Relationship between Optical CoherenceTomography Angiography Vessel Densityand Severity of Visual Field Loss in Glaucoma

Adeleh Yarmohammadi, MD,1 Linda M. Zangwill, PhD,1 Alberto Diniz-Filho, MD, PhD,1 Min Hee Suh, MD,1,2

Siamak Yousefi, PhD,1 Luke J. Saunders, PhD,1 Akram Belghith, PhD,1 Patricia Isabel C. Manalastas, MD,1

Felipe A. Medeiros, MD, PhD,1 Robert N. Weinreb, MD1

Purpose: To evaluate the association between vessel density measurements using optical coherence to-mography angiography (OCT-A) and severity of visual field loss in primary open-angle glaucoma.

Design: Observational, cross-sectional study.Participants: A total of 153 eyes from 31 healthy participants, 48 glaucoma suspects, and 74 glaucoma

patients enrolled in the Diagnostic Innovations in Glaucoma Study.Methods: All eyes underwent imaging using OCT-A (Angiovue; Optovue, Fremont, CA), spectral-domain

OCT (Avanti; Optovue), and standard automated perimetry (SAP). Retinal vasculature information was summa-rized as vessel density, the percentage of area occupied by flowing blood vessels in the selected region. Twomeasurements from the retinal nerve fiber layer (RNFL) were used: circumpapillary vessel density (cpVD) (750-mm-wide elliptical annulus around the optic disc) and whole-image vessel density (wiVD) (entire 4.5�4.5-mm scanfield).

Main Outcome Measures: Associations between the severity of visual field loss, reported as SAP meandeviation (MD), and OCT-A vessel density.

Results: Compared with glaucoma eyes, normal eyes demonstrated a denser microvascular network withinthe RNFL. Vessel density was higher in normal eyes followed by glaucoma suspects, mild glaucoma, andmoderate to severe glaucoma eyes for wiVD (55.5%, 51.3%, 48.3%, and 41.7%, respectively) and for cpVD(62.8%, 61.0%, 57.5%, 49.6%, respectively) (P < 0.001 for both). The association between SAP MD with cpVDand wiVD was stronger (R2 ¼ 0.54 and R2 ¼ 0.51, respectively) than the association between SAP MD with RNFL(R2 ¼ 0.36) and rim area (R2 ¼ 0.19) (P < 0.05 for all). Multivariate regression analysis showed that each 1%decrease in wiVD was associated with 0.66 decibel (dB) loss in MD and each 1% decrease in cpVD was asso-ciated with 0.64 dB loss in MD. In addition, the association between vessel density and severity of visual fielddamage was found to be significant even after controlling for the effect of structural loss.

Conclusions: Decreased vessel density was significantly associated with the severity of visual field damageindependent of the structural loss. Optical coherence tomography angiography is a promising technology inglaucoma management, potentially enhancing the understanding of the role of vasculature in the pathophysiologyof the disease. Ophthalmology 2016;123:2498-2508 ª 2016 by the American Academy of Ophthalmology

Glaucoma is a progressive optic neuropathy with unknowncause characterized by the degeneration of retinal ganglioncells and their axons, resulting in a characteristic appear-ance of the optic disc and visual field loss.1 There isincreasing evidence that optic nerve blood flowimpairment and microcirculatory deficiency may have arole in the pathogenesis of glaucoma,2e4 although the de-tails of this relationship have not been established pre-cisely.5e7 This is in part due to the instrumentation that hasbeen available and its difficulty in accurately measuringocular blood flow.8,9

In contrast to ocular blood flow, objective, accurate, andquantitative measurements of the optic nerve head (ONH)and macula can be obtained with optical coherence to-mography (OCT), which has become the standard for

2498 ª 2016 by the American Academy of OphthalmologyPublished by Elsevier Inc.

structural evaluation in glaucoma research and clinicalpractice. However, structural measurements have onlymoderate correlation with visual field loss.10e12

It recently has become possible to obtain noninvasiveimages to characterize retinal vasculature with OCT angi-ography (OCT-A),13,14 which provides reproducible quan-titative assessment of the microvasculature in the ONH,peripapillary retina, and macula.15e20 Recent studies usingOCT-A have suggested that this new technology might beuseful in the diagnosis, staging, and monitoring ofglaucoma.16,18e20 These measurements also may clarify therole of microcirculation and optic nerve blood flow in thepathogenesis of glaucoma.

This study evaluates the relationship between OCT-Aretinal vessel density parameters and functional

http://dx.doi.org/10.1016/j.ophtha.2016.08.041ISSN 0161-6420/16

Yarmohammadi et al � Vessel Density and VF Loss in Glaucoma

measurements and compares this with standard spectral-domain OCT (SD OCT) structural measurements.

Methods

This was an observational cross-sectional study including 153 eyesfrom 31 healthy participants, 48 glaucoma suspects, and 74 pa-tients with primary open-angle glaucoma enrolled in the DiagnosticInnovations in Glaucoma Study who underwent OCT-A (Angio-vue; Optovue Inc., Fremont, CA)13e20 and SD OCT ONH imaging(Avanti; Optovue Inc.).

The Diagnostic Innovations in Glaucoma Study eligibilitycriteria and methodological details have been reported in previouspublications.21 In brief, all participants completed a comprehensiveophthalmologic examination, including best-corrected visual acu-ity, slit-lamp biomicroscopy, intraocular pressure (IOP) measure-ment with Goldmann applanation tonometry, gonioscopy, dilatedfundus examination, stereoscopic optic disc photography, ultra-sound pachymetry, and standard automated perimetry (SAP) inboth eyes. Only participants older than 18 years of age with openangles on gonioscopy, and spherical refraction within �10 diopterswere included.

Written informed consent was obtained from all participants.The institutional review board at the University of California SanDiego approved all protocols, and the methods described were inagreement with the tenets of the Declaration of Helsinki and theHealth Insurance Portability and Accountability Act.

Healthy subjects were required to have an IOP of 21 mmHg orless with no history of elevated IOP, normal-appearing optic discs,intact neuroretinal rims and retinal nerve fiber layer (RNFL), andnormal visual field test results defined as a pattern standard devi-ation (PSD) within the 95% confidence limits and glaucomaHemifield test result within normal limits. Glaucoma suspects hadan IOP �22 mmHg or suspicious-appearing optic discs withoutevidence of repeatable glaucomatous visual field damage.

Glaucoma was defined by the presence of repeatable abnormalSAP results with a glaucoma Hemifield test outside normal limitsor PSD outside the 95% normal limits. Patients with glaucomawere additionally classified into 2 groups based on the severity oftheir visual field damage; mild glaucoma was defined as visual fieldmean deviation (MD) higher than �6 decibels (dB), and moderateto severe glaucoma was defined as a visual field MD lower than �6dB.22 To ensure comparability of age across study groups, onlysubjects aged �45 years were included.

