Efficacy and Safety of Intravitreal Therapy in Macular Edema Due to Branch and Central Retinal Vein Occlusion: a Systematic Review

Background Intravitreal agents have replaced observation in macular edema in central (CRVO) and grid laser photocoagulation in branch retinal vein occlusion (BRVO). We conducted a systematic review to evaluate efficacy and safety outcomes of intravitreal therapies for macular edema in CRVO and BRVO. Methods And Findings: MEDLINE, Embase, and the Cochrane Library were systematically searched for RCTs with no limitations of language and year of publication. 11 RCTs investigating anti-VEGF agents (ranibizumab, bevacizumab, aflibercept) and steroids (triamcinolone, dexamethasone implant) with a minimum follow-up of 1 year were evaluated. Efficacy: CRVO Greatest gain in visual acuity after 12 months was observed both under aflibercept 2 mg: +16.2 letters (8.5 injections), and under bevacizumab 1.25 mg: +16.1 letters (8 injections). Ranibizumab 0.5 mg improved vision by +13.9 letters (8.8 injections). Triamcinolone 1 mg and 4 mg stabilized visual acuity at a lower injection frequency (-1.2 letters, 2 injections). BRVO Ranibizumab 0.5 mg resulted in a visual acuity gain of +18.3 letters (8.4 injections). The effect of dexamethasone implant was transient after 1.9 implants in both indications. Safety Serious ocular adverse events were rare, e.g., endophthalmitis occurred in 0.0-0.9%. Major differences were found in an indirect comparison between steroids and anti-VEGF agents for cataract progression (19.8-35.0% vs. 0.9-7.0%) and in required treatment of increased intraocular pressure (7.0-41.0% vs. none). No major differences were identified in systemic adverse events. Conclusions Anti-VEGF agents result in a promising gain of visual acuity, but require a high injection frequency. Dexamethasone implant might be an alternative, but comparison is impaired as the effect is temporary and it has not yet been tested in PRN regimen. The ocular risk profile seems to be favorable for anti-VEGF agents in comparison to steroids. Because comparative data from head-to-head trials are missing currently, clinicians and patients should carefully weigh the benefit-harm ratio.


Introduction
Macular edema is the main cause of visual impairment in central retinal vein occlusion (CRVO) and branch retinal vein occlusion (BRVO) [1,2]. In recent years, anti-inflammatory and anti-angiogenic therapeutic strategies have been used to target vascular permeability and leakage to reduce macular edema and improve vision.
Among corticosteroids, triamcinolone acetonide, dexamethasone and fluocinolone have shown potential to reduce edema in RVO [3][4][5][6]. However, their drawbacks are known side effects, such as cataract progression and rise of intraocular pressure (IOP). The off-label triamcinolone acetonide is crystalline and is commercially available as Kenalog (Kenalog, 40 mg/mL, Bristol-Myers Squibb, Princeton, NJ), Trivaris (Pharm Allergan Inc., Irvine California) or prepared for injection with methocel [7]. Dexamethasone is more potent and soluble in comparison to triamcinolone and various attempts have been made to construct a slow-release device. Ozurdex™ (Pharm Allergan Inc., Irvine California) was the first dexamethasone implant that was approved by the Food and Drug Administration (FDA) for intraocular use in chronic uveitis and macular edema due to RVO in June 2009. Ozurdex shows an anti-edematous effect up to 6 months [8,9]. A fluocinolone implant is currently under investigation that is supposed to even last for up to 3 years [6].
