Detection of Glaucoma Progression by Assessment of Segmented Macular Thickness Data Obtained Using Spectral Domain Optical Coherence Tomography

Abstract

Purpose.: We evaluated the clinical use of segmented macular layer thickness measurement in terms of glaucoma diagnosis and the ability to detect progression, and to compare such outcomes to those by circumpapillary retinal nerve fiber layer (cRNFLT) and total macular thickness (TMT) measurements.

Methods.: The study included 141 glaucomatous and 61 healthy eyes. All glaucomatous eyes were subjected to at least four spectral domain optical coherence tomography (SD-OCT) examinations (mean follow-up, 2.13 years). Segmented macular layers were the macular nerve fiber layer (NFL), ganglion cell and inner plexiform layer (GCA), and outer retinal layer (ORL; from outer plexiform layer to retinal pigment epithelium). Areas under receiver operating characteristic curves (AUCs) discriminating healthy from glaucomatous eyes were determined in baseline measurements. The sensitivity and specificity of these parameters in terms of glaucoma progression detection were determined, with reference to assessment of optic disc/retinal nerve fiber layer (RNFL) photographs/visual field (VF) deterioration as standard(s).

Results.: GCA afforded the best diagnostic performance among three macular layers. The AUC of the GCA thickness (GCAT) was less than that of cRNFLT (0.869 vs. 0.953, P = 0.018), but superior to that of TMT (0.790, P = 0.05). Of the eyes, 38 showed progression during follow-up by standard methods. The sensitivities of TMT, GCAT, and cRNFLT values in terms of detection of progression were 14%, 8%, and 5%, respectively.

Conclusions.: Although baseline cRNFL measurement was optimal in terms of glaucoma diagnosis, the GCAT and TMT showed similar levels of sensitivity in progression detection.

Introduction

Glaucoma is characterized by a progressive loss of retinal ganglion cells (RGCs). It is well-known that 30% or more of RGCs are lost before any visual field (VF) abnormality is apparent. ¹ Other histologic studies have shown that abnormalities in RGC morphology and cell density precede clinically detectable VF loss. ^2–4 Thus, detection of RGC loss is crucial for early diagnosis of glaucoma and to detect glaucomatous progression. Retinal nerve fibers are the axons of the RGCs and emerge from the cells to form the optic nerve. Therefore, circumpapillary retinal nerve fiber layer (cRNFL) thickness has been used to assess the extent of glaucomatous damage. cRNFL thickness is determined with the aid of optical coherence tomography (OCT) and such tests currently are the principal means of glaucoma detection. cRNFL thickness determined by OCT affords good glaucoma diagnostic capability. ^5,6 However, as glaucoma eventually involves the cell body of the RGCs, measurement of RGC thickness (if possible), rather than of the RNFL alone, would help estimate the glaucomatous damage accurately. 

About 50% of RGCs are located in the macula. ⁷ In human and experimental primate models of glaucoma, RGC loss was evident around the fovea at early disease stages. ^8–10 Loss of RGCs necessarily causes atrophy of the ganglion cell layer (GCL). In the interval since the appearance of the first report from Zeimer et al., ¹¹ indicating that total macular thickness (TMT) was reduced in glaucoma patients, many studies using optical imaging instruments, including retinal thickness analyzers and time-domain optical coherence tomography (TD-OCT), have suggested that macular thickness serves as a surrogate parameter in evaluation of glaucoma. ^12–20  

Introduction of spectral-domain optical coherence tomography (SD-OCT) facilitated segmentation of the retina and quantification of data from all retinal layers. SD-OCT offers a higher scan resolution and a faster scan speed than were available in earlier instrumentation. Recent studies have found that glaucoma diagnostic accuracy is improved if SD-OCT macular measurements are focused on inner retinal layers. ^21–29  

However, it remains unclear whether the outer retinal layer (ORL), especially the photoreceptor layer, is involved in glaucoma. The histologic study of Panda and Jonas showed that photoreceptor count was significantly lower in eyes with angle-closure glaucoma occurring secondary to penetrating corneal injuries, ³⁰ whereas loss of photoreceptors was not found in other studies of primary open angle glaucoma (POAG) patients or in experimental models of glaucoma. ^31,32 Some electrophysiologic reports have suggested that the outer retina is involved in glaucoma, ^33–37 and recent studies claiming involvement of the foveal photoreceptor layer in glaucoma have appeared; the latter works used a foveal reflection analyzer and SD-OCT. ^38,39  