Eyes with a history of intraocular surgery (except for glaucomasurgery or uncomplicated cataract surgery), secondary causes ofglaucoma, nonglaucomatous optic neuropathies, vascular ornonvascular retinopathies, and other ocular or systemic diseasesknown to impair the visual field were excluded from theinvestigation.

Two blood pressure (BP) measurements obtained in a resting,seated position were taken at least 5 minutes apart using an OmronAutomatic (model BP791IT; Omron Healthcare, Inc., Lake Forest,IL) instrument. Mean arterial pressure (MAP) was calculated asMAP ¼ 1/3 systolic BP þ 2/3 diastolic BP, and mean ocularperfusion pressure (MOPP) was defined using the followingequation: MOPP ¼ 2/3 MAP-IOP.

Standard Automated Perimetry

All participants underwent visual field testing using the 24-2pattern Swedish interactive threshold algorithm on the HumphreyField Analyzer (Carl Zeiss Meditec, Dublin, CA) within 6 monthsof imaging. Only reliable tests (�33% fixation losses and false-negatives, and �15% false-positives) were included. The quality

of visual field tests was also reviewed by the Visual FieldAssessment Center23 staff to identify and exclude visual fields withevidence of inattention or inappropriate fixation, artifacts such aseyelid and lens rim artifacts, fatigue effects, and abnormal resultscaused by diseases other than glaucoma.

Optical Coherence Tomography Angiography

The OCT-A imaging system provides a noninvasive method forvisualizing the ONH and retinal vasculature. The image acquisitiontechnique is optimized for the Split-Spectrum Amplitude-Decor-relation Angiography algorithm described in detail by Jia et al.13

The Split-Spectrum Amplitude-Decorrelation Angiographymethod captures the dynamic motion of moving scatters, such asred blood cells in a flowing blood vessel, and computes a high-resolution 3-dimensional visualization of perfused vasculature.

The OCT-A characterizes vascular information at each retinallayer as an en face angiogram, a vessel density map (Fig 1), andquantitatively as vessel density (percentage), calculated as thepercentage area occupied by flowing blood vessels in theselected region.

For this study, we used vessel density measurements within theperipapillary RNFL in scans with a 4.5�4.5-mm field of viewcentered on the ONH. Vessel density within the RNFL wasmeasured from the internal limiting membrane to RNFL posteriorboundary using standard AngioVue software (version2015.1.0.90). Measurements were calculated in 2 areas. Whole-image vessel density (wiVD) was obtained over the entire 4.5�4.5-mm scan field, and circumpapillary vessel density (cpVD) wasmeasured in a 750 mmewide elliptical annulus extending outwardfrom the optic disc boundary, where the inner elliptical contour isobtained by fitting an ellipse to the disc margin on the OCT en faceretinal angiogram, and the ring width between inner and outerelliptical contour is defined as the circumpapillary region (Fig 1).

The Imaging Data Evaluation and Analysis Reading Centerestablished a standard protocol for OCT-A image quality review.Trained graders reviewed all images to identify poor-quality scans,defined as blurred images, scans with a signal strength index lessthan 48, residual motion artifacts visible as irregular vascularpattern or disc boundary on the enface angiogram, local weaksignal caused by floaters, and RNFL segmentation errors. Gradersalso reviewed the location of the optic disc margin for accuracy,and if needed the margin was adjusted manually and confirmed by2 graders.

Spectral-Domain Optical CoherenceTomography

Avanti SD OCT uses an 840-nm central wavelength, a 22-mm focalspot diameter, and a 70-kHz axial line scan rate that yields an axialresolution of 5 mm in tissue. The ONH map image acquisitionprotocol was used to obtain RNFL thickness measurements in a 10-pixel-wide band along a 3.45-mm-diameter circle centered on theONH and rim area measurements.

All participants had both SD OCT and OCT-A imaging per-formed on the same day. Participants with poor-quality ONH scansdefined by an SSI lower than 37 and scans with segmentationfailure or artifacts were excluded from the analysis.

A total of 351 eyes of 213 subjects had OCT-A and SD OCTimaging within 6 months of visual field testing and were poten-tially eligible for inclusion in the analysis. Fifty-four eyes wereexcluded because of poor-quality OCT-A scans, 16 eyes wereexcluded because of poor-quality SD OCT images, and 35 eyeswere excluded because of unreliable visual field tests. A total of246 eyes of 153 subjects had good-quality OCT-A, SD OCT, and

2499

Figure 1. Vessel density map of the retinal nerve fiber layer showing microvasculature in healthy (most dense), mild, moderate, and severe glaucoma (leastdense) eyes. Top: Vessel density extracted map with elliptical circumpapillary (circumpapillary vessel density) measurement region defined. Second row:Area vessel density color-coded map. Bottom row: Standard automated perimetry visual field results showing corresponding visual field defects.

Ophthalmology Volume 123, Number 12, December 2016

visual field tests. One eye of each of these 153 subjects wasrandomly selected to be included in the analysis.

Statistical Analysis

Descriptive statistics were calculated as the mean and standarddeviation, and categoric variables were compared using the chi-square test. Analysis of variance (ANOVA) was performed tocompare mean values among the healthy, glaucoma suspect, mildglaucoma, and severe to moderate glaucoma eyes.

Relationships between visual field parameters and OCT-Avessel density and SD OCT RNFL and rim area were evaluatedusing simple linear (y¼ax þ b) and second-order polynomial (orquadratic) models (y¼ax2þbxþc). Results were reported as R2

(coefficient of determination) with differences between the R2

values calculated using bootstrapping procedures to estimate the95% confidence intervals of the difference in coefficients ofdetermination. Akaike’s information criterion (AIC) was used tocompare the models for goodness of fit.24 The smaller the AIC, thebetter the model. Univariable linear regression models were builtusing visual field MD as the dependent variable and OCT-A pa-rameters, wiVD and cpVD, and SD OCT RNFL thickness and rim

2500

area measurements and other demographic and ocular character-istic variables as the independent variables. Multivariable modelsalso were used to evaluate the relationship between the visual fieldMD with vessel density and SD OCT RNFL and rim area whileadjusting for potential confounding parameters, such as age, IOP,central corneal thickness (CCT), and axial length.

All statistical analyses were performed with Stata version 14(StataCorp LP, College Station, TX) and JMP version 11.2.0 (SASInstitute Inc., Cary, NC). The alpha level (type I error) was set at0.05 for all comparisons.

