Beta and Gamma Peripapillary Atrophy in Myopic Eyes With and Without Glaucoma

Abstract

Purpose: To determine whether beta and gamma peripapillary atrophy (PPA) areas measured with optical coherence tomography (OCT) enhances glaucoma diagnosis in myopic subjects.

Methods: We included 55 myopic glaucoma patients and 74 myopic nonglaucomatous controls. Beta-PPA comprised the area external to the clinical disc margin, with absence of retinal pigment epithelium and presence of Bruch's membrane. Gamma-PPA comprised the area external to the disc margin, with absence of both RPE and Bruch's membrane. OCT scans colocalized to fundus photographs were used to measure PPA, choroidal thickness, border tissue of Elschnig configuration, optic disc area, and optic disc ovality.

Results: Beta-PPA area was larger in glaucoma patients compared with controls (median [interquartile range], 1.0 [0.66–1.53] mm² and 0.74 [0.50–1.38] mm², respectively), whereas gamma-PPA was smaller in glaucoma patients compared with controls (0.28 [0.14–0.50] mm² and 0.42 [0.17–0.74] mm², respectively). However, the distributions of both beta- and gamma-PPA in the two groups overlapped widely. The areas under the receiver operating characteristic curve of beta- and gamma-PPA areas were 0.60 and 0.59, respectively. Larger beta-PPA area was associated with larger disc area, thinner choroidal thickness, longer axial length, less oblique border tissue configuration, older age, and greater disc ovality. Larger gamma-PPA area was associated with greater disc ovality, more oblique border tissue configuration, and longer axial length.

Conclusions: Subclassifying PPA with OCT into beta and gamma zones reveals association with different covariates, but does not enhance the diagnostic performance for glaucoma in a population of predominantly Caucasians myopic subjects.

Beta peripapillary chorioretinal atrophy (beta-PPA) is defined as an area of visible sclera adjacent to the clinically visible optic disc margin,¹ corresponding to an area without RPE.^2,3 Although both the prevalence and size of beta-PPA is increased in glaucoma,^4–6 these features are also observed in myopia^7–9 and with increasing age.^4,8,10 Recent histologic¹¹ and spectral-domain optical coherence tomography (OCT)^12–16 imaging studies proposed a new classification for beta-PPA, dividing it into two subsets that are indistinguishable by clinical examination: an area with intact Bruch's membrane, still termed beta-PPA, and an area lacking Bruch's membrane, termed gamma-PPA. Recent studies showed that gamma-PPA had a higher association with myopia,^11–13,15 and could represent anatomical changes resulting from globe elongation, whereas beta-PPA with intact Bruch's membrane has a higher association with glaucoma,^11,14,15,17 reduced choroidal volume,¹⁷ and older age,^13,16,17 possibly representing degenerative changes. Therefore, this new subclassification of beta-PPA according to OCT findings could potentially be used to differentiate myopic eyes with and without glaucoma. 

Myopic eyes frequently pose a challenge in the diagnosis of glaucoma because several findings usually attributed to glaucoma can occur in nonglaucomatous myopic eyes, including visual field,^18,19 retinal nerve fiber layer,^20,21 and neuroretinal rim defects.^22,23 Additionally, as myopia prevalence rises to epidemic levels,²⁴ improving the diagnostic performance in these cases becomes increasingly relevant. The purpose of this study was to determine whether OCT-based quantification of PPA subsets enhances the glaucoma diagnosis in myopic subjects. 

Methods

Participants

Study participants included myopic glaucoma patients and myopic healthy controls. Patients were recruited consecutively from the glaucoma clinic at the Eye Care Centre, Queen Elizabeth II (QEII) Health Sciences Centre, Halifax, Canada. Controls were recruited consecutively from a local optometric practice. Study visits occurred from December 2012 to July 2013. For both groups, inclusion criteria were presence of peripapillary atrophy with clinical examination, best-corrected visual acuity of 20/40 or better, myopia greater than −2 diopters (D) (spherical equivalent) and astigmatism less than 4 D, and absence of degenerative myopia, or other retinal or optic nerve disease other than glaucoma. If both eyes were eligible, one eye was randomly selected as the study eye. The study was approved by the Ethics Review Board of QEII Health Sciences Centre and followed the tenants of the Declaration of Helsinki. Each subject gave written informed consent. 