Beside corticosteroids, progress in anti-angiogenic drug development provides us with multiple new therapeutic agents based on a concept of modified antibodies versus vascular endothelial growth factor (VEGF) and related molecules. For example, pegaptanib sodium, a 40-kDa RNA aptamer, binds to the isoform 165 of VEGF and was first approved for intravitreal therapy in wet age-related macular degeneration (AMD) [10,11]. A possible effect on visual acuity in macular edema due to CRVO and BRVO was investigated but pegaptanib did not receive approval for these indications [12,13]. Currently, the most frequently used anti-VEGF agents are ranibizumab (Lucentis™, Genentech, Inc., South San Francisco, CA, and Novartis Pharma AG, Basel, Switzerland) and bevacizumab (Avastin™, Genentech, Inc., South San Francisco, CA, and Roche, Basel, Switzerland). Ranibizumab is an antibody fragment with a high binding affinity towards all forms of VEGF. It has been approved for intravitreal therapy of wet AMD, diabetic macular edema and RVO by the FDA and European Medicines Agency (EMA) between 2006 and 2012 (depending on the indication). Bevacizumab is widely used as off-label intravitreal therapy. The cost difference to ranibizumab is striking. Large Head-to-head studies sponsored by the American and British public authorities have been conducted to compare both drugs in wet AMD (CATT [14] and IVAN [15], respectively). One and two year data of these trials showed no inferiority. Despite the approval of ranibizumab for macular edema in RVO, we assume that off-label bevacizumab is used as frequently in RVO as in AMD. This assumption is supported by a vast number of publications on the use of bevacizumab for RVO in clinical settings (e.g. [16][17][18]). The latest developed anti-angiogenic drug is aflibercept (VEGF-trap eye, Eylea™, Regeneron Pharmaceuticals, Inc., and Bayer Pharma AG, Berlin, Germany), a 115-kDa decoy receptor fusion protein, composed of the second domain of human VEGF receptor 1 and the third domain of VEGF receptor 2, which are fused to the Fc domain of human IgG1 [19,20]. A recent publication reported that the binding affinity for this drug is even higher than for ranibizumab and bevacizumab, respectively [21]. Aflibercept was first approved for intravitreal therapy of wet AMD in 2011. FDA approval for RVO was granted in September 2012. EMA approval is expected for 2013.
Most of this pharmaceutical research was conducted at the same time. Notably, the multicenter randomized controlled trials (RCTs) leading to FDA approval in RVO were conducted in parallel, which resulted in homogenous control groups using observation and/or sham injection in CRVO and sham injection and/or grid laser photocoagulation in BRVO [9,20,22,23]. These research proceedings, however, resulted in a lack of head-to-head studies, comparing the various drugs to each other. Currently the ophthalmologist can choose between different therapeutic options to treat vision loss due to macular edema in RVO. Some of the therapeutic options are approved, but others are used off-label. To support the ophthalmologist's decision between the different therapeutic options and to calculate the accompanying risks, we conducted a systematic review to investigate and compare the efficacy and safety of all therapeutic agents currently used in intravitreal therapy in RVO with a follow-up of at least 12 months.

Search Strategy
We searched Medline, Premedline, Embase and the Cochrane library from inception until November 2012. The search strategy was based on combinations of medical subject headings and keywords and was not restricted to specific languages or years of publications. The search strategy used in Medline is presented in Text S1. Search strategies for other databases were modified to meet the requirements of each database. The searches were supplemented by hand searching the bibliographies of included studies. Currently conducted RCT's on intravitreal treatment of macular edema in RVO were searched in the register for clinical trials (http:// clinicaltrials.gov) and in the WHO International Clinical Trial Registry Platform (http://www.who.int/ictrp/en/).

Inclusion and exclusion criteria
We included RCTs which investigated intravitreal ranibizumab, bevacizumab, pegaptanib, aflibercept, triamcinolone, dexamethasone, or fluocinolone as monotherapy in comparison to any non-surgical control group (observation or sham injection in CRVO, sham injection and/or grid laser photocoagulation in BRVO). Trials which compared one substance to another or to grid laser photocoagulation were also included. To address long-term efficacy and systematic adverse effects, one year follow-up data (or longer) had to be available.
Studies were excluded if participants received a combination of intravitreal pharmacological agents within the same eye (e.g. ranibizumab and triamcinolone), if both eyes of the same person were treated as study eyes, and if less than 10 patients per arm were enrolled. Studies which treated patients who had already been treated within 3 months using one of the substances were excluded. To account for the 2 different indications, study results had to be attributable to either BRVO or CRVO separately.

