessing changes of RGC bodies also are important. Ganglion cell densities reach 32,000 to 38,000 cells/mm² in a horizontally oriented elliptical ring 0.4 to 2.0 mm from the foveal center; and the human retina contains more than 1 million RGCs. ²⁴ Also, RGC bodies are more than 10 to 20 times the diameter of their axons, and the RGC layer in the macula is more than one cell thick. ^35,36 Because the macular region contains more than 50% of all RGCs, the macula is expected to be the best location for evaluating RGC changes. Although highly myopic patients tend to have thinner parafoveal and perifoveal thicknesses as well as different topographic profiles from those of nonmyopic subjects, ^37–39 recent studies have shown that another macular parameter, GCC, offers a comparable or superior glaucoma detection ability to peripapillary RNFL thickness in highly myopic patients. ^7,12,13 Notably, Shoji et al. ¹³ showed that the diagnostic accuracy of macular GCC does not decrease in highly myopic patients, unlike peripapillary RNFL thickness. But there is as yet no study that has compared macular GCIPL thickness with peripapillary RNFL for high myopia. 

Studies on the glaucoma diagnostic ability of macular GCC in highly myopic patients show conflicting results. Kim et al. ⁷ reported a comparable glaucoma detection ability of macular GCC thickness relative to peripapillary RNFL in highly myopic subjects. However, Shoji et al. ¹² found that global loss volume of macular GCC showed statistically significant better AUROC than average peripapillary RNFL thickness for perimetric glaucoma detection in highly myopic patients. Another study by Shoji et al. ¹³ showed that only peripapillary RNFL measurement had a decreased ability to detect glaucoma in a highly myopic group, whereas macular GCC measurements efficiently detected glaucoma in both highly myopic and emmetropic groups. The diagnostic ability of GCC thickness might actually be superior to total macular thickness measurement, because glaucoma preferentially affects the inner retinal layers; moreover, macular thickness decrease is believed to be due to the loss of RGC and RNFL. ^4,5 But, because the RNFL is included in the GCC, the diagnostic power of macular GCC is influenced by the RNFL. Unlike GCC measurements, the GCA algorithm segments the macular GCIPL without including the RNFL. So there is a theoretical possibility that macular GCIPL could offer a better diagnostic ability to reflect RGC change. However, previous studies with highly myopic patients have shown best macular GCC parameters AUROC ranging from 0.889 to 0.954, which are larger than our result (0.852 at inferotemporal GCIPL). However, because of the differences in the characteristics of the study subjects and anatomic structures that each measures, our GCIPL thickness results cannot be directly compared with those of macular GCC. Resolution of these issues must await further study comparing the diagnostic accuracy of GCC and GCIPL thicknesses in the same subjects. 

In our study, diagnostic accuracy was measured by means of ROC analysis. Previous studies have shown the commercial SD-OCT device's (Carl Zeiss Meditec) RNFL parameter AUROC ranged from 0.60 to 0.98, depending on the parameters and the characteristics of the participants. ^{15,16,40–42} Moreover, in those investigations, inferior RNFL thickness often demonstrated the best ability for discriminating normal eyes from glaucoma, with sensitivities between 38% and 82% for specificities of 90% or higher. ^15,40–47 And in other previous studies, macular GCIPL thickness showed AUROC of 0.64 to 0.99, with sensitivities ranging from 60% to 95%, ^15,16 which are results similar to those reported here for the present study. In our study, the best SD-OCT (Carl Zeiss Meditec) parameters with the largest AUROC were inferior RNFL thickness (AUROC = 0.906); the 7-o'clock sector (AUROC = 0.840); and the inferotemporal GCIPL (AUROC = 0.852) in the highly myopic group, and average RNFL thickness (AUROC = 0.920); the 6-o'clock sector (AUROC = 0.851); and the minimum GCIPL thickness (AUROC = 0.908) in the nonhighly myopic group. These results were consistent with our findings that the nasal and temporal peripapillary RNFL thicknesses showed no significant differences between normal and glaucomatous eyes, especially in the highly myopic group. 