Study Procedures

The study procedures were described in detail in a previous publication.²³ The study visit consisted of best-corrected distance visual acuity, slit lamp and fundus examination, ocular biometry (IOLMaster; Carl Zeiss Meditec, Dublin, CA, USA), stereo disc photography (Visucam Pro NM; Carl Zeiss Meditec), standard automated perimetry (SITA 24-2 strategy of Humphrey Field Analyzer; Carl Zeiss Meditec), and OCT imaging (Spectralis, Heidelberg Engineering GmbH, Heidelberg, Germany). All glaucoma patients and most myopic controls were perimetrically experienced. For individuals who were perimetrically naïve, a training visual field test was performed initially with the second test used for the study. All visual field examinations included in the study were judged reliable by the perimetrist and by the reliability indices (≤15% false positives, ≤30% fixation losses, and ≤30% false negatives). 

Two scan settings were used with OCT, both centered on Bruch's membrane opening (BMO) and aligned with the fovea-BMO center axis. Initially, the foveal pit and four BMO points were manually identified in two live B-scans. These positions were used by the device to automatically position the scans.²⁵ The first scan setting contained 24 radially equidistant B-scans, each subtending 15°. Data for each B-scan were averaged from 25 individual B-scans, with 768 A-scans per B-scan. The second scan setting was a 12° circular peripapillary B-scan with 768 A-scans, averaged 100 times. All scans were obtained with both standard and enhanced depth imaging²⁶ acquisition. 

Study Definition of Myopic Glaucoma Patients and Myopic Controls

The diagnosis of “myopic glaucoma” or “myopic control” was defined by consensus among three fellowship-trained glaucoma subspecialists who evaluated the visual fields and optic disc photographs from all participants independently, and were masked from all other demographic and clinical information. To minimize bias of defining glaucoma based on the amount of PPA, visual field appearance was primarily used for designating the diagnostic group of the participants. Subjects were included in the myopic control group if their visual field was graded as normal or with abnormalities consistent with myopia, but not glaucoma, independently by all three clinicians, irrespective of the grading given to their optic disc. If all three clinicians graded the visual field as having glaucomatous abnormalities, the participant was included in the myopic glaucoma group. In cases in which there was incomplete agreement in the visual field grading, the clinicians used their optic disc evaluation to obtain a consensus decision to place the participant in the myopic glaucoma or myopic control group. 

Optical Coherence Tomography Measurements

We used a prerelease version of the Spectralis device software (Spectralis Viewing Module 6.0.12.103), developed in collaboration with Heidelberg Engineering, to undertake image segmentation and quantitative measurements. Image scaling within the software adjusted measurements for magnification effects. The software automatically identified the inner limiting membrane and BMO in the radial scans, and the inner limiting membrane, the retinal nerve fiber, layer and the Bruch's membrane in the circular scans. The segmentation was checked and manually corrected when necessary. Global averages of BMO-minimum rim width (BMO-MRW) and circumpapillary retinal nerve fiber layer thickness (RNFLT) were exported for analysis. 

Optic Disc.

The color fundus photographs were imported into the software and automatically colocalized with the infrared image and OCT B-scans. The clinical disc margin, defined as the inner edge of scleral ring, was delineated in the color photographs. The longest and shortest diameters of the disc were estimated with an elliptical approximation of the disc margin and used to compute the disc ovality index²⁷ (the shortest diameter divided by the longest diameter, hence the smaller the ovality index, the more oval the disc). 

Peripapillary Atrophy.

