Abstract
Purpose.:
To profile macular thickness changes in glaucoma by using three-dimensional spectral-domain optical coherence tomography (3D-SD-OCT).
Methods.:
The study included 30 eyes with suspected glaucoma and preperimetric glaucoma (SGPPG) and 35 healthy eyes. The macular thickness, including those of the total retina, nerve fiber layer (NFL), and combined inner retinal layers (IRLs)—NFL+ganglion cell layer (GCL)+inner plexiform layer (IPL)—was measured by 3D-SD-OCT raster scans in a 6 mm2 region. The average and sectoral thicknesses were calculated on an Early Treatment of Diabetic Retinopathy Study (ETDRS) chart and a ETDRS chart with a 45° rotation (glaucoma sector chart, GSC).
Results.:
The mean IRL thickness was significantly less in the SGPPG eyes than in the healthy eyes, but the mean total retinal and macular NFL thicknesses were not. In the SGPPG eyes, the IRLs were thinner in the outer macula than in the inner macula, in the inferior hemisphere than in the superior hemisphere, and in the temporal hemisphere than in the nasal hemisphere. The significantly thinned sectors were nearly identical on the GSC but only slightly overlapped on the ETDRS chart. The IRLs in the inferior temporal outer sector (GSC) had the greatest area under the receiver operating characteristic curve, which was significantly greater than those for the IRLs over the entire macula, inferior hemiretinal region, and inferior outer hemicircular region (macular subfields), and that for the circumpapillary NFL in the inferior sectors (P = 0.001–0.036).
Conclusions.:
Macular IRL thickness measured by using 3D-SD-OCT is useful for profiling macular atrophy in SGPPG.
Glaucoma causes retinal ganglion cell (RGC) loss, which results from damage to the RGC axon within the optic nerve head and is observable as changes in the appearance of the optic nerve head and nerve fiber layer (NFL) defects. Because they precede clinically detectable visual field defects,
1 –6 it is important to identify the structural changes resulting from RGC loss as early as possible, to prevent vision damage.
In studies using time-domain optical coherence tomography (TD-OCT),
7 –14 two thickness changes related to RGC loss—specifically, decreased thickness of the macula (where 50% of the RGCs are concentrated) and decreased thickness of the circumpapillary NFL (cpNFL), were noted in eyes with glaucoma. These studies consistently showed that cpNFL thickness is a better measure than macular thickness to discriminate eyes with glaucoma from healthy eyes. In these studies, cpNFL thickness was measured by using a simple circular scan, which can sample almost all axons arising from RGCs. Moreover, cpNFL thickness measurements exclude the outer retinal layers that are unaffected by RGC loss in glaucoma. However, macular thickness measurements by using TD-OCT appear to have two disadvantages: sparse sampling density and nonselectivity. Six radial line scans by TD-OCT, which requires interpolation to fill the nonimaged areas, may not be adequate for studying the earliest macular structural changes in glaucoma. Further, macular thickness measurements include the outer retinal layers that are unaffected by RGC loss in glaucoma. These disadvantages appear to be more influential in earlier glaucoma, because two previous studies
8,9 using TD-OCT found significant thinning of macular thickness in eyes with perimetric (early and advanced) glaucoma but could not find thinning in any region in eyes with suspected glaucoma, although significant thinning of the inferior cpNFL was noted.
8
To overcome the latter disadvantage (i.e., nonselectivity), Tan et al.
15 used TD-OCT to measure the combined thickness of specific inner retinal layers (IRLs) such as the NFL, ganglion cell layer (GCL), and inner plexiform layer (IPL) (where RGC axons, cell bodies, and dendrites are located, respectively) and found that the combined thickness of the macular IRLs (NFL+GCL+IPL) and cpNFL thickness are comparable indicators of glaucoma in eyes with suspected and preperimetric glaucoma (SGPPG). Further, Ishikawa et al.
16 used TD-OCT to show that the thickness of the macular inner retinal layers (NFL+GCL+IPL+inner nuclear layer [INL] and GCL+IPL+INL) and cpNFL thickness are comparable indicators of glaucoma in eyes with perimetric glaucoma. However, the limited sampling density owing to the low imaging speed of TD-OCT in the previous studies remains to be overcome.