The Cirrus high definition OCT (HD-OCT), a commercial SD-OCT platform, has an inbuilt ganglion cell analysis (GCA; Carl Zeiss Meditec, Dublin, CA) algorithm, and it was shown recently that use of this algorithm allows successful segmentation of inner macular layers (the GCA; and a combination of the GCL and the inner plexiform layer [IPL]), and that thickness measurements of these layers were reproducible. ⁴⁰  

The purpose of the our study was to evaluate the diagnostic ability and capacity to detect glaucoma progression of segmented inner and outer retinal layer thickness values obtained from the macular area using the Cirrus HD-OCT GCA algorithm. We compared these outcomes to those obtained using conventional measures of cRNFL and TMT. To the best of our knowledge, this is the first report to explore the ability of segmented macular layer thickness measurements to detect glaucoma progression. 

Methods

Subjects

Glaucoma patients evaluated between September 2008 and October 2011 at the glaucoma clinic of the Asan Medical Center, Seoul, Korea, and who met our inclusion criteria, were included via retrospective medical record review. 

At initial testing, each participant received a comprehensive ophthalmologic examination, including a review of medical history; measurement of best-corrected visual acuity (BCVA), slit-lamp biomicroscopy, Goldmann applanation tonometry (GAT), gonioscopy, dilated funduscopic examination using a 90- or 78-diopter (D) lens, stereoscopic optic disc photography, retinal nerve fiber layer (RNFL) photography, a VF test (using a Humphrey Field Analyzer [HFA] running the Swedish Interactive Threshold Algorithm [SITA] 24-2; Carl Zeiss Meditec), and SD-OCT (Cirrus HD-OCT; Carl Zeiss Meditec). 

For inclusion, all participants had to meet the following criteria: a BCVA of 20/40 or better, with a spherical refractive error between −6.0 and +4.0 D, and a cylinder correction within +3 D, and presence of a normal anterior chamber and open-angle on slit-lamp and gonioscopic examinations. 

Our present study included healthy patients and patients with glaucomatous optic nerve heads. The latter patients presented with enlarged cupping, diffuse or focal neural rim thinning, disc hemorrhage, or RNFL defects that were confirmed and agreed upon by two glaucoma specialists (KRS and JHN), regardless of the presence of any VF abnormality. 

The glaucomatous eyes were divided into two subgroups: the perimetric glaucoma (PG) and preperimetric glaucoma (PPG) subgroups. PG eyes were those showing glaucomatous VF defects as confirmed on at least two initial reliable VF examinations. A glaucomatous VF defect was defined as a cluster of three points with a probability <5% on the pattern deviation map in at least one hemifield, including at least one point with a probability of <1%; or a cluster of two points with a probability of <1%, and a glaucoma hemifield test (GHT) result outside normal limits, or a pattern standard deviation (PSD) outside 95% of normal limits. Eyes were classified into the PPG group if normal VF results were obtained at baseline visits. Only reliable VF test results (false-positive errors <15%, false-negative errors <15%, and fixation loss <20%) were included in analysis. The first VF test result was excluded to obviate any learning effect. 

Patients with any other ophthalmic or neurologic condition that could result in a VF defect, or with a history of diabetes mellitus, were excluded. If surgical or laser treatment was performed during follow-up, only the data obtained in the period before such treatment were analyzed. 

All subjects with glaucoma were followed up at 6-month intervals via VF testing, stereoscopic optic disc photography, RNFL photography, and Cirrus SD-OCT scanning. All tests were performed at the same visit or within 2 weeks of that visit. All subjects were followed for at least 24 months. 

If both eyes of a subject met our study inclusion criteria, both eyes were included in either the PG or PPG group. If one eye was included in either group and the fellow eye was normal, the normal fellow eye was excluded from all groups. For the healthy group, one eye was selected at random. 

Institutional Review Board approval was obtained from the Asan Medical Center and our study design followed the principles of the Declaration of Helsinki. 

Optic Disc and RNFL Assessment

Progression of optic disc and RNFL defects was determined by evaluation of stereoscopic optic disc and red-free RNFL photographs. Serial stereoscopic optic disc and red-free photographs were displayed on an LCD monitor. Two glaucoma experts (KRS and JHN) independently assessed all photographs to estimate the extent of glaucoma progression between the first and last visits. Either grader was masked to the progression assessments rendered by the other, and to all clinical, OCT, and VF information. Photographs were presented in chronologic order, with masking of patient identification, age, and test date. Each grader viewed all photographs of each eye before making an assessment, and each was asked to determine disease progression by evaluating the glaucomatous features of the optic disc, or progression of RNFL defects, as revealed by an increase in the extent of neuroretinal rim thinning, enhancement of disc excavation, any widening or deepening of a pre-existing RNFL defect, and/or the appearance of a new disc hemorrhage. Each grader classified each glaucomatous eye as either stable or progressing. If the opinions of the two observers differed, a third examiner (YJK) made a final decision. 

