Trend-based Analysis of Ganglion Cell–Inner Plexiform Layer Thickness Changes on Optical Coherence Tomography in Glaucoma Progression

Trend-based Analysis of Ganglion CelleInner Plexiform Layer Thickness Changes on Optical Coherence Tomography in Glaucoma Progression Won June Lee, MD,1,2 Young Kook Kim, MD,1,2 Ki Ho Park, MD, PhD,1,2 Jin Wook Jeoung, MD, PhD1,2 Purpose: To evaluate the rate of thinning in ganglion celleinner plexiform layer (GCIPL) thickness by optical coherence tomography (OCT) in glaucomatous eyes and to use a trend-based approach to determine its diagnostic ability for detecting glaucoma progression. Design: Prospective, observational study. Participants: Sixty-five patients with primary open-angle glaucoma with a minimum 3-year follow-up involving serial spectral-domain OCT measurement of GCIPL thickness. Methods: Patients were divided into a nonprogressor group (n ¼ 38) and a progressor group (n ¼ 27) on the basis of serial red-free photography or visual field tests. The rates of GCIPL thinning in the global region, affected hemifield, and 6 macular sectors, and the minimum thickness, were determined by linear regression and compared between groups. The area under the receiver operating characteristic curves (AUCs) were calculated for each parameter. The GCIPL thinning rates were compared between affected hemifields and unaffected hemifields. Main Outcome Measures: The macular GCIPL thinning rates in the progressor and nonprogressor groups and the ability of the GCIPL thinning rate to diagnose glaucoma progression. Results: The GCIPL thinning rate was significantly faster in progressors than in nonprogressors in the global area (P < 0.001); in the affected hemifield (P ¼ 0.001); in the temporal, vertical, and nasal sectors of the affected hemifield (P ¼ 0.017, 0.032, and 0.030, respectively); and in the minimum GCIPL thickness (P < 0.001). In the temporal sectors, the GCIPL thinning rates were significantly faster in the affected than in the unaffected hemifield (P ¼ 0.013). The best GCIPL parameters were the global (AUC ¼ 0.791), minimum (AUC ¼ 0.755), inferior hemifield (AUC ¼ 0.708), and affected hemifield (AUC ¼ 0.702) thinning rates. The global circumpapillary retinal nerve fiber layer thinning rate correlated significantly with the global and inferotemporal sector GCIPL thinning rates (rho ¼ 0.259 and 0.366, respectively). Conclusions: The GCIPL thinning rate on OCT was significantly faster for patients with glaucoma with progression than for those without progression. The GCIPL thinning rate of the temporal sector was faster in the affected than in the unaffected hemifield, suggesting that the glaucomatous damage may progress locally in a specific sequence. Trend-based analysis of GCIPL thickness on OCT may be useful for assessing glaucoma progression objectively and quantitatively. Ophthalmology 2017;-:1e9 ª 2017 by the American Academy of Ophthalmology Supplementary material available at www.aaojournal.org.

Evaluation of glaucoma progression is essential in the management of patients with glaucoma. To date, optical coherence tomography (OCT) has been widely used for monitoring these patients and for detecting structural progression in glaucoma.1e4 Assessment of progressive changes in the optic disc and retinal nerve fiber layer (RNFL) is based on event analysis or trend analysis.3 Using Cirrus HD-OCT (Carl Zeiss Meditec Inc., Dublin, CA), guided progression analysis indicates the amount and location of significant changes by comparing individual pixels in follow-up images with the same pixels in the baseline image (event-based analysis). It also can perform linear regression ª 2017 by the American Academy of Ophthalmology Published by Elsevier Inc.

analysis of the RNFL thickness measurements over time (trend-based analysis).2,5 However, commercially available software and algorithms focus only on the RNFL thickness measurement and do not provide progression analysis of the ganglion celleinner plexiform layer (GCIPL). During the past few years, the value of assessing the macular inner retinal structure, including the GCIPL, for diagnosing glaucoma has been the focus of many studies, and these have demonstrated that GCIPL parameters show glaucoma diagnostic performances that are better than or comparable to those of the RNFL parameters.6e9 However, studies using GCIPL parameters for determining glaucoma progression http://dx.doi.org/10.1016/j.ophtha.2017.03.013 ISSN 0161-6420/17

1

Ophthalmology Volume -, Number -, Month 2017 have not been reported. To our knowledge, only 1 study has reported longitudinal spectral-domain OCTebased GCIPL measurements in patients with glaucoma.10 The purpose of this study was to evaluate the rate of change in GCIPL thickness, as measured by spectraldomain OCT, in patients with glaucoma. Using trend-based analysis, we evaluated the diagnostic performance of this GCIPL thinning rate in detecting glaucoma progression as observed on red-free photographs or visual field (VF) examination.

