Gap in Capillary Perfusion on Optical Coherence Tomography Angiography Associated With Persistent Macular Edema in Branch Retinal Vein Occlusion

Abstract

Purpose: To evaluate correlations between persistent macular edema associated with branch retinal vein occlusion (BRVO) and the macular perfusion status in the superficial capillary plexus (SCP) and deep capillary plexus (DCP) using optical coherence tomography angiography (OCTA).

Methods: Retrospective, case-control study. Twenty patients with BRVO followed for 12 months or more were enrolled. Persistent macular edema was defined as central retinal thickness exceeding 300 μm that persisted or recurred less than 3 months after the final treatment. We compared two groups (i.e., seven eyes with persistent macular edema and 13 eyes without macular edema). The macular perfusion status was evaluated using OCTA. We defined a gap vessel as a residual vessel in the SCP that existed simultaneously with capillary loss in the DCP. The gap vessels were determined by subtracting the vessel images of the DCP from the images of the SCP using an image processing technique.

Results: In eyes with persistent macular edema, the area with gap vessels was significantly (P = 0.0013) larger than in eyes without macular edema (14.34% vs. 8.02%). Other factors evaluated (i.e., the area of the foveal avascular zone, superficial vessel density, and deep vessel density) did not differ significantly (P = 0.66, P = 0.23, P = 0.34, respectively) between the groups.

Conclusions: The difference in capillary loss between the SCP and DCP can facilitate development of persistent macular edema in BRVO.

Branch retinal vein occlusion (BRVO) is the second most common retinal vascular disease after diabetic retinopathy,^1–4 and macular edema is the most common cause of visual loss in BRVO. Currently, anti-vascular endothelial growth factor (VEGF) agents are the primary therapies for macular edema associated with BRVO.^5–9 Although anti-VEGF agents have been used successfully to reduce macular edema due to BRVO, some patients have persistent macular edema after repeated treatments with anti-VEGF agents. 

Several reports have described the relationship between macular edema and retinal perfusion using fluorescein angiography (FA). Noma et al.¹⁰ reported a positive correlation between the severity of macular edema seen on optical coherence tomography (OCT) images and the size of the nonperfusion areas (NPAs), which can develop anywhere in the fundus. Prasad et al.¹¹ also reported that the incidence of macular edema is associated significantly with the NPAs on ultra-wide-field angiography, especially when the anterior hemisphere is involved. In contrast, Finkelstein¹² reported that ischemic macular edema appears to be transient with spontaneous improvement in the visual acuity (VA) and that perfused macular edema persists frequently with decreases in the VA. In addition, Sakimoto et al.¹³ reported that partially perfused capillary areas, and not the areas of complete capillary loss, are associated more frequently with macular edema. Despite publication of many reports about perfusion status and macular edema, the relationship between the two remains unclear. 

Optical coherence tomography angiography (OCTA), a recently developed technology, can visualize the microvascular abnormalities of each retinal capillary plexus. Several reports in which OCTA was performed in eyes with BRVO have described that retinal vascular abnormalities (i.e., capillary telangiectasia, microaneurysms, and NPAs) develop more frequently in the deep capillary plexus (DCP) than in the superficial capillary plexus (SCP).^14–16 In addition, Spaide¹⁷ also reported that macular edema in RVO occurs in association with the absence of flow in the deep vascular plexus. Therefore, the retinal perfusion status in the SCP and DCP often differs, and the difference in the retinal capillary loss between the two layers might contribute to the development of macular edema in BRVO. 

In the current study, we calculated the gaps in the capillary loss by subtracting the OCTA images of the DCP from those of the SCP and investigated the relation between the gap in capillary loss and persistent macular edema in BRVO. 

Methods

Patients

This retrospective observational study included 20 eyes of 20 patients who had been diagnosed with BRVO in the Department of Ophthalmology, Aichi Medical University Hospital, from March 1, 2006, to October 7, 2015. We reviewed the medical and ocular histories of these patients. The study adhered to the tenets of the Declaration of Helsinki. The institutional review board of Aichi Medical University approved the study protocol. 