SD-OCT Assessment

SD-OCT images were obtained using the Cirrus HD-OCT system. Our Cirrus HD-OCT platform is calibrated on a regular basis by a technician employed by the manufacturer. Pupil dilatation was performed if necessary. All accepted images exhibited a centered optic disc, were well-focused with even and adequate illumination, exhibited no eye motion within the measurement circle, and had a signal strength (SS) ≥7. For inclusion in progression analysis, at least four acceptable OCT images, taken at separate visits, were required. 

Three-dimensional macular cube OCT data were obtained using the Macular Cube 200 × 200 scan mode. TMT values were obtained using the cube average thickness data yielded by use of the conventional Macular Cube algorithm. 

To obtain segmented macular layer thickness data, the information was processed using a prototype algorithm (a pre-release version) that eventually will be incorporated into Cirrus version 6.0 software. The protocol performs 200 horizontal B-scans, comprising 200 A-scans per B-scan, 1024 times, within a cube measuring 6 × 6 × 2 mm, and is designed to allow analysis of retinal topography. ^40,41 The GCA algorithm processes data from either of the 3D volume scans performed by the Cirrus HD-OCT; both scan patterns cover the same physical field of view, namely 6 × 6 × 2 mm, but the dimensionality of the image data is either 512 × 128 × 1024 or 200 × 200 × 1024. Input image data initially are segmented using existing Cirrus inner limiting membrane (ILM) and retinal pigment epithelium (RPE) segmentation algorithms to create a region of interest within which lie the intraretinal layers. ^40,41 The algorithm identifies the RNFL layer (from the ILM to the outer boundary of the RNFL; the MNFL), the GCA layer (from the outer boundary of the RNFL to the outer boundary of the IPL; thus, a combination of the RGC layer and the IPL), and the ORL (from the outer boundary of the internal nuclear layer to RPE; thus, a combination of the outer plexiform layer [OPL], outer nuclear layer [ONL], photoreceptors, and RPE). The segmentation procedure operates in three dimensions and uses a graph-based algorithm to identify each layer. Eight thicknesses are measured within an annular area centered on the fovea; these are the average, minimum (lowest thickness over a single meridian crossing the scanning annulus), and six sectoral (superotemporal, superior, superonasal, inferonasal, inferior, and inferotemporal) values of the respective layers. 

cRNFL average thickness values also were included to allow comparisons to be made; these were obtained using the optic disc cube mode. From the data thus collected, the software automatically determines the center of the disc and next extracts a circumpapillary circle (1.73 mm in radius) from the data set, to allow cRNFL thickness measurements to be conducted along the perimeter. Details of the manner in which TMT and cRNFL thickness measurements are conducted have been reported previously. ^5,6,42 Glaucoma progression in each eye was determined by linear regression against patient age using serial TMT, cRNFL, and three segmented macular layer (MNFL, GCA, and ORL) thickness measurements. A significantly negative slope (P < 0.05) indicated progression as evidenced by analysis of that parameter. 

VF Assessment

VF progression was determined using two methods; event-based analysis (EA) and trend-based analysis (TA). To conduct EA, commercial software (HFA GPA; Carl Zeiss Meditec) was used. VF EA progression was defined as a significant deterioration from the baseline pattern deviation at three or more of the same test points evaluated on three consecutive examinations. ⁴³ The other VF progression criterion used linear TA regression using visual field index (VFI) data. A significantly negative slope (P < 0.05) indicated VF TA progression. 

Statistical Analysis

Baseline characteristics were compared among the three groups (the PG and PPG subgroups, and the healthy group) using a generalized estimating equation (GEE) approach. The Bonferroni correction was applied if multiple comparisons were conducted. To compare categorical data, the χ ² test was used. Areas under receiver operating characteristic curves (AUCs) were calculated to assess the ability of data from each macular layer to differentiate PG patients from healthy subjects. The 95% confidence intervals (CIs) of AUCs were estimated by bootstrapping method with resampling unit as each individual, which accounted for correlation between two eyes for the same individual. ⁴⁴  

