Risk of Intraocular Bleeding With Novel Oral Anticoagulants Compared With Warfarin: A Systematic Review and Meta-analysis

Key Points

Question  How does the intraocular risk of bleeding differ between patients receiving warfarin compared with the newer novel oral anticoagulants?

Findings  In this systematic review and meta-analysis of phase 3 randomized clinical trials involving 102 627 patients, use of novel oral anticoagulants was associated with a reduced risk of intraocular bleeding by one-fifth compared with warfarin.

Meaning  Patients requiring anticoagulation for atrial fibrillation or venous thromboembolism at high risk of spontaneous intraocular bleeding may benefit from novel oral anticoagulant therapy.

Importance  It is unclear if the risk of intraocular bleeding with novel oral anticoagulants differs compared with warfarin.

Objective  To characterize the risk of intraocular bleeding with novel oral anticoagulants compared with warfarin.

Data Sources  A systematic review and meta-analysis was undertaken in an academic medical setting. MEDLINE and ClinicalTrials.gov were searched for randomized clinical trials published up until August 2016. This search was supplemented by manual bibliography searches of identified trials and other review articles.

Study Selection  Studies were eligible for inclusion if they were phase 3 randomized clinical trials, enrolled patients with atrial fibrillation or venous thromboembolism, compared a novel oral anticoagulant (dabigatran, rivaroxaban, apixaban, or edoxaban) with warfarin, and recorded event data on intraocular bleeding. Data on intraocular bleeding were pooled using inverse-variance, weighted, fixed-effects meta-analysis.

Data Extraction and Synthesis  The PRISMA guidelines were used for abstracting data and assessing quality. Independent extraction was performed by 2 investigators.

Main Outcomes and Measures  Intraocular bleeding events and associated risk ratio for novel oral anticoagulants compared with warfarin.

Results  Twelve trials investigating 102 627 patients were included. Randomization to novel oral anticoagulants was associated with a 22% relative reduction in intraocular bleeding compared with warfarin (risk ratio, 0.78; 95% CI, 0.61-0.99). There was no significant heterogeneity observed (I² = 4.8%, P = .40). Comparably lower risks of intraocular bleeding with novel oral anticoagulants were seen in subgroup analyses, with no significant difference according to the indication for anticoagulation (P for heterogeneity = .49) or the novel oral anticoagulant type (P for heterogeneity = .15). Summary estimates did not differ materially when random-effects meta-analytic techniques were used.

Conclusions and Relevance  These results suggest that novel oral anticoagulants reduce the risk of intraocular bleeding by approximately one-fifth compared with warfarin. Similar benefits were seen in both patients with atrial fibrillation and venous thromboembolism. Our data have particular relevance for patients at higher risk of spontaneous retinal and subretinal bleeding. These findings may also have important implications in the perioperative period, in which the use of novel oral anticoagulants may be superior. Future studies are required to better characterize the optimal management of patients with both ophthalmic disease and cardiovascular comorbidities requiring anticoagulation.

Quiz Ref IDIn recent years, novel oral anticoagulants (dabigatran, rivaroxaban, apixaban, and edoxaban) have become widely used worldwide in patients with atrial fibrillation and in patients with venous thromboembolism. Several phase 3 randomized clinical trials^1-3 have established equal or superior antithrombotic efficacy of these agents and lower risk of intracranial hemorrhage compared with warfarin. Intraocular hemorrhage is an uncommon but potentially visually threatening adverse event that has been well documented in patients taking warfarin and other antithrombotics.^4-8 In the only previous meta-analysis⁹ on this topic to our knowledge, no significant difference was reported in the risk of intraocular bleeding between novel oral anticoagulants and other antithrombotic agents, such as warfarin. However, the authors acknowledged that this conclusion may have been susceptible to a lack of statistical power given the limited number of bleeding events. Moreover, analyses were performed only within disease indications, further reducing the ability to detect significant differences. Given this uncertainty, we sought to perform an updated meta-analysis on the risk of intraocular bleeding with novel oral anticoagulants compared with warfarin. We include additional data, analyze event data across disease indications, and discuss the implications of our findings in patients at high risk of spontaneous intraocular bleeding and in the perioperative setting.

This systematic review and meta-analysis was performed in an academic medical setting in accord with the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines.¹⁰ The PRISMA checklist is included in eTable 1 in the Supplement.