Data Extraction and Quality Assessment
Titles and abstracts were reviewed by using the above mentioned selection criteria. Full papers of appropriate studies were obtained for detailed evaluation. Data extraction and quality assessment were carried out after a modified evaluation tool of the Center for Reviews and Dissemination [24]. Data extraction of study characteristics comprised the following data: number of included patients, number of treated patients in each group, control treatment, follow-up time, treatment regimen and dosages, baseline visual acuity, and time between symptoms and/ or diagnosis of macular edema and treatment in months. Additionally, the percentage of patients included within 3 months of disease onset was noted, if reported. If duration was given in weeks or days, values were transformed to mean time in months. Visual acuity is given in number of letters (according to ETDRS scores) and was changed to letters, if reported in logMAR. Data extraction of study outcome comprised, if available, the following data: number of included patients, mean number of intravitreal injections within 12 months (or from month 12 to 24), percentage of patients with gain or loss of visual acuity ≥ 15 letters, and mean change in visual acuity from baseline to month 12 (from month 12 to 24) in letters. Despite the multitude of information modi provided, we extracted information on the percentage (and/or number of patients) of endophthalmitis, uveitis, retinal detachment, retinal tear, and traumatic lens damage, rise of intraocular pressure (IOP), medical treatment of IOP rise, glaucoma surgery, cataract progression and surgery, and vitreous hemorrhage. Given the underlying disease of BRVO or CRVO, we collected also all information on vitreous hemorrhage, neovascularization of the iris (NVI) and neovascular glaucoma (NVG), if reported. Regarding systemic adverse events we extracted information on death (any cause), myocardial infarction, cerebrovascular accidents, nonocular hemorrhage and infection if obtainable. Parameters to assess the methodological quality of RCTs and estimate risk of bias were information on blinding, transparency of patient flow, sample size calculation, intention-to-treat analysis, definition of expected adverse events, the method used to collect adverse events data, and comparability between groups at inclusion. All stages of study selection, data extraction and quality assessment were done independently by two reviewers (AP, CI or NF). Any disagreement was resolved by discussion and consensus.

Statistical Analysis
A narrative summary is provided because the identified studies are enormous heterogeneous in terms of intervention and sample size and, therefore, unsuitable for pooling.

Study characteristics
Triamcinolone -The SCORE study (CRVO/ BRVO) and Parodi 2008 (BRVO). The SCORE (Standard Care vs. Corticosteroid for Retinal Vein Occlusion) study compared intravitreal triamcinolone (Trivaris, Allergan) 1 mg or 4 mg with observation in CRVO [5] and with grid laser photocoagulation in BRVO [4]. In total, 271 patients were included in SCORE-CRVO and longest follow-up was 36 months. In the parallel SCORE-BRVO trial 411 patients were enrolled [4]. Treatment regimen allowed re-treatment with triamcinolone at PRN criteria using 4 months intervals [4,5]. The mean time from symptoms (or diagnosis) of CRVO to treatment initiation was 4 months, only 39% of the patients were treated within 3 months and 81% within 6 months of onset [5]. Correspondingly, the time span to treatment initiation in SCORE-BRVO was 4 months [4].
Dexamethasone Implant -The GENEVA study (CRVO/ BRVO). The GENEVA (Global Evaluation of Implantable Dexamethasone in Retinal Vein Occlusion With Macular Edema) study evaluated the safety and efficacy of an intravitreal dexamethasone implant 0.7 mg versus an implant 0.35 mg versus sham injection in both BRVO and CRVO. The primary outcome measure was time to visual improvement of 15 letters or more. This outcome was pooled for BRVO and CRVO patients [8]. GENEVA included 1256 patients in their open-label 6-month extension of which 1131 completed the trial. In total, 34.5% of patients presented with CRVO and 65.5% with BRVO in the GENEVA study. Time from symptoms Evaluation of Efficacy and Safety) study enrolled 392 patients and assigned them either to ranibizumab 0.3 mg or 0.5 mg monthly for 6 months followed by PRN treatment of ranibizumab 0.5 mg in all groups from month 6 to 12 [22]. Control treatment in this study consisted of sham injections until month 6. Mean duration of symptoms to treatment was 3.3 months. In contrast to SCORE, the percentage of patients receiving treatment within 3 months of CRVO onset was considerably higher (69% in CRUISE versus 39% in SCORE [5,22]).
The BRAVO (Ranibizumab for the Treatment of Macular Edema Following Branch Retinal Vein Occlusion: Evaluation of Efficacy and Safety) study assigned 397 patients to 3 treatment arms (corresponding to CRUISE): intravitreal ranibizumab 0.3 mg, intravitreal ranibizumab 0.5 mg and sham injections for 6 months followed by a PRN of ranibizumab 0.5 mg for all patients until month 12 [23]. Grid laser photocoagulation was only allowed as rescue therapy (to all groups), but not earlier than 3 months after inclusion [23]. 65% of BRVO patients were treated within 3 months of RVO duration, mean time from symptoms to treatment was 3.5 months [23].
Ranibizumab in the 2 nd year -The HORIZON trial (BRVO and CRVO). In cohort 2 of the open-label, single-arm, multicenter extension HORIZON trial, patients from BRAVO and CRUISE studies were included [30]. 304 patients of each trial continued treatment of macular edema with intravitreal ranibizumab 0.5 mg at PRN intervals (between 1 and 3 months) for additional 12 months. Grid laser photocoagulation was eligible as rescue therapy for BRVO patients. The HORIZON trial was discontinued after FDA approval for ranibizumab in RVO (in accordance to the study protocol). Finally, 181 of 304 CRVO patients and 205 of 304 BRVO patients were analysed. Regarding the safety analysis, data of all patients over the entire study duration of 24 months were included [30].