7 –16
Faster imaging (43–133 times faster than that with TD-OCT) is possible with spectral-domain OCT (SD-OCT). The high imaging speed enables the application of a three-dimensional (3D) raster scan protocol with high sampling density. Such imaging allows comprehensive observations of normal and pathologic macular structures,
17 –19 regional macular thickness profiling,
20 and highly repeatable thickness measurements of individual retinal layers in the macula.
21 –24 The mean thickness of IRLs measured over the whole macula by using ordinary SD-OCT instruments has been shown to be as accurate as measurement of cpNFL thickness for discriminating between eyes with and without glaucoma.
25,26 However, 3D imaging of these specific layers in eyes with glaucoma has, to our knowledge, not been reported. Herein, we report our findings in 3D-SD-OCT imaging with a raster scan protocol to measure the average and regional thicknesses of IRLs (NFL+GCL+IPL) in eyes with SGPPG.
The subjects for this prospective observational case study were all patients with SGPPG who were examined by the glaucoma service at Kyoto University Hospital (Kyoto, Japan) between June 2007 and December 2008. Volunteers with healthy eyes (intraocular pressure, ≤21 mm Hg) were also enrolled. All examinations were performed according to a protocol that complied with the tenets of the Declaration of Helsinki and that was approved by the Institutional Review Board and Ethics Committee of the Kyoto University Graduate School of Medicine. Informed consent for study participation was obtained from all subjects.
All patients underwent a comprehensive ocular examination, including automated refractor keratometer, measurements of uncorrected visual acuity and best corrected visual acuity with the 5-m Landolt chart, slit lamp biomicroscopy, gonioscopy, Goldmann applanation tonometry (IOL Master 500; Carl Zeiss Meditec AG, Jena, Germany), standard automated perimetry (Humphrey Field Analyzer; Carl Zeiss Meditec), dilated stereoscopic examination of the optic disc, stereo disc photography (3-Dx simultaneous stereo disc camera; Nidek Co., Ltd., Gamagori, Japan), and red-free fundus photography. All the examinations were performed on the same day.
Subjects with a neurologic disease or a history of diabetes or corticosteroid use were excluded. Eyes with any kind of retinal disease, opaque media, or a history of surgery were excluded. Eyes were also excluded if they had best corrected visual acuity worse than 20/25, refractive error outside the range −6.00 to +3.00 D, or did not have open angles confirmed by gonioscopy.
The patients and volunteers underwent standard automated perimetry (Humphrey 24-2 Swedish Interactive Threshold Algorithm [SITA]; Carl Zeiss Meditec, Dublin, CA) at least twice within 6 months of the comprehensive ocular examination; reliable visual field test results (fixation loss ≤ 20%, false-positive rate ≤ 15%, and false-negative rate ≤ 33%) were used. Visual field defects caused by glaucoma were defined, on the basis of these results, as abnormal range on glaucoma hemifield test or pattern SD of <5% of the normal reference (confirmed by two consecutive tests).
Eyes without visual field defects were categorized as healthy or as having SGPPG on the basis of the appearance of the optic nerve head on fundus photographs, including stereoscopic photographs, evaluated by three glaucoma specialists (YK, MH, and FH), who were blinded to other information on the eyes according to previously defined criteria.
27 Glaucoma was diagnosed if the eye had a vertical cup-to-disc ratio ≥ 0.9, vertical cup-to-disc ratio greater by ≥ 0.3 than that of the contralateral eye, or a rim-to-disc ratio ≤ 0.05. Glaucoma was suspected if the eye had a vertical cup-to-disc ratio between 0.7 and 0.9, vertical cup-to-disc ratio greater by 0.2 to 0.3 than that of the contralateral eye, or rim-to-disc ratio between 0.05 and 0.1, or if NFL defects were seen. NFL defects were defined as localized wedge-shaped defects in the NFL clearly visible on red-free fundus photographs. The three examiners classified each eye as having SGPPG or as healthy; if there was no agreement on a classification, they discussed the findings, by referring to the color fundus photographs, stereophotographs, and red-free fundus photographs, until a consensus was reached.
