Reproducibility of Thickness Measurements of Macular Inner Retinal Layers Using SD-OCT With or Without Correction of Ocular Rotation

Abstract

Purpose.: To evaluate the intervisit reproducibility of spectral-domain optical coherence tomography (SD-OCT) measurement of the macular retinal nerve fiber layer thickness (mRNFLT); combined ganglion cell layer and inner plexiform layer (GCL+IPL) thickness; and ganglion cell complex (GCC) thicknesses (sum of mRNFLT and GCL+IPL thicknesses) compared with that of circumpapillary RNFLT (cpRNFLT) and the effect of ocular rotation on reproducibility.

Methods.: SD-OCT imaging was performed twice on different days in one eye of 58 normal subjects and 73 glaucoma patients. The reproducibility was evaluated for the entire 4.8-mm × 4.8-mm macular area and subareas (upper and lower halves, 2 × 2, 4 × 4, and 8 × 8 grids), and the 360°, upper, and lower halves mean cpRNFLT with and without correction of ocular rotation.

Results.: The coefficients of variation (CVs) of GCL+IPL and GCC thickness measurements averaged below 1.0% for the entire and upper and lower half macular areas, and below 4.2% in the macular subareas in normal and glaucoma eyes, which were significantly smaller (P < 0.001) than those of mRNFLT measurements in the same areas of the same eyes. The CVs of mRNFLT measurements were significantly smaller than those of the cpRNFLT only in the lower half mean area in normal eyes. The reproducibility was minimally affected by correction of ocular rotation or presence of glaucoma.

Conclusions.: The reproducibility of the macular (GCL+IPL) and GCC thickness measurements was better than that of mRFNLT and cpRNFLT in normal and glaucoma eyes and minimally affected by correction of ocular rotation.

Introduction

Optical coherence tomography (OCT) is the leading method for imaging structural changes resulting from glaucoma because the technology shows continuous and marked developments. The short-term and intervisit reproducibility values of the average circumpapillary retinal nerve fiber layer thickness (cpRNFLT) measurements with spectral-domain (SD)-OCT have been reported as coefficients of variation (CVs) ranging from 1.0% to 2.9% ^1–10 and 1.8% to 4.0%, ^4,6,11 respectively, in normal eyes, and from 1.3% to 3.3% ^2–4,7,9,10 and 2.7% to 3.3%, ¹² in glaucoma eyes. In a time-domain (TD)-OCT study, the reproducibility of the macular retinal thickness measurements was reported to be favorably compared with that of the average cpRNFLT measurements in normal and glaucoma eyes. ¹³ Several previous SD-OCT studies ^8,14–17 of short-term reproducibility of macular thickness measurements have reported CVs ranging from 0.4% to 1.96% ^8,15,16 in normal eyes and a CV of 1.4% for the reproducibility of the macular retinal nerve fiber layer thickness (mRNFLT). ¹⁶ No study has reported the intervisit reproducibility of macular thickness measurement in normal eyes. In glaucoma eyes, the short-term and intervisit reproducibility values were reported as intraclass correlation coefficients (ICCs) of 0.998 and 0.995 for the macular retinal thickness and 0.975 and 0.878 for mRNFLT, ¹⁴ and the intervisit reproducibility was reported as a CV was 2.6% for the mRNFLT. ¹⁷  

Ishikawa et al. ¹⁸ introduced a segmentation algorithm for the macular inner retinal layer thickness with TD-OCT and showed that macular measurements can be used to diagnose glaucoma with accuracy equal to that of the cpRNFLT measurement. Although the ganglion cell layer (GCL), where glaucomatous structural change occurs, may not reliably be separated from the inner plexiform layer (IPL) at present, several previous studies have reported comparable capability of ganglion cell complex (GCC) thickness, defined as sum of the thickness of the mRNFLT, GCL, and inner plexiform layer (IPL) thickness, to cpRNFLT for detecting glaucoma. ^19,20 Mwanza et al. ¹⁷ evaluated the intervisit reproducibility of GCL+IPL complex with an OCT device (Cirrus HD-OCT; Carl Zeiss Meditec, Dublin, CA) in 50 patients with glaucoma and reported a CV of 1.8%. 

Measurements of the macular inner layer thickness might be advantageous over measurements of the cpRNFLT. Automatic determination of the center of the analysis area—that is, fovea in the macular area or barycenter of the optic disc area—which is an important benchmark for longitudinal follow-up, is technically easier in the macular area than in the optic disc area. The thickness measurements also are less affected by large retinal vessels in the macular area than in the peripapillary area. Thus, evaluation of the reproducibility of the macular inner layer thickness measurements of the GCC, GCL+IPL, and mRNFL, which are closely related to glaucomatous damage, should have clinical importance. However, there is little information about measurement of the intervisit reproducibility of the mRNFLT, GCC, and GCL+IPL thickness in normal eyes or of the GCC in glaucoma eyes. 