Methods This study was approved by the Institutional Review Board of Seoul National University Hospital, with informed consent obtained. The study design followed the tenets of the Declaration of Helsinki for biomedical research.

Subjects A total of 65 patients with glaucoma were enrolled and followed up at the Department of Ophthalmology of Seoul National University Hospital from October 2012 to September 2016. The subjects were enrolled in the Macular Ganglion Cell Imaging Study, an ongoing prospective study designed in 2011. All subjects underwent a complete ophthalmologic examination, including visual acuity tests, manifest refraction assessment, slit-lamp examination, intraocular pressure measurements using Goldmann applanation tonometry, gonioscopy, dilated fundus examination, axial length measurement (Axis II PR; Quantel Medical, Inc., Bozeman, MT), color disc photography and red-free RNFL photography (TRC-50IX; Topcon Corporation, Tokyo, Japan), Swedish interactive thresholding algorithm 30-2 perimetry (Humphrey Field Analyzer II; Carl Zeiss Meditec, Jena, Germany), and Cirrus HDOCT (Carl Zeiss Meditec, Inc., Dublin, CA). The inclusion criteria were age between 20 and 79 years, best-corrected visual acuity 20/40 in the study eye, refractive error within a 6.00-diopters equivalent sphere, and a 3.00-diopter astigmatism. Patients with a history of surgical therapy, such as glaucoma filtering surgery, in the study eye were excluded. Patients with any other ocular disease that could interfere with visual function, or any media opacity that would significantly interfere with OCT image acquisition, were excluded. Patients were excluded if a highquality image could not be obtained (i.e., if all of the OCT images showed a signal strength <6). Consecutive individuals were included if they had primary open-angle glaucoma and had localized RNFL defects that were clearly visible on the red-free fundus photographs and that were located in only 1 superior or inferior hemifield. Primary open-angle glaucoma was defined as the presence of glaucomatous optic disc changes with corresponding glaucomatous VF defects and an open angle confirmed by gonioscopic examination. Glaucomatous optic disc changes were defined as neuroretinal rim thinning, notching, excavation, or RNFL defects. Glaucomatous VF defects were defined as (1) glaucoma hemifield test values outside the normal limits; (2) 3 or more abnormal points with a <5% probability of being normal, of which at least 1 point has a pattern deviation of P < 1%; or (3) a pattern standard deviation of P < 5%. The VF defects were confirmed on 2 consecutive, reliable tests (fixation loss rate 20%, false-positive and false-negative error rates 25%). All patients underwent regular follow-up visits, 6 months apart, at which time patients underwent clinical examination, color disc photography, and red-free RNFL photography. Both eyes were imaged with Cirrus HD-OCT and were examined by standard

2

automated perimetry every 6 to 12 months for 36 months. For cases in which both eyes met all eligibility criteria, 1 eye was randomly chosen as the study eye before analyses.

Optical Coherence Tomography Imaging The optic disc cube scan and ganglion cell analysis (GCA) protocol for macular cube scanning (macular cube: 200200) was used for diagnosis and follow-up of glaucoma. The optic disc cube scan imaged the optic disc region in an area of 66 mm2 (200200 pixels). The RNFL thickness was measured at each pixel, and an RNFL thickness map was generated. A measurement circle 3.46 mm in diameter, consisting of 256 A-scans, was then automatically positioned around the optic disc. The global, superior, inferior, 12 clock-hour, and 4-quadrant circumpapillary RNFL thicknesses were analyzed. The macular cube scan generated 4 sets of 128 horizontal B-scans, each comprising 512 A-scans, centered on the 66-mm2 macular region. The built-in GCA algorithm (Cirrus HD-OCT software, version 6.0) detected and measured macular GCIPL thickness within a 662-mm cube in an elliptical annulus around the fovea. The GCA algorithm identified the outer boundary of the RNFL and the inner plexiform layer. The difference between the outer boundary segmentation of the RNFL and the inner plexiform layer represented the thickness of the GCIPL. The following GCIPL thickness measurements were analyzed: global, minimum, and 6 sectoral values (superonasal, superior, superotemporal [ST], inferotemporal [IT], inferior, and inferonasal). Satisfactory OCT quality was defined as (1) a well-focused image, (2) the presence of a centered circular ring around the optic disc, and (3) a signal strength 6. Any subject with less than satisfactory OCT image quality was excluded.

Calculation of Cirrus HD-OCT Ganglion CelleInner Plexiform Layer and Retinal Nerve Fiber Layer Thinning Rates Linear regression analysis versus time was performed for the GCIPL thickness for the global, superior, and inferior regions, minimum and for 6 sectors for each eye of each subject, in order to determine the rate of change in GCIPL thickness (expressed in micrometers per year). The rate of change in RNFL thickness was calculated in the same manner used in the global, superior, and inferior regions. Images with a signal strength <6 or those that did not focus on the fovea and cases of algorithm segmentation failure were excluded from the linear regression analysis.

