OCT angiography features associated with macular edema recurrence after intravitreal bevacizumab treatment in branch retinal vein occlusion

We aimed to evaluate the relationship between the capillary abnormalities including nonperfusion area (NPA) in optical coherence tomography angiography (OCTA) images and the recurrence of macular edema (ME) secondary to branch retinal vein occlusion (BRVO) after intravitreal injection of anti-vascular endothelial growth factor (anti-VEGF; bevacizumab). The records of 40 patients who underwent intravitreal bevacizumab injection for ME secondary to BRVO and had at least six months of follow-up were reviewed. Central retinal thickness (CRT; μm) and macular edema type were evaluated prior to treatment. After ME resolution, nonperfusion areas in the 1 mm (NPA1) and 1–3 mm (NPA3) zones on the Early Treatment Diabetic Retinopathy Study (ETDRS) circle within the superficial capillary plexus (SCP) and deep capillary plexus (DCP) were measured using OCTA images. Furthermore, other microvascular abnormalities in the both SCP and DCP were compared between groups. ME recurred in 25 of 40 (62.5%) eyes. The NPA1 of the SCP and DCP (p = 0.002, 0.004, respectively), NPA3 of the SCP and DCP (p = 0.002, 0.008, respectively), and initial CRT (p = 0.022) differed significantly between eyes with and without ME recurrence. In multivariate logistic regression analyses, the NPA1 of the DCP (OR: 344.718; p = 0.029) and NPA3 of the SCP (OR: 4.072; p = 0.018) were significantly associated with ME recurrence. Other microvascular abnormalities were not significantly different between two groups. The central NPA and parafoveal NPA of the SCP in OCTA images correlated strongly with ME recurrence in BRVO patients after intravitreal anti-VEGF injection.

With regard to the recurrence of ME in BRVO cases, our previous study of FA images showed that the degree of macular ischemia was associated with ME recurrence after intravitreal anti-VEGF injection 16 . Overall retinal vascular changes, such as macular capillary NPAs in BRVO, have been visualized by FA 17 , but specific evaluation of the vasculature at each capillary level is not possible with FA. Several studies in humans and animal models have shown that major arterioles and especially major venules can independently connect to the deep capillary plexus (DCP) without first communicating with the superficial capillary plexus (SCP) and also that the DCP could have a primarily venous role [18][19][20][21][22][23][24] . A recently developed optical coherence tomography angiography (OCTA) technique can noninvasively visualize and differentiate the SCP and DCP, allowing for analyses of the microvascular abnormalities of each retinal capillary plexus 25,26 . Some reports in which OCTA was performed in eyes with BRVO have concluded that retinal vascular changes, such as NPAs and microaneurysms, develop more frequently in the DCP than in the SCP [27][28][29] .
In the present study, we measured the vascular NPA of each capillary plexus using OCTA images of BRVO patients with ME. We also evaluated the ME type in optical coherence tomography (OCT) images before treatment and microvascular abnormalities in OCTA images after ME resolution. We evaluated which OCTA features are associated with ME recurrence after anti-VEGF treatment.

Materials and Methods
patients. The institutional review board (IRB) of Korea University Ansan Hospital approved this study, waiving the requirement for informed consent for study participation. All research protocols and methods of data collection adhered to the tenets of the Declaration of Helsinki.
This retrospective study included patients who underwent off-label intravitreal anti-VEGF (Avasin ® , bevacizumab; 1.25 mg in 0.05 ml) injection for ME secondary to unilateral BRVO at Korea University Ansan Hospital between January 2016 and March 2018. The study inclusion criteria were as follows: (1) unilateral BRVO onset within three months; (2) baseline examinations performed within the two-week period preceding the initial intravitreal anti-VEGF injection; (3) ME resolution after the first or second intravitreal anti-VEGF injection; (4) OCTA imaging at one or two months after ME resolution showing no major blocked signals associated with retinal hemorrhage; (5) OCT images with a signal-to-noise ratio of 0.6 or greater; and (6) a central retinal thickness (CRT) of ≥300 μm as measured by OCT at the first evaluation. The exclusion criteria were as follows: (1) a follow-up period of less than six months after intravitreal anti-VEGF injection; (2) the presence of other retinal disease, such as diabetic retinopathy or epiretinal membrane; (3) a history of other intraocular treatments, such as intravitreal triamcinolone injection or focal/grid laser photocoagulation; (4) CRVO or hemi-CRVO.