Physiologic ocular rotation often occurs during fundus imaging. The rotation angle varies among individuals and between fundus imaging examinations. The variations also affect the intervisit measurement reproducibility of the macular inner layer thickness or cpRNFLT. However, no previous study has evaluated the effect of physiologic ocular rotation on the measurement reproducibility. 

From a clinical standpoint, assessing the variations in thickness measurements that might occur on different days is more relevant for assessing disease progression. ²¹ The goals of the current study were to evaluate the intervisit reproducibility of SD-OCT measurement of the mRNFLT, GCL+IPL, and GCC thicknesses compared with that of the cpRNFLT and to investigate whether correcting ocular rotation improves the intervisit reproducibility in normal and glaucoma eyes. 

Materials and Methods

Participants

The data were acquired in six Japanese institutions: University of Tokyo, Tokyo; Kyoto University, Kyoto; Osaka University, Osaka; Kanazawa University, Kanazawa; Gunma University, Gunma; and Tajimi Municipal Hospital, Gifu. The institutional review board of each institution approved the study, which adhered to the tenets of the Declaration of Helsinki. Each subject provided written informed consent after the study was explained. Fifty-eight normal subjects and 73 patients with glaucoma underwent ocular examinations, including autorefractometry without cycloplegia; best-corrected visual acuity (BCVA) measurements; axial length measurement using partial coherence interferometry technology (Intraocular Lens Master; Carl Zeiss Meditec); slit-lamp examination; IOP measurement by Goldmann applanation tonometry; detailed fundus examinations; and visual field testing using the Humphrey 24-2 Swedish interactive testing algorithm standard strategy (Humphrey Field Analyzer; Carl Zeiss Meditec). The inclusion criteria for the normal subjects and patients with glaucoma were age 20 years or older; BCVA of 20/25 or better; refractive error between −6.0 and +5.0 diopters (D); no history of intraocular surgery; and no ocular diseases (except for glaucoma in the glaucoma eyes) that might affect the visual field or the OCT measurements. Subjects with an IOP of 22 mm Hg or higher, those with abnormal visual field testing results or optic disc finding, or an RNFL abnormality determined using red-free fundus photography were not included in the normal group. Glaucoma was diagnosed based on glaucomatous changes in the optic disc and/or RNFL and corresponding visual field defects. The glaucomatous fundus changes included an enlarged vertical cup-to-disc ratio, thinning of the rim width, and/or a RNFL defect. The definition of a glaucomatous visual field defect was the presence of at least one of the following according to the criteria of Anderson and Patella ²² : a pattern deviation probability plot showing a cluster of three or more points with a probability of less than 5% and at least one point with a probability less than 1% in an expected hemifield, a pattern standard deviation with a probability of less than 5%, or a glaucoma hemifield test that indicated that the field were outside the normal limits. To exclude the learning effect, results from the first session of visual field testing were not considered and eyes with unreliable test results (fixation loss or false-negative error >20%, false positive error >15%) were not included. After eligibility for study entry was confirmed, all subjects and patients underwent OCT. 

SD-OCT Imaging

Images of macular retina were obtained using an SD-OCT device (3D OCT-1000, version 2.13; Topcon, Inc., Tokyo, Japan). The SD-OCT device (Topcon, Inc.) is equipped with a nonmydriatic fundus camera function equivalent to a commercially available nonmydriatic fundus camera (TRC-NW200, Topcon, Inc.) to facilitate a direct comparison between the retinal images and the corresponding OCT tomograms. The OCT data sets were obtained using a three-dimensional (3D) scan protocol. Three-dimensional data sets centered on the fovea and optic disc were obtained with a raster scan of 512 × 128 axial scans (horizontal × vertical), which took 2.6 seconds and covered a 6 × 6-mm area with horizontal pixel spacing of 11 μm (6 mm/512) and vertical spacing of 47 μm (6 mm/128). To obtain accurate sizes on the obtained fundus images, the magnification effect in each eye was corrected according to the formula provided by the manufacturer (modified Littman's equation ²³ ), which is based on the measured refractive error, corneal radius, and axial length. First, a 3D scan centered on the optic disc was performed followed by a fundus photograph with a 45° visual angle. With intervals of approximately 1 minute, during which the subject removed his or her chin from the chinrest, the 3D scan centered on the macula was performed and a fundus photograph was taken. When acquiring the first 3D data from the peripapillary area, an experienced ophthalmologist (HH) drew the optic disc contour on the inner margin of the scleral ring (Elschnig's ring) on the color fundus photograph; the position of the optic disc contour was converted automatically into the 3D data set using the retinal vessels as guides. For the later measurements of the 3D scan in the same subject, the same position of the optic disc contour was used to ensure an identical measurement position during multiple measurements. 