It is known that among the six sectoral GCIPL parameters, the superonasal sector has the thickest GCIPL and the inferior sector the thinnest. ^16,23 This is consistent with other findings, which are that RGC densities in the nasal retina exceeded those at corresponding eccentricities in the temporal retina by more than 300%, and that superior exceeded inferior by 60%. ²⁴ Because glaucomatous optic disc change usually starts from the superior or inferior area ^48,49 —and RGCs in the superior or inferior temporal macula project their axons in an arcuate pattern to the superotemporal or inferotemporal portions of the ONH—it is reasonable that the inferotemporal GCIPL (RGC body) and the corresponding inferior peripapillary RNFL (RGC axon) showed the best AUROC in the highly myopic group. However, in the nonhighly myopic group, the average RNFL and the minimum GCIPL showed the best AUROC. Because the minimum GCIPL thickness is thought to represent a focal loss of RGC volume, it is possible that a focal loss of RGC would be easily detectable in a nonhighly myopic group, owing to diffuse thinning of the peripapillary RNFL and macular GCIPL in highly myopic eyes. This is consistent with the findings of studies by Mwanza et al. ¹⁵ and Takayama et al. ¹⁶ that indicate minimum and inferotemporal GCIPL thickness showed the best AUROC among GCIPL parameters. Also, previous studies have reported that inferior and temporal macular regions showed greater susceptibility to glaucomatous damage. ^50–52 We found that the diagnostic abilities of the GCIPL parameters were not superior to those of the peripapillary RNFL parameters. However, because it remains unknown whether axonal pathology precedes or follows RGC loss in glaucoma, ²³ and because macular GCIPL and peripapillary RNFL thickness represent the RGC body and axons respectively, these two parameters can be considered complementary in diagnosis of glaucoma. 

In our study, the average, superior, and inferior peripapillary RNFL along with all of the macular GCIPL thicknesses were significantly thinner in the highly myopic group than in the nonhighly myopic group. Also, the inferior peripapillary RNFL and all of the macular GCIPL thicknesses, excepting the superotemporal GCIPL, were thinner in the normal control of the highly myopic group than in the normal control of the nonhighly myopic group. However, the SD-OCT device (Carl Zeiss Meditec) used does not correct for axial length during acquisition or analysis, and so magnification errors can occur. In our study, the mean axial length in the highly myopic group was 26.63 (range: 24.52–29.29 mm). After adjustment by a factor of 1.10 (1.00–1.21) according to the Bennett formula, ⁵³ the inferior peripapillary RNFL and macular GCIPL thicknesses in the normal highly myopic group increased from 0% to 21% and lost their statistical significance relative to the normal control of the nonhighly myopic group. It is a limitation of this study that the decrease in the measured peripapillary RNFL and GCIPL thicknesses in the highly myopic group might be the effect of magnification on the area measured, which can overestimate the glaucoma diagnostic ability of high myopia. In this study, however, some of the macular GCIPL thicknesses were positively correlated with the spherical equivalent; and in fact, other studies have shown conflicting results with respect to the relationship between myopia and OCT parameters after magnification correction. ^32,54,55 Further studies assessing this relationship will be required. 

In this study, the highly myopic group was younger than the nonhighly myopic group. This was possibly due to the highly myopic group's many patients who had been referred to the glaucoma clinic due to a variation in the ONH shape. Also, because the highly myopic subjects had many other ophthalmic conditions such as peripheral retinal lattice degeneration or floater symptoms, they received their ophthalmic exams earlier and more frequently than the nonhighly myopic subjects. However, given that age showed no significant differences between normal and glaucomatous eyes in either the highly myopic or the nonhighly myopic group, the diagnostic ability of OCT parameters might not be influenced by age. Also, one of the limitations of this study is that the findings were obtained on Korean subjects lacking ethnic variety. Another limitation of this study is the small number of normal controls, especially in the highly myopic group and conducted only using an SD-OCT device (Carl Zeiss Meditec) for GCIPL measurement. Further large-cohort studies will be necessary in order to establish a normative database of highly myopic subjects. 

In conclusion, the glaucoma detection ability of macular GCIPL thickness in high myopia was found to be comparable with that of peripapillary RNFL thickness. Thus, the macular GCIPL thickness can be used as a complementary glaucoma diagnostic test. 

Footnotes

 The authors alone are responsible for the content and writing of the paper.

Footnotes

 Disclosure: Y.J. Choi, None; J.W. Jeoung, None; K.H. Park, None; D.M. Kim, None