Quiz Ref IDWe restricted our analysis to phase 3 randomized clinical trials that included patients with atrial fibrillation or venous thromboembolism who were randomly assigned to receive novel oral anticoagulants or warfarin. Studies were eligible for inclusion if data on intraocular bleeding were available as a predefined study outcome or from adverse event reports. We contacted study authors to obtain event data if this information was not readily available from the published articles. We did not use any data from phase 2 dose-ranging studies because of their small sample size and short follow-up.

We performed a systematic search of MEDLINE published up until August 2016. Search terms included atrial fibrillation, pulmonary embolism, deep vein thrombosis, deep venous thrombosis, venous thromboembolism, dabigatran, rivaroxaban, apixaban, edoxaban, factor Xa inhibitor, thrombin inhibitor, vitamin K antagonist, and warfarin. We also performed a search of ClinicalTrials.gov and manually searched bibliographies of identified trials, review articles, and meta-analyses to find relevant studies.

Two investigators (M.T.S. and C.X.W.) independently reviewed all identified studies for inclusion. The decision to include studies was hierarchical, initially based on the study title, followed by abstract and then full-text review of each remaining article. Disagreements were resolved by discussion between the 2 investigators.

The number of individuals in the treatment and control arms, number of intraocular bleeds in both arms, novel oral anticoagulant used, and indication for anticoagulation were recorded. A risk ratio from the total number of events was calculated for each trial.

The Cochrane Collaboration’s risk-of-bias tool was used to evaluate risk of bias for included trials.¹¹ Risk of bias was evaluated on the following 5 dimensions: selection bias (defined as random sequence generation and allocation concealment), performance bias (masking of both participants and investigators), detection bias (masking of evaluators), attrition bias (incomplete outcome data), and reporting bias (selective outcome reporting). Each dimension was judged to be of low, unclear, or high risk of bias. The bias in the effect estimate was assessed for each individual outcome, and the trial was then judged to be of low, unclear, or high risk of bias, depending on whether a trial resulted in a biased effect estimate and would have led to bias in the reported intraocular bleeding outcome. We also rated the quality of evidence using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) Working Group approach.¹²

Overall effect estimates were calculated using inverse-variance, weighted, fixed-effects meta-analysis with 95% CIs. Random-effects meta-analysis was performed as a sensitivity analysis. Heterogeneity was quantified using the I² statistic and tested between subgroups using Cochran Q statistic. The main analysis was performed to address whether novel oral anticoagulants reduce the risk of intraocular bleeding compared with warfarin. As prespecified secondary analyses, we stratified trials by whether they were conducted in patients with atrial fibrillation or venous thromboembolism and further stratified trials by novel oral anticoagulant type. All statistical analyses were performed using Stata version 13.0 (StataCorp LP) and R version 3.2.1 (R Project). Two-sided P < .05 was considered statistically significant.

The systematic search of electronic databases identified 12 070 reports, from which we identified 12 potentially relevant studies^1-3,13-21 for full-text review (Figure 1 and eFigure 1 in the Supplement). Five trials tested novel oral anticoagulants in patients with atrial fibrillation, and 7 trials tested novel oral anticoagulants in patients with venous thromboembolism. No additional trials were identified from a bibliography search of included studies. Additional data were provided by the authors of one study.¹⁵ Therefore, 12 trials investigating 102 627 patients were included in this analysis. The Table lists the baseline characteristics of each study. The mean age of patients was similar between trials, as was the proportion of men recruited. The median time in therapeutic range among patients receiving warfarin ranged from 55% to 65.3%. Included studies had an overall low risk of bias, and there was moderate-quality evidence for the assessed outcome (eTable 2 and eFigure 2 in the Supplement).

Quiz Ref IDRandomization to novel oral anticoagulants was associated with a 22% relative reduction in intraocular bleeding compared with warfarin (risk ratio, 0.78; 95% CI, 0.61-0.99) (Figure 2). The absolute risk decrease in intraocular bleeding associated with novel oral anticoagulant use compared with warfarin was 0.8 bleeds per 1000 person-years. There was no significant heterogeneity observed (I² = 4.8%, P = .40), and estimates did not differ materially when random-effects meta-analysis was used (eFigure 3 in the Supplement). The lower risk of intraocular bleeding with novel oral anticoagulants compared with warfarin was similar in patients with atrial fibrillation (risk ratio, 0.81; 95% CI, 0.62-1.07) and in patients with venous thromboembolism (risk ratio, 0.65; 95% CI, 0.37-1.14) (P for heterogeneity = .49) (Figure 3). Similarly, relative effects of novel oral anticoagulants compared with warfarin appeared to be similar regardless of the type of novel oral anticoagulant (P for heterogeneity = .15) (Figure 3). Estimates did not differ materially when random-effects meta-analysis was used (eFigure 4 in the Supplement).