Bevacizumab -Epstein 2012 (CRVO), Russo 2009 and Donati 2012 (BRVO).
Intravitreal bevacizumab in CRVO was evaluated in the RCT by Epstein and colleagues [26,31]. Patients received either intravitreal bevacizumab 1.25 mg or sham injection every 6 weeks over a period of 6 months. After this, an open label extension followed during which all patients received bevacizumab 1.25 mg every 6 weeks. Each group comprised 30 patients, who showed symptoms of CRVO for 1.9 months in the bevacizumab and 2.2 months in the sham/ bevacizumab group.
Russo et al. compared intravitreal bevacizumab 1.25 mg after PRN schemata versus grid laser photocoagulation in 30 patients with BRVO over 12 months [28]. Inclusion criterion was duration of macular edema of at least 3 months.
Donati et al. randomized 18 patients (1:1) to 3 intravitreal injections of bevacizumab 1.25 mg or intravitreal injections combined with grid laser photocoagulation (one week after first injection) [29]. Patients could receive further injections PRN. Follow-up visits were conducted at month 3, 6 and 12 (final visit). Thus the maximum number of injections per group was 5. Median duration of macular edema before treatment was 3.2 and 3.1 months, respectively [29].

Aflibercept -The COPERNICUS trial (CRVO).
Only recently, aflibercept has been investigated in patients with CRVO in the COPERNICUS (Controlled Phase 3 Evaluation of Repeated Intravitreal Administration of VEGF Trap-Eye in Central Retinal Vein Occlusion: Utility and Safety) trial [20]. 189 patients with CRVO were included and received intravitreal aflibercept 2 mg (aflibercept group) compared to sham treatment (sham/aflibercept group) at a randomisation ratio of 2:1 during the first 6 months followed by (monthly) PRN treatment of intravitreal aflibercept 2 mg in all patients. Primary endpoint was the proportion of patients with an improvement of BCVA ≥ 15 letters. Mean time from symptoms to treatment was relatively short with 2.7 months (aflibercept) and 1.9 months (sham/aflibercept). First treatment within 2 months occurred in 56.1% (aflibercept) and 71.2% (sham/aflibercept), respectively [20].
Study results: Efficacy Outcome (Table 2, Figure 2, 3) Systematic extraction of outcome measures was limited by different primary outcome measures reported in the studies.
The SCORE-BRVO study did not find a statistically significant difference of visual acuity gain after 12 months: 26% (triamcinolone 1 mg), 27% (4 mg) and 29% (grid laser photocoagulation) [4]. Mean change in BCVA was +4 to +5 letters in all groups. It was achieved with a mean treatment application of 2.2 injections (for triamcinolone 1 mg), 2.1 injections (for triamcinolone 4 mg), and 1.8 (for grid laser photocoagulations, if no dense macular hemorrhages were present at baseline; 0.7 applications for the group, that received deferred laser after resorption of macular hemorrhages) [4].
Parodi et al. found a visual acuity gain of more than 15 letters in 54% of the combination group (SGLT and triamcinolone) and 38% in controls (SGLT with sham injection) after a single treatment [27]. Mean visual acuity increased by +17 letters (intervention group) and +6.5 letters (control group), respectively (p=0.011 for mean change in visual acuity at month 12 compared to baseline in the combination group).
Primary outcome of the GENEVA open-label extension study until month 12 was safety [8]. Efficacy within the core study was measured in time to achieve a significant improvement (gain of more than 15 letters). Patients treated with a dexamethasone implant reached this goal significantly faster and in a higher percentage than sham treated patients (pooled data from BRVO and CRVO, p<0.001) [8,9]. Looking into BRVO and CRVO data separately, the GENEVA study did not find a significant difference in visual acuity at month 12 between the groups. Unfortunately, numbers are not provided within the publication, but figures indicate that change in BCVA was +2 letters in dexamethasone 0.7 mg/0.7 mg versus -1 letter in sham/0.7 mg in CRVO and +6 letters in BRVO in both groups [8]. Mean number of injections was 1.9 (0.7 mg/0.7 mg), 1.9 (0.35 mg/0.7 mg), and 0.8 (sham/0.7 mg).