On each B-scan image, the retinal surface and the interface between RNFL and GCL were drawn automatically with tools in the OCT software for both macular and circumpapillary analysis, and the interface between the IPL and inner nuclear layer was drawn automatically for macular analysis. ²⁴ The cpRNFLT was measured over approximately 1000 pixels located on the 3.4-mm circle centered on the barycenter of the optic disc. The combination of the mRNFL, GCL, and IPL was defined as the GCC. ¹⁹ An experienced examiner (AT) confirmed all segmentations. The criteria for acceptable SD-OCT images were the absence of blinking or ocular movements indicated by shifting of the vessels or disc images exceeding one diameter of the main vessels among the B-scan images or a straight line across the fundus OCT image, an image quality factor exceeding 60%, and no apparent mRNFL misalignment or GCL+IPL or GCC segmentation. 

In the current study, we also measured the ocular rotation angle, which was defined as the angle between a horizontal line including the barycenter of the optic disc and a line connecting the barycenter of the optic disc and the fovea on the fundus photographs automatically taken just after acquisition of the 3D data sets of peripapillary and macular areas, respectively (Figs. 1, 2). Eyes in which the 4.8 × 4.8-mm area centered on the fovea was included in the scan area after correction of rotation were included in the following analyses (Figs. 3A, 3B). The cpRNFLT was also evaluated after correction of rotation (Fig. 4). The SD-OCT measurements were repeated at the same time on a different day within 3 months after the initial measurement. 

Figure 1

 

The rotation angle (indicated by “a”) is defined as the angle with a horizontal line including the center of the optic disc and the line joining the center of the optic disc and the fovea on the fundus photograph obtained using OCT.

Figure 2

 

The rotation angle (indicated by “a”) is defined same as in Figure 1 for cpRNFLT analysis.

Figure 3

 

(A) 3D scan data centered on the fovea was obtained with 6 × 6-mm scan. For the reproducibility analysis, the 4.8 × 4.8-mm scan area was selected and the 4.8 × 4.8-mm scan area were divided into upper and lower half, grids of 2 × 2, 4 × 4, and 8 × 8. (B) With correction of the ocular rotation, the 4.8 × 4.8-mm scan area was rotated as shown. When the peripheral data became inadequate after the scan area was rotated, the eye was excluded from the reproducibility analysis.

Figure 4

 

The cpRNFLT was measured on the 3.4-mm circle centered on the barycenter of the optic disc. The 3.4-mm circle divided into upper- and lower-half circles. With correction of the ocular rotation, the 3.4-mm circle was rotated as shown.

Statistical Analyses

The intervisit reproducibility was evaluated using the CV, reproducibility coefficient (RC), and ICC. The RC was obtained using the following formula ²⁵ :  

 

where x_{i 1} and x_{i 2} indicate the thickness of the ith subject on the first and second measurements.  

Since structural changes due to open-angle glaucoma can be minute and limited in small locations, the reproducibility also was evaluated for subdivided areas or sectors. For macular analyses, the reproducibility for the entire area was evaluated (4.8 mm × 4.8 mm); and for subarea analyses, the entire area was divided into the upper and lower halves, with grids of 2 × 2, 4 × 4, and 8 × 8 (Fig. 2A). A Grid of 8 × 8 roughly corresponded to the grid of the central 10-2 program of Humphrey perimetry projected onto the fundus and the central four grids of the 8 × 8 grid were excluded. For circumpapillary analyses, the reproducibility was evaluated for the entire circumpapillary circle and then it was divided into two sectors of 180° each. The average CVs were compared among the results with and without rotation correction or among the results of mRNFL, GCL+IPL, GCC, and cpRNFL with the Wilcoxon test or ANOVA with Tukey's posthoc test. 

Multiple linear regression analyses were performed to identify the factors associated with reproducibility. The CVs for the average mRNFLT, GCL+IPL, and GCC thicknesses were the dependent variables and the independent variables were age, sex, axial length, IOP, average rotation angles measured on two occasions, intervisit difference in rotation angles, and mean deviation (MD) in patients with glaucoma. Statistical analyses were performed using a statistical software package (SPSS 15.0J; SPSS Japan, Inc., Tokyo, Japan). 