All patients underwent complete ophthalmic examinations prior to injection, including best-corrected visual acuity (BCVA) measurements (expressed as log MAR units), slit-lamp examination, dilated fundus examination, ultra-wide-field color fundus photography (WFP), FA, and OCT at the first visit (baseline assessment). All follow-up examinations, including OCT imaging, were repeated at intervals of 2-3 months after the 1-month follow-up visit. All patients received one initial intravitreal injection (IVR). And patients with persistent ME at a subsequent follow-up visit, received an additional intravitreal anti-VEGF injection 30 . If bevacizumab was injected again, the follow-up schedule like after 1st injection was reimplemented. BCVA measurement, slit-lamp examination, dilated fundus examination, WFP, OCT, and OCTA were performed one month after the initial intravitreal anti-VEGF injection. The development of ME was defined as a CRT ≥300 μm. Resolution of ME was defined as a CRT <300 μm after intravitreal anti-VEGF injection. Likewise, the recurrence of ME was defined as a CRT ≥300 μm after the resolution of ME or a CRT of <300 μm with a new retinal cyst or subretinal fluid development within six months of one or two intravitreal anti-VEGF injections. optical coherence tomography and optical coherence tomography angiography. OCT and OCTA images were obtained using a Spectralis OCT imaging device (version SP 6.7.21.0, 870 nm wavelength; Heidelberg Engineering, Heidelberg, Germany) at Ansan Hospital. All of the OCTA examinations were performed using a 3.0 mm by 3.0 mm volume scan pattern centered on the fovea according to the analysis protocol (FS-ADA) and variables. image analysis. The types of ME in the initial OCT images were evaluated and classified as follows 16 . Serous retinal detachment was defined as retinal thickening with subretinal fluid, and cystoid ME was defined as retinal thickening with intraretinal cystic areas of low reflectivity. The third type was mixed serous retinal detachment and cystoid ME, and the fourth type was sponge-like ME, defined as diffuse retinal thickening with retinal layers having a sponge-like appearance.
Using the default automated segmentation boundaries in the built-in software, separate flow maps of the SCP and DCP were obtained for each eye. The location of the segmentation line between the SCP and DCP in our module was as indicated in Table 1. The fovea-centered OCTA images were manually overlaid with the ETDRS circle. The foveal avascular zone (FAZ) was defined as a lesion devoid of retinal vessels within the boundary of the innermost the parafoveal. In our BRVO cases, some cases were an intact oval-shaped FAZ, but the others was not with capillary dropouts and anastomotic arcade disruptions ( Fig. 1) 31 . The NPA as identified in OCTA images was defined as the all dropout area of the retinal capillary beds with any dark or gray areas 32 , including the FAZ (Figs 1,  2). The two-layered capillaries at the border of the avascular zone ultimately join into a single-layered ring 20 , and the capillary plexus in the fovea was readily detected in the DCP using the Spectralis OCT device (Fig. 3). NPAs in the 1-mm zone (NPA1, including the FAZ) were measured at the CP by drawing the area contour point by point 33 using ImageJ software (1.51J8 version, National Institutes of Health, Bethesda, MD, USA); NPAs in the 1 mm to 3 mm (NPA3) zone (excluding the 1-mm zone) were measured within both the SCP and DCP by the same method ( Fig. 2) 31,34 . Measurement of the NPAs in the SCP and DCP was performed by two independent retina specialists (K.E.C., S.W.K.) who were both blinded to patient information. Other microvascular abnormalities (capillary telangiectasias, collateral vessels (venovenous drainage), and microaneurysms) were also compared between those with and without ME recurrence 27 , and OCT images before treatment were compared with OCTA images after ME resolution in all cases. Statistical analysis. Continuous variables are presented as mean ± SD and were analyzed using the Mann-Whitney test or Spearman correlation analysis. The other factors were compared between those with and without ME recurrence using the Pearson chi-square test, Fisher's exact test, or Wilcoxon signed rank test, as appropriate. Multivariate logistic analyses using the significant parameters identified in the univariate analyses as the independent variables were performed to identify factors associated with ME recurrence. Statistical analyses were performed using SPSS version 21.0.0.0 (SPSS, Chicago, IL, USA) or MedCalc V. 12    (b,c) In this case, the superior border of the foveal avascular zone (FAZ) was vague due to vascular dropout (blue arrow). Nonperfusion area 1 (NPA1) of (b) the superficial capillary plexus (SCP) and (c) the deep capillary plexus (DCP) was defined as the dropout-area in the 1-mm zone. The NPA1 of each capillary plexus was delineated as the gray area in the 1-mm zone.
