Greater Severity of Glaucomatous Damage in Eyes With Than Without Choroidal Microvasculature Dropout in Open-Angle Glaucoma

Abstract

Purpose: To assess whether open-angle glaucoma (OAG) eyes with choroidal microvasculature dropout (CMvD) have greater severity of glaucomatous damage compared to those eyes without CMvD.

Methods: In this retrospective case–control study, 80 eyes of 80 OAG patients with visual field (VF) defects confined to a superior hemifield (40 eyes with CMvD and 40 eyes without CMvD matched for age [≤10 years], axial length [≤1 mm], and VF loss [≤1 dB (decibel)]) and 43 healthy eyes were consecutively included. The circumpapillary retinal nerve fiber layer thickness (cpRNFLT), macular ganglion cell–inner plexiform layer thickness (mGCIPLT), circumpapillary vessel density (cpVD), parafoveal VD (pfVD), and VF mean sensitivity (VFMS) were measured. The relationships between CMvD angular extent and structural/VD/functional measures were assessed at both hemiretinae in OAG eyes with CMvD. Logistic regression analyses were performed to evaluate the associations between significant cpRNFLT reduction at perimetrically intact hemiretinae and relevant clinical variables.

Results: Sectoral cpRNFLT and mGCIPLT in the perimetrically intact hemiretinae of eyes with CMvD were significantly lower than those of eyes without CMvD (P < 0.05). There were significant correlations between CMvD angular extent and sectoral measures of structural/VD/functional parameters at perimetrically intact hemiretinae as well as perimetrically affected hemiretinae in OAG eyes with CMvD. The presence and extent of CMvD had a significant influence on cpRNFLT reduction at perimetrically intact hemiretinae (P < 0.05).

Conclusions: OAG eyes with CMvD showed significantly lower cpRNFLT and mGCIPLT than those without CMvD at the hemiretinae corresponding to intact hemifields, thus requiring more vigilant attention for greater disease severity.

Choroidal microvasculature dropout (CMvD), which may indicate choroidal vascular impairment within β-zone peripapillary atrophy (β-PPA) adjacent to the optic nerve head (ONH), has been observed in glaucomatous eyes using optical coherence tomography (OCT) angiography.^1–3 CMvD has been associated with more advanced glaucomatous visual field (VF) damage at initial presentation along with topographic correlation with structural damage, such as focal lamina cribrosa or retinal nerve fiber layer (RNFL) defects.^1,3–6 Moreover, CMvD detected during follow-up has been associated with faster rate of RNFL thinning, especially in open-angle glaucoma (OAG) eyes with disc hemorrhage (DH).⁵ Nonetheless, the clinical ramifications of CMvD still remain elusive, particularly with respect to its pathogenic role. 

Numerous studies have suggested that both mechanical and vascular factors may be involved in glaucomatous pathogenesis.^7–9 Choroidal layer microvasculature within the peripapillary area is likely closely related to blood flow in ONH as this area mainly receives its blood supply from the short posterior ciliary (SPC) artery, which also provides a vascular network to deep ONH structures, such as prelaminar and laminar tissues.^10–12 Since CMvD may represent an area of reduced or absent perfusion within the parapapillary choroidal area next to ONH,^2,5 it could be an important indicator for disease severity when present, considering that glaucoma pathogenesis is closely linked to vascular insufficiency. Thus, one of the crucial and unresolved questions related to CMvD is whether or not glaucomatous eyes with CMvD have worse disease characteristics than those without CMvD at a given stage of disease. 

In the present study, we hypothesized that the presence of CMvD may imply a negative impact on glaucomatous ONH, thereby rendering eyes with CMvD to have greater severity of glaucomatous damage compared to those without CMvD at similar stages of disease. To our knowledge, there is currently no information regarding the state of glaucomatous damage linked to the presence of CMvD during the early disease stage. Thus, the objectives of our study were as follows: (1) to determine whether eyes with CMvD are associated with a greater degree of glaucomatous damage using OCT/OCT angiography (OCT-A) measures compared to those without CMvD, which were matched for age, axial length (AL), and severity of VF damage in the hemiretinae corresponding to perimetrically intact hemifields, and (2) to explore the relationship between the size of CMvD and the extent of glaucomatous damage in both perimetrically affected and intact hemiretinae. 

Methods

Participants

This was a retrospective case–control study in which the participants consisted of consecutive OAG patients who visited the glaucoma clinic of Asan Medical Center, Seoul, Korea, from March 2017 to February 2018. This study was approved by the Institutional Review Board at the Asan Medical Center, University of Ulsan, College of Medicine, Seoul, Korea. All procedures conformed to the Declaration of Helsinki. 