HORIZON-BRVO: Patients with BRVO stabilized at the BCVA letter score gained at month 12. Mean BCVA change during the second year was -2.3 letters (0.3/0.5 mg), -0.7 letter (0.5/0.5 mg) and +0.9 letter (sham/0.5 mg). The mean difference in BCVA score between baseline and month 24 was +14.9 letters, +17.5 letters and +15.6 letters, respectively. Frequency of intravitreal injections was considerably lower than in the first year and even lower than in CRVO patients during year 2. Patients received 2.4 (0.3/0.5 mg), 2.1 (0.5/0.5 mg) and 2.0 (sham/0.5 mg) injections [30]. Caution: Numbers are probably that low because the HORIZON trial was terminated immediately after FDA approval of ranibizumab.
Bevacizumab letters in those treated with sham injections followed by bevacizumab (p<0.05) [26]. The percentage of BCVA gain of more than 15 letters was 60% (bevacizumab) vs. 33.3% (sham/ bevacizumab), respectively (p<0.05). Russo et al. found a change in BCVA of +15.5 letters in bevacizumab treated patients and +10 letters in grid laser photocoagulation (p<0.05), achieved by a mean number of 1.7 intravitreal injections of bevacizumab and 1.5 grid applications [28].
Donati et al. found an improvement in BCVA of +15 letters in patients treated with bevacizumab alone (median: 4 injections) compared to +20 letters in patients who received a combination of intravitreal bevacizumab (median 3 injections) and one grid laser photocoagulation (p<0.01 for visual acuity at month 12 versus baseline in each group, p=0.03 for difference between groups at month 12) [29]. Three of 9 bevacizumab patients gained more than 15 letters, in the group of patients who received combination therapy 7 of 9 patients reached this gain.

Aflibercept -The COPERNICUS trial (CRVO).
CRVO patients in the aflibercept 2 mg group gained +16.2 letters at month 12, compared to +3.8 letters in the sham/aflibercept (p<0.001) [20]. The percentage of patients with a gain of more than 15 letters was 55.3% (aflicercept) and 30.1% (sham/ aflibercept), respectively (p<0.001). After 5.8 aflibercept injections until month 6, visual acuity was stabilized with an additional 2.7 injections during PRN. In the control group, patients with a mean of 5.3 sham injections in the beginning received a mean of 3.9 injections of aflibercept 2 mg until month 12 [20].
Parodi et al. did not see any endophthalmitis in 11 patients treated with a single triamcinolone injection after 12 months [27]. The only ocular adverse event that occurred was a rise in IOP which was treated with antiglaucomatous medication in 54% of the triamcinolone plus SGLT group compared to 0% in the SGLT group [27].
Dexamethasone Implant -The GENEVA study (CRVO/ BRVO). Safety outcomes were not stratified into CRVO and BRVO patients but comparison was possible between the 6 different treatment arms (stratified by the number and dose of dexamethasone implant) at month 12 [8]. Haller et al. reported no endophthalmitis and a "rare" occurrence of retinal detachment and retinal tear without giving any numbers. Frequently observed adverse events were IOP rise and cataract progression. The increase of IOP ≥ 10 mmHg was most pronounced at day 60 with 32.8% over all injections (12.6% after the 1st injection and 15.4% after the 2nd, dexamethasone 0.7/0.7 mg group). Taken all dexamethasone treatment groups together, 1.3% of patients underwent glaucoma surgery compared to 0.0% of patient in the sham/ no implant group. 25.5% (dexamethasone 0.7/0.7 mg), 25.0% (dexamethasone 0.35/0.7 mg), and 28.1% (sham/ dexamethasone 0.7 mg) received medical treatment for IOP compared to 0% in 0.35 mg/no implant group and sham/no implant group [8]. Cataract progression varied between 5.7% (sham/no implant), 10.5% (sham/0.7 mg) and 29.8% (0.7/0.7 mg) dexamethasone treated patients. There was no safety information on vitreous hemorrhage or neovascularization [8].

Bevacizumab -Epstein 2012 (CRVO), Russo 2009 and Donati 2012 (BRVO).
Epstein et al. did not report any endophthalmitis, nor retinal tear or detachment in both groups [26]. In the sham group 16.7% developed NVI, but no one in the bevacizumab group. By month 12 no new development of NVI was found in either group. Furthermore, in all 16.7% NVI cases, neovascularization regressed completely until month 12 after treatment changed from sham to bevacizumab 1.25 mg [26]. Russo et al. and Donati et al. reported no ocular adverse events in 25 intravitreal injections of bevacizumab [28,29]. Neither group showed complications during grid laser photocoagulation [28].