The location of the NFL defects was described according to the 24-hour clock on red-free retinal NFL photographs (e.g., 7 o'clock was between 6:30 and 7:30; 8 o'clock was between 7:30 and 8:30). The center of each NFL defect at the disc margin was used as the representative location. The directional angle was assessed in the clockwise direction in the right eye and counterclockwise direction in the left eye. The location of localized rim thinning was similarly described.
Two 3D-SD-OCT protocols were conducted by using 3D-OCT-1000 (Topcon Corp., Tokyo, Japan). For macular thickness analysis, raster scanning over a 6-mm
2 area centered on the fovea was conducted with a scan density of 512 A-scans (horizontal) × 128 B-scans (vertical;
Fig. 1). For cpNFL thickness analysis, the scanning protocol consisted of seven circular scans of 2.2-, 2.5-, 2.8-, 3.1-, 3.4-, 3.7-, and 4.0-mm diameter, and 12 (one at each hour of the clock face) 6-mm-long radial scans, all centered on the optic disc. The cpNFL thickness measured on the 3.4-mm diameter circular scan was used for analysis.
The 3D-OCT-1000 in-built software automatically generated three boundary lines between the retinal layers by using a threshold algorithm that searches for changes in reflectivity. These boundary lines marked the vitreoretinal interface (
Fig. 1, red line), boundary between the NFL and the GCL (yellow line in
Fig. 1), and boundary between the retinal pigment epithelium (RPE) and the outer segment (OS; green line in
Fig. 1). The boundary line between the inner nuclear layer (INL) and the inner plexiform layer (IPL; blue line in
Fig. 1) was drawn manually in all the 128 B-scans per 3D scan with electronic calipers by the delineators certified in OCT image analysis by the Kyoto OCT Reading Center.
20 These boundary lines were used to measure the thickness of the total retina, macular NFL, and IRLs (NFL +GCL+IPL); the mean thicknesses in each sector were used.
Two sets of macular scans and two sets of cpNFL scans were obtained from each patient or volunteer eye at the same visit and used to assess the repeatability of measurements in that study eye.
Manual delineation of the boundary line was done by two independent masked delineators, using a standardized grading protocol.
20 Raw, black-and-white images, not processed images, were used for the measurements. We then assessed the interobserver agreement for measurements made on the same images by the delineators.
For this study, we used a glaucoma sector chart (GSC;
Supplementary Fig. S1) that divides the extrafoveal macula into eight sectors for measurements of retinal layer thickness in specific macular regions. The GSC, which was originally used by Wollstein et al.
11 to investigate the relationship of total macular thickness and visual field defects in eyes with glaucoma, is based on the Early Treatment Diabetic Retinopathy Study (ETDRS) chart, which divides the macula into inner and outer rings, with the inner ring being 1 to 3 mm and the outer ring being 3 to 6 mm from the foveal center. The ETDRS chart divides the macula further into superior, inferior, temporal, and nasal quadrants. However, the lines dividing these quadrants cross the horizontal meridian, which may misrepresent regions of structural change due to glaucoma; therefore, the GSC was created by rotating the ETDRS chart centered on the fovea by 45°. Thus, on the GSC, the line dividing the inferior and superior quadrants was horizontal. We used both the GSC and the ETDRS chart to calculate and compare the mean regional thicknesses of the retinal layers measured on the 3D-SD-OCT images of the SGPPG and healthy eyes.
Average macular thickness maps were created from the volume data set of segmented IRLs obtained by the 3D-SD-OCT 512 A-scans (horizontal) × 128 B-scans (vertical) (
Fig. 1). The measurement values on the raster scans were interpolated based on the thicknesses of the four neighboring A-scans to fill the 512 (horizontal) × 128 (vertical) pixels on the macular substructure segmentation map. Fractional deviation maps were calculated by subtracting the average IRL thickness maps of the SGPPG eyes from those of the healthy eyes and dividing the difference maps by the normal maps.