The inclusion criteria included a minimal follow-up period of 12 months regardless of treatment and OCTA measurements obtained at 12 months or later. The exclusion criteria included the presence of central or hemicentral RVO, epiretinal membranes, macular holes, diabetic retinopathy, and poor OCTA images with a signal strength of 50 or less. 

Persistent macular edema was defined as a central retinal thickness exceeding 300 μm and persistent or recurrent macular edema within 3 months after the final treatment. The absence of macular edema was defined as a central retinal thickness below 300 μm for more than 3 months after the final treatment. 

Optical Coherence Tomography Angiography

OCTA images were obtained using the RTVue XR Avanti (Optovue, Inc., Fremont, CA, USA) with a split-spectrum amplitude-decorrelation angiography algorithm as previously described.^18–21 The scans were taken from 3 × 3-mm cubes centered on the fovea. In eyes with persistent macular edema, OCTA was performed within 2 months after the last anti-VEGF treatment when the macular edema was suppressed. If considerable macular edema remained after anti-VEGF treatment, the cases were excluded because of the difficulty in precisely evaluating the vascular abnormalities.¹⁴ En-face images of the retinal vasculature were generated from the SCP and DCP based on automated layer segmentation performed by the OCT instrument software. Manual segmentation was performed if necessary to accurately assess the SCP and DCP. The en-face OCTA images of the SCP and DCP were analyzed by the following methods. 

Isolated Vessel Density in the SCP

We compared the macular perfusion in the SCP with that in the DCP and evaluated the perfusion gap in each retinal layer. We identified isolated vessels in the superficial layer (i.e., residual vessels in the SCP existing simultaneously with capillary loss in the DCP). To visualize the isolated vessels in the superficial layer, we used ImageJ software (version 1.50) provided in the public domain by the National Institutes of Health (Bethesda, MD, USA; http://imagej.nih.gov/ij/). OCTA images of the SCP and DCP were processed for binarization and subtraction (Figs. 1, 2, 3). The images were converted to 8 bits and assigned a value of 255 (complete white) to all pixels with a positive gray level and a value of 0 (complete black) to the others by the Niblack Auto Local Threshold method (radius, 75 pixels; parameter 1, default; parameter 2, default).^22,23 To visualize the isolated vessels, the binarized OCTA image of the DCP was subtracted from the image of the SCP. Binarization and subtraction made it possible to find very few or no isolated vessels in the subtracted image if the capillaries remained in both the SCP and DCP or was obstructed in both layers (Figs. 1, 2). When the capillaries remained only in the SCP, the isolated vessels were visible only in the subtracted images (Fig. 3). 

Figure 1

 

BRVO without macular edema in a 76-year-old patient. (A) An OCTA image of the SCP shows small capillary loss (arrowheads) in the upper right region. (B) An OCTA image of the DCP shows vessel telangiectasia and small capillary loss (arrowheads) in the upper right region. (C) A composite image of the SCP (blue) and the DCP image (red) show capillary loss in the SCP that corresponds to capillary loss in the DCP (arrowheads). (D) A binarized SCP image and (E) a binarized DCP image. (F) The binarized DCP image is subtracted from the binarized SCP image. No isolated vessels are seen in the subtracted image. (G) The horizontal and (H) vertical B-scan images showing the segmentation boundaries for en face OCTA images.

Figure 2

 

BRVO without macular edema in an 80-year-old patient. (A) An OCTA image of the SCP shows large capillary loss (arrowheads) in the lower left region. (B) An OCTA image of the DCP shows large capillary loss (arrowheads) in the lower left region. (C) A composite image of the SCP (blue) and the DCP (red) shows capillary loss (arrowheads) in the SCP that corresponds to capillary loss in the DCP. (D) A binarized SCP image and (E) a binarized DCP image. (F) The binarized DCP image is subtracted from the binarized SCP image. No isolated vessels are seen in the subtracted image. (G) The horizontal and (H) vertical B-scan images showing the segmentation boundaries for en face OCTA images.

