Article Text

Download PDFPDF

Regional vascular density–visual field sensitivity relationship in glaucoma according to disease severity
  1. Joong Won Shin1,
  2. Jiyun Lee2,
  3. Junki Kwon2,
  4. Jaewan Choi3,
  5. Michael S Kook2
  1. 1 Department of Ophthalmology, Asan Medical Center, Songpa-gu, Republic of Korea
  2. 2 Department of Ophthalmology, University of Ulsan, Asan Medical Center, Seoul, Republic of Korea
  3. 3 Glaucoma Service, Central Seoul Eye Center, Seoul, Republic of Korea
  1. Correspondence to Dr Michael S Kook, Department of Ophthalmology, University of Ulsan, Asan Medical Center, Seoul, 138-736, Republic of Korea; mskook{at}


Aims To study whether there are global and regional relationships between peripapillary vascular density (pVD) assessed by optical coherence tomography angiography (OCT-A) and visual field (VF) mean sensitivity at different glaucoma stages.

Methods Microvascular images and peripapillary retinal nerve fibre layer (pRNFL) thicknesses were obtained using a Cirrus OCT-A device in 91 glaucoma subjects. The pVD was measured at various spatial locations according to the Garway-Heath map, using a MATLAB software (The MathWorks, Natick, Massachusetts). VF mean sensitivity (VFMS) was recorded in the 1/L scale. Global and regional vasculature–function (pVD vs VFMS) relationships were assessed in separate patient groups at mild and moderate-to-advanced stages of glaucoma.

Results The pVDs at superotemporal and inferotemporal regions were significantly associated with corresponding VFMS in mild glaucoma (p<0.05). In moderate-to-advanced glaucoma, there were significant associations between pVD and VFMS, regardless of location. The association between global pVD and VFMS was significantly stronger than that between global pRNFL thickness and VFMS in moderate-to-advanced stage glaucoma (p <0.05).

Conclusion Global and regional pVD measured by OCT-A was significantly associated with corresponding VFMS in moderate-to-advanced glaucoma. OCT-A may be useful in monitoring glaucoma at various stages.

  • optical coherence tomography angiography
  • vascular density
  • vascular-functional relationship
  • advanced glaucoma

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


The recent advent of optical coherence tomography angiography (OCT-A) has allowed fast and noninvasive assessment of microvasculature in the optic nerve head (ONH), peripapillary retina and macula. In OCT-A studies, reduced peripapillary vascular density (pVD) was reported in glaucomatous eyes,1–4 and this spatially corresponded with damaged retinal nerve fibre layer (RNFL) bundles.5 6 Moreover, pVD was associated with the severity of global visual field (VF) damage, and its association was stronger than the RNFL–VF association.7 However, it remains unknown whether there are significant regional relationships between the pVD and visual function. In other words, the vascular–functional relationship should be tested and validated at various locations of VF since glaucomatous VF often starts and deteriorates regionally.8

Although it is well accepted that there is a moderate relationship between structure (ie, RNFL thickness) and function (ie, VF sensitivity) in glaucoma, it is still debated whether this relationship holds true at advanced stages of the disease. Currently, only limited information is available on the precise nature of the association between vascular parameters and functional measurements during the course of the glaucoma spectrum. With this in mind, the purpose of this study was to investigate the global and regional relationships between pVD derived from OCT-A, and VF mean sensitivity (VFMS) assessed by standard automated perimetry (SAP), in respect to glaucoma severity.