Determination of Glaucoma Progression Color disc photographs, red-free RNFL photographs, and VF tests were performed and assessed according to the patient’s regular follow-up schedule. Each patient’s progression status was determined from structural changes on the disc and RNFL photographs or functional changes on VF tests, as described previously.11e13 Progressive optic disc changes (i.e., focal or diffuse rim narrowing, neuroretinal rim notching, increased cup-to-disc ratio, adjacent vasculature position shift) were determined by comparing serial disc photographs and were regarded as indicating glaucomatous progression. Changes in an RNFL defect were determined from serial RNFL photographs and defined as the appearance of a new defect or an increase in the width or depth of an existing defect. These were regarded as indicative of structural progression.14 Two observers (W.J.L., J.W.J.), who were masked to all other patient information, independently evaluated all photographs. In cases of disagreement, a third glaucoma specialist (K.H.P.) served as an adjudicator.

Lee et al



Trend-Based Analysis of GCIPL Thickness

The VF tests were evaluated by means of standard automated perimetry. Patients were determined as having functional progression if any of the following conditions were fulfilled: (1) a mean deterioration of 3 decibels (dB) compared with 2 baseline values, observed at least twice during the follow-up period; or (2) a reproducible reduction in sensitivity of at least 10 dB in a cluster of 2 contiguous locations or a deterioration of at least 5 dB in a cluster of 3 contiguous locations, at least 1 of which had deteriorated by 10 dB on 2 consecutive VF tests, compared with 2 baseline values; or (3) “likely progression” according to the glaucoma progression analysis program of the Humphrey Field Analyzer.12,13 Serial VFs also were evaluated by 2 independent observers (W.J.L., J.W.J.) in a masked fashion.

Statistical Analyses All statistical tests were performed using PASW Statistics 18 (SPSS, Chicago, IL) and MedCalc (MedCalc Software, Ostend, Belgium). To compare the characteristics, including the rates of GCIPL thinning, the independent t test was used. To control for the location of the glaucomatous damage (RNFL defect), OCT parameters of the affected hemifield (hemifield with the RNFL defect) were analyzed, along with analysis of the original position. The ability of individual OCT parameters (GCIPL and RNFL, based on calculation) to differentiate glaucoma progressors from nonprogressors, according to trend-based analysis, was assessed using the area under the receiver operating characteristic curves (AUCs) and by calculating sensitivities at fixed specificities. Significant differences between AUCs were assessed using the method described by DeLong et al.15 To evaluate the correlation between the GCIPL thinning rate and the RNFL thinning rate, Spearman correlation analysis was used. P values < 0.05 were considered statistically significant. The values were recorded and are presented as mean  standard deviation.

Results The study involved 65 eyes with primary open-angle glaucoma that fulfilled the inclusion criteria for this study. Of these, 38 subjects were classified as nonprogressors and 27 were classified as

progressors, on the basis of evaluation of serial red-free fundus photographs and VF tests.

Clinical Demographics Table 1 shows the clinical demographics of all patients at the time of enrollment. Nonprogressors and progressors showed no significant differences in terms of age (51.112.2 vs. 55.910.1 years; P ¼ 0.100), intraocular pressure, VF indices, baseline RNFL thickness (84.29.4 vs. 82.48.4 mm; P ¼ 0.422), and GCIPL thickness (75.65.8 vs. 76.56.3 mm; P ¼ 0.567). The average number of GCIPL OCT scans and the follow-up periods were similar between the groups (P > 0.05).

Comparison of Ganglion CelleInner Plexiform Layer Thinning Rate between Nonprogressors and Progressors Table 2 and Figure 1 show a comparison of the rates of GCIPL and RNFL thinning and the AUCs for discrimination between nonprogressors and progressors. The rate of GCIPL thinning was significantly faster in progressors than in nonprogressors globally (P < 0.001); in the superior and inferior hemifields (P ¼ 0.043 and 0.001); in the superonasal, ST, IT, inferior, and inferonasal sectors (P ¼ 0.048, 0.004, 0.012, 0.012, and 0.035, respectively); in the minimum thickness (P < 0.001); in the affected hemifield (P ¼ 0.001); and in the affected temporal, vertical, and nasal sectors (P ¼ 0.017, 0.032, and 0.030, respectively). The rate of RNFL thinning was significantly faster in progressors than in nonprogressors globally (P ¼ 0.012), in the superior and inferior hemifields (P ¼ 0.037 and 0.015), and in the affected hemifield (P ¼ 0.019). The best GCIPL parameters of the Cirrus OCT when using trend-based analysis for discriminating between nonprogressors and progressors were the GCIPL thinning rate seen globally (sensitivity 63.0%; specificity 80%; AUC ¼ 0.791), in the minimum thickness (sensitivity 59.3%, at a specificity 80%; AUC ¼ 0.755), in the inferior hemifield (sensitivity 48.2%; specificity 80%; AUC ¼ 0.708), and in the affected hemifield (sensitivity 48.2%; specificity 80%; AUC ¼ 0.702) (Figs 1 and 2). Two sets of comparisons, 1 of eyes with RNFL defects in the superior hemifield and 1 of eyes with

