Factors associated with extremely poor visual outcomes in patients with central retinal vein occlusion

Here, we examined prognostic factors for extremely poor visual outcomes in patients with central retinal vein occlusion (CRVO) in actual practices. We included 150 consecutive eyes with treatment-naïve acute CRVO from four different facilities and observed them for over 24 months. Macular edema (ME) was treated with one or three monthly anti-vascular endothelial growth factor injections (1 or 3 + pro re nata). According to the final Snellen visual acuity (VA), we divided the patients into very poor VA (< 20/200) and control (≥ 20/200) groups and examined risk factors for poor final visual outcomes. The baseline Snellen VA was hand motion to 20/13. The mean number of anti-VEGF injections for ME was 5.3 ± 3.7 during the follow-up period. In total, 49 (32.7%) patients exhibited a very poor final VA; this group comprised significantly older patients with a significantly poorer baseline VA (P < 0.01 for both) than the control group. Comorbid internal carotid artery disease and diabetic retinopathy were significantly associated with a poor final VA. In actual clinical practice, visual outcomes may be extremely poor despite ME treatment in certain patients with CRVO, with advanced age, poor baseline VA, and comorbid internal carotid artery disease and diabetic retinopathy being significant risk factors.

. Characteristics of included 150 patients with central retinal vein occlusion. Data are expressed as mean ± standard deviation unless otherwise indicated. Patients with diabetes were included if they showed no diabetic retinopathy or mild to moderate nonproliferative diabetic retinopathy without macular edema.  www.nature.com/scientificreports/   www.nature.com/scientificreports/ initiation of the anti-VEGF therapy compared to that at baseline (P < 0.001 for all measuring points). Figure 3 is a scatter plot showing the logMAR BCVA at baseline and final examinations for each patient. In both groups, CRT at 6, 12, and 24 months was significantly lower than that at baseline (P < 0.001 for all) in the both two groups, with no significant between-group difference at any time point. The defect length in the EZ band was significantly greater in the very poor VA group than in the control group throughout the observational period (P < 0.001 for all).
In total, 22 (14.7%) eyes developed neovascular complications during the observational period; these included six eyes with rubeosis without IOP elevation and 16 eyes with NVG in the very poor VA group (Table 3). Vitreous hemorrhage occurred in 10 eyes, one of which belonged to the control group ( Table 3). The mean logMAR BCVA for eyes with NVG was 0.70 ± 0.60 (Snellen BCVA, hand motion to 20/25) at baseline and 1.50 ± 0.92 (Snellen BCVA, no light perception to 20/250) at the final examination. In the very poor VA group, the final BCVA was worse than counting fingers for six eyes, all of which had developed NVG.

