Anticoagulation in catheter ablation (CA) of atrial fibrillation (AF) is of paramount importance for prevention of thromboembolic events, and recent studies favor uninterrupted vitamin K antagonists (VKAs). We aimed to compare the efficacy and safety of new oral anticoagulants (NOACs) to uninterrupted VKAs for anticoagulation in CA by performing a meta-analysis. PubMed, EMBASE, the Cochrane Library, and Clinicaltrials.gov databases were searched for studies comparing NOACs with uninterrupted VKAs in patients who underwent CA for AF from January 1, 2000, to August 31, 2015. Odds ratio (OR) and Peto’s OR (POR) were used to report for event rates >1% and <1%, respectively. A total of 11,686 patients with AF who underwent CA in 25 studies were included in this analysis. There was no significant difference between NOACs and uninterrupted VKAs in occurrence of stroke or transient ischemic attacks (POR 1.35, 95% CI 0.62 to 2.94) and major bleeding (POR 0.87, 95% CI 0.58 to 1.31), which were consistent in subgroup analysis of interrupted and uninterrupted NOACs. A lower risk of minor bleeding was observed with NOACs (OR 0.80, 95% CI 0.65 to 1.00), and no major differences were observed for the risk of thromboembolic events, cardiac tamponade or pericardial effusion requiring drainage, and groin hematoma. NOACs, whether interrupted preprocedure or not, were associated with equal rates of stroke or TIA and major bleeding complications and less risk of minor bleeding compared with uninterrupted VKAs in CA for AF.
Catheter ablation (CA) is a well-established rhythm-control strategy for patients with drug-refractory, symptomatic atrial fibrillation (AF); however, there is substantial risk of thromboembolic (TE) events perioperatively. Rigorous anticoagulation is of paramount importance to prevent these deadly complications, whereas optimal anticoagulant strategy in the setting has not reached a consensus. Recent studies favored uninterrupted vitamin K antagonist (VKAs) therapy because of significantly less risk of TE events and bleeding complications observed with this strategy compared with interrupted VKAs. With the advent of new oral anticoagulants (NOACs), increasing numbers of patients with AF are on NOACs at the time a decision to proceed with CA is made. However, it is not clear whether NOACs still hold the same efficacy and safety compared with uninterrupted VKAs in ablation procedure; therefore, we aimed to systematically evaluate all available evidence in the setting in this study.
PubMed, EMBASE, the Cochrane Library, and Clinicaltrials.gov databases were searched to identify published reports comparing NOACs to uninterrupted VKAs in patients who underwent CA for AF from January 1, 2000, to August 31, 2015. The main key words we used were “atrial fibrillation,” “catheter ablation,” “anticoagulants,” “warfarin,” “dabigatran,” “rivaroxaban,” “apixaban,” “factor Xa inhibitor,” and “factor IIa inhibitor” (refer to Supplementary File 1 for details of search strategy). The report published in English language was restricted in our search. Besides, we also manually searched reviews, meta-analyses and reference lists of all retrieved reports for additional eligible studies not found in our initial electronic database search.
Two reviewers (SW and HBW) screened and identified studies that fit the following inclusion criteria: (1) studies published in English language, (2) either full-length report or conference abstract, (3) comparing NOACs with VKAs in patients with AF presenting for CA, (4) VKAs or warfarin were continuously used through the whole procedure, (5) reporting outcomes of interest: stroke, transient ischemic attack (TIA), major bleeding, TE events, minor bleeding, cardiac tamponade or pericardial effusion requiring drainage, and groin hematoma, (6) number of participants in studies >200. The exclusion criteria were as follows: (1) studies included patients with AF both for cardioversion and ablation but without reporting respective outcomes, (2) a lack of using VKA/NOAC as comparator, (3) VKAs were interrupted and bridged with low molecular weight heparin, or unfractionated heparin during the procedure, (4) studies evaluating antiplatelet agents, (5) starting of anticoagulant treatment only after ablation, (6) evaluation of efficacy restricted to laboratory or imaging end points without reporting of desired clinical outcomes, and (7) number of participants in studies <200. We compared reports from the same investigator and excluded those identical records from the same trial or study to minimize data duplication as a result of multiple reporting. All eligible clinical trials, including case-control studies, cohort studies, and randomized controlled trials (RCTs) were formally reviewed.
Two reviewers (SW and HBW) evaluated the study quality using Newcastle-Ottawa Quality Assessment Scale for observational studies and Delphi consensus criteria for RCTs. Baseline characteristics and numbers of events were extracted from the eligible studies by 2 independent reviewers (SW and HBW) and checked by another (YMY). Any disagreement was resolved through discussion with additional investigator (YMY). We abstracted and tabulated details for each study on senior investigator, year of publication, publication type, design, details of the anticoagulant protocols, follow-up data, baseline characteristics of participants, and information related to clinical outcomes.
The primary efficacy outcome was stroke or TIA. The primary safety outcome was major bleeding. Secondary outcomes included TE events, minor bleeding, cardiac tamponade or pericardial effusion requiring drainage, and groin hematoma. Stroke was defined as sudden onset of a focal neurologic deficit in a location consistent with the territory of a major cerebral artery, whereas outcomes as positive imaging finding on magnetic resonance imaging or computed tomography without clinical symptoms were excluded for this analysis. TE events included ischemic stroke, TIA, pulmonary embolism, and systemic embolism such as deep vein thrombosis and peripheral arterial embolism. Major bleeding was defined as excessive bleeding requiring transfusion or intervention, cardiac tamponade or pericardial effusion requiring drainage, intracranial hemorrhage, hemothorax, retroperitoneal bleeding, massive hemoptysis, and bleeding extending hospital stay. Minor bleeding was defined as the occurrence of a hematoma or any bleeding that did not require transfusion or other particular interventions or prolong hospitalization. The longest available follow-up data from individual studies were selected into analysis.