bvba, Ostend, Belgium). All statistics were two-tailed, and p values of <0.05 were considered statistically significant. The interclass correlation coefficient for each NPA measurement was calculated, and those mean values were used for the statistical analyses.

Results patient characteristics.
A total of 40 eyes were enrolled in this study. ME did not recur within six months after the initial intravitreal anti-VEGF injections (one or two) in 15 of the 40 eyes (37.5%), whereas in 25 of 40 eyes (62.5%) it did recur. 14 of the 40 patients were male, and the other 26 patients were female. The mean age of the patients was 62.15 ± 12.66 years. Mean disease duration was 1.09 ± 1.02 months, and mean follow-up duration was 12.33 ± 5.99 months. The mean number of injections before recurrence within 6 months after ME resolution was 1.38 ± 0.49 (25 patients received one injection, and 15 patients received two injections). At the initial visit, BCVA (logMAR) was 0.58 ± 0.38, and at the last visit, it was 0.25 ± 0.24. There was no significant difference in the initial BCVA assessment between eyes with ME recurrence and those without ME recurrence (log MAR: 0.69 ± 0.44 and 0.49 ± 0.47, respectively; p = 0.376). Other characteristics of patients were also not statistically different between two groups ( There was no case of only serous detachment type without cyst, and the type of ME before anti-VEGF injection did not differ significantly between the two groups (p = 0.495). The interclass correlation coefficients for the NPA measurements were >0.95. In each individual, there was no significant difference between the mean NPA3 of the DCP and that of the SCP (1.415 ± 0.708 mm 2 , 1.489 ± 0.797 mm 2 , respectively, p = 0.137 by Wilcoxon signed rank test). The mean NPA1 of the SCP was 0.441 ± 0.129 mm 2 in the nonrecurrence group and 0.610 ± 0.176 mm 2 in the recurrence group (p = 0.002). The mean NPA1 of the DCP was 0.473 ± 0.128 mm 2 in the nonrecurrence group and 0.629 ± 0.169 mm 2 in the recurrence group (p = 0.004). The mean NPA3 of the SCP was 1.005 ± 0.729 mm 2 in the nonrecurrence group and 1.779 ± 0.701 mm 2 in   In the analysis of all cases, there was no significant correlation between ME type on the OCT images before treatment and any of the vascular abnormalities on the OCTA images after ME resolution. In addition, there was no significant correlation between each vascular abnormality after ME resolution and the CRT at each follow-up, as determined by the Mann-Whitney U test. The only significant correlation was between the NPA3 of the DCP in the OCTA images after ME resolution and the initial CRT in the OCT images (Rho = 0.353, p = 0.026 by Spearman correlation analysis).