This prospective cross-sectional study enrolled glaucoma subjects from the glaucoma service of the Asan Medical Center in Seoul, Korea, between April 2016 and September 2016 in a consecutive manner. The institutional review board of the Asan Medical Center approved this study, and all procedures were executed in accordance with the principles of the Declaration of Helsinki. All participants provided written informed consent before enrolment. For inclusion in the study, all participants had to have glaucoma, regardless of intraocular pressure (IOP) level, with the following criteria: a best-corrected visual acuity of 20/40 or better, a spherical equivalent (SE) between −8.0 and +3.0 diopters (D) and a cylinder correction within +3 D, a normal anterior chamber and open angle on slit-lamp and gonioscopic examinations, and two reliable SAP (Humphrey Field Analyzer (HFA) with Swedish Interactive Threshold Algorithm (SITA) standard 24-2 test; Carl Zeiss Meditec, Dublin, CA) test results with a false-positive error <15%, false-negative error <15% and fixation loss <20%. Glaucoma was defined as the presence of glaucomatous ONH changes (eg, vertical cup-to-disc (C/D) ratio >0.7, focal or diffuse neural rim loss, disc haemorrhage or RNFL defects on red-free photography) with compatible glaucomatous VF defects.9 Patients with glaucomatous VF defects on the first 24-2 SITA test underwent a second round of HFA testing to minimise learning effects; the second HFA test was used in this analysis. The severity of glaucomatous damage was classified as mild (VF mean deviation (MD) ≥−6 dB) or moderate-to-advanced (VF MD <−6 dB).9 Subjects with any ophthalmic or neurologic disease known to affect ONH or VF sensitivity were excluded. Patients who were on topical antiglaucoma eye drops or systemic vasoactive medications were not excluded. When both eyes of a subject were eligible, one eye was selected at random.

RNFL thickness measurements

Using Cirrus spectral-domain optical coherence tomography (SD-OCT; Carl Zeiss Meditec, Dublin, CA; version 6.0), peripapillary RNFL thickness (pRNFLT) was measured using an Optic Disc Cube (ODC) in 200×200 scan mode. The Cirrus SD-OCT automatically detects the centre of the disc and then draws a peripapillary circle (3.46 mm diameter) from the cube data set for use in pRNFLT analysis. Poor-quality images with a signal strength less than 7, misalignment or overt decentration of the measurement circle location were discarded.


The commercial AngioPlex imaging system (Carl Zeiss Meditec, Dublin, California) is incorporated into the Cirrus SD-OCT 5000 instrument and generates high-resolution three-dimensional maps of the retinal microvasculature. This device operates at a central wavelength of 840 nm and a speed of 68 000 A-scans per second. An optical microangiography-complex (OMAGc) algorithm analyses the change of complex signal (both intensity and phase changes contained within sequential B-scans performed at the same position)10 and then generates en face microvascular images. AngioPlex also incorporates the FastTrac retinal-tracking technology to reduce motion artefacts. A 3×3 mm scan pattern composed of 245 A-scans and 245 B-scans around the optic disc was used. All scans were individually reviewed by two investigators (J.W.S. and M.S.K.) for evaluation of quality (ie, signal strength, segmentation error, loss of fixation or motion artefact). Eyes with poor image qualities were excluded on the basis of the following criteria10: (1) poor fixation resulting in a double vessel pattern and motion artefacts; (2) media opacity obscuring the vessel signal in the field of view or a signal strength index less than 7; (3) segmentation error resulting in poor outlining of vascular networks.

Vascular density measurements

Image processing and vascular density measurements were performed by a computer program written using MATLAB software (The MathWorks, Natick, Massachusetts). An automatic thresholding algorithm generated a binary image from the OCT-A microvasculature image around the 3×3 mm region of the optic disc. Blood vessels were defined as pixels having values above threshold level. Vascular density was calculated as the ratio of the total vessel area over the total area of the region of interest. pVD was calculated within a region defined as a 500 µm-wide elliptical annulus extending from the optic disc boundary. The optic disc boundary was determined from the optic disc photograph (AFC-210; Nidek, Aichi, Japan), which was overlaid and registered with the OCT-A image photograph according to the retinal blood vessels.