Results

Eyes were excluded if acceptable OCT macular or disc images could not be obtained or if the 4.8 × 4.8-mm area centered on the fovea was not included in the 6.0 × 6.0-mm measurement area at either visits. When both eyes were eligible, one eye was randomly selected. Fifty-one normal and 65 glaucoma eyes were included in the macular analysis. The demographic data from normal subjects and patients with glaucoma are shown in Table 1. The normal subjects and patients with glaucoma, respectively, were an average age of 45 and 56 years. The average refractions were −0.7 and −1.5 D and IOP, 13.7 and 13.8 mm Hg. Glaucoma eyes had a mean MD of −5.5 decibels (dB). There was a significant difference in age (P < 0.001, unpaired t-test) between normal subjects and patients with glaucoma, and glaucoma eyes were slightly more myopic (P = 0.03). 

Table 1. 

 

Demographic Data of Normal and Glaucoma Eyes

	Macular Analysis
	Normal	Glaucoma	P Value
Eyes, n	51	65
Age, y	45 ± 16	56 ± 11	<0.001
Refractive error, D	−0.7 ± 1.5	−1.5 ± 2.1	0.03
IOP, mm Hg	13.7 ± 2.3	13.8 ± 2.7	0.7
Axial length, mm	23.8 ± 0.8	24.2 ± 1.3	0.1
MD, dB	−0.1 ± 1.2	−5.5 ± 6.5	<0.001
Average rotation angle	7.05 ± 3.02	8.75 ± 3.10	0.004
Difference in rotation angle	1.10 ± 0.91	1.67 ± 1.47	0.02

 Data are expressed as the mean ± standard deviation. The mean deviation is based on the Humphrey Field Analyzer 24-2 SITA-S. The demographic data between normal subjects and glaucoma patients were compared using t-test and the results are expressed as a P value.

Patients with glaucoma had greater rotation angle (8.8°) and the difference in the rotation angle between the two measurements (1.7°), compared with those of normal subjects, was 7.1° and 1.1°, respectively (Table 1, Fig. 5). 

Figure 5

 

The distributions in the rotation angle and the two intervisit-differences in the rotation angles in macular analysis in normal and glaucoma eyes are shown as histograms.

Reproducibility of Macular Inner Layer Thickness Measurement

The CVs and RCs evaluated in normal and glaucoma eyes are shown in Table 2 and Table 3, respectively; the ICCs with 95% confidence intervals (CIs) are shown in Table 4. In normal eyes, the CVs of the average mRNFLT, GCL+IPL, and GCC thickness measurements were 1.7%, 0.5%, and 0.5%, respectively, with or without rotation correction, while the RCs ranged from 1.3 to 2.0 μm. With smaller grid sizes, the CVs increased, but the CVs were almost below 3.0%. The ICCs of the average GCL+IPL or GCC thickness measurement (0.997 and 0.996, respectively) over the whole area were significantly higher compared with that of the average mRNFLT measurements (0.968) and the similar results, which was also the case for the ICCs of GCL+IPL, GCC thickness, and mRNFLT measurements in the subdivided areas (Table 4). In glaucoma eyes, the CVs of the average mRNFLT, GCL+IPL, and GCC thickness measurements were 2.5%, 0.6%, and 0.6%, respectively, with and without rotation correction; the RC ranged from 1.3 to 2.0 μm. With smaller grid sizes, the CVs increased but were smaller than 4.2%, except for the mRNFLT measurement in the divided areas. The ICCs ranged from 0.995 to 0.997 for the average mRNFLT, GCL+IPL, and GCC thickness measurements with no significant interlayer difference. Glaucoma eyes had significantly higher ICCs for the total mean and half mean mRNFLT measurements compared with normal eyes with and without rotation correction. However, there were no significant differences in the ICCs of the GCL+IPL and GCC thickness measurements between normal and glaucoma eyes. 

Table 2. 

 

Averages of CVs and RC for Macular Inner Layer Thickness Measurement in 51 Normal Eyes

	CV, %			RC, μm
	Without Rotation Correction	With Rotation Correction	P Value	Without Rotation Correction	With Rotation Correction
mRNFL thickness measurement
Total mean	1.73 ± 1.17	1.75 ± 1.20	0.8	1.93	1.95
Upper-half mean	1.80 ± 1.38	1.92 ± 1.52	0.051	2.05	2.15
Lower-half mean	1.84 ± 1.30	1.76 ± 1.26	0.2	2.12	2.08
2 × 2	1.6–2.3	1.7–2.6	0.8	1.9–2.8	1.8–2.8
4 × 4	1.8–5.3	1.9–7.5	0.5	1.8–5.2	1.8–5.9
8 × 8	1.9–20.1	2.0–21.1	0.9	2.2–10.8	2.2–11.2
GCL+IPL thickness measurement
Total mean	0.46 ± 0.41	0.45 ± 0.40	0.9	1.35	1.33
Upper-half mean	0.51 ± 0.36	0.49 ± 0.38	0.8	1.36	1.35
Lower-half mean	0.56 ± 0.50	0.55 ± 0.51	0.7	1.64	1.63
2 × 2	0.6–0.7	0.6–0.7	0.9	1.8–2.0	1.7–1.9
4 × 4	0.6–1.5	0.5–1.6	0.9	1.6–4.2	1.4–4.6
8 × 8	0.7–2.9	0.6–3.0	0.9	2.2–8.2	2.0–8.0
GCC thickness measurement
Total mean	0.51 ± 0.36	0.52 ± 0.36	0.5	1.89	1.90
Upper-half mean	0.53 ± 0.43	0.56 ± 0.44	0.05	2.06	2.17
Lower-half mean	0.56 ± 0.42	0.53 ± 0.41	0.2	2.08	2.04
2 × 2	0.6–0.7	0.6–0.7	0. 8	2.3–2.5	2.3–2.7
4 × 4	0.6–1.4	0.6–1.4	0.9	2.7–5.1	2.5–5.3
8 × 8	0.6–2.5	0.6–2.8	0.9	3.0–7.2	2.9–7.1