Discussion
In the current study, the improvement in BCVA after the initial anti-VEGF injection(s) was transient in eyes with a final Snellen BCVA of < 20/200 (the difference was not statistically significant). Advanced age, poor baseline BCVA, and comorbid internal carotid artery disease and diabetic retinopathy were risk factors for extremely poor final visual outcomes. Moreover, all patients with neovascular complications except one with vitreous hemorrhage belonged to the very poor VA group.  www.nature.com/scientificreports/ In the present study, approximately a third of patients exhibited a final Snellen BCVA of < 20/200; this proportion may be considerably greater than that in previous clinical trials [9][10][11][12][13][14][15] . Some epidemiological studies and metaanalyses showed that glaucoma 18 and/or diabetes 19,20 are often accompanied by CRVO. Therefore, we included CRVO patients with diabetes and/or diabetic retinopathy or glaucoma as long as they met a certain criterion; this may be a reason for the high rate of patients with very poor visual outcomes in this study. We speculate that such severe cases were excluded from the previous clinical trials [9][10][11][12][13][14][15] .
The number of anti-VEGF injections required in the present study appeared to be smaller than that in other clinical trials 15,17 . However, CRT showed a significant decrease from baseline in both groups ( Fig. 1; Table 3). Therefore, the anti-VEGF therapies used in this study were considered effective for the suppression of retinal exudative changes to some extent. Retinal thinning due to macular ischemia could also be involved in the decrease in CRT. As previously shown, there may be a significant association between the resolution of retinal exudation and that of retinal nonperfusion 21,22 , because retinal perfusion could be the source of exudation. Inclusion of patients with severe CRVO in this study may be a reason for the smaller number of anti-VEGF injections relative to the number in other clinical trials [9][10][11][12][13][14][15] .
In the present study, neovascular complications, including vitreous hemorrhage, rubeosis, and NVG, occurred in 10 (6.7%), 22 (14.7%), and 16 (10.7%) eyes, respectively. These incidences are higher than those in the GAL-ILEO and CRYSTAL studies, which included patients with not only nonischemic CRVO but also ischemic CRVO [13][14][15] . However, the status of the retinal circulation in patients included in the previous studies was probably not very poor, because the incidence of neovascular complications was lower than that observed during the natural course of ischemic CRVO 13 . The RAVE trial for eyes with severe CRVO also reported that VEGF inhibition by ranibizumab injections could not ameliorate the risk of such complications 17 . In a recent clinical study of eyes with BRVO, 10% of the 58 patients showed retinal or disc neovascularization during the 24-month observational period 23 . Thus, anti-VEGF treatment following a pro re nata regimen for ME, as used in the present study, may only delay the occurrence of neovascular complications.
We found that old age and comorbid internal carotid artery disease and diabetic retinopathy were significantly associated with a final BCVA of < 20/200 (Tables 2, 3, 4); these findings were partly consistent with those of Rong et al. 24 . Changes in the retinal vasculature due to age, internal carotid artery disease associated with arteriosclerosis, and diabetic microangiopathy may accelerate the deterioration of the retinal microcirculation. Retinal microcirculation has an autoregulation system that maintains a relatively constant blood flow 25 . However, if patients with internal carotid artery disease and/or diabetic microangiopathy accompanying CRVO, the autoregulation of the retinal vasculature could be easily disrupted in the retina affected by the CRVO.
This study has several limitations. First, the anti-VEGF agents and treatment protocols were not strictly unified because of the retrospective study design. Second, some of the results may not completely represent the pathology of CRVO because we included CRVO patients with systemic and other ocular disorders. Third, we could not perform OCT angiography for the assessment of macular ischemia in all cases. Therefore, the association between the macular NPA and visual outcomes was not elucidated.
Currently, anti-VEGF agents are commonly available for the treatment of CRVO associated-ME. However, in actual practice, clinicians should be aware that a certain proportion of patients with CRVO (a third in the present study) present with extremely poor final visual outcomes despite anti-VEGF treatment, as shown in the present study. Moreover, advanced age, poor baseline BCVA, and comorbid internal carotid artery disease and diabetic retinopathy are potential risk factors for extremely poor visual outcomes in patients with CRVO. Neovascular complications are not always the cause of the poor final BCVA in such cases. Future studies with larger cohorts should prospectively study the effect of anti-VEGF therapies for not only ME but also NVG in patients with CRVO.

Methods
This retrospective study was approved by the institutional review board of Saneikai Tsukazaki Hospital (Hyogo, Japan), Kyoto University Graduate School of Medicine (Kyoto, Japan), Kagawa University Faculty of Medicine (Kagawa, Japan), and Tokushima University Faculty of Medicine (Tokushima, Japan). The procedures were in accordance with the ethical principles of the Declaration of Helsinki, and written informed consent was obtained from all included patients. We included all patients who presented at these four hospitals with ME due to acute CRVO with a symptom duration of less than 3 months before the initial treatment, between September 2013 and November 2016.
Patients. The inclusion criteria were as follows: treatment-naïve, symptomatic, acute CRVO with retinal hemorrhage spreading over four retinal quadrants, a central retinal thickness (CRT) of > 250 µm at the initial visit, a symptom duration of < 3 months prior to examination, and availability for follow-up for over 24 months. CRVO was diagnosed by retinal specialists at each institute on the basis of fundus examinations and fluorescein angiography (FA) findings of filing delay in the retinal veins in the four quadrants. Eyes with hemi-CRVO, BRVO, or other chorioretinal diseases, including uveitis, were excluded. Eyes that previously received any ocular treatments other than cataract surgery were also excluded. Eyes with glaucoma were included if the intraocular pressure (IOPs) was well controlled and the central visual field was intact. Patients with diabetes were included if they showed no diabetic retinopathy or mild to moderate nonproliferative diabetic retinopathy without ME.