All study participants underwent a comprehensive ophthalmologic examination, including assessment of past medical history, best-corrected visual acuity (BCVA), slit-lamp biomicroscopy, intraocular pressure (IOP) with Goldmann applanation tonometry (GAT), gonioscopy, axial length (AL) measurements with IOL master (Carl Zeiss Meditec, Dublin, CA, USA), central corneal thickness (CCT), dilated color fundus photography (Canon, Tokyo, Japan), stereoscopic photography of the ONH, and red-free RNFL photography (Canon). Participants also completed OCT-A (Angiovue; Optovue, Inc., Fremont, CA, USA), Cirrus HD spectral-domain optical coherence tomography (SD-OCT, Carl Zeiss Meditec), and Humphrey field analyzer Swedish Interactive Threshold Algorithm (SITA) 24-2 VF testing (Carl Zeiss Meditec). Systolic and diastolic blood pressures (BPs) were measured during outpatient clinic hours. Mean ocular perfusion pressure (OPP) was estimated as the difference between two-thirds of the mean arterial pressure (MAP) and IOP. MAP was derived as one-third systolic BP + two-thirds diastolic BP. 

Participants in our study included those with OAG as well as normal healthy subjects. All participants were older than 18 years of age and had a BCVA of 20/30 or better, with refractive error between +3 and −8 diopters (D) sphere and ±3 D cylinder. OAG patients were required to meet the following criteria: (1) typical glaucomatous ONH appearance regardless of IOP level, that is, focal or generalized narrowing or disappearance of the neuroretinal rim, DH, cup-to-disc asymmetry > 0.2, and not explained by optic disc size or an RNFL defect; (2) open anterior chamber angles on gonioscopy in both eyes; and (3) glaucomatous VF defects confined to the superior hemifield, corresponding to ONH appearance with a mean deviation (MD) greater than −10 dB (decibel) for the purpose of evaluating eyes with early-to-moderate glaucomatous damage.¹³ As CMvD develops preferentially at the inferotemporal sector, ^1–3we recruited the OAG patients with VF defects confined to a superior hemifield. To be defined as glaucomatous VF defects confined to superior hemifield, all of the following criteria had to be met: (1) three or more adjacent points with P < 0.05 on a pattern deviation (PD) probability map, or two or more test points with P < 0.02 or smaller on a PD probability map in a superior hemifield; (2) no clusters of three points with P < 0.05 and no clusters of two points with P < 0.02 on both the total deviation and PD probability maps in an inferior hemifield; (3) a glaucoma hemifield test (GHT) result that was outside normal limits¹³ and confirmed on two consecutive reliable SITA VF tests. In eyes with glaucomatous superior hemifield VF defects, the first perimetric result was excluded from the analysis to obviate learning effects. In OAG eyes with superior hemifield VF defects, therefore, inferior hemiretinae were regarded as perimetrically affected hemiretinae, while superior hemiretinae were regarded as perimetrically intact hemiretinae. The normal healthy group consisted of subjects from the general eye clinic who had no family history of glaucoma, an IOP of less than 21 mm Hg, normal anterior and posterior segments upon ophthalmologic examination, normal VF test results (defined as a pattern standard deviation within 95% confidence interval [CI] and a GHT result within normal limits), and nonglaucomatous optic discs assessed and confirmed by glaucoma specialists (Y.H.J., M.S.K.). 

Both OAG and healthy subjects were excluded if they had one or more of the following: excessively high myopia (spherical equivalent [SE] < −8 D) with severe myopic disc and fundus changes that impaired adequate ONH/VF evaluation for glaucoma; a history of intraocular surgery with the exception of uncomplicated cataract surgery; or a history of other ophthalmic diseases that could affect VF defects or ONH, including macular disease, diabetic retinopathy, or retinal vascular occlusive diseases. Unreliable VF results (fixation loss > 20%, false-positive error > 15%, and false-negative error > 15%) or poor-quality OCT/OCT-A images (with poor clarity or localized weak signal caused by artifact, motion artifacts, or segmentation error) were also excluded. If both eyes were eligible, one eye was randomly selected. 