Table 1. Clinical Demographic Characteristics of the Patients with Glaucoma

Age (yrs) Male gender IOP (mmHg) CCT (mm) Diabetes Hypertension Medications (no.) MD (dB) PSD (dB) RNFL (mm) GCIPL (mm) Minimum GCIPL (mm) Superior location of damage GCIPL OCT scan number Follow-up period (mos)

Total (n [ 65)

Nonprogressor (n [ 38)

Progressor (n [ 27)

P Value

53.111.5 34 (52.3) 14.53.8 546.730.1 3 (4.6) 14 (21.5) 0.910.61 2.463.19 4.603.34 83.59.0 76.06.0 67.110.1 15 (23.1) 4.10.9 37.77.2

51.112.2 22 (57.9) 14.14.5 546.030.5 3 (7.9) 8 (21.1) 0.790.58 2.333.23 4.453.44 84.29.4 75.65.8 67.68.7 10 (26.3) 4.20.9 37.76.6

55.910.1 12 (44.4) 15.12.6 547.530.3 0 (0) 6 (22.2) 1.070.62 2.643.20 4.803.25 82.48.4 76.56.3 66.411.9 5 (18.5) 4.10.9 37.78.0

0.100 0.285 0.232 0.858 0.135 0.910 0.061 0.703 0.680 0.422 0.567 0.656 0.462 0.838 0.986

Data are mean  standard deviation or no. (%). CCT ¼ central corneal thickness; dB ¼ decibels; GCIPL ¼ ganglion celleinner plexiform layer; IOP ¼ intraocular pressure; MD ¼ mean deviation; OCT ¼ optical coherence tomography; PSD ¼ pattern standard deviation; RNFL ¼ retinal nerve fiber layer. Comparisons were performed with the chi-square test for categoric variables and the independent t test for continuous variables.

3

4 Table 2. Ganglion CelleInner Plexiform Layer Thinning Rates in Patients with Nonprogressing and Progressing Glaucoma, Based on Area Under the Curve Values and Sensitivities at Fixed Specificities Rate of Thickness Change (mm/year) Nonprogressor

Progressor

P Value*

0.280.56 0.270.71 0.320.61 0.240.80 0.451.06 0.120.66 0.340.52 0.370.87 0.271.02 0.181.18 0.350.62 0.380.52 0.390.91 0.300.62

0.820.43 0.650.76 1.010.82 0.791.41 0.450.89 0.700.94 1.031.27 1.231.55 0.770.77 1.641.85 1.000.84 1.051.31 1.171.64 0.800.73

<0.001 0.043 0.001 0.048 0.988 0.004 0.012 0.012 0.035 <0.001 0.001 0.017 0.032 0.030

0.791 0.681 0.708 0.680 0.557 0.702 0.687 0.659 0.635 0.755 0.702 0.675 0.642 0.657

(0.672e0.882) (0.554e0.791) (0.582e0.814) (0.552e0.790) (0.428e0.680) (0.575e0.809) (0.560e0.796) (0.531e0.772) (0.506e0.751) (0.654e0.869) (0.576e0.809) (0.548e0.786) (0.514e0.757) (0.529e0.770)

22.2 14.8 33.3 3.7 3.7 14.8 33.3 37.0 7.4 44.4 33.3 37.0 33.3 7.4

(0.00e48.2) (0.00e45.8) (14.8e59.3) (0.0e22.2) (0.0e3.70) (0.0e29.6) (11.1e5.9) (3.7e63.0) (0.0e29.6) (11.1e66.7) (14.8e55.6) (18.5e63.0) (0.0e59.3) (0.0e29.6)

63.0 33.3 48.2 55.6 29.6 51.9 48.2 51.9 29.6 59.3 48.2 51.9 48.2 25.9

0.621.05 1.022.23 1.381.97 1.581.92

1.431.47 2.272.48 2.752.43 3.022.87

0.012 0.037 0.015 0.019

0.678 0.639 0.671 0.654

(0.551e0.789) (0.511e0.755) (0.543e0.783) (0.526e0.768)

11.1 14.8 14.8 22.2

(0.0e29.6) (0.0e40.7) (0.0e29.6) (7.4e45.1)

33.3 29.6 29.6 29.6

AUC

95% Specificity

80% Specificity

Rate of Change at Fixed Specificity 95% Specificity

80% Specificity

(29.6e85.2) (3.7e69.2) (26.5e66.7) (18.5e84.3) (0.0e55.6) (14.8e81.5) (25.9e70.4) (33.3e74.1) (3.7e48.2) (40.7e84.1) (29.6e70.4) (29.6e70.4) (29.6e70.4) (3.7e48.2)