The statistical analysis was performed according to the recommendations from Cochrane Collaboration using Review Manager, version 5.3, for Mac (The Nordic Cochrane Center, The Cochrane Collaboration, 2014, Copenhagen, Denmark). Outcomes were reported as odds ratio (OR) and their respective 95% CIs. We calculated Peto’s OR (POR) estimates and associated 95% CI using fixed effects models for event rates <1%. The Mantel–Haenszel method using fixed effects model and the DerSimonian–Laird method using random effects model were respectively used to combine results from individual studies for event rates >1%, depending on study heterogeneity. Subgroup analyses, according to different strategies of NOACs, that is, interrupted or not preprocedure, were also conducted for the primary outcomes. We used the Cochrane Q statistic and I 2 statistic to assess heterogeneity across studies. A p value <0.05 of Cochran Q statistic and I 2 statistic >50% were considered statistical significant to demonstrate heterogeneity for each outcome in this meta-analysis. Publication bias was assessed by visual inspection of the symmetry in Begg’s funnel plot and Egger’s linear regression test. All the p values were 2-tailed with statistical significance level at 0.05, and associated 95% CIs were also calculated.
Sensitivity analyses were performed by sequentially decreasing each individual study for the primary efficacy and safety outcomes. Further sensitivity analysis was carried out as follows: (1) full-text published studies, (2) studies with low/intermediate risk of bias, and (3) studies whose follow-up was at least 30 days. Stata, version 12.1, for Mac (StataCorp LP, College Station, Texas) was used for sensitivity analysis.
In addition, we used the trial sequential analysis (TSA) to calculate a diversity-adjusted required information size to avoid false positive and test for futility of this meta-analysis by establishing monitoring boundaries using O’Brien–Fleming method according to the required information size to detect or reject an intervention effect of a 150% increase in POR with control group event proportion (CEP) of 0.19% for stroke or TIA outcome, and intervention group event proportion of 1.6% and CEP of 1.1% for major bleeding outcome to challenge ORs estimated from conventional meta-analyses with a type II error of 20% (power of 80%). The TSA-adjusted 95% CIs were also calculated and provided. TSA, version 1.0, for Mac (Copenhagen Trial Unit, Center for Clinical intervention Research, Rigs Hospitalet, Copenhagen, Denmark) was used for TSA analyses.
The flowchart of detailed search process was illustrated in Figure 1 . From a total of 2,692 potentially relevant reports, 25 citations, involving 11,686 patients, fulfilled selection criteria and were finally included in this meta-analysis. Of the identified studies, only one was RCT, and the remaining were observational studies. Eighteen studies were published as full-text studies, whereas the rest were conference abstracts. Baseline data of selected studies, anticoagulation strategies periprocedure and clinical information of patients in the 2 groups were listed in Supplementary File 2 . Difference in terms of overall design and outcome definitions existed in these studies. NOACs were continuously used during ablation in 11 studies, whereas interrupted in 14 studies. Unbalanced clinical characteristics of patients between the 2 groups presented in part of those studies with relevant information reported. Particularly, 2 of 25 studies were considered as high risk of bias after quality assessment.
Quantitative synthesis indicated that NOACs had comparable efficacy to uninterrupted VKAs in prevention of ischemic stroke or TIA in patients presenting with CA for AF (0.28% vs 0.19%; POR 1.35, 95% CI 0.62 to 2.94, p = 0.44; I 2 = 38%; Figure 2 ). And the incidence of major bleeding was also well matched between the 2 groups (1.10% vs 1.60%; POR 0.87, 95% CI 0.58 to 1.31, p = 0.51; I 2 = 9%; Figure 3 ). Subgroup analyses showed that both interrupted and uninterrupted NOAC agents had similar efficacy and safety as VKAs as anticoagulant therapy in CA for AF, despite a higher heterogeneity existed in interrupted NOACs subgroup ( Figures 2 and 3 ).
For the outcome of TE events reported in 19 studies, there was no statistical difference between the NOACs group and VKAs group (0.31% vs 0.17%; POR 1.63, 95% CI 0.72 to 3.67, p = 0.24, I 2 = 0%; Figure 4 ). For the outcome of minor bleeding, a significantly lower risk was observed in the NOACs group (3.68% vs 4.38%; OR 0.80, 95% CI 0.65 to 1.00, p = 0.05, I 2 = 0%; Figure 4 ). Besides, no statistically significant difference was indicated in cardiac tamponade or pericardial effusion requiring drainage (0.85% vs 0.89%; OR 1.11, 95% CI 0.72 to 1.71, p = 0.64, I 2 = 14%; Figure 5 ) and groin hematoma (2.98% vs 3.14%; OR 0.90, 95% CI 0.71 to 1.15, p = 0.41, I 2 = 0%; Figure 5 ). All outcomes were summarized in Supplementary File 3 .
Sensitivity analyses failed to identify any of the individual study as having influenced the outcomes to a significant extent, and the results of subgroups were also concordant with the overall analyses ( Supplementary File 4 ). No evidence of publication bias existed by visual inspection of Funnel plots ( Figure 6 ) and Egger’s test ( t = 0.48, p = 0.640).