β-Zone Peripapillary Atrophy and Choroidal Microvascular Dropout Assessment

All patients' fundus photographs were reviewed by two glaucoma specialists (Y.H.J. and J.K.), masked to the patients' clinical information, in order to assess the presence and size of the β-zone PPA. The β-PPA was defined as an inner crescent of chorioretinal atrophy with visible sclera and choroidal vessels adjacent to the optic disc.^14,15 The β-PPA and clinical disc margin were manually demarcated based on the fundus photography and their areas were measured using ImageJ software (version 1.51; Wayne Rasband, National Institutes of Health, Bethesda, MD, USA) (Fig. 1A). The β-PPA area-to-optic disc area ratio was calculated in order to minimize the effect of photographic magnification error and to accurately represent the dimensions of the β-PPA.¹⁶ 

Figure 1

 

Measurement of the β-PPA area and angular extent of choroidal microvasculature dropout. (A) The β-PPA (yellow dotted outline) and clinical disc margin (blue outline) were manually demarcated on color fundus photography. (B) Area of CMvD demarcated by the red dashed line within the β-PPA. (C) Angular extent (α) of CMvD measured by two lines (red solid lines) connecting the disc center and circumferential margins of CMvD. The black line connecting the disc center to the fovea provides the horizontal reference used to determine the hemiretinal location of CMvD.

The CMvD was defined as focal complete loss of the choriocapillaris and choroidal microvasculature, as defined in previous studies.^1–4 On en face choroidal layer images CMvD was identified when the minimum angular width of vasculature dropout was 200 μm or greater at any location, based on the width of the central retinal vein in glaucomatous eyes¹ (Fig. 1B). 

Currently there is no universally accepted standard method to measure CMvD size quantitatively on a two-dimensional plane. In the current study, the angular extent of CMvD was used as a surrogate for CMvD size and was measured by drawing two lines connecting the disc center to the circumferential margins of CMvD from an OCT-A image overlaid onto the color fundus photograph using ImageJ (Fig. 1C).^3,17 The optic disc center was determined to be the point where the long and short axes of the disc crossed, as described in previous studies.^18–22 The two aforementioned glaucoma specialists independently assessed the presence and angular circumference of CMvD; any disagreements between these two specialists were resolved by a third adjudicator (M.S.K.). The average values from the two measurements were used in the analysis to minimize interobserver variation. Our method of measuring the angular circumference of CMvD has been validated in previous studies.^3,17 

In this case–control study, OAG eyes exhibiting a single CMvD confined to the inferior hemiretinae were consecutively enrolled from OAG eyes that met inclusion criteria. OAG eyes without CMvD [control eyes, CMvD(−) group] from the same database were matched to OAG eyes with CMvD [case eyes, CMvD(+) group] taking into account age (≤10 years), AL (≤1 mm), and the severity of VF loss (≤1 dB). In determining hemiretinal location, a reference line was drawn from the optic disc center to the center of the macula on a digital color fundus photograph (Fig. 1C). The center of the macula was defined as a punctate central reflex of the bandlike reflex on a color fundus photograph. This reference line was used to divide the hemiretinae into two areas (superior versus inferior). 

Cirrus Spectral-Domain Optical Coherence Tomography Imaging

The circumpapillary RNFL thickness (cpRNFLT) and macular ganglion cell–inner plexiform layer thickness (mGCIPLT) were measured using the Cirrus SD-OCT system. The ONH map protocol calculates cpRNFLT along a circle of 3.45 mm in diameter centered on the ONH. With Cirrus SD-OCT, cpRNFLT was measured globally and on each sector of four quadrant maps. Two quadrant measurements of cpRNFLT (superior and inferior) were used in our analysis to represent the corresponding sectoral values of cpRNFLT at the superior (perimetrically intact) and inferior (perimetrically affected) hemiretinae (Fig. 2A). 

Figure 2

 

Schematic illustration of the (A) retinal nerve fiber layer thickness map, (B) macular ganglion cell–inner plexiform layer thickness map, (C) circumpapillary vessel density map, (D) parafoveal vessel density map, (E) regional visual field mean sensitivity map according to Garway-Heath et al.,²⁵ and (F) regional parafoveal (within central 10°, 12 points) and peripheral (10°–24°, 40 points) VFMS map. These show perimetrically affected hemiretinae or hemifields in red and perimetrically intact hemiretinae or hemifields in gray.

The macular cube scan provided an mGCIPLT map of the annulus region centered on the fovea. The annulus region had outer and inner diameters of 4.8 and 1.2 mm horizontally, respectively, and outer and inner diameters of 4.0 and 1.0 mm vertically, respectively. Six regional measurements of mGCIPLT (ST, S, SN, IT, I, IN) were used in our analysis to represent the corresponding sectoral values of mGCIPLT at the superior (perimetrically intact) and inferior (perimetrically affected) hemiretinae (Fig. 2B). 