1.31 1.38 1.32 2.30 2.01 1.30 1.26 1.54 2.00 1.64 1.39 1.29 1.76 2.07

0.66 0.68 0.76 0.55 0.92 0.62 0.89 0.84 1.08 1.01 0.83 0.89 0.92 1.08

(14.8e74.1) (14.8e55.6) (3.7e55.6) (14.8e62.8)

2.54 4.71 5.00 4.97

1.76 2.67 3.09 3.39

AUC ¼ area under the receiver operating characteristic curve; GCIPL ¼ ganglion celleinner plexiform layer; I ¼ inferior sector; IN ¼ inferonasal; IT ¼ inferotemporal; RNFL ¼ retinal nerve fiber layer; S ¼ superior; SN ¼ superonasal; ST ¼ superotemporal. *The mean  standard deviation rates were compared using the independent t test. y The 95% confidence intervals are shown in parentheses.

Ophthalmology Volume -, Number -, Month 2017

GCIPL thickness (by linear regression) Global Superior Inferior SN S ST IT I IN Minimum Affected hemifield Affected temporal Affected vertical Affected nasal RNFL thickness (by linear regression) Global Superior Inferior Affected hemifield

Sensitivity at Fixed Specificityy y

Lee et al



Trend-Based Analysis of GCIPL Thickness

defects in the inferior hemifield, are shown in Tables S1 and S2 (available at www.aaojournal.org). Table 3 shows a comparison of the AUCs of the best parameters of the GCIPL thinning rate and the RNFL thinning rate for discriminating between nonprogressors and progressors. No significant differences were found, and the discriminating abilities among the parameters were comparable. Figure 3 demonstrates a representative case of the trend-based GCIPL analysis.

Comparison of Ganglion CelleInner Plexiform Layer Thinning Rate According to the Affected Hemifield Table 4 shows a comparison of the rate of GCIPL thinning according to the affected hemifield. The GCIPL thinning rate was significantly faster for the temporal GCIPL sectors of the affected hemifield than for those of the unaffected hemifield (0.660.98 vs. 0.330.80 mm/year; P ¼ 0.013).

Correlation between Ganglion CelleInner Plexiform Layer Thinning Rate and Retinal Nerve Fiber Layer Thinning Rate Table 5 and Figure S1 (available at www.aaojournal.org) show the correlation between the GCIPL thinning rate and the RNFL. The global GCIPL thinning rate correlated significantly with the global, superior, and inferior RNFL thinning rates (correlation coefficient, rho ¼ 0.259, 0.294, and 0.252, respectively). The GCIPL thinning rate of the ST sector was significantly correlated with the superior RNFL thinning rates (rho ¼ 0.262). The GCIPL thinning rate of the IT sector showed a significant correlation with the global RNFL (rho ¼ 0.366) and inferior RNFL (rho ¼ 0.319) thinning rates. The global RNFL thinning rate correlated significantly with the GCIPL thinning rates of the global and IT sectors (rho ¼ 0.259 and 0.366, respectively).

Discussion To evaluate the progression of glaucoma, it would be useful not only to evaluate new-onset structural or functional events but also to assess the rate of progression. Most previous studies that evaluated the rate of glaucoma progression focused on measuring the RNFL thickness, using software to do so automatically.1,2,5 Previous studies have reported that macular ganglion cell analysis, including the GCIPL, also exhibits good glaucoma-detecting ability, in a manner comparable to RNFL thickness.6,16e20 Our results demonstrate that eyes showing progression of glaucoma, as determined from red-free fundus photographs or VF tests, had a significantly higher GCIPL thinning rate over time than did nonprogressors. Several studies have evaluated the rate of macular thinning in glaucoma progression.21e23 Medeiros et al23 evaluated the changes of RNFL and macular thickness using the Stratus OCT (Zeiss) for detecting progressive structural damage in glaucoma, and no significant differences were found in the rates of changes for any of the macular thickness parameters. Sung et al21 evaluated the rate of progression of macular thickness with the Cirrus OCT in patients with advanced glaucoma. They enrolled patients with glaucoma with advanced disease, with a mean follow-up time of 2.2 years.21 Na et al22 also reported that the progression rate

Figure 1. Comparison of the area under the receiver operating characteristic curves (AUCs) for the ganglion celleinner plexiform layer (GCIPL) and retinal nerve fiber layer (RNFL) thinning rates, based on trend-based analyses, for discriminating between progressors and nonprogressors with openangle glaucoma. The AUCs of each parameter are shown in parentheses.