Study results: Safety Outcome -systemic adverse events (Table 5)
In general, the incidence of systemic adverse events was low and did not significantly differ between treatment groups in the multicenter trials [4,5,8,20,22,23,30]. The single center trials did not report any non-ocular adverse events [27][28][29].
Bevacizumab -Epstein (CRVO). Epstein et al. did not report serious non-ocular adverse effects [26,31]. However, one patient (3.3%) from the sham/bevacizumab group suffered from a transient ischemic attack and dropped out of the study.
Aflibercept -The COPERNICUS trial (CRVO). During the first 12 months of the ongoing COPERNICUS trial, no deaths were reported in the aflibercept 2 mg group. However, 2 deaths (2.7%) occurred in the sham/aflibercept group [20]. Both deaths had a vascular cause (myocardial infarction and cardiac arrhythmia). Non-lethal myocardial infarction was seen in 1 patient of each group (0.9% and 1.7%). The most frequent infections were nasopharyngitis ( [20]. The incidence of systemic adverse events did not differ significantly between groups. (Table 6) Adequate blinding over 12 months was reported in 3 trials (COPERNICUS, CRUISE, BRAVO) [20,22,23]. SCORE investigators were blind to the dosage (triamcinolone 1 mg versus 4 mg), but not to observation [4,5]. Two studies changed the status of blinding after the primary end point was reached from double-blind to open-label at month 6 (GENEVA, Epstein) [8,26]. HORIZON followed patients from CRUISE and BRAVO trials in an open-label design after month 12 [30]. However, interpretation of data is limited by the high drop-out rate in all groups and the early termination after approval for ranibizumab. Validity of efficacy was high in 6 trials based on adequate blinding, transparency of patient flow, comparability between groups, sample size calculation and ITT-analysis (SCORE-BRVO/ -CRVO, Epstein, COPERNICUS, CRUISE and BRAVO).

Methodological Quality and Risk of Bias
Based on our quality assessment, validity of safety was high in 4 trials (SCORE-BRVO/ -CRVO, GENEVA, and COPERNICUS) [4,5,8,20]. Four other trials showed moderate validity due to incomplete definition of expected adverse events or methods used to collect data on adverse events (Epstein, CRUISE, BRAVO, HORIZON) [22,23,26,30]. Information on ocular and systemic adverse events varied between very detailed reporting within most multicenter trials (including additional information available as online supplements) and, on the other hand, the overall notification of "no ocular and systemic adverse events" in a single-center trial.
The trials by Parodi et al., Russo et al. and Donati et al. substantially lack quality (unclear or no blinding, moderate (low) transparency of patient flow, lack of information on sample size calculation, data analysis, adverse events collection and reporting) [27][28][29]. Additionally, comparability of groups is impaired (due to small sample size) [27][28][29].

Discussion Principal Findings
Visual acuity. All anti-VEGF agents show a better gain of BCVA than steroids at month 12 (Figure 2 CRVO, 3 BRVO). A significant improvement of BCVA after 12 months of intravitreal anti-edematous therapy in CRVO is achieved with aflibercept 2 mg every 4 weeks for 6 months followed by a monthly PRN scheme (+16.2 letters, 8.5 injections) as well as continuous intravitreal injections of bevacizumab 1.25 mg every 6 weeks (+16.1 letters, 8 injections) [20,26]. In addition, ranibizumab (0.5 mg) leads to an improvement of VA of +13.9 letters in fixed plus monthly PRN dosage regimen as shown in the CRUISE study [22].
The SCORE-CRVO and GENEVA studies resulted in a stabilization of BCVA at month 12 with considerable less frequent injections (SCORE: -1.2 letters [2 injections], GENEVA: approximately +2 letters [1.8 injections]) [5,8]. Improvement of BCVA in triamcinolone treated patients was significantly better than in the observation group [5].
In BRVO best improvement of BCVA was found for ranibizumab in BRAVO at month 12 [23]. Treatment with 0.5 mg ranibizumab monthly for 6 months followed by monthly PRN resulted in +18.3 letters (8.4 injections) [23]. In comparison, bevacizumab seems to achieve similar results (+15 to 17 letters), but interpretation is limited by the very low numbers of participants in the included RCTs (9 to 15 patients per group) [27][28][29]. The RCT for aflibercept in BRVO is not yet published ("Study to Assess the Clinical Efficacy and Safety of VEGF Trap-Eye (Intravitreal Aflibercept Injection [IAI]), Also Commercially Known as EYLEA™ in Patients With Branch Retinal Vein Occlusion (BRVO)", NCT01521559, Regeneron Pharmaceuticals, source: clinicaltrials.gov). A multitude of non-RCTs is published on bevacizumab in RVO, but lack of quality, especially in adverse events documentation and reporting, low numbers of participants and multitude of therapeutic strategies lead to exclusion within this review.