Mapping vasculature (pVD) and structure (pRNFLT) to function (VFMS)

The MS at various VF sectors was defined as the average value of the differential light sensitivity (DLS) obtained at VF test locations corresponding to pVD sectors. The VFMS was expressed in unlogged 1/L scales (L, luminance measured in lamberts). The DLS at each tested location can be simply written as DLS (decibel (dB))=10×log10 (1/L). The non-logarithmic 1/L value at each tested location was calculated by dividing the decibel reading by 10, followed by derivation of the antilogarithm.

Vasculature–function and structure–function relationships were analysed by comparing pVD and pRNFLT, and the corresponding VFMS, according to the regionalization of Garway-Heath et al.11 The pVD and pRNFLT measurements were performed at six sectors: temporal (T, 316°–45°), superotemporal (ST, 46°–90°), superonasal (SN, 91°–135°), nasal (N, 136°–225°), inferonasal (IN, 226°–270°) and inferotemporal (IT, 271°–315°). HFA 24-2 testing points were grouped into six sectors: superonasal, nasal, inferonasal, inferotemporal, temporal and superotemporal, according to the structure–function map described by Garway-Heath et al.11 The regional pVD of each sector was measured by a computer program written using MATLAB software. The pRNFLT of each sector was estimated by integrating the clock hour pRNFLT from the Cirrus SD-OCT.12

Statistical analysis

Normally distributed data were compared using independent Student’s t-tests. Otherwise, Mann-Whitney U tests were used. Normality was tested using the Kolmogorov-Smirnov test. Categorical variables were compared using χ2 tests between the groups. In all regression tests assessing the vasculature–function or structure–function relationship, the VFMS was treated as the dependent variable and the corresponding pVD and pRNFLT as the independent variables. The linear regression better defines the structure–function relationship when VFMS is expressed in a linear unlogged 1/L scale.13 Therefore, the relationship between pVD or pRNFLT and VFMS was evaluated with linear regression analysis. Weighted correlation coefficients were calculated as the VFMS was derived by averaging a different number of VF test locations.14 Steiger’s test was used to assess the significance of differences between any two weighted correlation coefficients measured on the same individuals.15 Univariable linear regression models were built using VF MD as the dependent variable and patient’s demographics (age, IOP and SE), Cirrus SD-OCT parameters (disc area, rim area, RNFL thickness) and pVD derived from OCT-A as independent variables. Multivariable analysis was also performed to evaluate the relationship between the VF MD and independent variables while adjusting for potential confounding factors including structural parameters. p Values of 0.05 or less were considered statistically significant. Statistical analyses were performed using SPSS V.20.0 and the R statistical programming language (V.3.1.2; R Foundation, Vienna, Austria).


Of the 107 eyes of 107 glaucoma patients that initially met eligibility criteria, 12 eyes (11.2%) and 4 eyes (3.7%) were excluded because of poor-quality OCT-A images and SD-OCT ODC scans, respectively. Therefore, a total of 91 eyes from 91 glaucoma subjects were included in the final analysis. Fifty-one eyes in 51 subjects had mild glaucoma, and 40 eyes in 40 subjects had moderate-to-advanced glaucoma. Table 1 shows the demographic and clinical characteristics of the study subjects. There were statistically significant differences between groups in terms of use of some glaucoma medications (p<0.05), rim area (p<0.001), average pRNFLT (p<0.001), average pVD (p<0.001), mean MD (p<0.001) and mean pattern standard deviation (PSD) (p<0.001). Otherwise, there were no significant differences in age, gender, laterality and SE between the two groups (all p>0.05).