To assess the ability of measuring the IRL thickness in various macular sectors to discriminate between eyes with and without glaucoma, we calculated the area under the receiver operating characteristic (AROC) curve for the IRLs that were significantly thinned in the SGPPG eyes versus the healthy eyes. An AROC curve displays the relationship between sensitivity and 1 − specificity for the results of a given diagnostic test. An AROC curve of 1.0 indicates that the test perfectly discriminates between the presence and the absence of the condition, whereas an AROC curve of 0.5 represents chance discrimination. Commercially available software (MedCalc, ver. 9.3.8.0; MedCalc Software, Mariakerke, Belgium) was used to compare the AROC curves.
The statistical significance of differences between the SGPPG eyes and the healthy eyes was evaluated with the Mann-Whitney test.
The repeatability and interobserver reproducibility were calculated from measurements obtained at the two imaging sessions in the same visit. Repeatability was defined as 2.77 times the within-subject SD, which was calculated as half of the square root of the average variance of the difference between the sessions over subjects.
23 By definition, intersession differences for 95% of the subjects fall within twice the repeatability range. Interobserver reproducibility was defined as 2.77 times the interdelineator within-subject standard deviation, calculated as half the square root of the average variance of the difference in measurements obtained by the two delineators over subjects. Interobserver differences for 95% of the cases fall within the mean ± reproducibility range. The coefficient of variation (CV; 100 × within-subject SD/overall mean) and intraclass correlation coefficient (ICC) were computed. (SPSS for Windows, ver. 11.0J; SPSS, Inc., Chicago, IL was used for intergroup comparisons.)
P < 0.05 was considered statistically significant.
Table 1 shows that the patients did not differ in age or sex from the volunteers, but the refractive power and intraocular pressure were significantly higher in the SGPPG eyes than in the healthy eyes.
Among the SGPPG eyes, 13 had localized rim thinning and the rest (17 eyes) had diffuse rim thinning. The localized rim thinning was detected within the 7-o'clock sector in 10 eyes, 6-o'clock sector in two eyes, and 12-o'clock sector in one eye. Seven (23.3%) of the 30 SGPPG eyes had NFL defects on the color or red-free fundus photographs. All these defects were located between the 7- and 8-o'clock sectors. The center of the NFL defects at the disc margin was located within the 7-o'clock sector in six (85.7%) eyes and within the 8-o'clock sector in one (14.3%).
Figure 2 shows maps of the average IRL thickness of the SGPPS and healthy eyes, and
Figure 3 shows a fractional deviation maps.
Table 2 shows the mean thickness in various macular areas (macular subfields) of the total retina, IRLs, and macular NFL in the healthy and SGPPG eyes; all three layers were significantly thinned in the outer ring (3–6 mm from the foveal center) and in the outer ring of the inferior hemisphere in the latter group. The IRLs were also significantly thinned in all but the inner ring, either for the whole macula or just the superior or inferior inner hemicircular sector.
Table 3 shows the mean thickness of the total retina, IRLs, and macular NFL in the healthy and SGPPG eyes in each sector of the GSC. All three layers were significantly thinner in the SGPPG eyes in the inferior nasal outer and inferior temporal outer sectors, and all layers but the NFL were significantly thinner in the superior temporal outer sector.
Table 4 shows mean thicknesses in the macula of the total retina, IRLs, and NFL in healthy eyes and eyes with SGPPG in each sector of the ETDRS chart. All four layers were significantly thinner in eyes with SGPPG only in the inferior outer sector. In addition, they were significantly thinned in the superior outer sector (total retina, NFL) or temporal outer sector (total retina, IRLs). The IRLs were also thinned in the inferior and temporal inner sectors, and the IRLs and NFL were thinned in the nasal outer sectors.
The cpNFL was significantly thinner in the SGPPG eyes along the 6-o'clock sector and in the inferior quadrant (
Table 5).