Employing expert assessment of optic disc and RNFL photographs, and/or VF progression as the reference standard, the sensitivity and specificity of TM, cRNFL, and three segmented macular layer (MNFL, GCA, and ORL) thickness measurements in terms of glaucoma progression detection were determined. A GEE approach was used to estimate the sensitivity and specificity of each parameter, TMT, cRNFL, and three segmented macular layers (MNFL, GCA, and ORL) for adjusting correlation. The CIs for the sensitivity and specificity were constructed using 1000 times of bootstrapping. ^44,45 We used a Wilcoxon one-sample test for the bootstrapped samples of difference between two parameters. Using expert assessment of optic disc and RNFL photographs, and/or VF progression as the reference standard, the extent of agreement between such data and linear regression information obtained by analysis of macular layer thicknesses was estimated using Kappa statistics. The extent of agreement among TMT, cRNFL, and GCA thickness measurements also was estimated. Statistical analysis for AUCs, sensitivity, and specificity was performed using R.2.13.1 and other analyses were done using SPSS version 15.0 (SPSS Inc., Chicago, IL). 

Results

The study initially included 190 glaucomatous eyes of 112 subjects. Of the 190 eyes, 49 from 30 subjects were excluded due to macular segmentation failure (18 eyes); RNFL segmentation failure (7 eyes); poor quality of OCT images, such as low signal strength (SS <7), motion artifact, or decentration (20 eyes); and unacceptable optic disc/RNFL photography (4 eyes). Finally, a total of 141 glaucomatous eyes from 82 subjects was included. The mean ± SD duration of follow-up was 2.13 ± 0.33 years. The number of cRNFL and macular images analyzed in each eye ranged from 4–7 (mean 4.5 ± 1.4) and 4–7 (mean 4.6 ± 1.4), respectively. Of these 141 eyes, 38 (27%) showed glaucomatous VF abnormalities at baseline. A total of 61 healthy eyes was included. Table 1 shows the demographic and baseline characteristics of all participants. 

Table 1.  

 

Baseline Characteristics of Study Participants

	PG (n = 38)	PPG (n = 103)	Healthy (n = 61)	P Value
Age (y)	59.0 ± 10.4	56.3 ± 11.0	55.3 ± 8.1	0.014
				*0.011
				†0.131
				‡0.011
Visual acuity (decimal)	0.88 ± 0.17	0.92 ± 0.14	0.96 ± 0.14	0.083
				*0.038
				†0.343
				‡0.073
Spherical error (D)	−0.98 ± 1.91	−0.62 ± 2.02	−0.11 ± 1.45	0.068
				*0.018
				†0.142
				‡0.083
CCT (μm)	534.4 ± 27.4	540.6 ± 28.2	547.2 ± 23.9	0.272
				*0.109
				†0.392
				‡0.155
Initial IOP (mm Hg)	16.3 ± 3.3	17.0 ± 3.3	15.4 ± 3.1	0.018
				*0.005
				†0.243
				‡0.014
Median VFI (%, first, last quartiles)	94 (76, 97)	99 (98, 100)	99 (99, 100)	0.001
				*<0.001
				†0.009
				‡0.554
Median MD (dB, first, last quartiles)	−3.50 (−9.66, −1.51)	−0.28 (−1.22, 0.48)	0.17 (−0.56, 0.99)	<0.001
				*<0.001
				†<0.001
				‡0.015
Median PSD (dB, first, last quartiles)	4.57 (2.51, 9.33)	1.54 (1.36, 1.94)	1.51 (1.37, 1.69)	<0.001
				*<0.001
				†<0.001
				‡0.021

 Comparisons among the three groups were performed using the generalized estimating equation method.CCT, central corneal thickness; IOP, intraocular pressure; MD, mean deviation.

 * PG versus healthy.

 † PG versus PPG.

 ‡ PPG versus healthy.

Segmented Macular Layer Thickness Comparisons among the Three Groups

The baseline segmented macular layer thickness values for the MNFL, GCA, and ORL are shown in Tables 2–4, respectively. The values of GCA parameters differed significantly among the PG, PPG, and healthy eyes. However, MNFL thickness did not differ between PPG and healthy eyes. The ORL thickness was not significantly different among the three groups. The average TMT and cRNFL thicknesses (all in μm) were 280.6 ± 16.6 (PG), 287.9 ± 15.8 (PPG), and 295.3 ± 10.1 (healthy); and 77.8 ± 12.9 (PG), 86.9 ± 9.8 (PPG), and 98.8 ± 6.8 (healthy), respectively, and were significantly different in each group (all P values <0.001). 

Table 2.  

