Non‐vitamin K antagonist oral anticoagulants versus vitamin K antagonists in atrial fibrillation patients with previous stroke or intracranial hemorrhage: A systematic review and meta‐analysis of observational studies

Abstract Several observational studies have compared the effectiveness and safety outcomes between nonvitamin K antagonist oral anticoagulants (NOACs) and vitamin K antagonists (VKAs) in atrial fibrillation (AF) patients with a history of either stroke/transient ischemic attack (TIA) or intracranial hemorrhage. Therefore, our current meta‐analysis aimed to address this issue. The Cochrane Library, PubMed, and Embase databases were systematically searched until December 2020 for all relevant observational studies. We applied a random‐effects model to pool adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) for this meta‐analysis. A total of 10 studies were included. Among patients with a history of stroke/TIA, the use of NOACs versus VKAs was associated with decreased risks of stroke (HR, 0.82, 95% CI 0.69–0.97), systemic embolism (HR, 0.73, 95% CI 0.61–0.87), all‐cause death (HR, 0.87, 95% CI 0.81–0.94), major bleeding (HR, 0.77, 95% CI 0.64–0.92) and intracranial hemorrhage (HR, 0.54, 95% CI 0.38–0.77). Among patients with a history of intracranial hemorrhage, the use of NOACs versus VKAs was associated with reduced risks of stroke (HR, 0.81, 95% CI 0.68–0.95), all‐cause death (HR, 0.68, 95% CI 0.49–0.94), and intracranial hemorrhage (HR, 0.66, 95% CI 0.51–0.84). Compared with VKAs, the use of NOACs exhibited superior efficacy and safety outcomes in AF patients with previous stroke/TIA, and the use of NOACs was associated with reduced risks of stroke, all‐cause death, and intracranial hemorrhage in patients with a history of intracranial hemorrhage.


| INTRODUCTION
Oral anticoagulants (OACs) have been extensively prescribed for patients with atrial fibrillation (AF) as the first-line treatment to prevent stroke or other systemic embolisms. 1,2 Vitamin K antagonists (VKAs), predominantly warfarin, were the only available class of oral anticoagulants over the past decades until 2009 when nonvitamin K antagonist oral anticoagulants (NOACs, including dabigatran, apixaban, rivaroxaban, and edoxaban) were introduced. 3 Several studies have been conducted to compare the effectiveness and safety of NOACs versus VKAs among patients with AF.
Although it has been well-demonstrated by previously published meta-analyses that NOACs have an advantageous risk-benefit profile in stroke prophylaxis, intracranial hemorrhage, and all-cause mortality in comparison with warfarin among patients with AF, 3-6 the heterogeneity of the effectiveness and safety between NOACs and VKAs has not been revealed for a specific subgroup of patients, namely AF patients with a past medical history of stroke/transient ischemic attack (TIA), or intracranial hemorrhage. Given that patients in this subgroup are susceptible to recurrent cerebrovascular ischemic events 7 as well as intracranial hemorrhage, 8 which is the most lethal adverse effect of anticoagulants, 9 it is thus of significance to scrutinize their different responses to NOAC and VKA anticoagulant therapies.
In this meta-analysis of real-world studies, we address the discrepancies in effectiveness and safety outcomes between NOACs and VKAs in AF patients who have previously been diagnosed with a stroke/TIA, or intracranial hemorrhage.

| METHODS
Our current meta-analysis was performed based on the Cochrane handbook for systematic reviews. The results of this study were arranged based on the Preferred Reporting Items for Reporting Systematic Reviews and Meta-analyses (PRISMA). Given that no patients were involved in the establishment of the research question, the outcome measures, and the design or the implementation of this metaanalysis, ethical approval was necessary. The data, methods, and materials of this study are available to others for purposes of reproducing the results or replicating procedures by contacting the corresponding author.

| Search strategy
The three electronic databases (the Cochrane Library, PubMed, and Embase) were systematically searched up to December 2020 by two reviewers. We potentially included studies that evaluated the comparisons of effectiveness and safety between NOACs and VKAs in AF patients with a history of stroke and/or intracranial hemorrhage. We used the following search items including: (1) atrial fibrillation; AND (2) non-vitamin K antagonist oral anticoagulants OR direct oral anticoagulants OR dabigatran OR rivaroxaban OR apixaban OR edoxaban; AND (3) vitamin K antagonists OR warfarin OR coumadin OR acenocoumarol OR phenprocoumon OR fluindione OR phenindione OR anisindione. Cross-referencing was also applied to identify potentially missed studies. The reference lists of meta-analyses and systematic reviews were reviewed and retrieved for more studies. We applied no restrictions on language in the searches.