Table 6 shows the AROC curves for the sectors with significant thinning of the total retina, IRLs, or NFL in the SGPPG eyes versus the healthy eyes, as determined by comparison of the results based on the macular subfields, GSC, and ETDRS chart. In the macular subfield analysis of the mean thickness of the macular layers in the outer sectors (3–6 mm from the foveal center), the AROC curve for total retinal thickness (0.65 ± 0.07) was significantly less than that for IRL thickness (0.78 ± 0.06;
P = 0.004), which was significantly greater than that for NFL thickness (0.65 ± 0.07;
P = 0.018). In the inferior outer hemicircular sector, the AROC curve for IRL thickness was significantly greater than those for total retinal (
P = 0.004) and macular NFL (
P = 0.017) thicknesses. In the individual sectors of the GSC, the AROC curve for the IRL thickness (0.86 ± 0.05) was the greatest in the inferior temporal outer sector, and this value was significantly greater than the AROC curve in the same sector for the total retinal and NFL thickness (
P < 0.001 and
P = 0.001, respectively). Further, the AROC curve for the IRL thickness was significantly greater in the inferior temporal outer sector than in the inferior nasal outer sector (
P = 0.004). On the ETDRS chart, the AROC curve for the IRL thickness was significantly greater in the temporal outer sector than in the nasal outer sector and superior outer sector (
P = 0.003 and
P = 0.047, respectively), but not the inferior outer sector.
As shown in
Table 6, overall, the AROC curve for the IRL thickness in the inferior temporal outer sector (GSC) was significantly greater than for the IRL thicknesses over the whole macula (1–6 mm from the foveal center;
P = 0.005), in the outer ring (3–6 mm from the foveal center;
P = 0.03), in the inferior hemiretinal sector (
P = 0.013), and in the inferior outer hemicircular sector (
P = 0.036) of the macular subfields.
Table 7 shows the AROC curves with significant differences in the mean cpNFL thickness between the healthy eyes and the SGPPG eyes in the 6-o'clock sector (when 12-hour sectors were compared) and the inferior quadrant (when quadrants were compared). The AROC curve for the IRL thickness in the inferior temporal outer sector (GSC) was significantly larger than these sectors (
P = 0.001 and 0.009, respectively).
We calculated the variability of measurements of the total retinal and macular NFL thicknesses (automatic measurements) in 13 healthy eyes and 12 SGPPG eyes according to the macular subfields (two, four, or all sectors;
Table 8), GSC, ETDRS chart, or cpNFL thickness by 24-hour sectors or quadrants. Overall, the measurements tended to be less variable in the healthy eyes (macular subfields: CV of total retina = 0.26%–0.48%, CV of macular NFL = 2.02%–7.00%; GSC: CV of total retina = 0.46%–0.59%, CV of macular NFL = 2.47%–9.40%; ETDRS chart: CV of total retina = 0.57%–0.92%; CV of macular NFL = 2.61%–14.30%; and cpNFL: CV of average cpNFL =1.52%, CV of regional cpNFL = 2.25%–7.26%) than in the SGPPG eyes (macular subfields: CV of total retina = 1.01%–1.98%, CV of macular NFL = 2.88%–9.47%; GSC: CV of total retina = 0.88%–2.24%, CV of macular NFL = 2.84%–16.32%; ETDRS chart: CV of total retina = 0.30%–1.08%, CV of macular NFL = 2.49%–23.07%; and cpNFL: CV of average cpNFL = 1.60%, CV of regional cpNFL = 2.72%–11.19%). Thus, the measurements also tended to be less variable for total retina and cpNFL thicknesses than for macular NFL thickness in both the healthy eyes and the SGPPG eyes. Macular NFL thickness showed relatively high variability in the inner sectors (CV = 5.02%–9.47%) compared with the outer sectors (CV = 2.18%–3.21%) of the macular subfields, in the temporal sectors (CV = 4.59%–16.32%) compared with the nasal sectors (CV = 2.47%–10.06%) on the GSC, and in the temporal sectors (CV = 10.51%–23.07%), compared with the nasal sectors (CV = 3.53%–10.04%) on the ETDRS chart.
Table 9 shows the interobserver agreement for the IRL thickness measurements in various macular sectors. Overall, the IRL thickness measurements showed comparably small variability between the healthy eyes (CV = 0.38%–1.86%) and the SGPPG eyes (CV = 0.34%–1.29%).