Study examinations.
Each eye was examined every month in the first year and every 1 or 2 months in the second year at each facility. However, the interval between follow-up examinations and the follow-up duration were appropriately modified according to the disease severity.
At each visit, all patients underwent a comprehensive ophthalmic examination, including measurement of the best-corrected visual acuity (BCVA) using the Landolt chart, measurement of IOP, slit-lamp biomicroscopy, and spectral domain optical coherence tomography (OCT; Spectralis HRA + OCT, Heidelberg Engineering, Heidelberg, Germany or RS-3000, Nidek, 3D OCT-1, Topcon, Japan). Fluorescein angiography (Optos 200Tx imaging system, Optos PLC, Dunfermline, United Kingdom) was performed to assess the retinal perfusion status at the initial visit and additionally performed when deemed necessary.

Evaluation of CRT and the defect length in the ellipsoid zone (EZ) band.
For CRT measurement, a macular area in each eye was captured by spectral domain OCT. For volume OCT scanning, a whole retinal thickness map centered on the fovea was created. CRT was defined as the mean distance between the vitreoretinal interface and retinal pigment epithelium within the central subfield of the Early Treatment Diabetic Retinopathy Study grid.
To assess the integrity of the foveal photoreceptor layer, we quantified the disruption in the EZ band within the central 2-mm area on spectral domain OCT images that sectioned vertically and horizontally through the center of the fovea. The signal intensity of the EZ band was measured, and the band was quantified using the plot profile function in ImageJ software (National Institutes of Health, Bethesda, MD, USA) according to previous reports 21,26,27 . The average defect length in the EZ band was calculated from the defect lengths measured on the vertical and horizontal OCT sections.

Intravitreal anti-VEGF injection protocols.
For the treatment of ME and/or serous retinal detachment at the fovea, the patients received intravitreal anti-VEGF injections of ranibizumab (Lucentis; 0.5 mg/0.05 mL, Novartis Pharma AG, Basel, Switzerland) or aflibercept (Eylea; 2.0 mg/0.05 mL, Bayer Pharma AG; Berlin, Germany) according to different regimens in each institute. Twenty-nine eyes received one initial intravitreal anti-VEGF injection, and 121 eyes received three monthly intravitreal anti-VEGF injections. None of the patients received treatment for ME other than ranibizumab and aflibercept, such as bevacizumab injection, grid laser photocoagulation, steroid treatment, and surgical intervention.
After the initial injections in each facility, pro re nata anti-VEGF injections were administered when ME or serous retinal detachment was evident at the fovea on the OCT sections, and the patient's consent could be obtained. The same anti-VEGF agents were used for the initial and subsequent injections, without switching of the agents. Panretinal photocoagulation (PRP) was performed when NPA extended to the posterior pole or when rubeosis and/or NVG (rubeosis with an IOP increase of > 21 mmHg) were observed.

Medical interview and end points.
For all patients, we recorded the presumed time of disease onset; age; sex; present or past history of dialysis and smoking; and present or past history of systemic diseases, including systemic hypertension, hyperlipidemia, diabetes mellitus, internal carotid artery disease, coronary artery disease, dialysis, and smoking.
According to a classification of a previous study 8  www.nature.com/scientificreports/ respectively 28 . Comparisons between the very poor VA and control groups were performed using the unpaired t-test. Comparisons between baseline BCVA and those at other time points in each group were performed using the unpaired t-test. The significance of systemic and ocular manifestations as risk factors for a final Snellen BCVA of < 20/200 was evaluated by calculating the adjusted ORs and 95% CIs in the multivariate logistic regression analysis, in which, the variables of diabetes and age were excluded because they were strong confounding factors. A model with lower Akaike information criterion was therefore selected. Significant differences in the sampling distributions were determined using chi-square tests. A P value of < 0.05 was considered statistically significant.

Data availability
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.