Results

The study population consisted of 31 healthy subjects (mean age,69.0�7.7 years; SAP MD, 0.3�1.3 dB), 48 glaucoma suspects(mean age, 71.4�9.4 years; SAP MD �0.6�1.5 dB), 46 patientswith mild glaucoma (mean age, 72.9�10.7 years; SAPMD �3.0�1.8 dB), and 28 patients with moderate to severeglaucoma (mean age, 75.7�10.7 years; SAP MD, �13.6�6.6 dB)(Table 1). Healthy subjects tended to be younger than glaucomasuspects and glaucoma patients, but this difference was notstatistically significant (P ¼ 0.063, ANOVA).

Table 1. Demographics and Ocular Characteristics of Study Population

VariablesHealthy(n [ 31)

Glaucoma Suspect(n [ 48)

Mild Glaucoma(n [ 46)

Moderate andSevere Glaucoma (n [ 28) P Value*

Age (yrs) 69.0 (7.7) 71.4 (9.4) 72.9 (10.7) 75.7 (10.7) 0.063Gender (male/female) 11/20 18/30 20/26 17/11 0.181Ethnicity (AD/ED) 12/19 14/34 14/32 7/21 0.707BP (mmHg)Systolic 130.4 (15.8) 127.5 (19.6) 127.2 (13.1) 127.0 (15.4) 0.816Diastolic 84.3 (16.4) 78.3 (9.8) 78.6 (9.6) 77.8 (9.0) 0.076Mean 99.2 (12.0) 94.7 (11.4) 94.8 (9.0) 94.2 (9.4) 0.195

MOPP (mmHg) 56.3 (7.7) 52.1 (8.2) 54.4 (6.6) 54.6 (7.2) 0.101Heart rate (beats/min) 70.2 (13.1) 67.6 (11.4) 66.2 (12.1) 64.8 (8.2) 0.334Self-reported history of diabetes (%) 16.1 10.4 20.0 14.0 0.657Self-reported history of hypertension (%) 45.16 58.3 61.0 68.0 0.338IOP (mmHg) 14.5 (2.9) 16.5 (5.1) 13.2 (4.1) 12.3 (5.2) <0.001CCT (mm) 542.7 (43.4) 548.4 (38.1) 535.4 (37.0) 513.1 (36.1) 0.002Axial length (mm) 23.7 (1.2) 24.2 (1.0) 23.9 (1.1) 24.8 (1.7) 0.005Disc area (mm2) 2.0 (0.4) 2.1 (0.4) 2.1 (0.5) 2.0 (0.5) 0.596Rim area (mm2) 1.4 (0.3) 1.0 (0.3) 0.9 (0.4) 0.7 (0.4) <0.001Average RNFL thickness (mm) 96.6 (9.6) 86.8 (12.8) 78.6 (11.6) 65.2 (10.4) <0.001SAP MD (dB) 0.3 (1.3) �0.6 (1.5) �3.0 (1.8) �13.6 (6.6) <0.001SAP mean sensitivity (dB) 29.7 (1.6) 28.5 (1.8) 26.1 (2.0) 15.5 (6.6) <0.001SAP PSD (dB) 1.6 (0.4) 2.0 (0.7) 4.3 (2.2) 10.2 (2.4) <0.001OCT-A wiVD (%) 55.5 (3.2) 51.3 (4.6) 48.3 (4.2) 41.7 (5.5) <0.001OCT-A cpVD (%) 62.8 (3.9) 61.0 (4.7) 57.5 (4.4) 49.6 (6.9) <0.001

AD ¼ African descent; BP ¼ blood pressure; CCT ¼ central corneal thickness; cpVD ¼ circumpapillary vessel density; dB ¼ decibels; ED ¼ Europeandescent; IOP ¼ intraocular pressure; MD ¼ mean deviation; MOPP ¼ mean ocular perfusion pressure; OCT-A ¼ optical coherence tomography angi-ography; PSD ¼ pattern standard deviation; RNFL ¼ retinal nerve fiber layer; SAP ¼ standard automated perimetry; wiVD ¼ whole image vessel density.*Statistical significance tested by ANOVA.

Yarmohammadi et al � Vessel Density and VF Loss in Glaucoma

Qualitative Assessment

Healthy eyes generally appeared to have denser capillary networksin the RNFL layer compared with eyes with early glaucomatousoptic nerve damage, and a trend of a sparser microvascularnetwork could be detected with advancing stages of the disease(Fig 1).

Figure 2. Boxplots illustrating the distribution of circumpapillary vessel densitparticipants, glaucoma suspects, and those with mild and moderate to severe glError bars represent the interquartile range. OCT-A ¼ optical coherence tomo

Quantitative Assessment

Vessel density measurements were lower in more severe disease(Fig 2). Specifically, the mean wiVD in moderate to severeglaucoma eyes was significantly lower (41.7%�5.5%) than inmild glaucomatous eyes (48.3%�4.2%), glaucoma suspects(51.3%�4.6%), and healthy eyes (55.5%�3.2%) (P < 0.001,

y (cpVD) (left) and whole image vessel density (wiVD) (right) in healthyaucoma. The medians are represented by horizontal lines in the gray boxes.graphy angiography.

2501

Table 2. Age-adjusted Pearson Correlation Coefficient Matrix on Vessel Density, Visual Field, and Structural Variables

VariablewiVD (%)(n [ 153)

cpVD (%)(n [ 153)

Average RNFLThickness (mm)

(n [ 153)

SD OCT RimArea (mm2)(n [ 153)

SAP MD (dB)(n [ 153)

SAP PatternStandard Deviation(dB) (n [ 153)

cpVD (%) *0.91 (<0.001)Average RNFL thickness (mm) *0.83 (<0.001) *0.73 (<0.001)SD OCT rim area (mm2) *0.65 (<0.001) *0.44 (<0.001) *0.69 (<0.001)Visual field MD (dB) *0.71 (<0.001) *0.73 (<0.001) *0.59 (<0.001) *0.46 (<0.001)Visual field PSD (dB) *�0.63 (<0.001) *�0.61 (<0.001) *�0.60 (<0.001) *�0.48 (<0.001) *�0.79 (<0.001)Visual field mean sensitivity (dB) *0.73 (<0.001) *0.74 (<0.001) *0.62 (<0.001) *0.50 (<0.001) *1.00 (<0.001) *0.80 (<0.001)Visual field SAP meansensitivity (1/lambert)

*0.69 (<0.001) *0.65 (<0.001) *0.64 (<0.001) *0.55 (<0.001) *0.80 (<0.001) *0.79 (<0.001)

cpVD ¼ circumpapillary vessel density; dB ¼ decibels; MD ¼ mean deviation; PSD ¼ pattern standard deviation; RNFL ¼ retinal nerve fiber layer; SAP ¼standard automated perimetry; SD OCT ¼ spectral-domain optical coherence tomography; wiVD ¼ whole image vessel density.*Pearson’s r (P value to test r¼0).