In our analyses, we demonstrate that novel oral anticoagulants reduce the risk of intraocular bleeding by approximately one-fifth compared with warfarin. This finding has significant clinical implications for ophthalmologists managing patients receiving anticoagulation, particularly for those at high risk of intraocular bleeding. Furthermore, our results may have relevance for the perioperative management of patients receiving anticoagulation.

In the only previous meta-analysis on this topic, by Caldeira and colleagues,⁹ the risk of intraocular bleeding was analyzed within subgroups according to the indication for anticoagulation. As a result, the authors concluded that there was no difference in bleeding risk between novel oral anticoagulants and warfarin in patients with atrial fibrillation (risk ratio, 0.84; 95% CI, 0.59-1.19) and in patients with venous thromboembolism (risk ratio, 0.67; 95% 0.37-1.20). While the 95% CIs for these risk estimates cross unity, suggesting a lack of statistical significance, the actual point estimates and wide 95% CIs reported indicate that novel oral anticoagulants may have a moderate benefit (16% and 33% lower risk of intraocular bleeding, respectively). Therefore, their results do not provide reliable evidence of a lack of difference in intraocular bleeding. Indeed, our results taken as a whole suggest the converse. By pooling data across indications for anticoagulation and including additional data from other phase 3 randomized clinical trials, our analysis demonstrates that novel oral anticoagulants reduce the risk of intraocular bleeding by approximately one-fifth compared with warfarin. This difference was not only statistically significant, but we also show that this benefit appears to be similar regardless of the indication for anticoagulation (atrial fibrillation or venous thromboembolism) or the type of novel oral anticoagulant.

The exact mechanism for the reduced risk of intraocular bleeding with novel oral anticoagulants compared with warfarin is uncertain. Similar to the reduced risk of intracranial bleeding, this effect could be partially attributable to the fact that novel oral anticoagulants target only a single site in the coagulation cascade compared with multiple sites with warfarin.²² Furthermore, novel oral anticoagulants are known to have no direct effect on factor VIIa, which may add to their bleeding benefit. Finally, modest time in therapeutic range remains a considerable concern with warfarin, with unmonitored periods of supratherapeutic anticoagulation being another potential contributor to higher rates of bleeding.

Our findings have particular relevance for those at high baseline risk of ocular bleeding. Massive intraocular hemorrhage in patients with exudate age-related macular degeneration (AMD) is often associated with poor visual outcomes, and previous data have demonstrated an increased risk in patients receiving warfarin therapy.⁶ Tilanus and colleagues⁶ found that a patient having AMD with an associated massive intraocular bleed is 11.6 times more likely to be taking an oral anticoagulant and has significantly worse posthemorrhage visual acuity compared with controls. There is also an associated increased risk of subretinal hemorrhage in patients with AMD receiving anticoagulation, which is also known to confer a poorer visual prognosis.⁵ Furthermore, in a recent analysis of 1165 participants in a cohort study within the Comparison of Age-Related Macular Degeneration Treatments Trials, Ying and colleagues²³ found that there was an increased risk of retinal and subretinal hemorrhage in hypertensive patients receiving warfarin therapy, although this increased risk was not noted in normotensive patients receiving anticoagulation. To our knowledge, there are no previous reports in the literature investigating the risk of intraocular and subretinal bleeding in patients receiving novel oral anticoagulants. Therefore, our findings have significant implications for treating clinicians when considering patients with exudative AMD, especially in those who have hypertension. Further studies are required to better evaluate the role of novel oral anticoagulation in these settings.