 The CVs are expressed as the mean ± standard deviation or range of the mean. The P values indicated results of a Wilcoxon test between the CVs with and without rotation correction for total mean, upper or lower half (180°) mean and those of ANOVA for 2 × 2, 4 × 4, and 8 × 8 grids, respectively.

Table 3. 

 

Averages of CVs and RC for Macular Inner Layer Thickness Measurement in 65 Glaucoma Eyes

	CV, %			RC, μm
	Without Rotation Correction	With Rotation Correction	P Value	Without Rotation Correction	With Rotation Correction
mRNFL thickness measurement
Total mean	2.50 ± 2.11	2.46 ± 2.11	0.3	1.69	1.69
Upper-half mean	3.49 ± 8.38	3.84 ± 9.19	0.1	1.89	1.97
Lower-half mean	4.49 ± 5.08	3.91 ± 3.94	0.04	2.30	2.17
2 × 2	3.3–13.7	3.6–13.2	0.9	3.7–5.1	4.1–4.6
4 × 4	4.7–33.8	5.0–32.9	0.9	2.3–10.8	2.43–10.1
8 × 8	3.4–56.1	4.0–54.9	0.7	2.9–14.1	2.8–13.8
GCL+IPL thickness measurement
Total mean	0.55 ± 0.49	0.57 ± 0.49	0.09	1.32	1.35
Upper-half mean	0.63 ± 0.54	0.66 ± 0.52	0.5	1.52	1.55
Lower-half mean	0.78 ± 0.56	0.79 ± 0.59	0.9	1.67	1.74
2 × 2	0.8–1.0	0.9–1.1	0.9	2.3–2.6	2.4–2.7
4 × 4	0.9–2.2	0.9–2.2	0.9	2.4–5.0	2.3–5.1
8 × 8	1.0–4.2	1.1–3.5	0.9	2.7–83	2.7–9.1
GCC thickness measurement
Total mean	0.66 ± 0.52	0.63 ± 0.53	0.1	1.96	1.93
Upper-half mean	0.65 ± 0.54	0.71 ± 0.55	0.3	2.13	2.26
Lower-half mean	0.81 ± 0.76	0.78 ± 0.69	0.7	2.39	2.30
2 × 2	0.9–1.2	0.9–1.2	0.9	5.0–6.0	5.4–6.3
4 × 4	0.9–2.2	0.8–2.2	0.9	3.2–13.1	3.0–13.0
8 × 8	0.9–3.2	0.8–3.3	0.9	3.1–14.8	3.1–14.2

 The CVs are expressed as the mean ± standard deviation or range of the mean. The P values indicated results of a Wilcoxon test between the CVs with and without rotation correction for total mean, upper or lower half (180°) mean, and those of ANOVA for 2 × 2, 4 × 4, and 8 × 8 grids, respectively.

Table 4. 

 