Ophthalmology Volume 123, Number 12, December 2016

ANOVA; Tukey honestly significant difference P < 0.05 for allcomparisons) (Table 1). Mean cpVD values were significantlylower in moderate to severe glaucoma eyes (49.6%�6.9%)followed by mild glaucoma (57.5%�4.4%), glaucoma suspects(61.0%�4.7%), and healthy eyes (62.8%�3.9%) (P < 0.001,

Figure 3. Scatter plots illustrating the linear (grey line) and curvilinear (quadramean deviation and optical coherence tomography angiography (OCT-A) woptical coherence tomography (SD OCT) average retinal nerve fiber layer (RNR2 from the linear regression model.xR2: Adjusted R2 from the quadratic regres

2502

ANOVA; Tukey honestly significant difference P < 0.05 for allpairwise comparisons except between healthy and glaucomasuspect eyes ([P ¼ 0.322]). Standard structural and functionalmeasurements also showed statistically significant differencesamong groups (P < 0.001) (Table 1).

tic fit: dark lines) correlation between standard automated perimetry (SAP)hole-image vessel density, circumpapillary vessel density, spectral-domainFL) thickness, and rim area measurements. dB ¼ decibels. *R2: Adjustedsion model.

Figure 4. Scatterplots illustrating the linear (grey line) and curvilinear (quadratic fit: dark lines) correlation between standard automated perimetry (SAP)pattern standard deviation and optical coherence tomography angiography (OCT-A) whole-image vessel density, circumpapillary vessel density, spectral-domain optical coherence tomography (SD OCT) average RNFL thickness, and rim area measurements. dB ¼ decibels. *R2: Adjusted R2 from the linearregression model. xR2: Adjusted R2 from the quadratic regression model.

Yarmohammadi et al � Vessel Density and VF Loss in Glaucoma

The results of age-adjusted univariate linear regressions aresummarized in Table 2. Linear and curvilinear (quadratic)relationships between OCT-A vessel density and SD OCT struc-tural measurements with visual field MD and visual field PSD areillustrated in Figures 3 and 4, respectively. The strength of theassociations between vascular and structural measurements(expressed in linear scales) with different visual field indices(expressed in both decibels and linear scales) using both linearand quadratic regression models is summarized in Table 3. Weused AIC to compare the linear and curvilinear regressionmodels of OCT-A vessel density, and SD OCT RNFL thicknessand rim area with visual field MD, PSD, mean sensitivity (dB), andmean sensitivity in 1/lambert (Table 3). The quadratic model wasbetter than the linear model in assessing the relationship betweenvisual field measurements and vessel density parameters, as wellas between visual field measurements and structuralmeasurements (Table 3).

Significant differences were found comparing the strength ofthe associations between MD and both OCT-A vascular parameterswith the association between MD and RNFL and rim area mea-surements (P � 0.05 for all pairwise comparisons). The associationbetween MD and RNFL thickness also was significantly strongerthan between MD and rim area (P ¼ 0.001). The associationsbetween wiVD and cpVD with MD were similar (P ¼ 0.500). The

strongest associations with visual field PSD were with wiVD,cpVD (R2 ¼ 0.39 and 0.36, respectively), and RNFL (R2 ¼ 0.37)followed by rim area (R2 ¼ 0.23). Significant differences werefound between the associations of PSD with wiVD and rimarea (P ¼ 0.026) and between RNFL thickness and rim area(P ¼ 0.035). The linear associations between visual field meansensitivity were strongest with cpVD (R2 ¼ 0.55) followed bywiVD (R2 ¼ 0.53), RNFL thickness (R2 ¼ 0.37), and rim area(R2 ¼ 0.19). After converting mean sensitivity from logarithmic(dB) to linear units (1/lambert), a similar pattern was found: as-sociation with mean sensitivity (1/lambert) was highest for wiVDand cpVD (R2 ¼ 0.44 for both) followed by RNFL thickness (R2 ¼0.34) and rim area (R2 ¼ 0.18).

The strength of the associations between visual field MD withstructural and OCT-A measures also was compared using acurvilinear quadratic model. The associations between OCT-A andvisual field MD were significantly stronger than the associationsbetween visual field MD and RNFL and rim area (P < 0.05 for allpairwise comparisons using bootstrapping procedure).

Results from the univariate regression analysis for visual fieldMD as the dependent variable are summarized in Table 4.Multivariate linear regression analysis, while controlling for thepotentially confounding effect of age, IOP, CCT, and axial length,showed that each 1% decrease in cpVD was associated with a

2503

Table 3. Comparison of Linear and Quadratic Regression Models for Evaluating the Association among Visual Field, Structural, andVascular Measurements

Variables

wiVD (%) cpVD (%) RNFL Thickness (mm) Rim Area (mm2)

Linear Quadratic Linear Quadratic Linear Quadratic Linear Quadratic

*R2 yAIC *R2 yAIC *R2 yAIC *R2 yAIC *R2 yAIC *R2 yAIC *R2 yAIC *R2 yAIC

SAP MD (dB) 0.51 867 0.60 835 0.54 859 0.62 830 0.36 908 0.44 887 0.19 944 0.30 923SAP PSD (dB) 0.39 743 0.39 742 0.36 748 0.37 747 0.36 748 0.40 740 0.23 777 0.34 754SAP mean sensitivity (dB) 0.53 868 0.61 841 0.55 860 0.62 834 0.37 912 0.44 896 0.19 951 0.28 935SAP mean sensitivity (1/lambert) 0.44 2191 0.44 2190 0.40 2203 0.40 2202 0.34 2216 0.33 2214 0.16 2253 0.18 2249

AIC ¼ Akaike’s information criterion; cpVD ¼ circumpapillary vessel density; dB ¼ decibel; MD ¼ mean deviation; PSD ¼ pattern standard deviation;RNFL ¼ retinal nerve fiber layer; SAP ¼ standard automated perimetry; wiVD ¼ whole-image vessel density.*Adjusted R2 from the fitted model.yAkaike’s information criterion.

Ophthalmology Volume 123, Number 12, December 2016

0.64-dB loss in MD (P< 0.001) and that each 1% decrease in wiVDwas associated with a 0.66-dB loss in MD (P < 0.001).

Multivariate regression analysis that controlled for the effect ofpotential confounders (age, IOP, CCT, and axial length) andadjusted for the effect of RNFL thickness (Table 5) showed thatwiVD was independently associated with visual field MD.Similar results were found when cpVD was included in themodel instead of wiVD. The association between RNFL and MDwas no longer statistically significant when vessel density wasincluded in the model. The multivariate regression analysis alsowas completed using rim area instead of RNFL thickness, andthe results were similar. Each 1% decrease in wiVD wasassociated with a 0.71-dB loss in MD (P < 0.001), and theassociation between rim area and MD was no longer significant(P ¼ 0.285) when wiVD was included in the model.