Only images with a signal strength ≥7 were included. Images with motion artifacts, poor centering, or segmentation errors were checked and discarded by the operator, with rescanning performed during the same visit. 

Optical Coherence Tomography Angiography

All subjects underwent OCT-A imaging with the commercially available OCT-A system AngioVue (Angio Disc mode, Optovue, Inc.). It uses the split-spectrum amplitude-decorrelation angiography algorithm to capture the dynamic motion of red blood cells, and presents a three-dimensional angiogram of perfused retinal vasculature. The AngioVue provides vascular information at various user-defined retinal layers²³ qualitatively as a color-coded vessel density (VD) map, and quantitatively as VD (%) of the measured area.²⁴ 

In the current study, circumpapillary VD (cpVD) was acquired with a 4.5 × 4.5-mm scanning field centered on the optic disc. It was measured at a 750-μm-wide elliptical annulus extending from the optic disc boundary in RPC layer that extends from the internal limiting membrane (ILM) to the RNFL posterior boundary. The circumpapillary region was divided into six sectors based on Garway-Heath map: superonasal (SN), nasal (N), inferonasal (IN), inferotemporal (IT), temporal (T), and superotemporal (ST).²⁵ Among them, four regional measurements of cpVD (ST, SN, IT, IN) were used in our analysis to represent corresponding sectoral values of cpVD at the superior and inferior hemiretinae (Fig. 2C). Parafoveal VD (pfVD) was calculated from 6 × 6-mm scans centered on the fovea. It was measured in an annular region with inner and outer diameters of 1 and 3 mm, respectively, in a slab from the ILM to the posterior boundary of the inner plexiform layer (IPL).²⁶ The parafoveal region was divided into superior and inferior hemispheres. Two hemispheric measurements of pfVD were used to represent corresponding sectoral values of pfVD at the superior and inferior hemiretinae (Fig. 2D). Participants with poor-quality OCT-A images were excluded, which included those with a signal strength index of less than 48, poor image clarity, containing a motion artifact visualized as an irregular vessel pattern or disc boundary on the en face image, or RNFL segmentation error.¹ 

Visual Field Mean Sensitivity Measurement

In the current study, the six regional VF areas according to the Garway-Heath map were used to represent the regional VF mean sensitivity (VFMS) for 24-2 VF testing (Fig. 2E).²⁵ The VFMS was expressed in unlogged 1/L scales (L, luminance measured in Lamberts). The parafoveal VF defect was defined as a glaucomatous VF defect within 10° of fixation, with at least one point at P < 1% lying at the two innermost points on the PD plot, regardless of extension to a 10° to 24° region, as used in a previous study.¹⁷ The eyes without parafoveal VF defects consisted of those having VF defects only in the 10° to 24° VF area. The VFMS at each of the 52 individual test points was converted to the linear scale of 1/L.^27,28 Various regional VFMS values were calculated by averaging the anti-log absolute sensitivity values according to the Garway-Heath map (Fig. 2E)^25–28: parafoveal (within a central 10°, 12 points) and peripheral (10°–24°, 40 points) map (Fig. 2F).¹⁷ 

Statistical Analyses

All statistical analyses were performed with the SPSS software version 18.0 (SPSS, Inc., Chicago, IL, USA) and P values less than 0.05 (2-tailed) were considered statistically significant. Interexaminer agreements regarding the presence of β-PPA and CMvD, β-PPA area, and the angular extent of CMvD were assessed by κ statistics and intraclass correlation coefficients (ICCs). 

Normality of distribution was assessed using the Kolmogorov-Smirnov test. Differences in continuous variables among the three groups were analyzed using a 1-way analysis of variance with Tukey's post hoc test or the Dunnett T3 post hoc test, as appropriate. For categorical variables, the Fisher exact test or χ² test was used. The correlations between CMvD angular extent and OCT, OCT-A, and VFMS values at various sectors in both perimetrically affected and intact hemiretinae (hemifields) were determined by Pearson correlation coefficients (r). For the correlation between CMvD angular extent and VFMS, the mean linear scale values were log-transformed for dB scale. Both 1/L and dB scales were used in the correlation analyses. 

Univariate and multivariate logistic regression analyses were performed to determine the clinical factors associated with the presence of significant cpRNFLT reduction (P < 5% or P < 1%) determined by the internal normative database provided by Cirrus HD SD-OCT at superior quadrants in the perimetrically intact hemiretinae. After univariate analyses with each putative independent variable, including demographics, presence and extent of CMvD, and structural and VD measurements of the perimetrically affected hemiretinae, multivariate analyses including variables with P < 0.05 were performed in multiple ways to avoid multicollinearity using a backward elimination approach. 