 

Comparison of Baseline MNFL Thicknesses (μm)

	PG (n = 38)	PPG (n = 103)	Healthy (n = 61)	P Value
Average	28.46 ± 5.42	32.07 ± 4.96	33.00 ± 2.60	<0.001
				*<0.001
				†<0.001
				‡0.349
Minimum	12.08 ± 4.33	13.29 ± 3.42	15.28 ± 2.83	<0.001
				*<0.001
				†0.074
				‡0.001
Temporal superior	20.90 ± 4.94	22.14 ± 2.54	23.28 ± 2.37	0.005
				*0.005
				†0.077
				‡0.013
Superior	31.00 ± 7.16	34.34 ± 6.28	35.67 ± 3.84	0.003
				*0.001
				†0.004
				‡0.255
Nasal superior	34.10 ± 7.22	37.21 ± 7.99	37.31 ± 3.80	0.063
				*0.051
				†0.020
				‡0.785
Nasal inferior	34.62 ± 7.02	38.51 ± 7.73	39.20 ± 3.76	0.003
				*0.001
				†0.002
				‡0.694
Inferior	29.77 ± 9.14	35.81 ± 6.07	37.13 ± 3.20	<0.001
				*<0.001
				†<0.001
				‡0.183
Temporal inferior	20.54 ± 6.06	24.37 ± 3.96	25.51 ± 1.84	<0.001
				*<0.001
				†<0.001
				‡0.042

 Comparisons among the three groups were performed using generalized estimating equation method.

*  PG versus healthy.

†  PG versus PPG.

‡  PPG versus healthy.

Table 3.  

 

Comparison of Baseline GCA Thicknesses (μm)

	PG (n = 38)	PPG (n = 103)	Healthy (n = 61)	P Value
Average	73.03 ± 10.10	78.64 ± 7.24	84.93 ± 4.77	<0.001
				*<0.001
				†<0.001
				‡<0.001
Minimum	62.46 ± 15.53	74.52 ± 8.83	82.25 ± 5.33	<0.001
				*<0.001
				†<0.001
				‡<0.001
Temporal superior	71.62 ± 12.02	77.94 ± 7.55	83.41 ± 4.91	<0.001
				*<0.001
				†<0.001
				‡<0.001
Superior	74.05 ± 11.81	79.23 ± 8.12	86.20 ± 5.55	<0.001
				*<0.001
				†0.010
				‡<0.001
Nasal superior	78.44 ± 10.67	82.19 ± 8.81	87.44 ± 5.21	<0.001
				*<0.001
				†0.031
				‡<0.001
Nasal inferior	75.82 ± 11.59	79.24 ± 7.56	85.11 ± 5.56	<0.001
				*<0.001
				†0.009
				‡<0.001
Inferior	69.85 ± 14.97	75.74 ± 7.97	83.30 ± 4.81	<0.001
				*<0.001
				†0.005
				‡<0.001
Temporal inferior	68.33 ± 15.00	77.39 ± 857	84.13 ± 5.04	<0.001
				*<0.001
				†<0.001
				‡<0.001

 Comparisons among the three groups used generalized estimating equation method.

*  PG versus healthy.

†  PG versus PPG.

‡  PPG versus healthy.

Table 4.  

 

Comparison of Baseline ORL Thicknesses (μm)

	PG (n = 38)	PPG (n = 103)	Healthy (n = 61)	P Value
Average	126.26 ± 8.31	125.21 ± 7.95	126.49 ± 7.94	0.779
				*0.612
				†0.867
				‡0.481
Minimum	117.28 ± 11.34	115.27 ± 10.02	117.66 ± 10.76	0.486
				*0.845
				†0.395
				‡0.299
Temporal superior	129.18 ± 9.13	129.80 ± 8.87	129.79 ± 8.61	0.679
				*0.510
				†0.383
				‡0.886
Superior	130.49 ± 10.06	130.62 ± 9.37	131.72 ± 8.89	0.353
				*0.191
				†0.200
				‡0.703
Nasal superior	124.85 ± 11.49	123.41 ± 10.62	126.56 ± 10.27	0.316
				*0.328
				†0.641
				‡0.129
Nasal inferior	122.03 ± 9.76	120.61 ± 9.82	122.61 ± 9.67	0.589
				*0.796
				†0.449
				‡0.385
Inferior	123.62 ± 7.95	120.88 ± 8.71	121.80 ± 9.15	0.311
				*0.411
				†0.127
				‡0.606
Temporal inferior	127.44 ± 8.35	125.92 ± 9.00	126.43 ± 8.15	0.694
				*0.624
				†0.399
				‡0.628

 Comparisons among the three groups used generalized estimating equation method.

*  PG versus healthy.

†  PG versus PPG.

‡  PPG versus healthy.