For completeness, associations between clinical and ophthalmicfeatures and OCT-A vessel density also were evaluated. The OCT-Avessel density was significantly associated with RNFL and rim areameasurements (P < 0.001) (Table 2). Because structuralmeasurements, such as RNFL, ONH rim, and cup area, have beenshown to be associated with disc size, we also evaluated the effectof disc area on OCT-A vascular measurements. There were no sta-tistically significant correlations between disc area with wiVD andcpVD measurements in healthy eyes (R2 ¼ 0.005, P ¼ 0.696, and

Table 4. Association Between Visual Field Mean Deviation andVascular, Structural, Demographic, and Ocular Variables:

Univariable Analysis

Variables Coefficient R2 P Value

wiVD (per 1% lower) �0.666 0.511 <0.001cpVD (per 1% lower) �0.641 0.535 <0.001RNFL (per 1 mm thinner) �0.228 0.360 <0.001Rim area (per 0.1 mm2 lower) �0.603 0.190 <0.001Age (per 1 yr older) 0.121 0.042 0.011Gender (female) 0.945 0.007 0.322Race (African descent) 0.984 0.006 0.337IOP (per 1 mmHg higher) �0.374 0.092 <0.001CCT (per 10 mm thinner) �0.351 0.059 0.003Axial length (per 1 mm higher) �0.867 0.035 0.022Mean BP (per 10 mmHg higher) �0.160 0.001 0.721Hypertension (yes) �0.217 0.000 0.821Diabetes (yes) �0.320 0.000 0.809

BP ¼ blood pressure; CCT ¼ central corneal thickness; cpVD ¼ circum-papillary vessel density; IOP ¼ intraocular pressure; RNFL ¼ retinal nervefiber layer; wiVD ¼ whole-image vessel density.

2504

R2 ¼ 0.009, P ¼ 0.614, respectively). For this reason, disc area wasnot controlled for in the multivariable analyses. In addition, we didnot find a significant association between MOPP and cpVD (R2 ¼0.003; P ¼ 0.49) or wiVD (R2 ¼ 0.000; P ¼ 0.85).

Discussion

The results of this study demonstrate a significant relation-ship between vessel density and severity of visual fielddamage. Qualitatively, the OCT-A vessel density mapshowed sparser peripapillary vascular networks in moresevere glaucoma. Quantitatively, lower vessel densityvalues were associated with more advanced stages ofglaucomatous visual field damage. The principal finding ofthe study was a relatively strong association among cpVD,wiVD, and visual field loss expressed as MD (R2 ¼ 0.54 and0.51, respectively; P < 0.001 for both), suggesting thatreduced OCT-A vessel density is associated with more se-vere glaucoma. Our results also suggest the vascular-functional correlations were stronger than the standardstructural (RNFL and rim area)efunctional relationshipswhether comparing linear or nonlinear fitted models.Moreover, multivariate analyses indicated an independentrelationship between reduced vessel density and visual fieldloss, even after adjusting for the severity of structuraldamage measured by rim area and RNFL thickness.

Findings of a relatively strong correlation between vesseldensity measurements and visual field loss are in accordance

Table 5. Association Between Visual Field Mean Deviation andVascular, Structural, Demographic, and Ocular Variables:

Multivariable Analysis

Variables Coefficient 95% CI P Value

wiVD (per 1% lower) �0.676 (�0.88 to �0.48) <0.001RNFL (per 1 mm thinner) �0.007 (�0.08 to 0.09) 0.871Age (per 1 yr older) �0.058 (�0.13 to �0.02) 0.126IOP (per 1 mmHg higher) �0.220 (�0.37 to �0.07) 0.004CCT (per 10 mm thinner) �0.191 (�0.02 to 0.02) 0.832Axial length (per 1 mm higher) �0.112 (�0.68 to 0.46) 0.697

CCT ¼ central corneal thickness; CI ¼ confidence interval; IOP ¼intraocular pressure; RNFL ¼ retinal nerve fiber layer; wiVD ¼ wholeimage vessel density.

Yarmohammadi et al � Vessel Density and VF Loss in Glaucoma

with previous reports using OCT-A vessel density.16,18,20

However, it is worth mentioning that these reportsmeasured vessel density in a thick retinal slab from internallimiting membrane to retinal pigment epithelium, whereasour results focused on vessel density in a more superficiallayer, from the internal limiting membrane to the RNFLposterior boundary. Specifically, we found that glaucomaeyes had significantly sparser vessel density within theRNFL compared with both glaucoma suspects and healthyeyes, and that these vessel density measurements werelowest in the glaucoma eyes with the most severe disease.

To measure the vasculature within the RNFL, theboundaries were detected on the basis of conventionalretinal layer segmentation of the SD OCT images. Glau-coma is characterized by remodeling of the ONH tissue.However, even in the presence of tissue remodeling, theRNFL boundaries can be reliably detected on the basis ofthe backscattered intensity image from the SD OCT. Thereis spatial colocalization laterally and in depth between theSD OCT intensity and the OCT-A image.

Findings of decreased OCT-A vascular parameters inpatients with glaucoma also are in general agreement with thelarge body of evidence suggesting hemodynamicimpairments in the ONH, retina, choroid, and retrobulbarcirculations in glaucoma eyes.25e29 In addition, previousOCT-A reports compared the correlations between visualfield loss and vessel density with the correlations betweenvisual field loss and RNFL thickness.16,18 Consistent with ourresults, they reported stronger associations between visualfield damage and vessel density. However, the current studywent further and included 2 structural measures with vesseldensity in the multivariate model to demonstrate that vesseldensity was still strongly associated with visual field damageeven after controlling for standard structural measures in themodel. This finding is in agreement with Hwang et al,7 whoreported an independent relationship between visual fieldMD and another measure of retinal vascular integrity,namely, total retinal blood flow measured by Doppler OCT.

There are several possible explanations for the indepen-dent association of vessel density and visual field MD: First,it may be due, at least in part, to the existence of dysfunc-tional (i.e., preapoptotic) retinal ganglion cells that mayhave reduced blood flow and therefore lower vessel densityand poorer visual field sensitivity. Because these dysfunc-tional retinal ganglion cells have not yet atrophied, areduction in RNFL thickness and rim area may not bedetectable yet. Moreover, histologic studies also showedonly moderate agreement between RNFL thinning andretinal ganglion cell loss,30,31 suggesting that RNFL thin-ning does not completely reflect the functional status ofretinal ganglion cells. Therefore, the stronger correlationbetween vessel density and visual field damage might sug-gest that vessel density is a better reflection of retinal gan-glion cell functioning than structural loss.