Results

There were a total of 97 eyes from 97 OAG patients that met our initial inclusion criteria. Overall, 8 (8.2%) and 9 eyes (9.3%) of the 97 eyes were excluded from the final analysis due to unreliable VF testing and poor OCT-A image quality, respectively. In the subgroup analysis, 3 (6.3%) out of 48 eyes were excluded due to unreliable VF tests in the CMvD(+) group while 5 (10.2%) out of 49 eyes were excluded for the same reason in the CMvD(−) group. Five (10.4%) out of 48 eyes were excluded from the final analysis in the CMvD(+) group due to poor OCT-A image quality, while 4 (8.2%) out of 49 eyes were excluded from the CMvD(−) group. Therefore, our final analysis included 40 OAG eyes in the CMvD(+) group and 40 OAG eyes in the CMvD(−) group that were matched for age, AL, and VF severity. Forty-three healthy eyes were also consecutively enrolled for comparison in our final analysis. There were excellent interexaminer agreements regarding the detection of β-PPA (κ = 0.825, P < 0.001) and CMvD (κ = 0.927, P < 0.001). The interexaminer ICCs for the measurements of β-PPA area and CMvD angular extent were 0.971 (95% CI = 0.958–0.980, P <0.001) and 0.984 (95% CI = 0.977–0.989, P <0.001), respectively. 

Baseline and comparisons of demographics and clinical characteristics among healthy, CMvD(−), and CMvD(+) groups are shown in Table 1. Compared to those of healthy eyes, both OAG groups had significantly lower global cpRNFLT, mGCIPLT, and cpVD values (P < 0.001). Notably, these global structural/VD parameters of the CMvD(+) group were significantly lower than those of the CMvD(−) group (P < 0.05). Mean pfVD values of OAG eyes were lower than those of healthy eyes, while the mean pfVD values of CMvD(+) eyes were lower than those of CMvD(−), although the differences did not reach statistical significance (P > 0.05). Among the 40 OAG eyes with CMvD, 38 (95.0%) showed parafoveal VF defects, whereas 18 out of 40 eyes (45%) without CMvD exhibited parafoveal VF defects (P < 0.001). 

Table 1

 

Comparison of Baseline Characteristics Among Eyes With or Without CMvD Versus Healthy Eyes

 

Table 2 and Figure 3 show the differences in regional quantitative measurements of OCT, OCT-A, and VFMS among the three groups. The CMvD(+) group showed significantly lower mean cpRNFLT values not only in the inferior quadrant of the affected hemiretinae, but also in the superior quadrant of the intact hemiretinae versus those of CMvD(−) eyes. The mean mGCIPLT values were significantly lower at the ST sector of the intact hemiretinae as well as at the IN, I, and IT sectors of the affected hemiretinae in CMvD(+) compared with CMvD(−) groups (P < 0.05). The mean cpVD values at multiple sectors of inferior hemiretinae (IN and IT) in the CMvD(+) group were significantly lower than those of the CMvD(−) (P < 0.05). The mean cpVD values at superior hemiretina sectors (ST, SN) in the CMvD(+) group were also lower than those of the CMvD(−) group, although the differences did not reach statistical significance (P > 0.05). Within the VFMS analysis, there were statistically significant differences in superior parafoveal VFMS values between the CMvD(+) and CMvD(−) groups (P = 0.01), although there were no differences of VFMS values between the two groups at regional sectors (ST, SN, IT, and IN) corresponding to superior and inferior hemiretinae, according to the map by Garway-Heath et al.²⁵ (P > 0.05). 

Table 2

 

Regional Quantitative Comparison of OCT, OCT-A, and VF Data Among Three Groups

 

Figure 3

 

Comparison of circumpapillary retinal nerve fiber layer thickness (cpRNFLT), macular ganglion cell–inner plexiform layer thickness (mGCIPLT), and circumpapillary vessel density (cpVD) between healthy (blue) and glaucomatous eyes without (orange) and with (green) CMvD. In all sectors, values of eyes with CMvD are lower than those without CMvD or healthy controls. Asterisk indicates statistically significant difference between glaucomatous eyes with and without CMvD. Error bars indicate the standard error of the mean.

The angular extent of CMvD showed strong correlations with sectoral values of cpRNFLT, mGCIPLT, cpVD, pfVD, and VFMS (linear scale) for the perimetrically affected hemiretinae (Table 3) in glaucomatous eyes with CMvD. Moreover, it also showed strong correlations with the sectoral values of cpRNFLT, mGCIPLT, cpVD, pfVD, and VFMS (linear scale) for the perimetrically intact hemiretinae (Table 3). The angular extent of CMvD also showed strong correlations with sectoral values of VFMS on the dB scale at both hemiretinae (Table 3). 