ICCs of Macular Analysis

	Normal Eyes, n = 51		Glaucoma Eyes, n = 65
	Without Rotation Correction	With Rotation Correction	Without Rotation Correction	With Rotation Correction
mRNFL thickness measurement
Total mean	0.968 (0.944–0.982)	0.968 (0.944–0.982)	0.995 (0.993–0.997)	0.995 (0.993–0.997)
Upper-half mean	0.968 (0.944–0.982)	0.966 (0.940–0.980)	0.996 (0.993–0.997)	0.995 (0.992–0.997)
Lower-half mean	0.967 (0.942–0.981)	0.969 (0.946–0.982)	0.995 (0.992–0.997)	0.996 (0.993–0.997)
2 × 2	0.929 (0.877–0.960) − 0.980 (0.964–0.988)	0.922 (0.864–0.955) − 0.978 (0.962–0.988)	0.979 (0.965–0.987) − 0.985 (0.976–0.991)	0.976 (0.960–0.985) − 0.983 (0.973–0.990)
GCL+IPL thickness measurement
Total mean	0.997 (0.994–0.998)	0.997 (0.994–0.998)	0.998 (0.996–0.998)	0.997 (0.996–0.998)
Upper-half mean	0.997 (0.994–0.998)	0.997 (0.994–0.998)	0.998 (0.997–0.999)	0.998 (0.996–0.999)
Lower-half mean	0.995 (0.992–0.997)	0.995 (0.992–0.997)	0.996 (0.993–0.997)	0.995 (0.993–0.997)
2 × 2	0.993 (0.988–0.996) − 0.995 (0.991–0.997)	0.993 (0.988–0.996) − 0.996 (0.992–0.997)	0.991 (0.985–0.994) − 0.995 (0.992–0.997)	0.987 (0.978–0.992) − 0.995 (0.993–0.997)
GCC thickness measurement
Total mean	0.996 (0.993–0.998)	0.996 (0.993–0.998)	0.998 (0.997–0.999)	0.998 (0.997–0.999)
Upper-half mean	0.996 (0.993–0.998)	0.995 (0.992–0.997)	0.999 (0.998–0.999)	0.998 (0.997–0.999)
Lower-half mean	0.995 (0.992–0.997)	0.995 (0.992–0.997)	0.998 (0.997–0.999)	0.998 (0.997–0.999)
2 × 2	0.992 (0.987–0.996) − 0.996 (0.993–0.998)	0.991 (0.984–0.995) − 0.996 (0.992–0.997)	0.989 (0.982–0.993) − 0.994 (0.990–0.996)	0.988 (0.980–0.992) − 0.993 (0.988–0.996)

 The ICCs are expressed with the 95% CIs in parentheses.

We also compared the CVs among the macular inner layer thickness. For all segments, the CVs of mRNFL measurements were significantly greater than those of GCL+IPL measurements (P < 0.001), except for the upper-half mean in glaucoma eyes. Also for all segments, the CVs of GCL+IPL measurements were not significantly different from those of GCC measurements. 

CVs and ICCs showed no significant difference between the analyses with rotation correction and without it, both in normal and glaucoma eyes. 

Multiple regression analysis showed that CV of the average mRNFLT measurement was significantly associated with the difference in the rotation angle in normal eyes (P = 0.04) and with the MD (P = 0.004) and axial length (P = 0.009) in glaucoma eyes. However, no factors were significantly associated with the CVs of the average GCL+IPL and GCC thickness measurements in normal and glaucoma eyes. 

Comparison of Measurement Reproducibility in the Macular and Circumpapillary Areas

Eyes in which the circumpapillary and macular measurement results fulfilled the inclusion criteria were selected and 30 normal eyes and 31 glaucoma eyes were included. The CVs of the cpRNFLT measurements were significantly higher than those of GCL+IPL and GCC measurements (P < 0.001, Wilcoxon test) in both normal and glaucoma eyes (Table 5). No significant differences were found between the CVs of the cpRNFLT and mRNFLT, except for the analysis of inferior half mean in normal eyes (P < 0.003, Wilcoxon test). 

Table 5. 

 

Comparison of CVs of cpRNFLT and Macular Inner Layer Thickness Measurement

Eyes	Without Rotation Correction				With Rotation Correction
	CV, %	CV, % (P Value)			CV, %	CV, % (P Value)
	cpRNFL	mRNFL	GCL+IPL	GCC	cpRNFL	mRNFL	GCL+IPL	GCC
Normal (n = 30)
360° mean	2.9 ± 1.9	1.5 ± 1.1 (0.01)	0.38 ± 0.35 (<0.001)*	0.44 ± 0.34 (<0.001)*	2.9 ± 1.9	1.6 ± 1.1 (0.01)	0.38 ± 0.35 (<0.001)*	0.46 ± 0.35 (<0.001)*
Superior-half mean	2.8 ± 2.2	1.6 ± 1.3 (0.1)	0.46 ± 0.34 (<0.001)*	0.50 ± 0.45 (<0.001)*	2.7 ± 2.3	1.8 ± 1.4 (0.14)	0.46 ± 0.36 (<0.001)*	0.56 ± 0.47 (<0.001)*
Inferior-half mean	3.3 ± 2.6	1. 7 ± 1.3 (0.003)*	0.54 ± 0.38 (<0.001)*	0.47 ± 0.39 (<0.001)*	3.4 ± 2.6	1. 6 ± 1.3 (0.003)*	0.52 ± 0.41 (<0.001)*	0.57 ± 0.42 (<0.001)*
Glaucoma, n = 31
360° mean	3.2 ± 3.0	2.6 ± 1.9 (0.6)	0.59 ± 0.53 (<0.001)*	0.67 ± 0.58 (<0.001)*	3.1 ± 2.8	2.6 ± 1.9 (0.6)	0.61 ± 0.51 (<0.001)*	0.65 ± 0.54 (<0.001)*
Superior-half mean	2.8 ± 2.6	2.7 ± 2.4 (0.9)	0.69 ± 0.59 (<0.001)*	0.66 ± 0.58 (<0.001)*	2.8 ± 2.5	2.9 ± 3.1 (0.9)	0.63 ± 0.52 (<0.001)*	0.63 ± 0.51 (<0.001)*
Inferior-half mean	4.2 ± 3.8	4.3 ± 4.2 (0.9)	0.86 ± 0.54 (<0.001)*	0.86 ± 0.83 (<0.001)*	4.3 ± 4.2	4.2 ± 4.1 (0.9)	0.88 ± 0.64 (<0.001)*	0.90 ± 0.72 (<0.001)*