The relationship between detectable structural and func-tional damage and change in glaucoma is complex,32 andour results suggest that the relationship between OCT-Avessel density and visual field measures also is complexand influenced by many factors. Previous reports on stan-dard structure-function relationships suggested that the

strength of these relationships depends on the methods andscales of visual field expression, the type of structuralmeasuring device, and the characteristics of the studiedpopulation.32 Therefore, several global visual field indices,including MD, PSD, and mean sensitivity, were assessedand compared using both linear and curvilinearassociations between SAP functional measurements withvessel density, RNFL thickness, and rim areameasurements.

Because the nature of the association between newvascular parameters and functional measures is not wellestablished, we also investigated whether the relationshipbetween vessel density and visual field damage was linear orcurvilinear. Previous reports on OCT-A demonstrated alinear relationship between vessel density and visual fieldmeasurements.16,18 Our findings, based on AIC analysis,suggest that a quadratic model provides a somewhat betterfit to the relationship between vessel density and visualfunction than a linear model. However, AIC values are onlyused for comparison between a set of candidate models anddo not suggest the adequacy of the preferred model.24

Future studies are required to explore the exact nature ofthese relationships.

In our study, SAP MD, PSD, and mean sensitivitymeasured in a logarithmic scale reported in dB and meansensitivity converted to a linear scale reported in 1/lambertwere significantly associated with vessel density measure-ments (P < 0.001 for all). The associations among cpVD,wiVD, and MD (R2 ¼ 0.54 and 0.51, respectively) werehigher than with PSD (R2 ¼ 0.36 and 0.39, respectively; P<0.001). Reports comparing the strength of the associationbetween OCT-based vascular measurements and differentvisual field summary measures are inconsistent. Hwang et al7

showed that total retinal blood flow measured by DopplerOCT was highly correlated with MD, but its relationshipwith PSD did not reach statistical significance. AnotherDoppler OCTebased study33 investigating hemisphericretinal blood flow measurements in eyes with glaucomatousvisual field damage confined to a single hemifield reportedsignificant differences in blood flow measurements betweenthe affected and unaffected retinal hemispheres in patientswith glaucoma compared with healthy age-matched sub-jects, but failed to find an association between hemisphericretinal blood flowmeasurements and visual field mean retinalsensitivity measured as 1/lambert in the correspondinghemifield. In a recent OCT-A study, Liu et al18 reported thatvessel density measured in a circumpapillary ring was morestrongly correlated with visual field PSD compared withMD. Finally, more consistent with our findings, Wanget al20 demonstrated that optic disc OCT-A vessel densitycorrelated with both SAP MD and PSD, but was morestrongly correlated with MD.

The conflicting results in recent OCT-based Doppler andOCT-A studies investigating the relationship betweenvascular measurements and visual field abnormalities couldbe due, in large part, to different aspects of retinal vascu-lature that were measured. Moreover, the investigated studypopulations vary regarding the patients’ risk profiles, theseverity and pattern of glaucoma damage, and the systemicfactors that might have an effect on ocular hemodynamics.

2505

Ophthalmology Volume 123, Number 12, December 2016

Reduced vessel density can be the result of capillarydropout or very low or absent flow. If vessel density is lowerin more advanced glaucoma eyes, it would indicate one orboth scenarios may be occurring. Vessel density measure-ment within the RNFL is not a direct quantification of theblood flow. But if flow in certain vascular structures isreduced to a level below the detection limit of OCT-A indiseased eyes while clearly detectable in normal eyes, it ispossible that lower vessel density may be a surrogate indi-cator of decreased blood flow in the diseased eyes. Com-parison of validated measures of blood flow such as DopplerOCT with OCT-A vessel density in longitudinal studies mayhelp answer this important question.

Although there is general consensus that ocular bloodflow is reduced in glaucoma,3,34 studies investigating thisissue have been hampered by the lack of a reproduciblemethod to measure aspects of ocular circulation. Opticalcoherence tomography is a widely available tool for struc-tural assessment in glaucoma management that is now usedextensively in standard clinical care. Using OCT-A tech-nology for ocular hemodynamic evaluations not only offersthe advantage of providing a quantitative assessment ofocular circulation at a level of precision that has not beenachieved with previous instruments that measured bloodflow9,18 but also its feasibility from a clinical standpointsuggests that OCT-A may be a useful modality that mayreflect hemodynamic considerations relevant to glaucomamanagement. Moreover, the device’s ability to visualize thevascular networks in easily interpretable images and densitymaps (Fig 1) may provide new clinically relevantinformation that can easily be incorporated into the routinemanagement of patients with glaucoma.

The current OCT-A device detects vessels on the basis ofamplitude decorrelation, which results from blood flow;however, it does not directly quantify flow within thesevessels. In fact, decorrelation is linearly related to bloodflow only over a limited range above a certain threshold ofmotion. In other words, vessels with very slow or absentflow below the detection threshold of the instrument will notbe detected, and for the vessels detected, the current in-strument does not differentiate a vessel with faster flow fromone that has slower flow.35 For these reasons, the term“vessel density” is used as a quantitative summarymeasure of the vascular structures detected that reflects theproportion of area occupied by flowing vessels. Anotherknown limitation of OCT-A technology is projection arti-facts, which result from ghost images of anterior vesselsprojecting posteriorly.13,36 However, projection artifacts donot affect the measurements in the current study becausemicrovasculature within the RNFL is the most anteriorretinal vasculature and therefore is not affected by projectionartifact. Although results of the present study and previousreports using OCT-A15,16,18e20 suggest lower vessel densityin glaucoma and glaucoma suspect eyes may be relevant tothe pathophysiology of the disease, the concept of vesseldensity is not well understood. In an earlier study with asimilar technology, OCT microangiography, Zhi et al37

documented disappearance of signal from the peripapillarymicrovasculature in rats. However, there was persistenceof some signal in large vessels in this region despite high

2506

IOP (100 mmHg). Comparison of in vivo assessment ofOCT-A vessel density with histopathologic studies isneeded to clarify what vessel density is measuring.

Study Limitations

This study has several other limitations. We used visual fieldindices to reflect the severity of glaucoma. However, othernonglaucomatous factors, such as refractive error andlenticular and media opacities,38 also may contribute to thevisual field MD, PSD, and mean sensitivity. Moreover, ourstudy population included few patients with advancedglaucoma. Recent reports suggest that standard structuralmeasures such as RNFL thickness reach a floor effect andthat visual field tests are highly variable in eyes withadvanced glaucoma.11,39e42 It is important to evaluatewhether OCT-A has a sufficient dynamic range to provideclinically relevant information across the full spectrum ofglaucoma severity. In addition, we did not evaluate thepossible confounding impact of various systemic conditions,BP and perfusion pressure, glaucoma eye drops, and sys-temic medications on vessel density and its relationship tostandard structural and functional measures. It also shouldbe noted that the cross-sectional design of this study limitsthe determination of the temporal relationship betweenOCT-A vessel density loss and glaucomatous structural andfunctional damage. Longitudinal studies are necessary toevaluate the topographic and temporal relationship betweenchanges in OCT-A vessel density and glaucomatouschanges in standard structural and functional measures inhealthy participants, glaucoma suspects, and those withglaucoma.