BCVA results at 12 months were less favorable for steroids in BRVO. No significant effect was found for triamcinolone in SCORE-BRVO versus grid (2.1 injections) [4]. BRVO patients treated with dexamethasone implant deteriorated to approximately +6 letters after temporary visual improvement at month 12 (1.8 injections) [8].
Grid laser photocoagulation. Despite the gold standard of grid laser photocoagulation implemented after the BVOS [2], we found heterogeneous control groups ranging from sham injection in BRAVO to grid laser photocoagulation in SCORE-BRVO [4,23]. BRAVO allowed grid as rescue therapy, which complicates the indirect comparison of the trials. Only one trial compared grid laser photocoagulation in combination to intravitreal bevacizumab 1.25 mg to bevacizumab alone, and found even better results in the combination group (+20 letters [median 3 injections]) [29]. There are currently 4 trials registered that compare intravitreal ranibizumab with macular laser photocoagulation and/ or a combination therapy (RABAMES, NCT00562406; BRIGHTER, NCT01599650; NCT01189526, clinicaltrials.gov, and EUCTR2008-007175-24-HU, www.who.int/ictrp/en/), and 1 RCT comparing triamcinolone versus grid versus combination (ChiCTR-TRC-05000661, www.who.int/ictrp/en/). A major limitation of all studies with grid laser photocoagulation is the heterogeneity of the treatment itself and the lack of evaluation of grid by the reading centers.
Two year data. Two year data are only published for ranibizumab within the course of the HORIZON trial [30]. But interpretation of data is limited by the high drop-out rate in all groups and the early termination after approval for ranibizumab. Patients received significantly less intravitreal injections during the second year (CRVO 3.5 injections, BRVO 2.1) and lost little visual acuity BCVA (CRVO -4 letters, BRVO -0.7 letters) [30]. Two years follow-up of COPERNICUS [20]  and GALILEO [33] will provide us with more evidence regarding anti-VEGF in CRVO. Clinically significant gain in BCVA (more than 15 letters). Analysing SCORE, COPERNICUS, CRUISE, BRAVO and Epstein studies, anti-VEGF treatment of macular edema leads to a clinically significant visual improvement in almost twice as many participants as triamcinolone or grid laser photocoagulation. Approximately 25% responders were found for triamcinolone [4,5], while anti-VEGF trials found 50% [22], 55% [20] and 60% [26] in CRVO ( Figure 2) and 60% in BRVO [23] (Figure 3). Results at 12 months in control groups are better than after 6 months and the differences between drug and control groups smaller, because patients were treated open-label with the investigational product for months 6 to 12 in all multicenter trials (CRUISE, BRAVO, Epstein, COPERNICUS, GENEVA) [8,20,22,23,26]. Our analysis of 12 months data still shows a striking difference over all anti-VEGF trials between the groups, which were treated with the individual therapeutic agent from the beginning and the control/PRN groups, which had a delay of 6 months before therapy started. In conclusion, the effect of the anti-VEGF agents on macular edema in RVO is more pronounced if treatment starts early after onset of macular edema.
Time between symptoms and treatment. On basis of these results, one can assume that time between occlusion and treatment is a critical factor both for patient choice in the context of clinical trials but also with regard to the therapeutic effect and the interpretation of study data. In practice, studies including and treating patients shortly after RVO onset might find better visual acuity outcome and/ or response to treatment than trials including old RVO with persistent macular edema, resulting in false negatively reduced outcome. Therefore, we specifically depicted time of occlusion in the list of study characteristics. Both, mean time from occlusion and percentage of patients with ≤ 2-3 months occlusion time differed between the included RCTs. One explanation is the choice of time span for inclusion criteria. In CRVO trials, mean time of occlusion ranged from 1.9 months (bevacizumab, Epstein 2012), 1.9 and 2.7 (aflibercept, COPERNICUS), 3.3 (ranibizumab, CRUISE), to 4.2 (triamcinolone, SCORE) and 5.2 months (dexamethasone implant, GENEVA) (Table 1) [4,5,8,20,22,26]. We found that the time to treatment is considerably longer in the trials investigating steroids compared to anti-VEGF trials. This is also expressed in high percentages of "fresh" RVO ≤ 2 months of 56-71% in COPERNICUS, RVO ≤ 3 months of 69% in CRUISE and only 15-18% in GENEVA (Table 1) [8,20,22]. The same discrepancy was found in BRVO trials, the percentage of "fresh" RVO was 65% in BRAVO and 37% in SCORE [23,34]. This might put into perspective the differences in visual acuity outcome between the RCTs.
