Clinical characteristics of branch retinal vein occlusion with increased retinal haemorrhage during treatment for macular oedema

This study was performed to evaluate clinical characteristics of branch retinal vein occlusion (BRVO) patients with increased retinal haemorrhage during macular oedema (MO) treatment. Patients were divided into increased and non-increased retinal haemorrhage groups. The former group was sub-divided based on the degree of increase: < 50% or ≥ 50%. Baseline characteristics, clinical data, and best vision achieved before and after increased retinal haemorrhage were assessed. Sixty-eight eyes of 68 patients were included. Forty-six eyes were non-increased, 11 eyes experienced < 50% increase in retinal haemorrhage, and 11 eyes had ≥ 50% increase. Ischaemic BRVO was related to increased haemorrhage based on the multivariate analysis. The ≥ 50% increase group exhibited higher baseline central subfield macular thickness (CSMT), poorer baseline best corrected visual acuity (BCVA), and longer mean periods between the final intravitreal injections and the time increased retinal haemorrhages were observed, compared to the < 50% group. The best vision achieved before and after increased haemorrhage was significantly worse in the ≥ 50% group. In conclusion, the ischaemic type of BRVO is related to increased retinal haemorrhage during MO treatment, and a ≥ 50% increase in haemorrhages is associated with higher CSMT and poorer BCVA at baseline, with poor visual gain.

www.nature.com/scientificreports/ BRVO patients who were followed up for at least 6 months after MO treatment from January 2016 to December 2018. MO was defined as central subfield macular thickness (CSMT) ≥ 300 µm. The exclusion criteria were as follows: severe media opacities that prevented the interpretation of images, a history of pars plana vitrectomy or cataract surgery within the 6 months preceding the start of the trial, the presence of other retinal diseases that could cause decreased BCVA or MO, such as epiretinal membrane, a grade exceeding severe non-proliferative diabetic retinopathy, rhegmatogenous retinal detachment, age-related macular degeneration, and uveitis. Patients who did not show increase in retinal haemorrhages during the follow-up period and who were followed up for at least 6 months were allotted to the non-increased group. Increased retinal haemorrhage was defined as an increase, detectable by the human eye, based on the fundus colour images; patients who were followed up for at least 6 months after the detection of increased haemorrhage were allotted to the increased group. The increase in retinal haemorrhages was measured using the National Institutes of Health's image-analysis software (ImageJ 1.44p; National Institutes of Health, Bethesda, MD, USA) 10,11 . Colour images were obtained at each of three timepoints (timepoint 1: initial baseline; timepoint 2: a visit prior to the increase in retinal haemorrhage; timepoint 3: the day the increased retinal haemorrhage was detected). The images were converted into grayscale and were thresholded (local Otsu with a radius of 18 pixels) using ImageJ. The perimeter of the retinal haemorrhagewas drawn by a clinician (YHL) using the polygon selection tool based on the fundus colour image, and the number of pixels in this area with a value of zero was counted (Fig. 1). The ratio of the percent increase in the count of zero value pixels at timepoint 3 (count of zero value pixels at timepoint 3 minus count at timepoint 2) to the percent decrease at timepoint 2 (count at timepoint 1 minus count at timepoint 2) was determined. The increased retinal haemorrhage group was further divided into two groups (< 50%, Fig. 2, and ≥ 50% increase, Fig. 3) for subgroup analysis. clinical data collection. At the initial visit, demographic characteristics, including the age, sex, laterality, presence of hypertension, diabetes, cerebrovascular attack, and hyperlipidaemia were assessed. Also, the baseline logMAR BCVA, intraocular pressure (IOP) measured by a non-contact tonometer (Canon TX-20, Canon Inc, Kanagawa, Japan), and spherical equivalent values were recorded. The location of the BRVO was categorized as either superior or inferior. The major type of BRVO and the macular type of BRVO were categorized based on the site of where the retinal vascular occlusion was seen in the fundus colour images; the major type was defined as an occlusion of a temporal arcade vein or branch extending to the peripheral retina beyond the retinal vascular arcades; the macular type was defined as an occlusion confined between the superior and inferior retinal temporal vascular arcades 12,13 . The baseline CSMT was defined as the average thickness at the centre of a 1 mm diameter circle of the Early Treatment Diabetic Retinopathy Study grid, while subfoveal choroidal thickness (SCT) was defined as the thickness from the outer line of the retinal pigment epithelium to the chorioscleral interface. The CSMT and SCT were measured using OCT (DRI OCT-1; Topcon, Tokyo, Japan), and all patients underwent fluorescein angiography (FA;HRA-2; Heidelberg Engineering, Heiderberg, Germany) to identify the occlusion site. The occlusion distance was defined as the distance (µm) between the optic disc margin and the occlusion site. Ischaemic BRVO was defined as a nonperfusion exceeding a five-disc diameter based on FA. The treatment of MO secondary to BRVO included an intravitreal injection of either bevacizumab (1.25 mg/0.05 mL) or triamcinolone (4.0 mg/0.05 mL). The risk of glaucoma and the possibility of systemic side effects were considered in the choice of the intravitreal injection and performed in the pro re nata regimen. The patients' follow-up interval was between 1 and 3 months. Statistical methods. Data were calculated as means ± standard deviation (SD), or n (%; e.g., number of eyes). Statistics were performed using Statistical Package for the Social Sciences (SPSS) version 12.0 (IBM, Chicago, IL, USA). The between-group differences in age, BCVA, IOP, CSMT, SCT, and occlusion distance were compared between groups using independent t tests. The categorical variables, such as sex, affected eye, the presence of systemic diseases, the location of the BRVO, the type of BRVO (ischaemic versus non-ischaemic, and major type versus macular type), and baseline treatment (bevacizumab versus triamcinolone) were compared using chi square tests. Binary logistic regression analysis was used to identify factors associated with increased retinal haemorrhages. The subgroup analysis to compare differences between the those with increased retinal haemorrhages of < 50% and ≥ 50% was performed using Mann-Whitney U tests for continuous variables and www.nature.com/scientificreports/ Fisher's exact test for categorical variables. The comparison of best vision achieved before and after an increase in retinal haemorrhages was performed using Wilcoxon signed-rank tests.