In conclusion, OCT-A vessel density measurements aresignificantly associated with the severity of visual fielddamage. These associations are generally stronger thanstandard structural measures such as RNFL and rim area.Moreover, OCT-A vessel density measurements are stillsignificantly associated with the severity of visual field losseven after adjusting for standard structural measurements.For these reasons, OCT-A is a promising technology thatwill allow clinical monitoring of vascular changes in glau-coma, and it could potentially enhance our understanding ofthe pathophysiology of the disease, specifically its under-lying vascular mechanism.

References

1. Weinreb RN, Aung T, Medeiros FA. The pathophysiology andtreatment of glaucoma: a review. JAMA. 2014;311:1901-1911.

2. Flammer J. The vascular concept of glaucoma. Surv Oph-thalmol. 1994;38(Suppl):S3-S6.

3. Weinreb RN, Harris A, eds. Ocular Blood Flow in Glaucoma.Amsterdam, The Netherlands: Kugler Publications; 2009.

4. Kornzweig AL, Eliasoph I, Feldstein M. Selective atrophy ofthe radial peripapillary capillaries in chronic glaucoma. ArchOphthalmol. 1968;80:696-702.

5. Rankin SJ, Drance SM, Buckley AR, Walman BE. Visual fieldcorrelations with color Doppler studies in open angle glau-coma. J Glaucoma. 1996;5:15-21.

Yarmohammadi et al � Vessel Density and VF Loss in Glaucoma

6. Sato EA, Ohtake Y, Shinoda K, et al. Decreased blood flow atneuroretinal rim of optic nerve head corresponds with visualfield deficit in eyes with normal tension glaucoma. GraefesArch Clin Exp Ophthalmol. 2006;244:795-801.

7. Hwang JC, Konduru R, Zhang X, et al. Relationship amongvisual field, blood flow, and neural structure measurements inglaucoma. Invest Ophthalmol Vis Sci. 2012;53:3020-3026.

8. Harris A, Kagemann L, Cioffi GA. Assessment of humanocular hemodynamics. Surv Ophthalmol. 1998;42:509-533.

9. Schuman JS. Measuring blood flow: so what? JAMA Oph-thalmol. 2015;133:1052-1053.

10. Bowd C, Zangwill LM, Medeiros FA, et al. Structure-functionrelationships using confocal scanning laser ophthalmoscopy,optical coherence tomography, and scanning laser polarimetry.Invest Ophthalmol Vis Sci. 2006;47:2889-2895.

11. Leite MT, Zangwill LM, Weinreb RN, et al. Structure-functionrelationships using the Cirrus spectral domain optical coher-ence tomograph and standard automated perimetry.J Glaucoma. 2012;21:49-54.

12. Nilforushan N, Nassiri N, Moghimi S, et al. Structure-functionrelationships between spectral-domain OCT and standardachromatic perimetry. Invest Ophthalmol Vis Sci. 2012;53:2740-2748.

13. Jia Y, Tan O, Tokayer J, et al. Split-spectrum amplitude-decorrelation angiography with optical coherence tomogra-phy. Opt Express. 2012;20:4710-4725.

14. Spaide RF, Fujimoto JG, Waheed NK. Optical coherence to-mography angiography. Retina. 2015;35:2161-2162.

15. Jia Y, Morrison JC, Tokayer J, et al. Quantitative OCT angi-ography of optic nerve head blood flow. Biomed Opt Express.2012;3:3127-3137.

16. Jia Y, Wei E, Wang X, et al. Optical coherence tomographyangiography of optic disc perfusion in glaucoma. Ophthal-mology. 2014;121:1322-1332.

17. Yu J, Jiang C, Wang X, et al. Macular perfusion in healthyChinese: an optical coherence tomography angiogram study.Invest Ophthalmol Vis Sci. 2015;56:3212-3217.

18. Liu L, Jia Y, Takusagawa HL, et al. Optical coherencetomography angiography of the peripapillary retina in glau-coma. JAMA Ophthalmol. 2015;133:1045-1052.

19. Yarmohammadi A, Zangwill LM, Diniz-Filho A, et al. Opticalcoherence tomography angiography vessel density in healthy,glaucoma suspect, and glaucoma eyes. Invest Ophthalmol VisSci. 2016;57:451-459.

20. Wang X, Jiang C, Ko T, et al. Correlation between optic discperfusion and glaucomatous severity in patients with open-angle glaucoma: an optical coherence tomography angiog-raphy study. Graefes Arch Clin Exp Ophthalmol. 2015;253:1557-1564.

21. Sample PA, Girkin CA, Zangwill LM, et al. The AfricanDescent and Glaucoma Evaluation Study (ADAGES): designand baseline data. Arch Ophthalmol. 2009;127:1136-1145.

22. Hodapp E, Parish II RK, Anderson DR. Clinical Decisions inGlaucoma. St. Louis, MO: Mosby; 1993:52-61.

23. Racette L, Liebmann JM, Girkin CA, et al. African Descentand Glaucoma Evaluation Study (ADAGES): III. Ancestrydifferences in visual function in healthy eyes. Arch Oph-thalmol. 2010;128:551-559.

24. Mwanza JC, Warren JL, Hochberg JT, et al. Combining fre-quency doubling technology perimetry and scanning laserpolarimetry for glaucoma detection. J Glaucoma. 2015;24:561-567.

25. Flammer J, Orgul S. Optic nerve blood-flow abnormalities inglaucoma. Prog Retin Eye Res. 1998;17:267-289.

26. Weinreb RN, Bartsch DU, Freeman WR. Angiography of theglaucomatous optic nerve head. J Glaucoma. 1994;3(Suppl 1):S55-S60.

27. Piltz-seymour JR, Grunwald JE, Hariprasad SM, Dupont J.Optic nerve blood flow is diminished in eyes of primaryopen-angle glaucoma suspects. Am J Ophthalmol. 2001;132:63-69.

28. Grunwald JE, Piltz J, Hariprasad SM, DuPont J. Optic nerveand choroidal circulation in glaucoma. Invest Ophthalmol VisSci. 1998;39:2329-2336.

29. Tobe LA, Harris A, Hussain RM, et al. The role of retro-bulbar and retinal circulation on optic nerve head and retinalnerve fibre layer structure in patients with open-angle glau-coma over an 18-month period. Br J Ophthalmol. 2015;99:609-612.

