Intravenous Fibrinolytic Therapy in Central Retinal Artery Occlusion: A Patient-Level Meta-analysis

Importance  Central retinal artery occlusion (CRAO) is an ophthalmologic emergency that can result in blindness. At present, no proven therapy for CRAO exists. Treatment with fibrinolytic agents has shown promise but remains of unproven benefit.

Objectives  To assess the efficacy of systemic fibrinolytic therapy for patients with CRAO and to define a time window of efficacy.

Data Sources  We systematically queried PubMed, Web of Science, and Scopus using the following index terms: “retinal artery occlusion” OR “retinal ischemia” AND “thrombolysis” OR “fibrinolysis” OR “tissue plasminogen activator” OR “streptokinase” OR “urokinase.” Search was not limited by year of publication or language and was conducted in August 2014. In addition, we evaluated the references from relevant review articles.

Study Selection  We assembled observational studies reporting on visual acuity outcomes after CRAO. Inclusion criteria were complete reporting of visual outcomes after CRAO (with or without fibrinolytic therapy) and a series of more than 5 patients for fibrinolysis treatment or more than 20 cases when untreated or treated with conservative modalities.

Data Extraction and Synthesis  Patient-level data were sought for studies reporting outcomes of treatment with fibrinolysis. Summary statistics were obtained for conservative treatment and natural history studies. The studies were weighted by the inverse of variance and merged in a random-effects model.

Main Outcomes and Measures  Rate of visual recovery (defined as improvement of visual acuity from 20/200 or worse at presentation to 20/100 or better) was calculated for patients treated with fibrinolytic and conservative therapies and those who received no treatment.

Results  We obtained summary statistics from 7 studies that included 396 patients who received no treatment after CRAO and from 8 studies that included 419 patients treated with ocular massage, anterior chamber paracentesis, and/or hemodilution (conservative treatment). Patient-level data were obtained for 147 patients treated with systemic fibrinolysis. We found that fibrinolysis was beneficial at 4.5 hours or earlier after symptom onset compared with the natural history group (17 of 34 [50.0%] vs 70 of 396 [17.7%]; odds ratio, 4.7 [95% CI, 2.3-9.6]; P < .001). Absolute risk reduction was 32.3%, with a number needed to treat of 4.0 (95% CI, 2.6-6.6). We also found that conservative treatment significantly worsened visual acuity outcomes and recovery rates after CRAO compared with the natural history group (31 of 419 [7.4%; 95% CI, 3.7%-11.1%] vs 70 of 396 [17.7%; 95% CI, 13.9%-21.4%]; P < .001; number needed to harm, 10.0 [95% CI, 6.8-17.4]).

Conclusions and Relevance  Our analysis suggests that a clinical trial of early systemic fibrinolytic therapy for CRAO is warranted and that conservative treatments are futile and may be harmful.

Central retinal artery occlusion (CRAO) is an ophthalmologic emergency and a cause of acquired blindness. This acute, painless condition is typically the result of thrombosis or embolism leading to ischemia of the retina and optic nerve head with profound loss of vision. The source of thrombosis or embolism in CRAO is thought to arise predominantly from atherosclerotic plaques, carotid stenosis, inflammatory vascular disease, or cardiac abnormalities. Once the central retinal artery is occluded, survival of the retina depends on the degree of collateralization and the duration of retinal ischemia before the offending embolus or thrombus is dislodged or autolyzed; however, most patients develop blindness. An experimental study of CRAO in rhesus monkeys found that complete ischemia to the retina lasting more than 4 hours resulted in severe neuronal loss.¹ Current standard treatment options for acute CRAO management include sublingual isosorbide dinitrate, topical timolol maleate, systemic pentoxifylline, inhalation of 10% carbon dioxide, hyperbaric oxygen, ocular massage, intravenous acetazolamide, mannitol, anterior chamber paracentesis, hemodilution, and corticosteroids. None has been shown to be more effective than placebo.² Fifty years of experience with various fibrinolytic treatments for acute CRAO has included systemic and local or intra-arterial delivery. These treatments have demonstrated promising results, but the first randomized clinical trial, the European Assessment Group for Lysis in the Eye (EAGLE) trial,³ evaluated intra-arterial delivery of fibrinolysis from 4.5 to 12.0 hours after the onset of symptoms and did not demonstrate improved visual outcome with treatment. Systemic fibrinolytic therapies are simpler and faster to deploy and may be safer. Rapid administration of systemic fibrinolytic agents has proved to be a feasible and effective treatment for acute ischemic stroke but only within 4.5 hours of symptom onset^4,5; systemic fibrinolysis may improve the efficacy of treatment for CRAO. For this reason, the aim of this analysis is to assemble the published literature on systemic fibrinolysis in CRAO at the patient level and to compare the results against the natural history of this illness to determine a time window of maximal effectiveness of fibrinolytic interventions for use in future clinical trials.