Figure 3

 

BRVO with persistent macular edema in a 75-year-old patient. (A) An OCTA image of the SCP shows capillary loss (arrowheads) in the upper left region. (B) An OCTA image of the DCP shows large capillary loss (arrowheads) in the upper left region. (C) A composite image shows capillary loss in the DCP but capillaries remaining in the SCP (blue vessels) (arrowheads) in the upper region. (D) A binarized SCP image and (E) a binarized DCP image. (F) Isolated vessels (arrowheads) are seen in the upper region. (G) The horizontal and (H) the vertical B-scan images showing the segmentation boundaries for en face OCTA images.

OCTA Image Analysis

To quantify the OCTA images, we measured the area of the foveal avascular zone (FAZ) in the SCP and the vessel density in each layer. Measurements of the FAZ area were calculated using the nonflow function in the OCTA internal software built (version 2016.1.0.26; Optovue, Inc.). The vessel density was calculated as the ratio of the areas of the pixels representing the vessels as previously described.^24,25 The vessel density was calculated for the affected sector of the grid (Fig. 4). 

Figure 4

 

A binarized and subtracted OCTA image. The vessel density is calculated for the affected sector surrounded by the green line (a hemi-ring-like region between the central 1-mm circle and the 3-mm-diameter circle).

Statistical Analysis

All data were analyzed using JMP version 11.0.0 (SAS Institute, Inc., Cary, NC, USA). The data are expressed as the average ± standard deviation. The best-corrected visual acuity (BCVA) was measured with Landolt C acuity charts, and the decimal BCVA was converted to the logarithm of the minimal angle of resolution units for the statistical analyses. Pearson's χ² test, Fisher's exact test, t-test, and Wilcoxon rank-sum test were performed. P < 0.05 was considered significant. 

Results

Patient Demographic Data

Twenty eyes of 20 patients (8 men, 12 women) with BRVO were included in this study. The cases included seven eyes with persistent macular edema and 13 eyes without macular edema at the final visit. The average follow-up duration from the first visit to the OCTA examination was 40.4 ± 33.7 months (range, 12–124 months). Table 1 shows the patient demographic data. There were no significant differences in age (P = 0.81), sex (P = 0.36), follow-up duration (P = 0.58), initial BCVA (P = 0.16), final BCVA (P = 0.16), or incidence of macular edema at the initial visit (P = 0.52) between the eyes with persistent macular edema and those without macular edema. There were no significant (P > 0.05 for each comparison) differences in the mean numbers of interventions including sub-Tenon's triamcinolone acetonide injections, laser photocoagulation, or pars plana vitrectomy, but the mean number of total anti-VEGF injections administered was significantly (P = 0.032) higher in the group with persistent macular edema group than in the group without macular edema (3.9 vs. 2.0, respectively). 

Table 1

 

Patient Demographic Data From Eyes With and Without Persistent Macular Edema

 

OCTA Measurements With and Without Persistent Macular Edema

Table 2 shows the comparison of the OCTA parameters between patients with BRVO with and without persistent macular edema. There was no significant (P = 0.66) difference in the area of the FAZ between patients with and without macular edema (0.41 mm² versus 0.49 mm², respectively). No significant differences in the vessel densities were seen in the SCP and DCP (30.67% vs. 30.6%, respectively, in the SCP [P = 0.23] and 46.57% vs. 46.3% in the DCP, respectively [P = 0.34]). However, the densities of the isolated vessels (the gap between the SCP and DCP) were significantly (P = 0.0013) greater in the group with persistent macular edema (14.34% vs. 8.02%; Table 2). 

Table 2

 

Comparisons of OCTA Parameters in BRVO With or Without Persistent Macular Edema

 

Discussion

In the current study, we evaluated the isolated vessels in the superficial layer using OCTA and a threshold technique. To our knowledge, this is the first study to analyze objectively the correlation between the interlayer perfusion status and macular edema in patients with BRVO. We found that the gap between the SCP and DCP (i.e., specifically the presence of isolated vessels in the SCP existing simultaneously with capillary loss in the DCP) was associated with persistent macular edema. 