| Study eligibility
Eligible studies were included if they met all of the following criteria: (1) observational studies focusing on nonvalvular AF patients with a history of stroke/TIA or patients with a history of intracranial hemorrhage; (2) comparisons of outcomes between NOACs (dabigatran, rivaroxaban, apixaban, or edoxaban) and VKAs (warfarin, coumadin, phenprocoumon, acenocoumarol, fluindione, phenindione, or anisindione); (3) the effectiveness outcomes included stroke, systemic embolism, and all-cause death, while the safety outcomes included major bleeding, intracranial hemorrhage, and gastrointestinal bleeding.
We only included studies that reported the effect estimates: adjusted hazard ratios (HRs) and 95% confidence intervals (CIs). If one study reported adjusted HRs in multiple models, the most adjusted model was included.
We applied the following exclusion criteria: (1) studies were duplicated publications; (2) publications had no relevant data (e.g., reviews, case reports, editorials, and comments); (3) studies focused on AF patients after interventions, such as cardioversion, ablation, or leftatrial appendage closure.

| Study selection and data collection
The retrieved pieces of literature were imported into the NoteExpress V3.0 software (Beijing Aegean Sea Music Technology Co., Ltd.; http://www.inoteexpress.com/aegean/), and we deleted duplicate records. Two reviewers independently screened the titles and abstracts of the retrieved records. Then, the full texts of the records were reviewed to identify all potentially eligible studies. Any disagreement was addressed by discussion or was resolved by an additional reviewer. For each included study, we mainly collected the following data: the first author and publication year, study design, data source, inclusion period, type of NOACs and VKAs, the reported effectiveness and safety outcomes, and the adjusted effect estimates. Two reviewers independently collected the data and compared the results to ensure coherence.

| Quality assessment
For the observational studies, two reviewers independently evaluated the methodological quality using the Newcastle-Ottawa Scale (NOS) tool. This tool involved three blocks: the selection of cohorts, the comparability of cohorts, and outcome evaluation. A study can be

| Statistical analysis
The statistical analyses were performed using Review Manager 5.  and three studies 20-22 focused on AF patients with a history of stroke/TIA, and a history of intracranial hemorrhage, respectively, and one study assessed these two populations. 17 The baseline characteristics of the included studies are shown in Table 1 and Supplemental   Tables 1-2. Regarding the quality assessment, the 10 observational studies exhibited acceptable quality.

| Sensitivity analysis
After we excluded one study at a time, the corresponding results of this meta-analysis were stable. In addition, the corresponding results did not change substantially when we reperformed the analyses by using a fixed-effects model.

| DISCUSSION
To compare the effectiveness and safety outcomes between NOACs and VKAs, our meta-analysis pooled the data from seven [13][14][15][16][17][18][19] and stroke/TIA or intracranial hemorrhage. 11 Previous studies have elaborated that this superiority is predominantly attributed to significant prevention against hemorrhagic stroke. Given that hemorrhagic stroke is included in both stroke and intracranial hemorrhage, NOACs consequently reduce their risk profiles by halving the risk of hemorrhagic stroke. 3 As the most lethal complication of anticoagulant treatment, intracranial hemorrhage is a well-recognized factor in risk-benefit assessment for ischaemic stroke prophylaxis among patients with AF. 8 34 and low-dose edoxaban. 35 Our meta-analysis provides more robust estimates to detect differences in secondary outcomes and subgroups.

| Limitations
Some limitations have been identified in our study. First, our statistical analysis was performed without individual participant data for all the included observational studies. Although each study's methodological quality was evaluated by using the NOS tool, we pooled the data from these studies, of which the quality and robustness were unavoidably variable and inconsistent. Second, the severity, imaging, and functional disabilities of prior stroke/TIA or intracranial hemorrhage were not addressed and adjusted, which might have con-

| CONCLUSION
Compared with VKAs, the use of NOACs exhibited superior efficacy and safety outcomes in AF patients with a history of stroke/TIA, and the use of NOACs was associated with reduced risks of stroke, allcause death, and intracranial hemorrhage in patients with a history of intracranial hemorrhage.

CONFLICT OF INTEREST
All authors declare that they have no potential conflicts of interest that might be relevant to the contents of this review.

DATA AVAILABILITY STATEMENT
Availability of data and materials have been described in the manuscript.They are freely available to any scientist who wishes to use them without breaching participant confidentiality.