This analysis was performed in accordance with the Meta-analysis of Observational Studies in Epidemiology (MOOSE) guidelines.⁶ The literature search strategy was not limited by year of publication or language. Data collection occurred from August 1 through 30, 2014. We systematically queried the PubMed, Web of Science, and Scopus databases for the following variations of keywords: “retinal artery occlusion” OR “retinal ischemia” AND “thrombolysis” OR “fibrinolysis” OR “tissue plasminogen activator” OR “streptokinase” OR “urokinase.” All studies reporting visual outcomes after CRAO (with or without fibrinolytic therapy) were collected. We also evaluated the reference lists from the identified studies and relevant review literature to identify additional relevant studies. Studies reporting fibrinolytic treatment in more than 5 patients were considered for inclusion in the treatment group (case reports and smaller series were excluded to reduce the risk for selection bias in favor of positive outcomes). Studies reporting outcomes with any other standard treatment or with no treatment in a series of more than 20 cases were included in the conservative treatment group and the group receiving no treatment (natural history group). Studies reporting outcomes after treatment with hyperbaric oxygen, intra-arterial fibrinolysis, and acetylcholine injection were not included; however, if these studies reported on a control group, that group was considered for inclusion. We did not include CRAO related to a primary large-vessel vasculitis in this analysis because this condition is treated primarily with immunomodulatory medications. Branch retinal artery occlusions and retinal vein occlusions were excluded. Exclusion criteria also included insufficient reporting of the data; specifically, the time to treatment and pretreatment and posttreatment visual acuity were required for the treatment group.

The visual recovery rate was the primary outcome assessed in this analysis, which we defined as an initial visual acuity of at least 1.0 logMAR or the Snellen equivalent of 20/200 or worse that improved to 0.7 logMAR or less (Snellen equivalent, 20/100 or better). Patients with visual acuity better than 1.0 logMAR (Snellen equivalent, 20/200) on initial evaluation were excluded because they frequently experience spontaneous visual recovery,⁷ likely as a result of some degree of reperfusion or collateral blood supply. Inclusion of these cases was likely to bias the results in favor of fibrinolysis because many patients in the observational studies were first encountered at later time points. In addition, this criterion emphasizes a recovery of functional vision in patients meeting the legal definition of blindness (in the United States) at the time of presentation. In the natural history group, reporting of visual acuity at follow-up (≥48 hours after onset) was required, and the relevant summary statistics, specifically, the total number of patients presenting with visual acuity of 1.0 logMAR or greater (Snellen equivalent, 20/200 or worse) and the percentage of cases with spontaneous recovery of vision, were extracted from the identified studies. For the intravenous or systemic fibrinolysis studies, tissue plasminogen activator (tPA), urokinase, and streptokinase were all considered appropriate treatments. Patient-level data, including time from symptom onset to treatment and pretreatment and posttreatment visual acuity, were extracted. In those studies in which these data were not available, the authors were contacted directly to obtain these data. In most studies, visual recovery was assessed at the end of the acute hospitalization or shortly thereafter. In those studies explicitly stating the timing, follow-up ranged from 1 day to a few weeks after fibrinolytic treatment, with most ranging from 3 to 7 days. Exclusion criteria applied to studies in which the essential data could not be obtained or were reported in insufficient detail.

Data analysis occurred from August 1 through 30, 2014. For the fibrinolytic treatment group, we obtained patient-level data for pretreatment and posttreatment visual acuity and time from symptom onset to treatment. The data were dichotomized to identify patients who recovered and those who did not; the percentage with spontaneous visual recovery was reported along with the binomial SD. The patients were divided into those treated at 4.5 hours or earlier, longer than 4.5 to 12.0 hours, longer than 12.0 to 24.0 hours, and longer than 24.0 hours. The 4.5-hour time point was selected as an interval of interest based on animal literature indicating that the retina has a tolerance to ischemia for a little more than 4 hours¹ and that fibrinolysis is helpful in ischemic stroke within 4.5 hours.⁵ The other time points divide the remaining cases into equivalently powered epochs. We calculated the visual recovery rate for each group. For a comparison group, studies of the natural history of CRAO and conservative therapies were aggregated by meta-analysis using the Mix software (version 1.7; BioStatXL), and the data were weighted by the inverse of variance. Heterogeneity was assessed with the Q statistic and I² test.⁸ Because of the presence of heterogeneity in the conservative treatment group, this group was believed to be inadequate for comparison, and only those studies reporting the natural history of the disease (or minimal treatments) were used for comparison with intravenous fibrinolysis treatment. The hypothesis that the probability of visual recovery was related to delay of treatment was assessed with the Kruskal-Wallis test analyzing treatment in each of the 4 time windows and spontaneous recovery in the natural history group (5 groups). We then compared the frequency of successful treatment at each interval with the frequency of spontaneous recovery in the natural history analysis as a Bernoulli trial.⁹ We accepted α < .05 as significant. Odds ratios, relative risk reduction, and number needed to treat are reported.