AUCs of Data Obtained Using Information from the Three Segmented Macular Layers, and TMT and cRNFL Thickness Values

Analysis of GCA data yielded higher AUCs than were afforded by examination of data on other layers in all sectors evaluated, and the minimum parameters afforded the highest diagnostic abilities of all parameters at baseline (Table 5). The AUCs of average GCA values were inferior to that of average cRNFL thickness (0.869 vs. 0.953, P = 0.018). However, the minimum GCA parameters were comparable statistically, in terms of diagnostic utility, with cRNFL thickness assessment (0.889 vs. 0.935, P = 0.144). TMT analyses yielded lower AUCs than did evaluation of average GCA values (0.790 vs. 0.869, P = 0.05, Fig. 1). 

Table 5.  

 

The Glaucomatous Diagnostic Capabilities of Macular Thickness Parameters in Eyes with PG

	MNFL		GCA		ORL
	AUC (95% CI)	Cut-Off Value	AUC (95% CI)	Cut-Off Value	AUC (95% CI)	Cut-Off Value
Average	0.754 (0.658–0.835)	28	0.869 (0.722–0.946)	79	0.511 (0.409–0.611)	130
Minimum	0.719 (0.620–0.804)	10	0.889 (0.811–0.943)	69	0.505 (0.404–0.607)	120
Temporal superior	0.655 (0.553–0.747)	22	0.792 (0.699–0.867)	75	0.495 (0.393–0.596)	135
Superior	0.726 (0.627–0.810)	33	0.807 (0.716–0.879)	77	0.514 (0.412–0.615)	123
Nasal superior	0.649 (0.547–0.742)	38	0.769 (0.674–0.847)	85	0.523 (0.421–0.624)	120
Nasal inferior	0.694 (0.594–0.783)	33	0.740 (0.643–0.823)	81	0.505 (0.403–0.607)	111
Inferior	0.745 (0.648–0.827)	31	0.801 (0.710–0.874)	78	0.562 (0.459–0.661)	113
Temporal inferior	0.721 (0.622–0.806)	22	0.841 (0.754–0.906)	75	0.537 (0.435–0.638)	130

 The cut-point was determined as optimal by using the Youden index, which maximizes the sum of sensitivity and specificity.

Figure 1. 

 

Comparisons of receiver operating characteristic (ROC) curves created using average thickness of the GCA, TMT, and average cRNFL thickness.

Progression as Revealed by Optic Disc/RNFL Photography or VF Assessment

Two experts showed the same opinion in 124 eyes (87.9%) in the assessment of serial optic disc/RNFL photographs. The rest of them (17 eyes) required adjudication by a third examiner. 

Of the 141 glaucomatous eyes, 38 (27%, PG 17 and PPG 21) showed progression during follow-up as revealed by expert assessment of optic disc/RNFL photographs and/or VF testing. Of the 38 eyes with progression, 19 (50%) showed progression on optic disc/RNFL photographs only, 7 (18%) by VF GPA only, and 8 (21%) using both methods. The remaining four eyes (11%) showed progression only by VFI trend analysis. 

Progression as Revealed by Thickness Analysis of Each of the Three Segmented Macular Layers, and TMT and cRNFL Thicknesses

A total of 3, 5, and 5 eyes showed progression upon analysis of average data from the MNFL, GCA, and ORL layers, respectively. Also, 10 and 10 eyes progressed as determined by average TMT and cRNFL thickness measurements, respectively. The sensitivity and specificity of each parameter, in terms of progression detection, using either expert assessment of optic disc stereo photographs and/or VF progression as standards, are shown in Table 6. The sensitivities of TM, GCA, and cRNFL thickness measurements were 14.0%, 8.0%, and 5.0%, respectively. There was no statistically significant difference between TMT and cRNFL thickness (P = 0.129), nor between GCA and cRNFL thickness (P = 0.613). In the meantime, 2, 7, 0, 2, and 1 eye showed a significant positive trend (P < 0.05) by linear regression analysis against patient age using serial TM, cRNFL, and three segmented macular layer (MNFL, GCA, and ORL) thickness measurements, respectively. 

Table 6.  

 

The Sensitivity and Specificity of Average Parameters in Terms of Glaucoma Progression Detection

	% Sensitivity (95% CI)	% Specificity (95% CI)
MNFL	5.0 (0 ∼ 13.0)	99.0 (97.0 ∼ 100)
GCA	8.0 (0 ∼ 17.0)	98.0 (95.0 ∼ 100)
ORL	5.0 (0 ∼ 14.0)	97.0 (93.0 ∼ 100)
TMT	14.0 (5.0 ∼ 29.0)	95.0 (92.0 ∼ 99.0)
cRNFL	5.0 (0 ∼ 13.0)	92.0 (87.0 ∼ 97.0)

Agreement among the Various Progression Criteria

MNFL measurements rarely agreed with data from both the GCA and the ORL (κ = −0.027, P = 0.737; −0.027, P = 0.737). GCA data were in slightly better agreement with those from the ORL (κ = 0.171, P = 0.043) than with measurements on other layers. 