In this study, we used SD-OCT with a raster scan protocol to create 3D images of the macula for measuring the thickness of the IRLs in various macular sectors and found that eyes with SGPPG have significant IRL thinning compared with healthy eyes and that the IRL thickness is a more accurate indicator of glaucoma than the macular total retina and cpNFL thicknesses. The key findings in the macular sector analysis were greater thinning of the total retina and IRLs in the outer ring than in the inner ring, in the inferior hemisphere than in the superior hemisphere, and in the temporal hemisphere than in the nasal hemisphere, which are consistent with the results of previous studies.
8,11,13 –15,25 In our study, the greatest AROC curve (0.86) was for the IRL thickness in the inferior temporal outer sector of the GSC.
The macula comprises multiple layers with different configurations and symmetry. Glaucoma is associated with morphologic changes that affect some macular layers, particularly the NFL and GCL. These layers are not anatomically uniform: although the GCL is nearly three-dimensionally symmetric, the NFL is symmetric only vertically. Glaucomatous morphologic changes in the macula are not necessarily identical in each of these layers or in each sector. The 3D raster scan we used can be expected to precisely profile the complex macular structures and their glaucomatous thinning. In most previous studies, the slow imaging speed of TD-OCT caused low sampling density; a radial scanning protocol consisting of six radial B-scans required interpolation to fill areas between the scans.
7 –16 Using 3D-SD-OCT, we were able to sample the macula with greater density in a short time and our raster scan protocol did not require interpolation. In fact, our protocol of 512-A scans (horizontal) × 128 B-scans (vertical) provides horizontal pixel spacing of 11.7 μm (6 mm/512 A-scans) and vertical pixel spacing of 46.9 μm (6 mm/128 A-scans). This raster scan protocol would theoretically allow almost infinitely detailed sampling.
In this study, on the 3D-SD-OCT images, the mean retinal thickness over the entire macula was a less accurate indicator of glaucoma than was the IRL thickness. This finding is consistent with the findings by Tan et al.
15,25 (who used TD-OCT with six radial scans and also SD-OCT with RTVue-100 [Optovue, Inc., Fremont, CA] to obtain serial vertical scans) and Ishikawa et al.
16 (who used TD-OCT with six radial scans). This closer association of change in IRL thickness compared with total retinal thickness in glaucoma is expected, because 65% to 70% of the total retinal thickness is due to the outer layers, which are unaffected by glaucoma.
Sectoral differences in the reduction in macular thickness were revealed on 3D-SD-OCT imaging; these differences may be attributable, at least in part, to the location of the glaucomatous damage.
11 Localized rim thinning was detected in the inferior temporal regions (7-o'clock sector) in a large part of the eyes in this study. This localization may be associated with the greatest AROC curve for the IRL thickness in the inferior temporal outer sector on the GSC. However, it was difficult to study the relationship between the sectoral differences and the location of localized rim thinning further, because the number of eyes with localized rim thinning in each clock-hour location was small except for the 7-o'clock sector among our subjects. A larger scale study will be required to correlate the sectoral differences with the location of glaucomatous damage on the optic disc appearance.
In our study, the average macular IRL thickness was a good indicator of glaucoma, whereas, unexpectedly, we did not find significant thinning of the average cpNFL thickness in the SGPPG eyes compared with the healthy eyes. Previous studies found that the average macular IRL and cpNFL thicknesses measured by either SD-OCT or TD-OCT have comparable glaucoma discrimination ability.
15,16,25 We do not know why cpNFL thickness was a less accurate indicator of SGPPG in our study but it may be attributable to differences in the study population: the criteria for enrollment in earlier studies were based on qualitative assessment of the optic nerve head appearance, and we used quantitative assessment.