The external margin of the peripapillary atrophy was defined as the edge of visible sclera in the color photograph, corresponding to the end of the hyperreflective area in the infrared image and the end of the RPE in the OCT B-scans. The examiner used simultaneous visualization of these three modalities (color, infrared, and OCT images) to increase the precision in identifying the PPA margins. The total PPA area was measured as the area inside the PPA margin and outside the disc margin. This total PPA area was divided into gamma-PPA and beta-PPA, corresponding to the subset areas inside and outside BMO, respectively (Fig. 1, Supplementary Video S1). 

Figure 1

 

Example of measurement of beta- and gamma-PPA areas. (A) Color photograph of the optic disc. (B) Same image as (A), with the optic disc margin (light blue dots), Bruch's membrane opening (red dots), and peripapillary (PPA) margin (dark blue dots); shaded areas represents beta-PPA (blue) and gamma-PPA (red); dashed green line indicates the location of OCT scan. (C) Optical coherence tomography image (along the dashed green line in [B]) showing the corresponding location of disc margin (light blue vertical line), BMO (red dot) and PPA margin (dark blue vertical line); the blue horizontal arrows indicate beta-PPA and the red horizontal arrow indicates gamma-PPA.

Border Tissue Obliqueness.

The obliqueness of the border tissue of Elschnig was measured in each radial B-scan using the following methodology: (1) a line starting at each BMO point was manually positioned to segment the border tissue (Figs. 2B, 2E); (2) the internal angles between the two border tissue surfaces and a line connecting the two BMO points were measured, resulting in two border tissue angles per radial scan, corresponding to opposite positions in the optic nerve head (ONH) (Figs. 2C, 2F); (3) for each radial scan, the largest of the two angles was subtracted by the smallest angle, and the result was used as a border tissue obliqueness index for the corresponding radial scan (Figs. 2C, 2F); (4) the above steps were repeated for the 24 radial scans of each eye, and the average border tissue obliqueness index was used for analysis. 

Figure 2

 

Measurement of border tissue obliqueness in two different eyes of different subjects (A–C and D–F, respectively). (A), (D) Optical coherence tomography radial scan of the ONH. (B), (E) Two yellow lines are positioned to delineate the border tissue of Elschnig, starting at BMO. (C), (F) The red line connects both BMO points and is used as reference to measure border tissue obliqueness angles. In an eye with an oblique optic nerve insertion (A–C columns), one border tissue angle (163°) is larger than the other (25°), resulting in a high border tissue obliqueness index (163° − 25° = 138°). In an eye with nonoblique optic insertion (D–F), both angles tend to be similar, resulting in a low border tissue obliqueness index (150° − 130° = 20°).

Choroidal Thickness.

The peripapillary choroidal thickness was measured in the enhanced depth imaging circular scans, as the distance between Bruch's membrane and the manually segmented posterior limit of vascular choroid (Fig. 3).²⁸ The software measures the thickness at each of the 768 A-scans which compose the circular scan and their average was used for analysis. 

Figure 3

 

Example of peripapillary choroidal thickness measurement, in OCT circular scan. Red line indicates the Bruch's membrane and green line indicates the posterior limit of vascular choroid.

Data Analysis

While performing the segmentations and measurements, the examiner was masked to subjects' diagnosis. To evaluate intraobserver reproducibility, segmentations were performed twice in 15 randomly selected subjects by the same examiner. Another examiner independently repeated the segmentations in the same 15 subjects to evaluate interobserver reproducibility. Reproducibility was assessed by average within-subject SDs obtained from analyses of variance.²⁹ 

Group comparisons were made with the Mann-Whitney U test for continuous variables and Fisher's exact test for categorical variables. The relationships between beta-PPA and age, gamma-PPA, disc area, disc ovality index, border tissue obliqueness index, peripapillary choroidal thickness, spherical refractive error, axial length, and presence of glaucoma were evaluated with univariate and multivariate linear regression. For the multivariate model, we included all variables and performed backward stepwise selection based on statistical significance (P < 0.05). The same regression analyses were performed with gamma-PPA as response variable and beta-PPA as one of the explanatory variables. The diagnostic accuracy of beta- and gamma-PPA areas was determined by analysis of the area under the receiver operating characteristic curves (AUC) and sensitivities at 90% specificity. Confidence intervals for AUC and sensitivities were estimated with 2000 bootstrap replicates. We also analyzed the relationship between beta- and gamma-PPA areas and visual field mean deviation (MD), used as a quantification of glaucomatous damage. Statistical analyses were performed with open-source software R (Version 3.2.2)³⁰ using package pROC (Version 1.8).³¹ 

