Comparison of Major Bleeding Events of Uninterrupted Novel Oral Anticoagulants Versus Uninterrupted Vitamin K Antagonist During Catheter Ablation of Atrial Fibrillation: A Meta-analysis of Randomized Controlled Trials

Background: Previous meta-analyses comparing ecacy and safety of uninterrupted novel oral anticoagulants (NOACs) versus uninterrupted vitamin K antagonist (VKA) during catheter ablation (CA) of atrial brillation (AF) had no consensus in major bleeding, and didn’t perform subgroup analyses for different types of major bleeding events. This meta-analysis was performed to comprehensively evaluate the risk of major bleeding events of these two anticoagulant strategies during CA of AF. Methods: We searched online databases for randomized controlled trials that compared major bleeding events of uninterrupted NOACs and VKA during CA of AF up to January 2021. A xed-effect model was used if the chi-squared test P-value was > 0.10 and I 2 was < 50%, otherwise a random- effect model was used. Results: Six published studies including 2392 patients were identied for inclusion in the analysis. The overall incidence of major bleeding events was lower in the NOACs group than in the VKA group (OR = 0.56, 95% CI = 0.34 – 0.93, I 2 = 38%, P = 0.15). Subgroup analyses showed that the incidence of severe puncture site complications was lower in the NOACs group than in the VKA group (OR = 0.53, 95% CI = 0.30 – 0.96, I 2 = 16%, P = 0.32). But the incidence of cardiac tamponade (OR = 0.53, 95% CI = 0.23 – 1.26, I 2 = 0%, P = 0.46), intracranial (OR = 0.25, 95% CI = 0.03 – 2.23, I 2 = 0%, P = 0.82) and gastrointestinal bleeding (OR = 0.98, 95% CI = 0.18 – 5.39, I 2 = 0%, P = 0.43) had no statistically signicant differences between the two groups. Conclusion: This meta-analysis suggests that compared to uninterrupted VKA, uninterrupted NOACs are superior in major bleeding and severe puncture site complications during CA of AF, but are not


Background
Catheter ablation (CA) is an effective therapeutic option for patients with atrial brillation (AF) [1] . Cardiovascular guidelines recommended performing AF ablation with uninterrupted vitamin K antagonist (VKA) or uninterrupted novel oral anticoagulants (NOACs) as anticoagulant strategies [2,3] . NOACs don't need to monitor international normalized ratio (INR) and are more convenient than VKA, so they are increasingly used in recent years. At present, two latest meta-analyses including the same studies suggested that in AF patients undergoing CA, there was no difference between uninterrupted NOACs and uninterrupted VKA with regards to stroke and transient ischemic attack, however they had different conclusions with regards to major bleeding events [4,5] . In addition, neither of them did subgroup analyses for speci c major bleeding events. We therefore performed this meta-analysis to compare the overall and different types of major bleeding events of uninterrupted NOACs versus uninterrupted VKA during CA of AF.

Data extraction and quality assessment
We extracted the following information from each included study: (1) study population sample size and characteristics; (2) information on anticoagulant therapy; (3) the number of different types of major bleeding events from the start of CA to the end of follow-up. We used Cochrane Risk of Bias Tool [6] to assess the risk of bias of the included studies. One reviewer (QY) abstracted the data, and then the other (YD) checked the documentation. They nally reached an agreement on the data by consensus.

Statistical analysis
We performed the statistical calculations with RevMan version 5.3 (The Cochrane Collaboration, Oxford, UK). We calculated the odds ratio (OR) and 95% con dence interval (CI) in each trial separately, and for combinations of studies according to xed-effect or random-effect models. We used the chi-squared test to assess heterogeneity and I 2 to quantify heterogeneity. If the chi-squared test P-value was > 0.10 and I 2 was < 50%, we analyzed the data using a xed-effect model (the Mantel-Haenszel method), otherwise we used a random-effect model [7,8] .

Characteristics of the included studies
Six published studies [9][10][11][12][13][14] including 2392 patients were identi ed for inclusion in the analysis. The process of study selection is summarized in Figure 1. The target INR of 2.0-3.0 was used in the warfarin arm in all studies. The characteristics of the studies included in the meta-analysis were shown in Table 1.

Results of the meta-analysis
The major bleeding events of uninterrupted NOACs and uninterrupted warfarin groups were reported in all included studies. Five studies [9][10][11]13,14] applied the International Society on Thrombosis and Hemostasis (ISTH) de nition [15] for major bleeding, except the ASCERTAIN study [12] . There was a moderate heterogeneity, so a xed-effect model was used. The overall incidence of major bleeding events was lower in the NOACs group than in the VKA group (OR = 0.56, 95% CI = 0.34 -0.93, I 2 = 38%, P = 0.15). When the ASCERTAIN study [12] was excluded, the incidence of major bleeding events still had no statistically signi cant differences between the two groups (OR = 0.53, 95% CI = 0.32 -0.88, I 2 = 43%, P = 0.13). Results are shown in Figure 2.