Discussion
In the present study, we focused on the relationship between each NPA in the OCTA images and the recurrence of ME after intravitreal anti-VEGF injection in eyes with BRVO. Other possible risk factors such as microaneurysms in OCTA images did not differ significantly between the ME recurrence and nonrecurrence groups. The recurrence rate of ME within six months of the initial intravitreal anti-VEGF injection was 62.5% (25 of 40 eyes). We were unable to adequately evaluate the FAZ in the SCP using the Spectralis OCT device (Heidelberg) (Fig. 1). Single-layered vessels adherent to the FAZ were identified in our OCTA machine by comparing those images with images of the corresponding inner retinal capillary plexus layer (range: internal limiting membrane to basement membrane) (Fig. 3). In all cases, the DCP presented as a single layer of vessels near the FAZ, but the SCP in 6 cases (15%) presented as a partially intact single layer of vessels.  www.nature.com/scientificreports www.nature.com/scientificreports/ ME recurrence was correlated with the extent of the NPA1 (including the FAZ) of the DCP, NPA1 of the SCP, NPA3 of the DCP, NPA3 of the SCP, the initial CRT, and three-month CRT in both the Mann-Whitney test and univariate logistic regression analyses. In the multivariate logistic regression analysis, the extent of the NPA3 in the SCP and the NPA1 in the DCP were significantly associated with ME recurrence in BRVO patients. In our study, the extent of the NPA1 (including the FAZ) appeared to be consistent with the central retinal nonperfusion status, and the extent of the NPA3 was consistent with the parafoveal nonperfusion status. However, central hypoperfusion and parafoveal hypoperfusion were prominent in both the SCP and DCP in all cases, and there was no significant difference between the SCP and DCP. The NPA3s of the SCP and DCP were strongly correlated (Rho = 0.764, p < 0.001), and the NPA1s of the SCP and DCP were also strongly correlated (Rho = 0.876, p < 0.001).
Previous FA studies have demonstrated a relationship between ME and retinal nonperfusion in eyes with BRVO 35,36 . In addition, ME recurrence after intravitreal anti-VEGF injection in eyes with BRVO is associated with the initial extent of macular capillary nonperfusion as assessed by FA and foveal thickness as measured by OCT 16 . Recently, several OCTA studies have also suggested that, although both the superficial and deep vascular networks are severely affected in BRVO, microvascular abnormalities, including capillary nonperfusion, are more common in the DCP than in the SCP 27,28,32,37,38 . In previous OCTA studies, ME caused by retinal vein occlusion has been shown to occur in association with an absence of blood flow in the DCP or the difference in retinal capillary loss between the SCP and DCP 39,40 . On the other hand, one study showed that retinal sensitivity over the NPAs on both the SCP and DCP decreased, though the decrease was larger in the SCP 41 . Furthermore, at both CRVO and BRVO, changes in vascular density and complexity in the SCP were more severe than in the DCP 42 .
The central retinal NPA (NPA1, including the FAZ) was also a prominent factor correlated with ME recurrence in our BRVO cases. Interestingly, some ME cases with large parafoveal NPAs (NPA3) did not recur, whereas some cases with moderate parafoveal NPAs did recur. Notably, in the nonrecurrence group, cases with large parafoveal NPAs had small central retinal NPAs, whereas in the recurrence group, cases with moderate parafoveal NPAs had large central retinal NPAs, which were found close in size of the 1-mm zone (Fig. 4).
After treatment of retinal vein occlusion with anti-VEGF agents, both short-term and long-term studies have found no significant differences in the FAZ area between CRVO and BRVO patients 43,44 . Adhi et al. showed that the FAZ area was larger in eyes with RVO than in their fellow eyes and controls, and in the fellow eyes of RVO patients when compared to controls 32 , Rispoli et al. reported similar results 37 . Yoo et al. showed that the degree of macular ischemia (not including the FAZ) as assessed by FA was associated with ME recurrence after intravitreal anti-VEGF injection 16 .