Results
Demographic characteristics. A total of 68 eyes of 68 patients were included in this study. The number of increased retinal hemorrhages was noted in 22 eyes of 22 patients. Among the increasedhaemorrhage group, an increase of ≥ 50% was visible in 11 eyes of 11 patients. Demographic data of patients displaying an increase in retinal haemorrhages is shown in Table 1. The mean age was 68 ± 10 years in the increased haemorrhage group and 61 ± 13 years in the non-increased group; there was a significant difference in age between the two groups (p = 0.032). The affected eye, sex, and systemic diseases such as hypertension, diabetes mellitus, hyperlipidaemia, and cerebrovascular accident did not differ significantly between the groups (p > 0.05).  Table 2. A longer total follow-up period was observed in the increased group (14 ± 3 months); however, there was an insignificant difference when compared with the non-increased group (13 ± 3 months, p = 0.137). The total number of intravitreal injections showed an insignifanct difference between the increased and non-increased groups (4 ± 1 versus 3 ± 1, p = 0.093, respectively). The baseline BCVA was not significantly different (increased group versus non-increased group, 0.71 ± 0.25 logMAR versus 0.60 ± 0.33 logMAR, p = 0.158). The difference in the mean IOP was not significant between the increased group (14.36 ± 2.63 mmHg) and the non-increased group (

Analysis of the factors associated with increased retinal haemorrhages.
In binary logistic regression analysis, the ischaemic type of BRVO was the only factor that was significantly associated with increased retinal haemorrhages during the treatment of MO (p = 0.007; Table 3).
Subgroup analysis based on the increased amount of retinal haemorrhages. The demographics of increased retinal haemorrhages between the < 50% and the ≥ 50% increase groups are shown in Table 4. There were no significant differences between the two groups (p > 0.05). The comparison of clinical characteristics between the two groups is shown in Table 5. The average percent increases in the amount of retinal haemorrhages were 10.18 ± 7.32 in the < 50% group and 73.76 ± 13.37 in the ≥ 50% group (p < 0.001). The interval between the baseline and the time the increased retinal haemorrhages were detected showed insignificant difference between www.nature.com/scientificreports/ the < 50% group (5 ± 2 months) and the ≥ 50% group (7 ± 2 months, p = 0.088). The mean period between the last intravitreal injection and the time the increased retinal haemorrhages were detected was 59.27 ± 10.53 days in the < 50% group and 103.27 ± 26.85 days in the ≥ 50% (p < 0.001). The baseline BCVA was significantly worse in the ≥ 50% group (0.88 ± 0.15 logMAR) than in the < 50% group (0.60 ± 0.23 logMAR, p = 0.015). Moreover, a thicker baseline CSMT was observed in the ≥ 50% group compared to the < 50% group (561.46 ± 69.89 versus 461.82 ± 106.17 μm, p = 0.013). As the baseline CSMT showed significant difference between the two groups, the ratio of the CSMT at the date the increased retinal haemorrhages were observed and the baseline CSMT was also assessed; the ratio was also significantly higher in the ≥ 50% group (78.03 ± 13.30 versus 66.02 ± 13.57, p = 0.019).
Other characteristics such as the superior segment involvement, ischaemic type, macular type, initial intravitreal bevacizumab injection, baseline IOP, baseline SCT and ratio of SCT (after increased retinal haemorrhages was  www.nature.com/scientificreports/ observed) to baseline SCT, baseline spherical equivalent, and occlusion distance from the optic disc margin did not significantly differ between groups (p > 0.05, Table 5).The best vision achieved before and after the increased retinal haemorrhages were observed was significantly worse in the ≥ 50% group (0.24 ± 0.19 versus 0.39 ± 0.21 logMAR, p = 0.007, Table 6); this was not observed in the < 50% group (0.29 ± 0.19 versus 0.35 ± 0.18 logMAR, p = 0.145, Table 6).