27 For example, in our criteria for diagnosis of preperimetric glaucoma, a patient with a vertical cup-to-disc ratio ≥ 0.9 or rim-to-disc ratio ≤ 0.05 without visual field defects is more likely to have a myopic tilted disc or disc anomaly such as inferior tilted disc. However, we did not include eyes with refractive errors greater than −6 D and included only two eyes with myopic tilted discs. Another possible reason is that the location of the NFL defects in the SGPPG eyes affected the superiority of the macular structures to cpNFL thickness in glaucoma discrimination ability. For example, if the NFL defects are located closer to the fovea and within the macula, the results may favor macular thickness rather than cpNFL thickness. Unfortunately, in our study, only 7 (23.3%) of 30 SGPPG eyes had photographic localized NFL defects. A larger scale of study is required to answer to this question. A third reason is that 3D-SD-OCT imaging with a raster scan increases the glaucoma discrimination ability of macular IRL thickness compared with the radial or serial vertical scans used in the previous studies. However, this is uncertain, because the AROC curve for the mean IRL thickness over the whole macula in our study (0.74) was comparable to the AROC curve for the IRL thickness (0.78) measured on SD-OCT serial vertical scans in eyes with preperimetric glaucoma in a previous study.
25
Tan et al.
25 suggested that the thickness of both the macular IRLs and the cpNFL should be considered to be able to detect glaucoma in a complementary fashion. In their study, both parameters were abnormal in only 33% of the eyes with preperimetric glaucoma, but 12% and 11% of the eyes with preperimetric glaucoma had abnormal thinning of only the cpNFL or only the GCC (comparable to the IRL thickness in our study), respectively. Measurements of cpNFL thickness, obtained on circle scans, theoretically represent the sum of the thickness of all RCG axons, whereas measurements of macular IRL thickness can be expected to identify localized areas of RCG loss.
25 This specificity of each parameter may be responsible for the complementary glaucoma detection ability and may underline the different results with regard to the superiority of the macular IRL and cpNFL thickness in glaucoma discrimination ability between the previous studies and our study.
15,16,25
Other studies showed that the average cpNFL thickness has higher or comparable glaucoma discrimination ability to cpNFL thickness in any quadrant, including the inferior one, or in any clock-hour sectors, including the 6-o'clock one in perimetric glaucoma.
12 –14 In our study, there were no significant differences between the healthy eyes and the SGPPG eyes in the average cpNFL thickness or sectoral cpNFL thickness except for the inferior quadrant and 6-o'clock sector, which is seemingly inconsistent with the previous studies. Guedes et al.
8 reported using a prototype TD-OCT in which the average cpNFL thickness had comparable or better glaucoma discrimination ability to the inferior and superior cpNFL thicknesses in eyes with early and advanced glaucoma but lower glaucoma discrimination ability than the inferior cpNFL thickness in eyes with suspected glaucoma and ocular hypertension. Thus, the earliest glaucomatous changes, particularly those associated with localized rim thinning, in eyes with SGPPG may be more difficult to detect in terms of the average cpNFL thickness than with the sectoral cpNFL thickness. As already mentioned, the average cpNFL thickness is a measure of the global thickness of the NFL and therefore is probably most effective in differentiating relatively severe cpNFL damage in perimetric glaucoma from healthy eyes, but it is less effective for differentiating less cpNFL damage in SGPPG, because the cpNFL changes in this early condition are presumably small and localized. In contrast, it has been shown that macular thickness in the inferior outer sector has higher glaucoma discrimination ability than the average macular thickness and any other sectoral macular thickness,
13,14 consistent with our finding that the macular IRL thickness in the inferior temporal outer sector of the GSC has the highest glaucoma discrimination ability. Thus, sectoral analysis of cpNFL and macular IRL thicknesses may be useful, particularly for detecting SGPPG.
Leung et al.
28 used Stratus (TD)-OCT 3.4-mm diameter circular scans (the so-called fast RNFL scans; Carl Zeiss Meditec, Inc.) to evaluate thinning of the macular NFL in eyes with suspected or perimetric glaucoma. In eyes with perimetric glaucoma, but not in those with suspected glaucoma, they found a significant reduction in macular NFL thickness, but they found no advantage of these measurements over the total macular thickness measurements for detecting perimetric glaucoma. In our study, we found significant thinning of the NFL not in the entire macula but in the outer ring in the SGPPG eyes. However, the NFL thickness in the outer ring had only weak diagnostic power, consistent with the results of the previous study.