of macular volume and thickness, as measured by Cirrus OCT, may be helpful in identifying glaucoma progression. However, all of these studies evaluated only the macular thickness parameters, not GCIPL thickness. In this study, we analyzed the GCIPL thinning rate based on linear regression in order to evaluate glaucoma progression. The GCIPL thinning rate was significantly faster in progressors than in nonprogressors. The GCIPL thinning rate showed a good ability to discriminate glaucoma progression, comparable to that of the RNFL thinning rate. Nonprogressors in our study showed a GCIPL thinning rate, with age-related GCIPL thinning, similar to that reported in a previous study by Leung et al10 (0.318 mm/year). However, the GCIPL thinning rate of progressors was faster than the normal age-related thinning rate (0.82 mm/year for global GCIPL thinning rate; Table 2). In this study, the discriminating ability of the GCIPL thinning rate was not statistically different from that of the RNFL. Statistical analysis revealed that the difference between the 2 values was not significant, but it is noteworthy that the AUCs of GCIPL thinning were higher for some parameters compared with those of RNFL thinning. This may be because a single circular measurement with a 3.46-mm diameter was analyzed for RNFL, and the RNFL analysis contained the nasal and temporal sectors, where early glaucomatous changes are rarely observed.10 Virtually all parameters of the GCIPL thinning rate, including the global, minimum, and affected hemifield, performed well in terms of the ability to discriminate glaucoma progression. In addition, the thinning rate of the temporal GCIPL sectors (affected hemifields, ST, and IT)

5

Ophthalmology Volume -, Number -, Month 2017

Figure 2. The area under the receiver operating characteristic curves (AUCs) for the ganglion celleinner plexiform layer (GCIPL) thinning rate according to the sectors for discriminating between progressors and nonprogressors with open-angle glaucoma.

showed good discrimination ability for glaucoma progression, and these temporal sectors correlated well with RNFL thinning rates. These findings could suggest a spatial relationship or continuum between the affected RNFL and the related temporal GCIPL sectors. Hood et al24 recently proposed that early glaucomatous damage involves the

macula. The glaucomatous macular damage is typically arcuate and is often associated with local RNFL thinning in a narrow region of the disc, called the “macular vulnerability zone.” Kim et al25 also reported the topographic relationship between RNFL and GCIPL defects using spectral-domain OCT. Our results are in

Table 3. P Values for Testing Difference in Area Under the Curve Values between Rates of Ganglion CelleInner Plexiform Layer and Retinal Nerve Fiber Layer Thinning by Sector GCIPL Thinning

GCIPL Global Affected hemifield Affected temporal Minimum RNFL Global Affected hemifield

RNFL Thinning

Global

Affected Hemifield

Affected Temporal

Minimum

Global

Affected Hemifield

NA 0.1210 0.1483 0.7905

0.1210 NA 0.6904 0.1441

0.1483 0.6904 NA 0.0874

0.7905 0.1441 0.0874 NA

0.1945 0.7787 0.9737 0.2787

0.1139 0.5762 0.8087 0.1567

0.1945 0.1139

0.7787 0.5762

0.9737 0.8087

0.2787 0.1567

NA 0.7197

0.7197 NA

GCIPL ¼ ganglion celleinner plexiform layer; NA ¼ not available; RNFL ¼ retinal nerve fiber layer. P values are shown. The method described by DeLong et al15 was used for AUC comparisons.

6

Lee et al



Trend-Based Analysis of GCIPL Thickness

Figure 3. Representative case of the trend-based ganglion celleinner plexiform layer (GCIPL) thinning rate analysis. Serial red-free fundus photographs of a progressor show the overlay of serial retinal nerve fiber layer (RNFL) thickness profiles, the RNFL progression deviation map, serial GCIPL thickness profiles, GCIPL deviation maps, GCIPL sector color code maps, and linear regression graphs of global, superior, and inferior RNFL, and global, superior, and inferior GCIPL during the follow-up period. Localized RNFL and GCIPL thinning are seen in the inferior hemifield. The rate of GCIPL thinning was faster in the affected hemifield (inferior hemifield) than in the unaffected hemifield (superior hemifield).

close agreement with those of earlier studies that reported the relationship of glaucomatous lesions in the macula and peripapillary regions. Moreover, our study demonstrated that the GCIPL thinning rate of the temporal sector was significantly faster in the affected hemifield than in the unaffected hemifield, indicating that glaucomatous damage may progress locally with a specific sequence, at least in subjects who had a localized RNFL defect in 1 hemifield. Study Limitations Several points need to be considered when interpreting the results of the current study. First, this study had a small sample size and included only subjects who had a localized

RNFL defect in a single hemifield. Therefore, the rates of GCIPL thinning for patients with multiple RNFL defects or diffuse RNFL damage remain to be determined. Second, the number of GCIPL images acquired was limited (4.10.9). For this reason, we cannot set 2 consecutive GCIPL examinations as the baseline, as is automatically the case with Cirrus Guided Progression Analysis software. Instead, we used the first examination as the baseline and then performed linear regression analysis in both the GCIPL and the RNFL, as in other previous studies using trend-based analysis.1,3 Because the rate of change is related to the baseline thickness, the baseline setting also is important in trend-based analysis.3 Thus, further studies acquiring more GCIPL images may be required to evaluate the