Discussion
The current study retrospectively analysed the baseline characteristics of BRVO patients treated for MO, and subgroup analysis was performed based on the amount by which the retinal haemorrhages increased. At baseline, the increased retinal haemorrhages group was older than the non-increased retinal haemorrhages group. Young et al. 14 reported that the veins stiffen with age, which could lead to higher intraluminal pressureassociated with an increasedrisk of retinal haemorrhage in BRVO patients. The higher baseline CSMT with the higher proportion of ischaemic type BRVO seen in the increased retinal haemorrhages group implies that the severity at the time of presentation is related to increased retinal haemorrhages during MO treatment. A shorter occlusion distance from the optic disc margin was also found in the increased retinal haemorrhage group. However, the ischaemic type of BRVO was the only factor that was associated with increased retinal haemorrhage in the multivariate analysis.
The subgroup analysis revealed that the ≥ 50% group exhibited significantly higher baseline CSMT and poorer baseline BCVA values. Moreover, the comparison of the best vision achieved before and after an increase in retinal haemorrhage was significantly worse in the ≥ 50% group. Bayat et al. 15 . reported that the baseline BCVA is a predictor of visual gain, while baseline CSMT is a predictor of anatomical gain in BRVO patients treated with intravitreal ranibizumab injection for MO. In addition, our subgroup analysis highlighted that the poor visual prognosis in those with low baseline BCVA and CSMT may have resulted from a ≥ 50% increase in retinal haemorrhages during the MO treatment. The shorter mean period between the last intravitreal injection and the increased retinal haemorrhages was observed in the < 50% group. It is known that the single intravitreal bevacizumab injection for BRVO induced ME lasts for 6 weeks to 2 months 16,17 . Interestingly, most patients were treated with intravitreal bevacizumab injection in this study, and the mean period was 59.27 ± 10.53 days in the < 50% group. The lower CSMT ratio (date of increased retinal haemorrhages compared to baseline) was also observed in this group. One previous study hypothesized that bevacizumab reduces VEGF-mediated permeability and steroids reduce inflammation and limit extravasation from the blood vessels 18 . From these observations, we suggest that the < 50% increased retinal haemorrhage resulted from an increase in VEGF-mediated permeability as the effect of intravitreal bevacizumab decreases. On the contrary, the longer period between the last intravitreal injection and the increased retinal haemorrhages with the higher CSMT ratio (date of increased retinal haemorrhages compared to baseline) might have resulted from other causes. One previous report demonstrated a rebound effect after intravitreal bevacizumab injection and a recurrence of BRVO with massive retinal haemorrhages 19 . As bevacizumab cannot prevent the recurrent occlusion of BRVO, we believe that the recurrent occlusion might be related to the ≥ 50% increase in retinal haemorrhages. Previous studies reported a BRVO recurrence rate of 3.3-14.4% 20,21 . In our study, however, 16.2% patients showed ≥ 50% increased retinal haemorrhages. This is because we evaluated MO-treated BRVO patients and assessed retinal haemorrhage changes that were limited to the primary area of BRVO. We hypothesis that the small increase in retinal haemorrhages was related to the increase in VEGF-mediated vascular permeability, while the large increase in retinal haemorrhages was related to the recurrent occlusion in the first BRVO. Further studies with FA at the time of increased retinal haemorrhages will be beneficial.
The current study had some limitations. First, the retrospective design may have resulted in a patient selection bias. Second, the increased retinal haemorrhages were first detected by the human eye, so patients with minimal increases of retinal haemorrhages may have been overlooked. Third, this study included only a small number of patients in subgroup analysis. Fourth, the colour images of the fundus obtained from the OCT did not provide wide-field images. Retinal haemorrhages beyond the colour images were not measured in this study. One previous study demonstrated that most BRVO (particularly major BRVO) were ischaeminc in nature 22 . Therefore, further studies measuring the quantitative amount of ischaemia in the wide-field-based 7-standard field images might provide more information. Finally, we measured the amount of retinal haemorrhages within a given 2-dimensional area, so any changes in the third dimensions would have been overlooked and the actual amount of retinal haemorrhages may have been underestimated. However, we believe that this is the first study to assess the clinical characteristics of increased retinal haemorrhages during MO treatment in cases of BRVO.
In conclusion, the ischaemic type of BRVO is related to increased retinal haemorrhages during the treatment of MO in those with BRVO. Among the patients who experienced an increase in retinal haemorrhages, Table 6. Comparison of best vision achieved between before and after increase in retinal haemorrhages (≥ 50% versus < 50%) in branch retinal vein occlusion patients. *Statistically significant as calculated using Wilcoxon signed rank test.

Parameters
Before increase in retinal haemorrhages After increase in retinal haemorrhages P value www.nature.com/scientificreports/ an increase of ≥ 50% was associated with higher CSMT and poorer BCVA values at baseline, as well as poorer visual gain.

Data availability
The datasets generated and/or analysed during the present study are available from the corresponding author on reasonable request.
Received: 1 April 2020; Accepted: 5 June 2020 Scientific RepoRtS | (2020) 10:10272 | https://doi.org/10.1038/s41598-020-67395-7 www.nature.com/scientificreports/ Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creat iveco mmons .org/licen ses/by/4.0/.