Each year, approximately 10% of patients receiving anticoagulant agents require treatment interruption for surgical interventions.²⁴ For routine ophthalmic operations, the continuation of anticoagulation agents has not been found to significantly increase the risk of severe sight-threatening hemorrhagic complications; hence, cessation of these agents is generally not recommended.^25-27 However, there are a few studies that have investigated the implications of novel oral anticoagulant therapy during ocular surgery; while the rapid onset and offset action of these agents is potentially advantageous, an understanding of the optimal cessation and recommencement timing is critical to balance bleeding and thrombotic risks.²⁸ The Randomized Evaluation of Long-term Anticoagulation Therapy (RE-LY) study²⁹ comparing dabigatran with warfarin was the only trial included in this meta-analysis that reported bleeding outcomes in patients who underwent cataract surgery. Of the 800 patients treated with dabigatran and 400 patients treated with warfarin, 9.3% underwent cataract surgery during the trial period, and neither group had associated significant bleeding complications, although this study was not specifically designed or powered to evaluate periprocedural bleeding. A small study³⁰ of 36 patients undergoing vitreoretinal surgery maintained on a regiment of novel oral anticoagulants found no associated increased risk of perioperative complications, while high-risk oculoplastic procedures, such as dacryocystorhinostomy and deep orbital and extensive eyelid surgery, require cessation of novel oral anticoagulant therapy 48 hours before surgery.^28,31 Data from the present analyses raise the possibility that novel oral anticoagulants may also be superior to warfarin in the perioperative period. Furthermore, the recent development of previously unavailable reversal agents for novel oral anticoagulants now counters this one advantage that warfarin once had.³² However, additional clinical experience and data are still required to characterize the optimal management strategies when novel oral anticoagulant–associated intraocular bleeding occurs.

Quiz Ref IDOur study has some limitations that warrant recognition. First, intraocular bleeding is uncommon; therefore, the total number of events in any individual trial was low. The pooling of trials in this meta-analysis was thus required to demonstrate a significant lower risk of intraocular bleeding with novel oral anticoagulants compared with warfarin. Second, we did not have access to individual patient data, which may have allowed for an exploration into specific subgroups in which the benefits of novel oral anticoagulants may be more pronounced, such as those with preexisting ophthalmic disease. Third, the short follow-up of included trials precludes any assessment of the long-term intraocular bleeding risks of novel oral anticoagulants. Observational data from registries may be more useful in this regard. Fourth, although we observed no statistically significant difference in the risk of intraocular bleeding according to the indication for anticoagulation or the type of novel oral anticoagulant, we cannot exclude the possibility of a small difference that may be demonstrable with greater statistical power. Fifth, the protocol for this systematic review and meta-analysis was not formally registered; therefore, the objectives and methods are not publicly and explicitly prespecified.

Our data suggest that novel oral anticoagulants reduce the risk of intraocular bleeding by approximately one-fifth compared with warfarin. Similar benefits were seen in both patients treated for atrial fibrillation and venous thromboembolism. Our data have particular implications for patients at higher risk of spontaneous retinal and subretinal bleeding. These findings may also have important implications in the perioperative period, in which the use of novel oral anticoagulants may be superior. Future studies are required to better characterize the optimal management of patients with both ophthalmic disease and cardiovascular comorbidities requiring therapy with anticoagulants.

Accepted for Publication: May 22, 2017.

Corresponding Author: Michelle T. Sun, MBBS, South Australian Institute of Ophthalmology, The University of Adelaide and Royal Adelaide Hospital, Level 8, East Wing, Adelaide, South Australia, Australia 5000 (michelle.sun@adelaide.edu.au).

Published Online: July 6, 2017. doi:10.1001/jamaophthalmol.2017.2199

Author Contributions: Drs Sun and Wong had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Sun, Chan, Selva, Sanders, Wong.

Acquisition, analysis, or interpretation of data: Wood, Casson, Wong.

Drafting of the manuscript: Sun, Wood, Wong.

Critical revision of the manuscript for important intellectual content: Sun, Chan, Selva, Sanders, Casson, Wong.

Statistical analysis: Sun, Wong.

Administrative, technical, or material support: Wood, Casson.

Study supervision: Sun, Chan, Selva, Sanders, Casson, Wong.

Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Dr Sanders reported having served on the advisory board of Biosense-Webster, Medtronic, and St Jude Medical; reported having received lecture or consulting fees from Biosense-Webster, Medtronic, and St Jude Medical; and reported having received research funding from Medtronic, St Jude Medical, Boston Scientific, Biotronik, and Sorin. Dr Wong reported having received lecture or travel funding from Novartis, Servier, Boehringer-Ingelheim, and Medtronic. No other disclosures were reported.

Funding/Support: The study was sponsored by The University of Adelaide, South Australian Institute of Ophthalmology, and Centre for Heart Rhythm Disorders. Dr Sun is supported by an Australian Postgraduate Award. Dr Sanders is supported by a Practitioner Fellowship from the National Health and Medical Research Council of Australia (NHMRC) and The Heart Foundation of Australia. Dr Wong is supported by a Rhodes Scholarship and a Neil Hamilton Fairley Fellowship from the NHMRC.

Role of the Funder/Sponsor: The design, approval, and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, and approval of the manuscript; and the decision to submit the manuscript for publication was executed through the authors as employees of The University of Adelaide and overseen by it.