The extent of agreement, in terms of glaucoma progression, among GCA, TMT, and cRNFL thickness values was determined. When GCA and TMT data were compared, the κ value was 0.370 (P < 0.001). The GCA versus cRNFL comparison yielded a κ value of −0.050 (P = 0.529), whereas the TM versus cRNFL comparison was associated with a κ of 0.031 (P = 0.710, Fig. 2). Agreement in terms of glaucoma progression among data obtained using standard methods on one hand, and GCA, TMT, or cRNFL thickness values on the other hand was poor (κ = 0.082 [P = 0.090], 0.155 [P = 0.015], and −0.033 [P = 0.607], respectively; Fig. 3). The extent of agreement between values obtained using the standard method and TMT data was slightly better than other data. 

Figure 2. 

 

Area-proportional Venn diagram shows agreement in terms of glaucoma progression when GCA, TMT, and cRNFL average thickness values were used to this end.

Figure 3. 

 

Area-proportional Venn diagram shows agreement in terms of glaucoma progression when standard methods (optic disc stereophotography and/or VF assessment), and (a) GCA or (b) TM or (c) cRNFL average thickness values were used to this end.

Discussion

We found that the thickness at baseline of a segmented RGC layer (the GCA) differed significantly between glaucomatous and healthy eyes. A significant difference also was evident between healthy and preperimetric eyes, suggesting that GCA thickness might serve as an early indicator of glaucomatous structural damage. However, MNFL thickness did not differ significantly between PPG and healthy eyes. The NFL is thickest in the superior and inferior peripapillary areas, becoming gradually thinner as the distance from the disc margin increases. Because macular scan covers the 6 × 6 × 2 mm of retina with fovea as the center, most of peripapillary area, namely superotemporal and inferotemporal areas, wherein glaucomatous RNFL loss first appears can be covered. Nevertheless, a certain proportion of peripapillary RNFL bundle running outside the 6 × 6 × 2 mm field of macular scan is not included in our macular measurements. That may explain why MNFL did not differ significantly between PPG and healthy eyes. 

Baseline ORL thickness did not differ among the three groups of our present study, in line with data of a previous report by Vajaranant et al. ⁴⁶  

When we assessed the glaucoma diagnostic capacities of baseline measurements, we found that GCA thickness was reasonably useful in this respect. However, neither MNFL nor ORL measurements afforded good diagnostic capabilities. When this is considered together with the fact that ORL and MNFL thicknesses did not differ between healthy and glaucomatous eyes, it must be concluded that measurements on these two layers are not useful in terms of glaucoma diagnosis. 

Conventionally, among OCT parameters, cRNFL thickness has been used most frequently to diagnose glaucoma. Therefore, we compared the AUCs obtained using GCA measurements, and cRNFL and TMT thicknesses. Baseline average GCA data outperformed average TMT information in terms of discriminating between healthy and glaucomatous eyes. However, the average GCA data yielded a smaller AUC than did that obtained using average cRNFL information. This is in line with the findings of previous reports. ^21,23–27 When it is considered that our glaucoma patients were in relatively early stages of disease, including the preperimetric stage, we concluded that GCA measurements are not superior to cRNFL data when early structural diagnosis is required. This is because glaucomatous change usually begins in the superior and inferior temporal areas, and certain regions of these areas are not included in our macular measurements as mentioned earlier. 

If a glaucoma diagnostic strategy is to be useful in clinical practice, it should detect progression. This is as important as is cross-sectional glaucoma diagnosis because, if progression is confirmed, treatment should be modified or enhanced to prevent further irreversible loss of visual function. Glaucoma can progress structurally and/or functionally. Glaucomatous damage may be generalized and/or localized. Accordingly, detection of progression can be rather variable, depending on the detection strategy used; this problem is well-recognized. ^47–53 Hence, we used several relevant clinical criteria to define progression. Upon expert examination of photographs of the optic disc and RNFL, or using VF GPA, or via VFI TA, 38 eyes were determined to have progressed. These 38 eyes served as the reference when we calculated the sensitivities and specificities of various OCT-derived parameters. 