Table 1

Demographic and clinical characteristics of glaucoma subjects at various stages

Overall (n=91), the association between pVD and VFMS was generally stronger than that between pRNFLT and VFMS without statistical significance except for nasal sector (p=0.018). However, the vasculature–function and structure–function relationships showed different characteristics according to the stage of glaucoma. The pVD and pRNFLT at superotemporal and inferotemporal regions were significantly associated with corresponding VFMS in the mild stage of glaucoma (Table 2, all p <0.05). There were no statistically significant differences in the correlation coefficients between pVD–VF and pRNFLT–VF associations in mild glaucoma, regardless of location. In the moderate-to-advanced stage of glaucoma, there were statistically significant associations between pVD and VFMS regardless of location (Table 2 and Figure 1). However, the relationship between pRNFLT and VFMS was statistically significant only in the superotemporal, inferotemporal and temporal regions in moderate-to-advanced glaucoma. The global association between pVD and VFMS was significantly stronger than that between pRNFLT and VFMS in moderate-to-advanced glaucoma (p=0.040).

Table 2

Weighted correlation coefficients (p value) between peripapillary vascular density (pVD), retinal nerve fibre layer (RNFL) thickness and visual field mean sensitivity (VFMS)

Figure 1

Scatter plots of regional peripapillary vascular density (pVD) and corresponding visual field (VF) mean sensitivity in mild and moderate-to-advanced glaucoma in a linear scale (1/L). MD, mean deviation.

The results from the univariable and multivariable regression analysis for glaucoma severity determined with MD as the dependent variable are summarised in Table 3. Multivariable linear regression analysis, while controlling for the confounding effect of age, rim area and pRNFLT, demonstrated that pVD was independently associated with VF MD (p=0.010).

Table 3

Univariable and multivariable regression analysis to determine the factors associated with visual field mean deviation


The best way to evaluate the structure–function relationship in glaucoma is to compare local VF sensitivity to local structural measurements.16 Most OCT-A studies compared vascular parameters with VFMS data averaged over a large region (eg, hemifield or full-field),1 4 7 with the regional vasculature–function relationship having not been studied as yet. In the present study, the vasculature–function relationship was assessed by matching specific VF locations to corresponding sectors of the circular pVD, in accordance with the regionalization of Garway-Heath et al.11 The vasculature–function relationship was significant in moderate-to-advanced glaucoma regardless of the VFMS location. However, in mild glaucoma, the vasculature–function relationship was only significant in the superotemporal and inferotemporal sectors. These regions are the ones that are most vulnerable to glaucomatous damage at an early stage.17 Therefore, it can be postulated that vascular damage begins with localised regions in the early stage of glaucoma and then expands to the whole peripapillary area as the disease advances. Further longitudinal study is needed to confirm our speculation on the progressive vascular changes in glaucoma.

In moderate-to-advanced glaucoma, although the vasculature–function relationship was significant in all sectors, the structure–function relationship was only significant in the superotemporal, inferotemporal or temporal sector. Consequently, the association between global pVD and VFMS was significantly stronger than that between global pRNFLT and VFMS in moderate-to-advanced glaucoma. In addition, the correlation coefficients between regional pVD and VFMS were generally higher than those between regional pRNFLT and VFMS. Therefore, our findings suggest that, in terms of detecting VF changes occurring at the advanced stage of glaucoma, the monitoring of pVD change could be more useful than that of pRNFLT change. Figure 2 shows representative cases in which VF deterioration is more strongly correlated with pVD reduction than it is with pRNFL thinning.

Figure 2

Representative cases of moderate-to-advanced stages of glaucoma. As glaucoma progresses to more advanced stage, regional capillary dropouts (arrows) and quantitative measurements of peripapillary vascular density (pVD) show good correlations with progressive visual field defects (asterisks), whereas regional peripapillary retinal nerve fibre layer (pRNFL) thinning and average thickness nearly reached a floor.

Mwanza et al. 18 reported that the average pRNFLT reached a floor at a relative VF sensitivity of −10.4 dB (range, −3.3 to −17.5) in simple linear regression. Considering the range of VF sensitivity at which a RNFL floor begins, VF tests have been preferred to pRNFLT measurements at the moderate-to-advanced stage of glaucoma for follow-up.12 However, the VF test sometimes shows large fluctuations in patients with advanced glaucoma, rendering it difficult to define actual glaucomatous change.19 By contrast, structural/vascular assessment is not dependent on patient response and may thus yield less variable results if the technology used offers good reproducibility. OCT-A provides objective information on retinal vasculature with high intra-visit repeatability and inter-visit reproducibility.20 Our investigation provides evidence to indicate that, along with the VF test, the pVD could be another candidate for following glaucoma in the advanced stage, as there is a significant vasculature–function association regardless of location.