Table 3

 

Correlations of Global and Regional cpRNFLT, mGCIPLT, VD, and VFMS Versus CMvD Angular Extent in OAG Eyes With CMvD

 

Table 4 shows the results of the univariate and multivariate logistic regression analyses for determining clinical factors associated with the presence of significant cpRNFLT reduction (P < 0.05, yellow or red) based on the normative database at superior quadrants in the perimetrically intact hemiretinae in 80 glaucomatous eyes with or without CMvD. In univariate logistic analyses, CMvD presence, larger CMvD angular extent, thinner global and inferior quadrant cpRNFLT, and lower global and cpVD IN were significantly associated with the presence of significant superior quadrant cpRNFLT reduction. To determine which parameters were independently associated with the presence of superior quadrant cpRNFLT reduction (P < 0.05, yellow or red), separate multivariate models were constructed to avoid the effect of multicollinearity. The presence of CMvD was significantly associated with superior quadrant cpRNFLT reduction (P < 0.05) at perimetrically intact hemiretinae in model 1 while the CMvD angular extent was significantly associated with superior quadrant cpRNFLT reduction (P < 0.05) in model 2. Similar results of the univariate and multivariate logistic regression analyses for determining clinical factors associated with the presence of significant cpRNFLT reduction (P < 0.01, red) at superior quadrants in the perimetrically intact hemiretinae were also noted. 

Table 4

 

Univariate and Multivariate Logistic Regression Analysis for the Association Between Relevant Clinical Factors and the Presence of Significant cpRNFLT Reduction (P < 5%) (Upper Section) and (P < 1%) (Lower Section) of the Superior Quadrant in Perimetrically Intact Hemiretinae in 80 OAG Eyes With or Without CMvD

 

Representative cases are shown in Figures 4A and 4B. A 70-year-old woman with OAG and AL of 22.73 mm exhibited CMvD at inferior hemiretina on the choroidal map of OCT-A imaging and had a single-hemifield VF defect (MD −5.25 dB) confined to the superior hemifield. She had localized RNFL defects at both the inferotemporal and superotemporal regions in the right eye, as shown in the red-free fundus photography. She showed a significant cpRNFLT reduction (P < 0.01) at the superior quadrant of perimetrically intact superior hemiretina as well as at the inferior quadrant of perimetrically affected inferior hemiretina, according to SD-OCT imaging analysis. The cpVD reduction was also apparent on the en face image and color-coded map of OCT-A at both superior and inferior hemiretinae (Fig. 4A). In the matched CMvD(−) case for age, AL, and VF severity, a 65-year-old OAG man without CMvD on the choroidal map of OCT-A imaging had an AL of 23.18 mm and single-hemifield VF defect (MD −4.59 dB) confined to superior hemifield. He had a localized inferotemporal RNFL defect in the left eye as determined by red-free fundus photography. However, he showed relatively healthy RNFL distribution in the perimetrically intact superior hemiretina. The RNFLT and color-coded maps of SD-OCT showed significant reduction of RNFLT only at the inferior quadrant of perimetrically affected inferior hemiretina. The cpVD reduction was only apparent on the en face image and color-coded map of OCT-A at the perimetrically affected inferior hemiretina (Fig. 4B). 

Figure 4

 

Representative cases (A) with CMvD and (B) without CMvD. In each case, the choroidal en face image of OCT-A (top row, left), the visual field (VF) test (top row, right), red-free photography showing retinal nerve fiber layer (RNFL) defects (indicated by yellow arrows) (middle row, left), RNFLT thickness map and color-coded map of spectral-domain optical coherence tomography (SD-OCT) (middle row, right), the retinal peripapillary capillary (RPC) layer of en face image and color-coded map of OCT-A (bottom row) showing circumpapillary vessel density (cpVD) (the decreased cpVD region indicated by white arrows) were assessed.

Discussion

In previous reports including a recent study by Yarmohammadi et al.,²⁹ glaucomatous eyes with VF damage confined to a single hemifield showed diminished cpRNFT, mGCIPLT, and cpVD compared to healthy eyes, even in the retinal hemispheres corresponding to the perimetrically intact hemifield.^30–33 However, to our knowledge, there are no reports assessing the impact of CMvD in terms of severity of glaucomatous damage in the perimetrically intact hemiretinae when OAG patients present with localized VF damage confined to a single hemifield. The current study clearly demonstrates that more severe structural damages in the form of significant cpRNFLT and mGCIPLT reduction were found in CMvD(+) compared to CMvD(−) eyes at the perimetrically intact hemiretinae. Nonetheless, the pathophysiology as to how this occurs remains to be elucidated with future longitudinal studies. 