Results

We recruited 131 subjects for the study. All clinicians independently agreed that 42 subjects were glaucomatous and 72 were nonglaucomatous based on visual field assessment (i.e., complete agreement in 114 [87%] subjects). Of the remaining 17 (13%) subjects, consensus classification after the optic disc evaluation was reached in 16 subjects, whereas the remaining 1 subject in whom consensus was not obtained was excluded from the study. An additional subject, in whom consensus was obtained, was excluded due to insufficient OCT image quality that precluded segmentation. Hence, we analyzed data from 55 myopic glaucoma patients and 74 myopic controls whose demographic and ocular data are shown in Table 1. 

Table 1

 

Group Characteristics

 

The inter- and intraobserver reproducibility of the measured beta-PPA area, gamma-PPA area, disc area, disc ovality index, border tissue obliqueness index, and peripapillary choroid thickness are shown in Table 2. 

Table 2

 

Reproducibility of Measured Parameters

 

Myopic glaucoma patients were older, had thinner BMO-MRW and RNFLT and worse MD than myopic controls (Table 1). Beta-PPA area was nonsignificantly larger in the myopic glaucoma patients compared with the myopic controls, (medians, 1.0 and 0.74 mm², respectively, Table 1, P = 0.05), whereas the gamma-PPA area was nonsignificantly smaller in the glaucoma patients (medians, 0.28 and 0.42, respectively, Table 1, P = 0.09). The distribution of beta- and gamma-PPA in both groups was right-skewed and overlapped widely (Fig. 4). 

Figure 4

 

Distribution of beta- and gamma-PPA in the myopic glaucoma (upper histogram) and myopic controls (lower histogram) groups. Proportions are relative to the number of subjects in each group. Dashed lines indicate median values.

In univariate analysis, larger beta-PPA area was associated with larger disc area, less oblique border tissue configuration, thinner choroidal thickness, and longer axial length (Table 3; Fig. 5). These covariates remained significantly associated with larger beta-PPA in the multivariate analysis, in addition to older age and greater disc ovality (Table 4). Presence of glaucoma was not significantly associated with beta-PPA area in the univariate or multivariate analysis. 

Table 3

 

Beta-PPA Area Univariate Linear Regression Models

 

Figure 5

 

Scatterplots showing the relationships of beta- and gamma-PPA areas with the three most statistically significant (highest R²) covariates for each PPA subset. Data are shown for the myopic glaucoma and myopic control groups. Dashed lines represent univariate linear models.

Table 4

 

Beta-PPA Area Multivariate Regression Model

 

In univariate analysis, larger gamma-PPA area was associated with more oblique border tissue configuration, greater disc ovality, increased myopia, and longer axial length (Table 5; Fig. 5). The same covariates were significant in multivariate analysis (Table 6), with the exception of degree of myopia. Like beta-PPA, the presence of glaucoma was not significantly associated with gamma-PPA area. 

Table 5

 

Gamma-PPA Area Univariate Linear Regression Models

 

Table 6

 

Gamma-PPA Area Multivariate Regression Model

 

Schematics of the optic disc, and beta- and gamma-PPA areas of all study subjects are shown in Figure 6, whereas Figure 7 shows the corresponding average configurations of these parameters in each group. 

Figure 6

 

Schematics of optic disc and PPA of all studied subjects. Using OCT radial scans, the polar coordinates of the disc margin, BMO, and PPA margin were exported for plotting (all eyes were converted to right-eye format). S, superior; T, temporal; N, nasal; I, inferior.