When GCA data were compared with cRNFL or TMT thickness measurements, GCA information categorized fewer eyes as having progression. In terms of the TMT thickness comparison, this is explained by the fact that the GCA is much thinner than is the TMT (by about 67%), and the amount of change in GCA thickness, thus, would be less than that in the TMT. Therefore, the extent of change in GCA thickness may be too low to permit detection, or may be beyond the level of resolution of current OCT technology. We also would speculate that progression may be evident in the ORL in some proportion of glaucomatous eyes. In our present study, we found that agreement was evident, in some eyes, between ORL and GCA data when these were used to detect progression. If this is true, concurrent thinning of the ORL and GCA may mean that TMT thickness measurements can be used to detect eyes showing more severe progression. Panda and Jonas found that, histologically, photoreceptor counts were lower in specimens from glaucomatous eyes. ²⁸ If, indeed, photoreceptor numbers become reduced in vivo as glaucoma progresses, this would lead to a reduction in ORL thickness. Notably, glaucoma principally is a disease of the RGC and ORL measurements were not useful in terms of diagnosis. However, if gradual thinning of the GCA and ORL layers occurs in some eyes as glaucoma progresses, it may become possible to detect glaucoma by conducting TMT thickness measurements. This could reveal progressing eyes not identified using GCA thickness data. In fact, although ORL thickness did not differ among PG, PPG, and healthy eyes, some eyes did show significant progression when ORL data were examined. Turning to the fact that GCA data detected fewer eyes as having progression, compared to the use of cRNFL thickness measurements, we cannot currently explain this result. 

Overall, the sensitivities of segmented macular parameters, and those of TMT and cRNFL measurements, in terms of detecting glaucoma progression evident on optic disc photographs and/or upon VF assessment, were not good. This also has been shown in previous reports. ^47–53 The reason why all tested strategies showed low sensitivity in terms of glaucoma progression detection may be attributable to a relatively short follow-up period (2.13 ± 0.33 years) and few OCT exams (cRNFL 4.5 ± 1.4, macular 4.6 ± 1.4) during follow-up. Outcome of regression analysis with fewer test points may be affected by outliers and, thus, show reduced sensitivity. 

The sensitivities afforded by use of macular parameters (GCA and TMT data) were better than that of cRNFL thickness measurements; however, these differences were not statistically significant. Agreement between TMT data with those derived using standard methods was slightly better than that obtained using cRNFL thickness measurements. Therefore, we would summarize our findings by observing that, although use of baseline cRNFL data is optimal in terms of glaucoma diagnosis, employment of macular parameters afforded similar level of sensitivities, and the data were in better agreement with those derived from use of standard methods than were cRNFL measurements. When GCA and TMT data were compared, use of TMT information detected more eyes as progressed. 

The reason why cRNFL measurements did not show better sensitivity than GCA or TMT measurement in terms of progression detection, although it was superior in cross-sectional glaucoma diagnosis, can be explained as below. First, cRNFL thickness loss is an early indicator of glaucomatous damage and, thus, some glaucomatous eyes already have substantial loss of RNFL at baseline. Change detection of RNFL thickness in such eyes with thin baseline RNFL may not be easy in some cases. Second, more eyes showed a significant positive trend in the linear regression analysis when assessed by cRNFL thickness measurement. When we consider that true increase in the thickness does not occur in real life, this result may suggest higher measurement variability of cRNFL thickness at different visits. Higher variability may reduce the progression detection capability. 

Our study had several limitations, including a short follow-up period and a small sample size. Additionally, we used disc and RNFL photography, or VF analysis, as reference standards. These two tests may not be perfect in terms of detection of progression, although both methods currently serve as reference standards. Also, all glaucomatous and healthy eyes were Asian (Korean) and, thus, our data may not be generalizable automatically to other races. 

In summary, compared to TMT information, GCA data were superior in terms of glaucoma diagnostic capability, but the GCA data were inferior to those derived from cRNFL analysis. However, in terms of detection of progression, cRNFL analysis did not outperform GCA nor TMT data. Macular and cRNFL algorithms seek to measure different areas and layers. Hence, variability in the use of such data for detection of progression, thus, is perfectly understandable. Given that various measurement strategies are possible using OCT, some may be helpful clinically for glaucoma diagnosis and to detect progression. Therefore, we suggest that macular parameters, such as those of the GCA and TMT, measured using OCT, can be used to detect glaucoma progression. 

Footnotes

 Disclosure: J.H. Na, None; K.R. Sung, None; S. Baek, None; Y.J. Kim, None; M.K. Durbin, Carl Zeiss Meditec, Inc. (E); H.J. Lee, None; H.K. Kim, None; Y.H. Sohn, None