Our study had some limitations. First, all data were gathered retrospectively from one clinic using a particular kind of OCTA device. There are quite prominent differences among different OCTA devices with respect to tracing the superficial and deep capillary networks within the macular 1-mm zone (Table 1). Different OCT devices use different techniques to differentiate blood vessels by depicting changes in the OCT signal induced by moving blood cells 45 . In our OCTA study using the Spectralis (Heidelberg) imaging system, the FAZ size in the SCP was larger than in the DCP, and vascular complexes adherent to the FAZ were readily detected only within the DCP. Previous studies have generally presented OCTA analyses in BRVO using other OCA devices, such the RTVue XR 100 (Optovue) 9,25,[27][28][29][30][31][32]34,[37][38][39][40]42,44,[46][47][48][49][50] , and AngioPlex (Zeiss) 33 , and found that the FAZ of the SCP was smaller than the FAZ of the DCP. The differences between the settings used to identify the interface between the SCP and DCP provided by each OCTA software package could affect the distinction of the en-face image (Table 1). Furthermore, although we tried to minimize the effect of artifacts by excluding poor-quality scans and OCTA images with significant artefacts 46 , image quality could have affected the vascular density measurements.
Second, we did not assess preoperative OCTA information, which is important to consider. Although some studies showed no significant differences in the OCTA findings of BRVO with ME after short-term anti-VEGF treatment 40,43,44 , other studies showed that the foveal microstructure could be altered following anti-VEGF or steroid treatment 50,51 . The obstructed capillaries might not be completely re-perfused even after successful treatment of the ME in cases where deep capillary ischemia persists 32,52 . Nevertheless, we did not routinely evaluate OCTA images before the intravitreal injections, because retinal hemorrhage following BRVO diminished image quality. Valerie Mane et al. showed that cystoid spaces were located within capillary dropout areas in chronic diabetic macular edema (DME). The en-face of OCTA could detect cyst lesions in the DME cases 52 , but also in the BRVO cases 29 www.nature.com/scientificreports www.nature.com/scientificreports/ is more difficult than in DME cases because intraretinal hemorrhage following BRVO obscured the OCTA signal from the retinal layers involved in and beneath the hemorrhage 53,54 . In those situations, detecting the flow signal in OCT and OCTA images before treatment was also difficult. Fortunately, Kia Tarassoly et al. showed that vessel density and FAZ did not differ significantly between groups with and without cystoid change 55 . The impact of microvascular abnormality findings in the initial OCTA images on the risk of recurrence of BRVO-related ME remain to be clarified.
Third, the number of injections differed between the recurrence and nonrecurrence groups, although that difference was not statistically significant (p = 0.376 by Mann-Whitney U test, 0.898 by multivariate regression analysis). The treatment protocol of our study varies from the established protocol of 3 monthly anti-VEGF injections followed by PRN dosing. A further study with protocol of 3 monthly anti-VEGF injections for both groups could support our results.
In conclusion, our study demonstrates the utility of OCTA in quantifying vascular abnormalities, especially vascular nonperfusion in each of two capillary plexuses in BRVO eyes with ME recurrence. We found that nonperfusion status was strongly correlated with ME recurrence, especially in central foveal and parafoveal nonperfusion. Given its noninvasive nature and detailed depth-resolved vascular mapping ability, OCTA could become a preferred tool for the diagnosis and longitudinal evaluation of BRVO and other vascular occlusive diseases. Future studies comparing NPAs among various OCTA images and evaluating retinal vascular changes over time in response to therapy might be needed.

Data Availability
The datasets generated and analyzed during the current study are available from the corresponding author upon reasonable request. The NPA3 (yellow area) and NPA1 (gray area plus blue area) of the DCP in this recurrence case were 0.571 mm 2 and 0.531 mm 2 , respectively. (c) The NPA3 (yellow area) and NPA1 (gray area plus blue area) of the SCP in this nonrecurrence case were 1.585 mm 2 and 0.359 mm 2 , respectively. (d) The NPA3 (yellow area) and NPA1 (gray area plus blue area) of the DCP in this nonrecurrence case were 1.435 mm 2 and 0.281 mm 2 , respectively.