 The comparisons between cpRNFL and mRNFL, GCL+IPL, or GCC thickness measurement reproducibility are expressed as the P value obtained using the Wilcoxon test.

*   P < 0.05 with Bonferroni correction.

Discussion

Several recent studies have reported on macular assessment for glaucoma, and the GCC (combined mRNFLT, GCL+IPL) thickness assessed by SD-OCT has diagnostic capability comparable with that of the cpRNFL thickness. ^19,20 Mwanza et al. ²⁶ also showed the clinical usefulness of the combined thickness of GCL and IPL, which represents the size of the retinal ganglion cell population in the macula using an OCT device (Carl Zeiss Meditec). 

Only two previous reports on glaucoma eyes have evaluated the intervisit reproducibility of the macular inner layer thickness measurements. Sakamoto et al. ¹⁴ measured the mRNFLT in 13 patients with glaucoma using an SD-OCT device (Topcon, Inc.) and reported ICCs of 0.88. Mwanza et al. ¹⁷ measured the average thickness of the mRNFLT and the GCL/IPL complex using an OCT device (Carl Zeiss Meditec) in 50 glaucoma eyes, and reported CVs of 2.6% and 1.8%, respectively. In the current study, the CVs of mRNFLT measurements (2.5%) were comparable with those reported by Mwanza et al., ¹⁷ while the CVs of GCL+IPL measurements (0.6%) were smaller than those reported by Mwanza et al. ¹⁷ Although the reason for this difference is unclear, the difference in the segmentation method between IPL and inner nuclear layer may be partly responsible. However, the ICCs of average mRNFLT in glaucoma eyes differed considerably from that reported by Sakamoto et al. ¹⁴ (0.995 vs. 0.88). Since both studies used the same SD-OCT instrument, the reason for this difference is unclear; however, those authors measured the reproducibility in only 13 patients, and this may partially explain the discrepancy. We found the higher ICC for mRNFLT measurements of glaucoma eyes (0.995) compared with that of normal eyes (0.968), while calculated CV and RC indicated better reproducibility in normal eyes (1.7% and 1.93 μm, respectively) compared with those in glaucoma eyes (2.5% and 1.69 μm, respectively). This apparently paradoxical result may be attributable to the feature which ICC is calculated to be larger when the values ranged widely. In this study, the SD of mRNFLT measurements in normal eyes was smaller (1.2) than that in glaucoma eyes (2.1), indicating that the measurement values were more widely varied in glaucoma eyes, resulting in the higher ICCs for mRNFL measurements in glaucoma eyes compared to those in normal eyes. 

The intervisit reproducibility of the average cpRNFLT measurement has been well studied with various SD-OCT instruments; CVs ranging from 2.1% to 4.0% and ICCs ranging from 0.90 to 0.99 in normal eyes and CVs of 2.7% to 3.3% and ICCs of 0.97 to 0.99 in glaucoma eyes have been reported, ^7,11,12 and our results agreed with those reports. There were no reports comparing the measurement reproducibility of cpRNFLT with macular inner layer thickness; our study is the first. The reproducibility showed no significant difference between mRNFLT and cpRNFLT measurements except for inferior half mean in normal eyes. It is noteworthy that the reproducibility values for the GCL+IPL and GCC thickness measurements were better than those for the mRNFLT and the cpRNFLT measurements. The reproducibility for the GCL+IPL and GCC thickness measurements did not differ significantly except for the 8 × 8 sector in normal eyes (P = 0.001). 