Figure 7

 

Schematics of average configuration of optic disc and PPA in the myopic glaucoma and myopic control groups. For each angular coordinate (48 positions per eye, 2 from each radial scan), the average radial coordinates of disc margin, BMO, and PPA margin were computed in the myopic glaucoma and myopic control groups. The resulting values are shown in right-eye format to illustrate the average configuration of these structures.

The AUC of beta- and gamma-PPA areas were 0.60 (95% confidence interval [CI] 0.50–0.70) and 0.59 (95% CI 0.49–0.69), respectively (Fig. 8). The respective sensitivities at 90% specificity were 15% (95% CI 4%–27%) and 9% (95% CI 3%–21%). Neither beta nor gamma-PPA areas were associated with MD in myopic glaucoma patients (P > 0.29; Fig. 9). 

Figure 8

 

Receiver operating characteristic curves of beta- and gamma-PPA areas.

Figure 9

 

Scatterplots showing the relationships of beta- and gamma-PPA areas with visual field MD in myopic glaucoma patients. Dashed lines and numbers inside the plots represent univariate linear models.

Discussion

The results of the current study indicate that beta- and gamma-PPA as determined by OCT do not help discriminate between myopic glaucoma patients and myopic healthy controls. Myopic glaucoma patients had larger beta-PPA areas than myopic controls, approaching statistical significance; however, the distributions of both beta- and gamma-PPA in the two groups overlapped too widely to have any practical utility for discrimination, as indicated by the low AUC and sensitivity values. Additionally, although the configuration of optic disc, and beta- and gamma-PPA varied considerably among subjects, the average configuration in both groups was nearly identical. 

Our results contrast with a report by Dai et al.,¹⁴ who found that myopic glaucoma patients had significantly larger beta-PPA areas and smaller gamma-PPA areas than myopic controls. The mean PPA areas in their study differed considerably with those in the current study; they reported smaller beta-PPA areas (1.14 mm² in glaucoma and 0.56 mm² in controls, compared with 1.34 mm² in glaucoma and 1.12 mm² in controls in the current study) and larger gamma-PPA areas (1.25 mm² in glaucoma and 1.0 mm² in controls, compared with 0.41 mm² in glaucoma and 0.55 mm² in controls in the current study). The differences between their study and ours could be due to differences in study populations; our study included mostly (93%) Caucasian subjects, whereas Dai et al.¹⁴ presumably included mostly Chinese subjects, raising the possibility that racial differences could influence the extent of beta- and gamma-PPA. There is an absence of research exploring the relationship of PPA and race to elucidate this matter. Additionally, because subjects in their study were younger (mean age 42 years compared with 60 years in the current study), a smaller beta-PPA would be expected because of its association with age.^13,17 Another possible explanation for the differences between the studies is the variability in criteria used to define the diagnostic category that could influence the characteristics of the study groups. Of note, the average axial length and refractive error were similar in both studies. 

It is plausible that the difference in beta-PPA areas between myopic glaucoma and myopic controls could have been statistically significant with a larger sample size. However, to detect a difference of 0.5 mm² between the mean beta-PPA area of two groups, with SD of 0.6 mm² (the mean and SD values reported by Dai et al.¹⁴), a sample of 32 subjects in each group would be enough to achieve 90% power. More importantly, a statistically significant difference in averages of a parameter between two groups does not ensure clinical utility. The considerable overlap observed in the distribution of beta-PPA area in myopic glaucoma patients and myopic controls results in low discrimination ability (15% sensitivity at 90% specificity). This overlap is unlikely to change with larger sample sizes and suggests the lack of diagnostic utility of subclassification of PPA areas. As a comparison, in a previous study on the same myopic subjects, the sensitivity at 90% specificity was 71.4% for both RNFLT and BMO-MRW.²³ 

Beta- and gamma-PPA areas were mostly associated with different covariates, and not related to each other, reinforcing the hypothesis that they could be the result of different physiological mechanisms.^11–14 Gamma-PPA was significantly associated with oval discs and oblique configuration of the border tissue of Elshnig, similar to what was previously reported.^13,14 Kim et al.³² documented development of PPA and progressive tilting of the optic disc in children with myopic shift, demonstrating that PPA might be caused by globe elongation. Even though they did not have OCT images to differentiate between beta- and gamma-PPA in every case in that study, it is reasonable to assume that increasing axial length would cause traction in the ONH tissues, resulting in sliding of the BMO³³ and stretching of the border tissue. The stretched border tissue without Bruch's membrane would constitute gamma-PPA, as shown in the example from Kim et al.³² in whom OCT images were available. 