Subgroup analysis
Six included studies all reported different types of major bleeding events, and these were shown in Table 2. Subgroup analyses of cardiac tamponade, severe puncture site complications, intracranial bleeding and gastrointestinal bleeding were performed to reduce the heterogeneity and provide some evidences for clinical work. These subgroup analyses all had a low heterogeneity, so a xed-effect model was used.

Intracranial bleeding
Two [11,13] studies reported intracranial bleeding. The ELIMINATE-AF study [14] also reported a case of intracranial bleeding in the NOACs group, but it happened before CA. The incidence of intracranial bleeding had no statistically signi cant differences between the NOACs and VKA groups (OR = 0.25, 95% CI = 0.03 -2.23, I 2 = 0%, P = 0.82). Results are shown in Figure 5.

Quality assessment and bias assessment
All the included studies are RCTs, so they have high quality. Risk of bias was assessed using the methods described in Cochrane collaboration's handbook and results are summarized in Figure 7.

Discussion
At present, CA of AF as an important rhythm control strategy is widely used in clinic [1,2] . Cardiovascular guidelines recommended performing AF ablation with uninterrupted VKA or uninterrupted NOACs in order to reduce the risk of thromboembolism complications [2,3] . The main complication of uninterrupted anticoagulation during CA is bleeding, and major bleeding events, such as cardiac tamponade, can be very serious even life-threating [17] . So it is very important to optimize perioperative anticoagulant therapy to reduce major bleeding events. The metaanalysis by Romero et al. [5] including six RCTs showed that uninterrupted NOACs appeared to be safer than uninterrupted VKA with a decreased rate of major bleeding events. However the other meta-analysis by Brockmeyer et al. [4] including the same RCTs showed that uninterrupted NOACs was not superior to VKA with regards to major bleeding. We found that the two different results were due to their different de nitions of major bleeding. In additon, neither of them did subgroup analyses for speci c bleeding events, so we performed this meta-analysis.
Our meta-analysis also included six RCTs [9][10][11][12][13][14] . Different from the previous two meta-analyses, we used a xedeffect model rather than a random-effect model, for the reason that the chi-squared test P-value was > 0.10 and I 2 was < 50%. Except the the ASCERTAIN study [12] , they all used the ISTH standard to de ne major bleeding [15] . No matter whether the ASCERTAIN study was included, the result showed that uninterrupted NOACs was superior to VKA, especially dabigatran (1.6% vs. 6.9%) [11] . Then we did subgroup analyses of speci c major bleeding events, which further reduced heterogeneity. Cardiac tamponade is a relatively infrequent and potentially fatal complication, and sometimes needs emergency surgical intervention [17,18] . Five [10][11][12][13][14] studies reported cardiac tamponade, which only needed pericardial drainage and administration of protamine and vitamin K. Subgroup analysis showed that uninterrupted NOACs was not superior to uninterrupted VKA. However the incidence of cardiac tamponade was relatively lower in the NOACs group than in the warfarin group (0.7% vs. 1.3% ), especially in dabigatran (0.3% vs. 1.9% ) and apixaban (0.6% vs. 1.6% ) groups. The largest sample size of these included studies just exceeded 600. The effective of NOACs may emerge as the number of large scale RCTs increasing.
Puncture site complications are most common during CA of AF. Five [9,[11][12][13][14] studies reported severe puncture site complications, which were classi ed as major bleeding events. Subgroup analysis showed that uninterrupted NOACs, especially dabigatran, were superior to VKA in severe puncture site complications. The intracranial and gastrointestinal bleeding are infrequent complications. Both of them were reported only in two studies. Subgroup analyses showed that uninterrupted NOACs was not superior to uninterrupted VKA in intracranial and gastrointestinal bleeding during CA of AF.

Limitations
This meta-analysis has the following limitations. Only published RCTs were included in our meta-analysis, so publication bias was unavoidable. The sample size of the included studies was limited, and the number of studies in some subgroups were limited. These were the main limitations of our analysis, so the results were less persuasive.

Conclusion
Compared to uninterrupted VKA, uninterrupted NOACs are superior in major bleeding and severe puncture site complications during CA of AF, but are not superior in cardiac tamponade, intracranial and gastrointestinal bleeding.

Declarations Acknowledgments
None.
Authors' contributions QY conceived the study, participated in the design, collected the data, performed statistical analyses and drafted the manuscript. YD conceived the study, collected the data, and helped to draft the manuscript. XFC and JLZ performed statistical analyses and helped to draft the manuscript. All authors read and approved the nal manuscript.

Funding
None.

Availability of data and materials
The datasets generated and analyzed during the current study are available from the corresponding author on reasonable request.