Fourteen potentially relevant studies^10-23 were excluded because visual acuity data were not presented or were presented in insufficient detail to include in this analysis. We identified 7 studies^24-30 that addressed the natural history of CRAO. Most of these studies^24,26-28,30 used no treatment; 1 study²⁹ treated with corticosteroids, and 2 studies^25,29 treated with topical intraocular pressure–lowering drops and/or acetazolamide. None of the treatments used have any demonstrated effect on recovery of visual acuity. We found no statistically significant heterogeneity in this analysis (Q = 4.3; P = .75; I² = 0% [95% CI, 0%-67.6%]); 396 patients were included, and 70 patients (17.7%) had spontaneous visual recovery (95% CI, 13.9%-21.4%).

Eight studies^7,31-37 reported visual acuity outcomes in 419 patients after conservative therapies, including ocular massage, anterior chamber paracentesis, and hemodilution. These interventions frequently were used concurrently or in sequence, making it difficult to analyze their effect independently of each other. In the studies using these more aggressive nonfibrinolytic therapies, the recovery rate was substantially lower than in the natural history control group (31 of 419 [7.4%; 95% CI, 3.7%-11.1%] vs 70 of 396 [17.7%; 95% CI, 13.9%-21.4%]; P < .001; number needed to harm, 10.0 [95% CI, 6.8-17.4]). We found evidence of modest heterogeneity in these studies (Q = 15.0; P = .04; I² = 52.3 [95% CI, 0%-79.0%]) that was contributed by the earliest study in the group,³¹ which behaved as an outlier. Exclusion of this study from the analysis would further support the conclusion that these treatments are harmful.

Of 13 studies^38-50 we identified in which patients with CRAO were treated with systemic fibrinolysis, we were able to obtain patient-level data for 9 (Table 1).^38-46 Two of these studies^38,39 used intravenous tPA; 4, streptokinase^40-43; and 3, urokinase.^44-46 This list represents an 80% capture of the total number of patients in the published studies; data from 4 studies^47-50 that included 27 patients could not be obtained. Demographics for the 147 included patients are provided in Table 2. Age and starting visual acuity were similar in each group (no significant differences were determined by Kruskal-Wallis analysis). In this group of studies, 46 treated patients (31.3%; 95% CI, 24.3%-39.3%) had visual recovery. This recovery rate is significantly higher than that for the natural history cohort or the conservative treatment cohort (P < .001 for each comparison; Figure 1). We found no significant between-study heterogeneity in this group of studies (Q = 5.6; P = .69; I² = 0% [95% CI, 0%-64.8%]); inclusion of the studies for which patient-level data were not obtained did not introduce significant heterogeneity or alter the conclusions of this analysis.

We found a significant effect of time to fibrinolytic administration on visual recovery after CRAO (P < .001) (Figure 2). Systemic fibrinolysis within the first 4.5 hours after symptom onset resulted in recovery of vision in 17 of 34 patients (50.0%; 95% CI, 32.4%-67.6%). This rate of spontaneous recovery is nearly 3 times that in the natural history cohort (odds ratio, 4.7 [95% CI, 2.3-9.6]; P < .001), with a 32.3% absolute risk reduction and a number needed to treat of 4.0 (95% CI, 2.6-6.6). We found no significant difference in the recovery rate after fibrinolysis compared with the natural history cohort for those patients treated in any of the epochs after 4.5 hours (P = .07, P = .22, and P = .11 for >4.5 to 12.0, >12.0 to 24.0, and >24.0 hours, respectively). Two studies^38,43 reported long-term visual acuity outcomes after fibrinolysis. Rumelt and colleagues⁴³ observed patients for 1 to 10 years and found that only 1 of the 8 successfully treated patients experienced subsequent deterioration of their visual acuity owing to a cataract. Kattah and colleagues³⁸ observed patients for 3 months and found 1 of 4 patients with a response to fibrinolysis had subsequent deterioration that was caused by glaucoma. Both studies concluded that improvements in visual acuity were durable. Serious hemorrhagic events occurred in 5 of 147 patients (3.4%) of the total fibrinolysis cohort, for a number needed to harm based on these data of 30.0 (95% CI, 15.8-212.3). The adverse effect profile of streptokinase is inferior to that of tPA, particularly regarding hemorrhagic complications.⁵¹ Among the patients treated with streptokinase, 5 serious hemorrhagic complications occurred (4 of them were fatal, including 3 fatal intracerebral hemorrhages and 1 fatal hemorrhage from the liver). No major hemorrhages occurred after administration of urokinase or tPA in this analysis.