30. Munguba GC, Galeb S, Liu Y, et al. Nerve fiber layer thinninglags retinal ganglion cell density following crush axonopathy.Invest Ophthalmol Vis Sci. 2014;55:6505-6513.

31. Chauhan BC, Stevens KT, Levesque JM, et al. Longitudinalin vivo imaging of retinal ganglion cells and retinal thicknesschanges following optic nerve injury in mice. PLoS One.2012;7:e40352.

32. Leung CK, Chong KK, Chan WM, et al. Comparative study ofretinal nerve fiber layer measurement by StratusOCT and GDxVCC, II: structure/function regression analysis in glaucoma.Invest Ophthalmol Vis Sci. 2005;46:3702-3711.

33. Sehi M, Goharian I, Konduru R, et al. Retinal blood flow inglaucomatous eyes with single-hemifield damage. Ophthal-mology. 2014;121:750-758.

34. Flammer J, Orgul S, Costa VP, et al. The impact of ocularblood flow in glaucoma. Prog Retin Eye Res. 2002;21:359-393.

35. Tokayer J, Jia Y, Dhalla AH, Huang D. Blood flow velocityquantification using split-spectrum amplitude-decorrelationangiography with optical coherence tomography. Biomed OptExpress. 2013;4:1909-1924.

36. Spaide RF, Fujimoto JG, Waheed NK. Image artifacts in op-tical coherence tomography angiography. Retina. 2015;35:2163-2180.

37. Zhi Z, Cepurna WO, Johnson EC, et al. Impact of intra-ocular pressure on changes of blood flow in the retina,choroid, and optic nerve head in rats investigated by op-tical microangiography. Biomed Opt Express. 2012;3:2220-2233.

38. Bengtsson B, Heijl A. A visual field index for calculation ofglaucoma rate of progression. Am J Ophthalmol. 2008;145:343-353.

39. Rao HL, Senthil S, Choudhari NS, et al. Behavior of visualfield index in advanced glaucoma. Invest Ophthalmol Vis Sci.2013;54:307-312.

40. Gardiner SK, Swanson WH, Goren D, et al. Assessment ofthe reliability of standard automated perimetry in regions ofglaucomatous damage. Ophthalmology. 2014;121:1359-1369.

41. Hood DC, Anderson SC, Wall M, Kardon RH. Structureversus function in glaucoma: an application of a linear model.Invest Ophthalmol Vis Sci. 2007;48:3662-3668.

42. Hood DC, Kardon RH. A framework for comparing structuraland functional measures of glaucomatous damage. Prog RetinEye Res. 2007;26:688-710.

2507

Ophthalmology Volume 123, Number 12, December 2016

Footnotes and Financial Disclosures

Originally received: February 22, 2016.Final revision: August 22, 2016.Accepted: August 26, 2016.Available online: October 7, 2016. Manuscript no. 2016-371.1 Hamilton Glaucoma Center, Shiley Eye Institute, Department ofOphthalmology, University of California San Diego, La Jolla, California.2 Department of Ophthalmology, Haeundae Paik Hospital, Inje UniversityCollege of Medicine, Busan, South Korea.

Financial Disclosure(s):The author(s) have made the following disclosure(s): L.M.Z.: Researchsupport e Optovue Inc., Carl Zeiss Meditec, Heidelberg Engineering,National Eye Institute, Topcon, Nidek.

F.A.M.: Financial support e Alcon, Allergan, Bausch & Lomb, Carl ZeissMeditec, Heidelberg Engineering, Merck, Reichert, Sensimed, Topcon;Research support e Alcon, Allergan, Carl Zeiss Meditec, National EyeInstitute, Reichert; Consultant e Allergan, Carl Zeiss Meditec, Novartis.

R.N.W.: Research support e Carl Zeiss Meditec, Genentech, HeidelbergEngineering, National Eye Institute, Optovue, Topcon; Consultante Alcon,Allergan, Bausch & Lomb, ForSight, Unity.

Supported in part by National Institutes of Health/National Eye InstituteGrants EY011008 (L.M.Z.), EY14267 (L.M.Z.), and EY019869 (L.M.Z.),Core Grant P30EY022589, an unrestricted grant from Research to PreventBlindness (New York, NY), and grants for participants’ glaucoma medi-cations from Alcon, Allergan, Pfizer, Merck, and Santen.

2508

Author Contributions:

Conception and design: Yarmohammadi, Zangwill, Diniz-Filho, Medeiros,Weinreb

Data collection: Yarmohammadi, Zangwill, Diniz-Filho, Suh, Yousefi,Saunders, Belghith, Manalastas, Medeiros, Weinreb

Analysis and interpretation: Yarmohammadi, Zangwill, Diniz-Filho, Suh,Yousefi, Saunders, Belghith, Manalastas, Medeiros, Weinreb

Obtained funding: Not applicable

Overall responsibility: Yarmohammadi, Zangwill, Diniz-Filho, Suh,Saunders, Medeiros, Weinreb

Abbreviations and Acronyms:AIC ¼ Akaike’s information criterion; ANOVA ¼ analysis of variance;BP ¼ blood pressure; CCT ¼ central corneal thickness;cpVD ¼ circumpapillary vessel density; dB ¼ decibels; IOP ¼ intraocularpressure; MAP ¼ mean arterial pressure; MD ¼ mean deviation;MOPP ¼ mean ocular perfusion pressure; OCT ¼ optical coherence to-mography; OCT-A ¼ optical coherence tomography angiography;ONH ¼ optic nerve head; PSD ¼ pattern standard deviation;RNFL ¼ retinal nerve fiber layer; SAP ¼ standard automated perimetry;SD OCT ¼ spectral-domain optical coherence tomography;wiVD ¼ whole-image vessel density.

Correspondence:Robert N. Weinreb, MD, Hamilton Glaucoma Center and Department ofOphthalmology, University of California, San Diego, 9500 Gilman Drive,La Jolla, CA 92093-0946. E-mail: rweinreb@ucsd.edu.

Pictures & Perspectives

Atypical Presentation of IgG4-Related Disease of the EyelidA 37-year-old HIV-positive woman presented with a 1-month history

of progressive, painless, full-thickness destruction of the left upper,lower, and medial canthal eyelid with madarosis and symblepharon (Fig1A). Incisional biopsy revealed a dense polytypic plasma cell infiltrate(Fig 1B), fibrosis, and phlebitis. Immunohistochemistry showed >100IgG4 plasma cells per high-power field (Fig 1C) and a ratio of IgG4:IgG(Fig 1D) of >50% consistent with IgG4-related disease.

MAMTA SHAH, BS1

CHARLES SHAO, MD2

ROMAN SHINDER, MD, FACS1

1Ophthalmology, SUNY Downstate Medical Center, Brooklyn, New York;2Pathology, SUNY Downstate Medical Center, Brooklyn, New York