Ocular adverse events. The rate of endophthalmitis was very low in all RCTs (<1%) regardless of the substance class. SCORE-BRAVO, GENEVA, Epstein, and CRUISE did not find any endophthalmitis in approximately 6000 intravitreal injections (Table 3) [4,8,22,26]. These findings in RVO are in accordance to safety results and analysis of ranibizumab and bevacizumab in AMD [35]. Due to the different frequency of injection between anti-VEGF (8 injections/ 12 months) and steroids (2 injections/ 12 months) the risk of endophthalmitis for the individual patient is higher in an anti-VEGF treatment regimen than in triamcinolone or dexamethasone.
IOP increase and medical treatment of IOP occurred more frequently in SCORE and GENEVA than in any anti-VEGF trial ( Table 4). Looking into SCORE data, IOP increase seems to be dose-dependent with maximum rates of 35% (CRVO) and 41% (BRVO) for triamcinolone 4 mg [4,5]. GENEVA found approximately 25% in all dexamethasone implants after 6 months. It remains unclear, whether the 10% rate reported for 6-12 months data are in addition to these 25% [8,9]. No need for medical or surgical IOP control was reported in Epstein, CRUISE, BRAVO and HORIZON (Table 4) [22,23,26,30]. COPERNICUS found IOP increase in all groups of > 10%, which needed no medical treatment [20,36]. Rates of glaucoma surgery was occasional in GENEVA (1.3%, all dexamethasone implants), low in SCORE (2.2% CRVO triamcinolone 4 mg) and COPERNICUS (2.7% sham group, 1.7% sham/aflibercept 2 mg), and not reported for bevacizumab or ranibizumab trials [5,8,20,22,23,26]. It is remarkable, that in SCORE and GENEVA trials only participants with treatment of steroids developed glaucoma which needed surgery, while in anti-VEGF trials participants did not need surgery under ranibizumab or bevacizumab treatment [4,5,8,22,23,26]. Additionally, in COPERNICUS, surgery had to be performed in the sham/aflibercept group, probably due to secondary neovascular glaucoma [20].
Rates of cataract and surgery were found more frequently in SCORE and GENEVA with 20 -35% cataract progression, compared to none in anti-VEGF trials (Table 4) [4,5,8].
Vitreous hemorrhages, either due to the injection procedure or because of progression to neovascular RVO, was found seldom in all treatment groups. We found a tendency towards higher rates in control groups in CRVO, which received treatment 6 months later. This difference was most pronounced in COPERNICUS (0.9% [aflibercept] versus 7.1% [sham/ aflibercept]), and less striking in CRUISE (5.3% [ranibizumab 0.3/0.5 mg], 5.4% [0.5/0.5 mg] versus 8.8% [sham/0.5 mg]) [20,22]. GENEVA did not report vitreous hemorrhages [8,9]. In BRVO no difference between treatment and control groups was found [23,30]. This finding may be explained by the higher risk of CRVO for vitreous hemorrhage compared to BRVO due to the bigger area of tissue affected [1,2].
In conclusion, physicians have to take these differences in ocular risk profiles between anti-VEGF and steroids into account to choose the best therapy for their individual patients.
Systemic adverse events. The systemic risk seems to be comparable low between all compounds given intravitreally in RVO. Rates of death were low (0-3%) over all groups and no statistically significant difference was found ( Table 5).
Comparison of systemic adverse event rates was biased by different reporting and difficulty in data extraction. None of the RCTs investigated in this review found significant differences within cerebrovascular events, non-ocular hemorrhages or infections. Infections were found in up to 7.9% in anti-VEGF trials (all groups, all substances), but more often in SCORE patients (10-19% all groups) [4,5,20,30]. Rates of infections are ranibizumab 0.5 mg in BRVO. Cataract and glaucoma are the main draw-backs in the use of triamcinolone and dexamethasone, while low injection frequency is in favor of steroids. To our knowledge, this review is the first to report 12 months data of five different therapeutic intravitreal agents currently used in RVO. Therefore, these findings may be of great value for all ophthalmologists.
Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) checklist. (DOC)