7

Ophthalmology Volume -, Number -, Month 2017 Table 4. Comparison of the Rates of Changes in Affected and Unaffected Hemifields Affected Hemifield

Unaffected Hemifield

P Value*

0.620.78 0.510.93 0.711.31 0.660.98

0.410.75 0.441.14 0.460.93 0.330.80

0.099 0.663 0.220 0.013

2.182.44

1.312.15

0.013

Rate of Changes GCIPL Overall hemifield Nasal sector Superior or inferior sector Temporal sector RNFL Overall hemifield

GCIPL ¼ ganglion celleinner plexiform layer; RNFL ¼ retinal nerve fiber layer. *The mean  standard deviation rates were compared with the paired t test.

effect of the baseline setting in determining the GCIPL thinning rate. Third, the variability in the estimates of the rate of GCIPL and RNFL changes is relatively large. We speculate that this large variability could be caused by outliers, which may be due to testretest variability or measurement error. Fourth, the subjects enrolled in this study were patients with early to moderate glaucoma. Thus, the results of this study may not be valid in patients with advanced glaucoma, who have a markedly thinner RNFL and GCIPL at baseline. Further studies may be needed to evaluate the rate of GCIPL thinning in patients with advanced glaucoma. In conclusion, the GCIPL thinning rate was significantly faster in patients with progressive glaucoma, as diagnosed based on red-free fundus photographs or VF tests, than in those without progressive disease. This study suggests that trend-based analysis of GCIPL thickness is potentially useful in evaluating glaucoma progression and may be a valuable approach for monitoring patients with glaucoma. Table 5. Correlation between Ganglion CelleInner Plexiform Layer Thinning Rate and Retinal Nerve Fiber Layer Thinning Rate GCIPL Thinning Rate

GCIPL Global ST IT RNFL Global Superior Inferior

RNFL Thinning Rate

Global

ST

IT

Global

Superior

Inferior

NA 0.636y 0.430y

0.636y NA 0.333y

0.430y 0.333y NA

0.259* 0.219 0.366y

0.294* 0.262* 0.156

0.252* 0.205 0.319y

0.259* 0.294* 0.252*

0.219 0.262* 0.205

0.366y 0.156 0.319y

NA 0.690y 0.626y

0.690y NA 0.206

0.626y 0.206 NA

GCIPL ¼ ganglion celleinner plexiform layer; IT ¼ inferotemporal sector; NA ¼ not available; RNFL ¼ retinal nerve fiber layer; ST ¼ superotemporal sector. Correlation coefficients (rho) are shown. Spearman correlation was used for evaluating correlation. *P < 0.05. y P < 0.01.

8

References 1. Lee EJ, Kim TW, Weinreb RN, et al. Trend-based analysis of retinal nerve fiber layer thickness measured by optical coherence tomography in eyes with localized nerve fiber layer defects. Invest Ophthalmol Vis Sci. 2011;52: 1138-1144. 2. Leung CK. Diagnosing glaucoma progression with optical coherence tomography. Curr Opin Ophthalmol. 2014;25:104-111. 3. Leung CK, Cheung CY, Weinreb RN, et al. Evaluation of retinal nerve fiber layer progression in glaucoma: a study on optical coherence tomography guided progression analysis. Invest Ophthalmol Vis Sci. 2010;51:217-222. 4. Wollstein G, Schuman JS, Price LL, et al. Optical coherence tomography longitudinal evaluation of retinal nerve fiber layer thickness in glaucoma. Arch Ophthalmol. 2005;123: 464-470. 5. Leung CK, Yu M, Weinreb RN, et al. Retinal nerve fiber layer imaging with spectral-domain optical coherence tomography: patterns of retinal nerve fiber layer progression. Ophthalmology. 2012;119:1858-1866. 6. Jeoung JW, Choi YJ, Park KH, Kim DM. Macular ganglion cell imaging study: glaucoma diagnostic accuracy of spectral-domain optical coherence tomography. Invest Ophthalmol Vis Sci. 2013;54:4422-4429. 7. Mwanza JC, Durbin MK, Budenz DL, et al. Glaucoma diagnostic accuracy of ganglion cell-inner plexiform layer thickness: comparison with nerve fiber layer and optic nerve head. Ophthalmology. 2012;119:1151-1158. 8. Jeong JH, Choi YJ, Park KH, et al. Macular Ganglion Cell Imaging Study: covariate effects on the spectral domain optical coherence tomography for glaucoma diagnosis. PLoS One. 2016;11:e0160448. 9. Lee SY, Jeoung JW, Park KH, Kim DM. Macular ganglion cell imaging study: interocular symmetry of ganglion cell-inner plexiform layer thickness in normal healthy eyes. Am J Ophthalmol. 2015;159:315-323.e2. 10. Leung CK, Ye C, Weinreb RN, et al. Impact of age-related change of retinal nerve fiber layer and macular thicknesses on evaluation of glaucoma progression. Ophthalmology. 2013;120:2485-2492. 11. Kim KE, Jeoung JW, Kim DM, et al. Long-term follow-up in preperimetric open-angle glaucoma: progression rates and associated factors. Am J Ophthalmol. 2015;159:160-168. e1-e2. 12. Suh MH, Park KH, Kim H, et al. Glaucoma progression after the first-detected optic disc hemorrhage by optical coherence tomography. J Glaucoma. 2012;21:358-366. 13. Kim SH, Park KH. The relationship between recurrent optic disc hemorrhage and glaucoma progression. Ophthalmology. 2006;113:598-602. 14. Suh MH, Kim DM, Kim YK, et al. Patterns of progression of localized retinal nerve fibre layer defect on red-free fundus photographs in normal-tension glaucoma. Eye (Lond). 2010;24:857-863. 15. DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics. 1988;44:837-845. 16. Kim KE, Ahn SJ, Kim DM. Comparison of two different spectral domain optical coherence tomography devices in the detection of localized retinal nerve fiber layer defects. Jpn J Ophthalmol. 2013;57:347-358. 17. Kim YJ, Kang MH, Cho HY, et al. Comparative study of macular ganglion cell complex thickness measured by