The cpVD reduction may be an epiphenomenon or secondary change of RNFL loss in glaucomatous eyes and it therefore displays a topographic association with glaucomatous structural damage, such as focal lamina cribrosa or RNFL defects.^1,3 In the same location where cpRNFLT reduction was found, our study revealed that the mean cpVD of CMvD(+) eyes was significantly lower at inferior sectors (IT, IN) in the perimetrically affected hemiretinae versus those of CMvD(−) or normal eyes in the matched hemiretinae (P < 0.05). Likewise, the mean cpVD of CMvD(+) eyes was significantly lower at superior sectors (ST, SN) in the perimetrically intact hemiretinae versus those of normal or CMvD(−) eyes (P < 0.05), although the difference in the mean cpVD between the CMvD(+) and CMvD(−) groups did not reach statistical significance. One possible explanation for our finding is that the glaucomatous eyes in our study are in relatively early stages of disease (mean MD = −3.67 dB), such that although there is a greater amount of cpRNFLT reduction in CMvD(+) compared to CMvD(−) eyes in the perimetrically intact superior hemiretinae based on OCT imaging, this structural change has not been sufficiently reflected in the loss of cpVD within the RNFL layer of the superior hemiretinae, resulting in a lack of statistical significance between the two groups. 

Regardless of the presence of glaucomatous VF defects, there were also no significant differences of pfVD measurements in the corresponding hemisectors among the three groups. Previously, Rao et al.³⁴ reported that the pfVD was lower in the POAG group versus the control group, but the diagnostic ability for pfVD was poor with an area under the curve of 0.63 (CI, 0.48–0.75). The discrepancies between the studies regarding pfVD measurements may be explained by differences in study design, patient populations, or the OCT-A devices used. For instance, we included glaucomatous eyes at relatively early disease stages with possible sparing of central VF defects, particularly in the eyes without CMvD, and this might have resulted in minimal changes of the measured pfVD values in the macula. 

Of note, strong correlations between CMvD angular extent and cpRNFLT, mGCIPLT, cpVD, and pfVD at inferior sectors of the perimetrically affected hemiretinae suggest that CMvD size may be well associated with the degree of glaucomatous injury even at an early stage of disease, as noted in our data (MD = −3.67 dB), and angular extent can be a potential biomarker for monitoring glaucoma progression even at this preliminary stage of disease. Moreover, cpRNFLT, mGCIPLT, cpVD, and pfVD measurements showed significant correlations with CMvD angular extent even at the superior sectors of the perimetrically intact hemiretinae in CMvD(+) eyes. In fact, our current findings indicate that such correlations not only exist between CMvD angular extent and structural/VD parameters, such as cpRNFLT, mGCIPLT, and cpVD, but also between CMvD angular extent and VFMS at the superior sectors (ST and SN) of the perimetrically intact hemiretinae in CMvD(+) eyes. This implies that the CMvD(+) eyes may have worse overall VFMS in the seemingly normal-appearing hemifields than those without CMvD. Furthermore, the greater the angular extent of CMvD, the worse the VFMS as well as cpRNFL and mGCIPL loss becomes in the hemiretina, corresponding to the seemingly normal-appearing visual hemifield. 

Our multivariate analyses revealed that the presence of significant cpRNFLT loss at the superior quadrant of the hemiretinae corresponding to perimetrically intact hemifields, in which cpRNFLT values are outside the 95% and 99% limits of normal variability, was significantly associated with the presence and angular extent of CMvD at initial presentation. In other words, our results suggest that significant cpRNFLT loss may have already occurred in the hemiretinal segments corresponding to the normal hemifields in eyes with CMvD, and the amounts of structural damage in these sectors were far greater than in those eyes without CMvD. Moreover, the larger the angular extent of CMvD, the greater the chance becomes of having significantly reduced cpRNFLT (P < 0.05 or 0.01) at the perimetrically intact hemiretinae. 