Although the exact pathogenesis of the macular edema in BRVO has not been established, the development of macular edema has been hypothesized to be a result of fluid flux from vessels to tissue according to Starling's law. Breakdown of the blood-retinal barrier in the tight junctions of the capillary endothelial cells causes secretion of various factors into the vitreous, resulting in retinal edema.^10,26,27 Previous studies of BRVO generally have used FA to grade the perfusion status. However, FA cannot clearly visualize the perfusion status in the separate retinal layers, especially in the DCP, but OCTA can visualize the different layers separately.²¹ Spaide and colleagues^17,28 have studied the OCTA findings in RVO and forwarded a new theory (i.e., a normal DCP helps remove interstitial fluid from the retina, and deteriorated flow in the DCP might cause macular edema). Spaide¹⁷ also reported that cystoid edema does not occur when capillary loss is seen in both the SCP and DCP. These observations agreed well with the current study. When capillary loss occurs in both the SCP and DCP, the retina becomes atrophic and thin as a result of ischemia, and the macular edema no longer progresses. In contrast, in the area with isolated vessels in the superficial layer or where a gap in the capillaries exists, persistent edema is seen frequently. We speculated that the presence of isolated vessels is related closely to macular edema followed by BRVO. 

We developed a new technique to evaluate the capillary gaps between the SCP and DCP using OCTA and an image processing technique, the main purpose of which is to visualize the isolated vessels using OCTA. This method showed that the density of the isolated vessels is significantly greater in eyes with persistent macular edema, whereas the vessel densities in the SCP and DCP are not correlated. Therefore, we believe that analyzing the isolated vessel is more important to evaluate persistent macular edema than analyzing the vessel density separately in the two layers. In addition, our protocol eliminates subjectivity. We believe that the current method is advantageous over previous methods, since the gap between the SCP and DCP is associated directly with persistent macular edema. 

The vessels in the SCP and DCP indicated the area of vascular flow around the macula. The vessel density might be correlated with the NPA. Previous studies have reported relationships between macular edema and increased NPAs. The discrepancy between those results and our findings might be attributable to different observational periods. Intraretinal hemorrhages and macular edema associated with BRVO generally resolve within 6 to 12 months.^29,30 Hayreh and Zimmerman³¹ reported that in the natural history of BRVO approximately 40% of the macular edema resolves 15 months after the beginning of the follow-up period. The current study included BRVO cases with a minimal follow-up of 12 months. However, Noma et al.¹⁰ reported that the mean duration of BRVO at the time of examination was 3.8 ± 3.0 months (range, 1–10 months), and Prasad et al.¹¹ reported 24.5 months (range, 0–216 months). Thus, the latter study might have included both cases with spontaneous resolution and those with persistent macular. Evaluation of macular perfusion is difficult in early-phase BRVO because of the presence of intraretinal hemorrhages, hard exudates, and retinal edema. Therefore, the current study found different results in the relationship between the perfusion status and macular edema compared to previous studies. 

The limitations of the current study were its retrospective nature and small sample size. In addition, the OCTA technology has several limitations. First, our instrument's software cannot resolve projection artifacts. Hwang et al.³² reported that projection-resolved OCTA enables evaluation of three retinal vascular plexus images, which may provide a better understanding compared to evaluation of two retinal plexus images. Second, to obtain accurate flow maps in OCTA, patients must have good fixation to prevent significant motion artifacts. Some patients with BRVO have poor vision, which limits the usefulness of OCTA, and were excluded from the analysis. Despite these limitations, this study objectively analyzed the interlayer perfusion status and for the first time provided a method to quantify the gap in the perfusion status between the SCP and the DCP. 

In conclusion, we found that the isolated vessels were important findings in persistent macular edema in BRVO. The vessels were well outlined by OCTA. The gaps in capillary loss between the SCP and DCP might be useful markers for predicting the prognosis in patients with macular edema. This new insight might help facilitate the understanding of the mechanism of persistent macular edema in BRVO and contribute to the development of novel treatments for this disease. 

Acknowledgments

Disclosure: K. Tsuboi, None; Y. Ishida, None; M. Kamei, None