Because most modern fibrinolytic therapy consists of tPA, we performed a subgroup analysis of cases treated specifically with tPA. This analysis confirmed a comparable significant benefit of early treatment compared with untreated patients in the window of 4.5 hours or earlier, which is not present in the window of longer than 4.5 to 12 hours; 8 of 13 patients recovered if treated within 4.5 hours after onset of symptoms (P < .001); 4 of 23, after 4.5 to 12.0 hours after onset (P = .40).

We examined the effectiveness of systemic fibrinolytic therapy in the acute treatment of CRAO using assembled, primarily retrospective data in this meta-analysis. We obtained patient-level data from 147 patients who received fibrinolytic therapy and analyzed these data with a particular focus on defining a time window in which fibrinolytic therapy is effective. The effective window appears to be within the first 4.5 hours after symptom onset. A similar result was obtained from a subanalysis of only those patients treated with intravenous tPA. Five major hemorrhages occurred in this series, including 4 treatment-related fatalities that all occurred in 1 early study that used streptokinase infusion,⁴¹ and the dosing used in that study was not clear. In the literature at large, 1 case of a symptomatic intracerebral hemorrhage has been reported in a patient with CRAO treated with intravenous tPA at the recommended dose for ischemic stroke.⁵⁰ Experience with tPA infusion in stroke mimics suggests that, in the absence of cerebral ischemic injury, intracerebral hemorrhage is a rare complication.⁵² To our knowledge, no reports of intraocular hemorrhage after fibrinolytic therapy exist (even when anterior chamber paracentesis was performed concomitantly).

We found no convincing evidence from the literature that any conservative treatment modality (specifically, ocular massage, hemodilution, and/or anterior chamber paracentesis) is effective, and the results from this meta-analysis suggest that these modalities may be harmful. The recovery rate in patients treated in this fashion was less than half that of patients receiving no treatment. Our conclusions on this topic were limited by the fact that most of these studies reported outcomes of patients treated with multiple interventions; however, this multimodal treatment may reflect a typical clinical approach to this disease that, based on this evidence, should be discouraged.

Based on these results, we believe that a clinical trial of systemic fibrinolysis within 4.5 hours of CRAO is warranted. Because of the successful application of systemic fibrinolysis in the setting of acute stroke and because of the similarities between CRAO and stroke, consideration of a similar treatment in acute CRAO is logical. However, differences in the vascular anatomy and metabolic characteristics of the retina preclude direct extrapolation from the stroke literature. Fibrinolytic drugs have been used to treat acute retinal vascular occlusion in observational studies for more than 5 decades; however, convincing randomized data demonstrating the efficacy of this treatment strategy are lacking. The first large clinical trial³—the EAGLE study—examined intra-arterial delivery of tPA and was terminated owing to the futility of the experimental treatment. All patients in that trial were treated from at least 4.5 to 24.0 hours after symptom onset, and this delay to treatment may account for the negative results. The conservative treatment group in the EAGLE trial also had better-than-expected outcomes (visual acuity improving by 3 lines in 60% of untreated cases), which may have contributed also. Intra-arterial delivery of fibrinolytics is a higher-morbidity intervention and requires more time for drug delivery compared with intravenous methods. Given this delay, this technique does not appear to have much applicability in CRAO.

The major strengths of this analysis are the thoroughness of the literature review, the low levels of heterogeneity in the groups of studies, and the analysis of patient-level data. We show that systemic fibrinolysis is significantly effective only within the first 4.5 hours of symptom onset. These data should be interpreted cautiously in the clinical application of fibrinolytics because numerous limitations apply to the current analysis, including the retrospective and nonrandomized nature of the data and variability in specific treatment procedures between and within the studies. However, the design of a randomized controlled trial of tPA treatment in acute CRAO should take these data into account.

Systemic fibrinolysis for CRAO has not yet been evaluated in an adequate clinical trial, although the results of this meta-analysis are promising. Conservative treatments of CRAO are futile and may be harmful. Therefore, a clinical trial of early systemic fibrinolytic therapy for CRAO is warranted.

Accepted for Publication: June 1, 2015.

Corresponding Author: Matthew Schrag, MD, Department of Neurology, Yale University, 15 York St, Floor LCI-9, New Haven, CT 06510 (matthew.schrag@vanderbilt.edu).

Published Online: August 10, 2015. doi:10.1001/jamaneurol.2015.1578.

Author Contributions: Dr Schrag had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Schrag, Schindler.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Schrag, Youn, Schindler, Kirshner.

Critical revision of the manuscript for important intellectual content: Schrag, Schindler, Kirshner, Greer.

Statistical analysis: Schrag, Schindler.

Administrative, technical, or material support: Schrag, Schindler, Greer.

Study supervision: Schindler, Kirshner, Greer.

Conflict of Interest Disclosures: None reported.