Mansoori et al 21 reported that pRNFLT measurements for nasal quadrant showed higher variability than superior and inferior quadrants in normal and glaucomatous eyes (mean MD=−14.19 dB). This may partially explain our finding that a significant structure–function relationship was absent in the nasal sectors in moderate-to-advanced cases. The clinical implication of the stronger association between pVD and VFMS than between pRNFLT and VFMS, at all sectors of ONH, including nasal and temporal sectors in moderate-to-advanced glaucoma, is that vascular density may be a better biomarker than pRNFLT for monitoring of glaucoma in advanced cases.

In the present study, univariable linear regression analysis demonstrated VF loss to be correlated with both reduced pVD and structural loss such as rim area and RNFL thickness. However, multivariable analysis revealed that reduced pVD was an independent predictor of VF loss after adjusting for standard structural parameters such as rim area and pRNFLT. One explanation for our findings is that global association between pVD and VFMS was significantly stronger than that between pRNFLT and VFMS in moderate-to-advanced cases in the present study. This may explain the stronger association between pVD and VF MD compared with that between pRNFLT and VF MD found in our multivariable regression analysis. This finding is consistent with a previous study that used a different OCT-A device7 and reported that decreased pVD was significantly associated with the severity of VF damage, independent of the structural loss.

There are several limitations to the present study. Vascular dysfunction has been proposed as one of the contributing factors in the development and progression of glaucoma, especially in normal-tension glaucoma (NTG) patients.22 Since most of the glaucoma subjects (77%) in the present study were NTG, the possibility of overestimation of the vasculature–function relationship should be noted. However, in our separate analysis using NTG (n=70) and high-tension glaucoma patients (n=21), we found similar trends of vasculature–function relationship regardless of the type of glaucoma (data not shown). Antiglaucoma eye drops or systemic vasoactive medications may affect the retinal vascular diameter. Therefore, the possible confounding effect of these medications on vascular density should be taken into account when interpreting the relationship between pVD and VF in the present study. In the current study, we used the boundary from the optic disc photography as this method has been used in the past and published in the previous study.23 However, for more accurate determination of optic disc boundary, using Bruch’s membrane opening lines based on SD-OCT measurements is preferred.24 The participants were ethnically homogeneous Koreans, and data from a single ethnic group may not be generalizable to other races. The small number of patients in each group may also limit the generalizability of our findings to the general population with glaucoma. A cause-and-effect relationship between vascular damage and VF loss could not be determined by our cross-sectional study, and further longitudinal studies are needed.

In conclusion, global and regional pVD measured by OCT-A was significantly associated with corresponding VFMS in moderate-to-advanced glaucoma, while there were regional associations in superotemporal and inferotemporal sectors in the mild stage of glaucoma. The global vasculature–function relationship was significantly stronger than the conventional structure–function relationship using pRNFLT in moderate-to-advanced glaucoma. Monitoring changes in vascular density using OCT-A could be a useful modality for clinically following glaucoma patients with moderate-to-advanced disease.



  • Contributors JWS designed the conception of the work, analysed data, and drafted and revised the paper. JL monitored the acquisition of data. JK monitored the acquisition of data, and analysed data. JC designed the conception of the work and monitored the acquisition of data. MSK designed the conception of the work, monitored the acquisition of data, analysed data, and drafted and revised the paper.

  • Competing interests None declared.

  • Ethics approval The institutional review board of the Asan Medical Center.

  • Provenance and peer review Not commissioned; externally peer reviewed.