On the GSC, the sectors with significant thinning of the total retina, IRLs, and macular NFL in the SGPPG eyes were almost identical, whereas on the ETDRS chart, there was only partial overlap of the sectors with significant thinning of the total retina, IRLs, and macular NFL. In addition, the AROC curve for the IRL thickness in the inferior temporal outer sector on the GSC was greater than those of any other sector of the macular subfields or the ETDRS chart. We used GSC because its sectors appear to match the areas susceptible to thinning by glaucoma. Macular subfields do not assess the differences between the temporal and the nasal parts. The ETDRS chart, which was originally designed for evaluating diabetic macular edema and is not optimized for glaucoma detection; this chart includes four sectors that cross the horizontal meridian, which does not appear to reflect the anatomy of the NFL.
From the results of histologic studies of enucleated eyes of patients with glaucoma by Quigley et al.
29 and Kerrigan-Baumrind et al.
30 and eyes of primates with experimental glaucoma by Harwerth et al.,
31 it appears that between 30% and 50% of the RGCs in specific areas of the macula are lost by the time visual field defects are detectable. Consistent with the results of these studies, our study showed significant thinning of RGC-related macular structures before visual field defects were detectable. However, we found that the mean IRL thickness over the whole macula in the SGPPG eyes was reduced to only 93.3% of the mean thickness in the healthy eyes. Even in the inferior temporal sectors on the GSC, where the IRLs were the most thinned, the mean IRL thickness was 88.1% of that in the healthy eyes. There may have been lesser changes in IRL thickness in our study compared with the reduction in RGC populations in the histologic studies because the IRLs include the IPL, which appears to be minimally, if at all, affected by glaucoma. Another possible explanation for the difference in the proportion of RGCs lost versus the decrease in the mean IRL thickness is that residual glial tissue may partially compensate for some of the space for dead RGCs, which can cause lesser changes in IRL thickness compared with reduction in RGCs, as speculated on the basis of a case of end-stage visual field defect (−35 dB) in which the fractional GCC (comparable to the IRLs in our study) loss was approximately 50%.
25
The automated measurements of macular NFL thickness were highly variable in some sectors. This variability is partly attributable to the failure of automated segmentation to properly detect the outer boundary of the NFL in some sectors, whereas with the IRL thickness measurements combined with manual segmentation of the outer boundary of the IPL, the variability dramatically decreased, suggesting that manual segmentation is more reproducible. In addition, the high variability of the macular NFL thickness measurements in some sectors could also be attributed to the very thin macular NFL, which is thinner than the axial resolution of the SD-OCT instrument (6 μm); the macular NFL is extremely thin in the vicinity of the foveal center and temporal raphe, even in healthy eyes and could also be thin in severely damaged areas due to glaucoma. In fact, the macular NFL thickness showed relatively high variability in the inner sectors compared with the outer sectors on the macular subfields, in the temporal sectors compared with the nasal sectors on the GSC, and in the temporal sectors compared with the nasal sectors on the ETDRS chart. Furthermore, the macular NFL thickness tended to show higher variability in the SGPPG eyes than in the healthy eyes.
In conclusion, 3D-SD-OCT imaging allowed us to measure the IRL thickness in macular sectors defined by using the ETDRS chart or GSC. Comparisons of the AROC curves for various parameters showed that thinning of the IRLs is more indicative of glaucoma than that of the total retina in the macula. IRL thinning was also a more accurate indicator of glaucoma in our study than cpNFL thinning. In the macular sector analysis, the regional IRL thickness in the inferior temporal outer sector on the GSC showed the greatest AROC curve. These results indicate that 3D profiling of the macular IRLs is useful for detecting early glaucomatous changes.
Supported in part by Grant-in-Aid for Scientific Research 20592038 from the Japan Society for the Promotion of Science (JSPS) and Topcon Corp. (Tokyo, Japan).
Disclosure:
Y. Kotera, None;
M. Hangai, Nidek Co., Ltd. (C), Topcon Corp. (C);
F. Hirose, None;
S. Mori, None;
N. Yoshimura, Nidek Co., Ltd. (C), Topcon Corp. (C)