Lee et al



Trend-Based Analysis of GCIPL Thickness

spectral-domain optical coherence tomography in healthy eyes, eyes with preperimetric glaucoma, and eyes with early glaucoma. Jpn J Ophthalmol. 2014;58:244-251. 18. Kim YK, Yoo BW, Kim HC, Park KH. Automated detection of hemifield difference across horizontal raphe on ganglion celleinner plexiform layer thickness map. Ophthalmology. 2015;122:2252-2260. 19. Seong M, Sung KR, Choi EH, et al. Macular and peripapillary retinal nerve fiber layer measurements by spectral domain optical coherence tomography in normaltension glaucoma. Invest Ophthalmol Vis Sci. 2010;51: 1446-1452. 20. Ye C, To E, Weinreb RN, et al. Comparison of retinal nerve fiber layer imaging by spectral domain optical coherence tomography and scanning laser ophthalmoscopy. Ophthalmology. 2011;118:2196-2202.

21. Sung KR, Sun JH, Na JH, et al. Progression detection capability of macular thickness in advanced glaucomatous eyes. Ophthalmology. 2012;119:308-313. 22. Na JH, Sung KR, Lee JR, et al. Detection of glaucomatous progression by spectral-domain optical coherence tomography. Ophthalmology. 2013;120:1388-1395. 23. Medeiros FA, Zangwill LM, Alencar LM, et al. Detection of glaucoma progression with Stratus OCT retinal nerve fiber layer, optic nerve head, and macular thickness measurements. Invest Ophthalmol Vis Sci. 2009;50:5741-5748. 24. Hood DC, Raza AS, de Moraes CG, et al. Glaucomatous damage of the macula. Prog Retin Eye Res. 2013;32:1-21. 25. Kim KE, Park KH, Yoo BW, et al. Topographic localization of macular retinal ganglion cell loss associated with localized peripapillary retinal nerve fiber layer defect. Invest Ophthalmol Vis Sci. 2014;55:3501-3508.

Footnotes and Financial Disclosures Originally received: October 28, 2016. Final revision: March 6, 2017. Accepted: March 6, 2017. Available online: ---.

Analysis and interpretation: Lee, Kim, Park, Jeoung Obtained funding: Not applicable Manuscript no. 2016-680.

1

Department of Ophthalmology, Seoul National University College of Medicine, Seoul, Korea. 2 Department of Ophthalmology, Seoul National University Hospital, Seoul, Korea. Financial Disclosure(s): The author(s) have no proprietary or commercial interest in any materials discussed in this article. Author Contributions: Conception and design: Lee, Jeoung Data collection: Lee, Kim, Park, Jeoung

Overall responsibility: Lee, Jeoung Abbreviations and Acronyms: AUC ¼ area under the receiver operating characteristic curve; dB ¼ decibels; GCA ¼ Ganglion Cell Analysis; GCIPL ¼ ganglion celleinner plexiform layer; IT ¼ inferotemporal; OCT ¼ optical coherence tomography; RNFL ¼ retinal nerve fiber layer; ST ¼ superotemporal; VF ¼ visual field. Correspondence: Jin Wook Jeoung, MD, PhD, Department of Ophthalmology, Seoul National University Hospital, Seoul National University College of Medicine, 101 Daehak-ro, Jongno-gu, Seoul 03080, Korea. E-mail: neuroprotect@ gmail.com.

9