It is also noteworthy that the reproducibility of the GCL+IPL and GCC thickness measurements was better than that of mRNFLT measurement in smaller areas such as 4 × 4 and 8 × 8 grids, which roughly corresponded to the grid size of the Humphrey perimetry central 10-2 test program. Recently, Raza et al. ²⁷ reported the clinical usefulness of the central 10-2 test result correlating the result of the SD-OCT-determining thickness of inner retinal layers of macular area. These findings indicated that the GCL+IPL or GCC thickness measurement should be more useful for diagnosing and following glaucoma than the RNFLT measurement in circumpapillary or macular areas. There is also a possible effect of the anatomy on dependence of variation caused by ocular rotation: while the RCG+IPL is rotationally almost symmetric, the GCC is less so and the mRNFL is not. In healthy eyes, this may reduce the variation caused by rotation for the RGC+IPL and, to a lesser amount, the GCC. In glaucoma eyes, this effect would depend on the amount and distribution of the damage. For example, in eyes with focal paracentral defects, the variance would depend more strongly on ocular rotation. 

In the current study, we also studied the effect of rotation correction on the measurement reproducibility. The average rotation angles in normal and glaucoma eyes were approximately 7.1° and 8.8°, and the difference between measurements was 1.1° and 1.7°, respectively. However, contrary to our expectation, the rotation correction showed no significant improvement in measurement reproducibility in the macular inner retinal layers even in divided areas or sectors. This is probably because the intervisit difference in the rotation angle was relatively small (1.1°∼1.7°). Furthermore, the limitation of the current SD-OCT measurement was the insufficient lateral resolution for detecting the minute difference resulting from the change in ocular rotation angle. The average change in rotation angle as well as the SD of this change was small (1°∼2°), which indicates very small displacements within the image. For the most peripheral parts (2.4 mm off center), the average displacement by a change in the rotation angle would be approximately 60 μm, while in 1.2-mm eccentricity, this was only 30 μm. Thus, it appears unlikely that the structural defect of this magnitude is to be reliably detected. 

We also investigated factors affecting the reproducibility and found that the mRNFLT measurement reproducibility was significantly aggravated along with the MD—that is, glaucoma progression—which agrees with Garas et al. ⁴ and Li et al., ²⁸ who reported that increasing disease severity negatively affects the reproducibility of the cpRNFLT measurement using SD-OCT. Further, in glaucoma eyes, the mRNFLT measurement reproducibility was significantly aggravated along with longer axial length. 

While the significant relationship between the intervisit difference in the rotation angle and reproducibility in normal eyes in mRNFLT measurement was found, no factors significantly affected the measurement reproducibility of GCL+IPL or GCC thickness, which may be another advantage of observing the GCL+IPL or GCC thickness over time. 

A limitation of this study was that only Japanese subjects were included. Since the RNFL thickness differs among different ethnicities ²⁹ and the RNFLT can affect the reproducibility, ⁴ there may be racial differences in the reproducibility of the RNFLT measurements. Another limitation was that the SD-OCT RNFLT measurements were obtained using only one instrument (Topcon, Inc.). Significant differences in RNFLT measurements among different SD-OCT instruments have been reported. ³⁰ Therefore, the results of the present study should be confirmed in other countries and ethnicities and using other SD-OCT instruments. 

In conclusion, we studied the intervisit reproducibility of the mRNFLT, GCL+IPL, and GCC thickness with SD-OCT in normal and glaucoma eyes, and compared the results with those of cpRNFLT in the same eyes. Better and excellent reproducibility of the macular GCL+IPL and GCC thickness compared with that of the RNFLT measurement in the macular and circumpapillary areas indicated that evaluation of the GCC or GCL+IPL thickness in the macular area in normal and glaucoma eyes might be useful for early detection and longitudinal follow-up of glaucomatous changes by SD-OCT. We also found that correction of ocular rotation had no substantial effect on the reproducibility of the SD-OCT measurements in normal and glaucoma eyes. 

Acknowledgments

Supported by Grant-in-Aid for Scientific Research by the Ministry of Health, Labor and Welfare of Japan (H18-Sensory-General-001) and Topcon, Inc. (Tokyo, Japan). The authors alone are responsible for the content and writing of the paper. 

Disclosure: H. Hirasawa, Topcon, Inc. (F, R); M. Araie, Topcon, Inc. (F, C, R); A. Tomidokoro, Topcon, Inc. (F, R); H. Saito, Topcon, Inc. (F, R); A. Iwase, Topcon, Inc. (F, R); S. Ohkubo, Topcon, Inc. (F, R); K. Sugiyama, Topcon, Inc. (F, R); T. Ootani, Topcon, Inc. (F, R); S. Kishi, Topcon, Inc. (F, R); K. Matsushita, Topcon, Inc. (F, R); N. Maeda, Topcon, Inc. (F, R); M. Hangai, Topcon, Inc. (F, R); N. Yoshimura, Topcon, Inc. (F, C, R)