Although the exact mechanism underlying greater amount of glaucomatous damage observed in eyes with compared to those without CMvD remains unclear, one may speculate that CMvD could be an indicator for generalized vascular insufficiency, leading to greater severity of glaucomatous damage at perimetrically intact as well as affected hemiretinae, in OAG eyes with single-hemifield VF defects. Parapapillary choroid is closely linked to ONH perfusion since it is proximal to the ONH and receives its blood supply from the SPC artery, which also perfuses the deep ONH structures.^10–12 Suh et al.¹ have reported that the presence of CMvD may be associated with.decreased total choroidal thickness (CT) in the parapapillary region of OAG eyes. Therefore, the presence of CMvD may be closely linked to thinner choroid in glaucoma patients, which can result in decreased blood flow to ONH or decreased OPP.³⁵ Diminished ONH blood flow or OPP may play a significant role in greater severity of glaucomatous damage in OAG eyes with CMvD, given that decreased blood flow to ONH or OPP has been shown to be an independent predictor of glaucoma progression.^36,37 

A recent study has found that when CMvD was found in OAG eyes using an OCT-A device, particularly in association with DH, it was significantly associated with progressive RNFL thinning,⁵ suggesting that CMvD may have a prognostic significance. The major limitation of this study was that OCT-A imaging was performed during follow-up and not at baseline of the study, rendering it impossible to explore whether a cause-and-effect relationship was evident between CMvD and disease progression. Our study is unique in that in comparison with OAG eyes without CMvD, the greater degree of cpRNFL and mGCIPL damage discovered in the seemingly normal-appearing visual hemifield in OAG eyes with CMvD and single-hemifield VF defects could indicate important clinical relevance with respect to disease prognosis. However, the exact role of CMvD in terms of disease prognosis can be assessed only using a longitudinal study design with sufficient follow-up duration. 

There are several limitations to our study. First, the use of a homogeneous Korean population could be a biasing factor. Since our patients represent a subgroup of glaucoma patients with and without CMvD in Korea, data from a single ethnic group may not be generalized to other races or the general population. Selection bias may be another limitation as our patients were enrolled and matched in a retrospective manner at a large university practice instead of within population-based settings; thus they may not possess the same characteristics as similar patients in the general population. Since SD-OCT uses a built-in internal normative database that is not based on a Korean population, determination of the cpRNFLT and mGCIPLT reduction using a built-in normative database might have been affected by the participants' ethnicity. In addition, the current detection of CMvD using OCT-A technology has certain limitations. For example, DH or large retinal vessels may project to en face choroidal layer images, inducing projection artifacts and rendering it difficult to detect or define the CMvD boundary. As a result, detection of CMvD and measurement of its angular extent may be biased to the subjectivity of examiners. To mitigate this subjectivity, we used two independent examiners for CMvD detection and measurement of CMvD angular extent,^16,17 which was met with excellent interexaminer agreement. One should consider that CMvD has volumetric measure in a three-dimensional plane. Nonetheless, the current OCT-A device does not allow three-dimensional analysis of CMvD size using multiple planes. Therefore, the angular extent of CMvD obtained from a two-dimensional en face choroidal image might not fully represent the accurate CMvD size. Circle diameter and scan dimensions for VD measurements are different in each individual with different ALs due to the ocular magnification effect. Unfortunately, current OCT-A devices do not automatically adjust for the magnification effect from different ALs on the measurements of VDs at various retinal locations. Therefore, the use of VD measures derived from OCT-A remains a possible source of error in our analyses. In order to mitigate this error, both groups were matched for AL (≤ 1 mm). Although 24-2 VF testing is routinely used in glaucoma patients at relatively early stages, as in our study patients, a VF 10-2 test may increase retinal sensitivity to detect parafoveal VF defects^38,39 as well as association between CMvD and parafoveal VF defects. Finally, since OCT-A is a relatively new imaging device, it is currently difficult to perform long-term longitudinal studies on CMvD in terms of its prognostic role in glaucoma. Due to the cross-sectional study design, we could not assess the relationship between CMvD and glaucoma progression longitudinally in either affected or intact hemiretinae. 

In conclusion, we have demonstrated that in OAG eyes with single-hemifield VF defects, eyes with CMvD showed significantly greater loss of cpRNFLT and mGCIPLT than those without CMvD in the hemiretinae corresponding to the intact visual hemifields. In addition, the size of CMvD as defined by angular extent was significantly correlated with the structural/VD/functional severity of glaucoma at both hemiretinae with and without VF defects. Finally, profound structural damage in the superior hemiretinae corresponding to the perimetrically intact hemifields was significantly associated with the presence and size of CMvD in OAG eyes with single-hemifield VF defects. 

Acknowledgments

Disclosure: Y.H. Jo, None; J. Kwon, None; K. Shon, None; D. Jeong, None; M.S. Kook, None