Beta-PPA was significantly associated with reduced peripapillary choroidal thickness. Similarly, Sullivan-Mee et al.¹⁷ reported a significant association between juxta-papillary choroidal volume and beta-PPA. The loss of RPE in the beta-PPA could be caused by insufficient blood supply due to a thinned choroid.^3,17 However, an alternative hypothesis could be that the loss of RPE and photoreceptors would cause choroidal thinning, by reduction of RPE produced trophic factors.^34,35 Additionally, beta-PPA was related to axial length in ours and previous studies.^14,16,17 Eyes with longer axial length might have thinner³⁶ or more atrophic RPE, more susceptible to degeneration. Also in accordance with previous studies,^13,17 beta-PPA was related to older age in our multivariate analysis, whereas gamma-PPA was not. Beta-PPA was associated with structural characteristics of the ONH, including larger disc area, less oblique border tissue configuration, and greater disc ovality. It is not clear through which mechanisms these associations may occur and further studies are required to explore them. 

There is a considerable number of previous studies evaluating the association of PPA with glaucoma. However, because most of them did not use BMO-based definitions of beta- and gamma-PPA, it is difficult to compare them with our study. Two recent studies in glaucoma patients associated larger beta-PPA areas with faster rates of visual field loss¹⁶ and RNFL loss,¹² whereas larger gamma-PPA areas were associated with slower rates.^12,16 Both of these studies used a different methodology to measure beta- and gamma-PPA, only quantifying the temporal PPA, while we measured the PPA around the whole ONH. Additionally, these studies were not restricted to myopic subjects. Although evaluating specifically myopic subjects with PPA was the goal of our study, it is possible that because these subjects already have increased beta-PPA (e.g., due to longer axial length or thinner choroid), they would have a weaker association of beta-PPA and glaucoma than the general population. 

Some limitations should be considered when interpreting the results of this study: (1) Measurements were performed in fundus photographs, which are flattened projections of the curved surface of the posterior pole and therefore could lead to errors. (2) Our sample was not ideally suited to elucidate factors related to the PPA subtypes; for example, most of our patients were between 50 and 70 years old and this narrow range could weaken the observed relationship between beta-PPA and age. (3) The diagnosis of glaucoma in myopic subjects is likely to be less accurate than in nonmyopic subjects. To attempt to mitigate this limitation, we used consensus classification agreement among three experts. 

In summary, our findings suggest that subclassifying PPA with OCT reveals two zones (beta- and gamma-PPA) that are related to different factors, but do not assist in the diagnosis of glaucoma in myopic subjects, at least in Caucasian individuals. Further research may reveal other clinical utilities of this PPA subclassification, such as estimating risk of progression,^12,16 or assisting in the diagnosis of glaucoma in other populations. 

Acknowledgments

The authors thank David Dobbelsteyn, OD, and the staff at the Insight Optometry Group, Halifax, Nova Scotia, Canada, for their assistance with recruiting healthy volunteers for this study. 

Supported by (1) a grant from Canadian National Institute for the Blind–Canadian Glaucoma Clinical Research Council; (2) equipment and unrestricted research support from Heidelberg Engineering, Heidelberg, Germany; and (3) a Mathers Fellowship award, Halifax, Nova Scotia. 

Disclosure: J.R. Vianna, None; R. Malik, None; V.M. Danthurebandara, None; G.P. Sharpe, None; A.C. Belliveau, None; L.M. Shuba, None; B.C. Chauhan, Allergan (C), Heidelberg Engineering (C, F), Topcon (F); M.T. Nicolela, Alcon (C), Allergan (C)