Introduction

Catheter ablation (CA) of atrial fibrillation (AF) is an effective therapeutic option in symptomatic, drug-refractory AF (1). CA is a technically difficult procedure with potentially serious complications, such as stroke, transient ischemic attack, or systemic embolism, with reported incidence of adverse events as high as 4.6% (2).

The COMPARE (Role of Coumadin in Preventing Thromboembolism in Atrial Fibrillation (AF) Patients Undergoing Catheter Ablation) study established uninterrupted warfarin as the standard of care for patients undergoing CA of AF (3). This randomized controlled trial (RCT) demonstrated that performing this procedure without interrupting oral anticoagulation with warfarin was associated with a decreased risk of stroke and minor bleeding complications.

After the introduction of direct oral anticoagulants (DOACs), several studies tried to establish their noninferiority to vitamin K antagonists (VKAs) in CA of AF, including some RCTs. The VENTURE-AF (Uninterrupted rivaroxaban vs. uninterrupted vitamin K antagonists for catheter ablation in non-valvular atrial fibrillation) was the first RCT that compared an uninterrupted DOAC (rivaroxaban) to an uninterrupted VKA (4). After this study, several others came out comparing outcomes between dabigatran versus VKA and rivaroxaban versus VKA during CA of AF (5–10). Based on these studies, the Heart Rhythm Society/European Heart Rhythm Association published their updated expert consensus statement regarding catheter and surgical ablation of AF in 2017 (1). The current standard of care incorporates the use of uninterrupted VKAs (international normalized ratio goal: 2.0 to 3.0), uninterrupted dabigatran (Class I, Level of Evidence: A), or uninterrupted rivaroxaban (Class I, Level of Evidence: BR).

Despite these studies providing evidence for individual DOACs, data regarding potential benefits of DOACs (as a class) over VKAs during CA of AF are scant. The present meta-analysis builds on the foundation laid by a previous meta-analysis by Romero et al. (11), which included RCTs only, and showed a trend toward fewer major bleeding events (MBEs) in the uninterrupted DOACs group during CA of AF. We thus decided to update our meta-analysis, with the data obtained by the ASCERTAIN (Asymptomatic Cerebral Infarction During Catheter Ablation for Atrial Fibrillation) trial and the recently published ELIMINATE-AF (Uninterrupted edoxaban vs. vitamin K antagonists for ablation of atrial fibrillation) trial (12,13). Our goal was to determine the existence of any statistically significant benefit of uninterrupted DOACs over uninterrupted VKAs for CA of AF.

Methods

The present meta-analysis was performed according to Cochrane Collaboration and Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statements (14). This meta-analysis was registered in PROSPERO (International prospective register of systematic reviews) with registration number CRD42018089183. Updates that reflect new available evidence incorporated into the present analysis were also registered.

Search strategy

We searched PubMed, Embase, and Cochrane Central Register of Clinical Trials (Cochrane Library, Issue 02, 2017) databases from January 2008 to April 2019 to identify RCTs that compared uninterrupted DOACs versus uninterrupted VKAs for CA of AF.

We used the following terms: (“direct oral anticoagulants” OR DOAC OR dabigatran OR rivaroxaban OR apixaban) AND (warfarin OR “vitamin K antagonists” OR VKA) AND (“auricular fibrillation” OR “atrial fibrillation”) AND (ablation OR “catheter ablation”). No language restriction was applied. The reference lists of identified articles were also reviewed for additional sources.

Eligibility criteria

Studies with the following characteristics were considered eligible to be included in this meta-analysis: 1) RCTs that compared uninterrupted DOACs versus uninterrupted VKAs for CA of AF; and 2) Studies that compared the event rates of MBEs and/or minor bleeding and/or thromboembolic events (TEs) and/or incidence of silent cerebral infarction (SCI) detected by diffusion-weighted imaging (DWI) brain magnetic resonance imaging (MRI) between the 2 groups.

Case reports, editorials, reviews, and expert opinions were excluded from our analysis. Abstracts presented at major international conferences that were not published as full papers were also not considered.

Primary and secondary outcomes

The primary outcomes of this study were: 1) MBEs; 2) minor bleeding events; and 3) TEs. The secondary outcome was SCI as detected by post-procedural DWI brain MRI.

MBEs were defined by using the Bleeding Academic Research Consortium (BARC) criteria, with a scale ≥2 being considered major bleeding (see Online Appendix: Material, Definition of Major Bleeding Events). In the case the reported events by the individual studies were reported using different criteria, these were reclassified using BARC, by consensus of the authors of this paper.

Minor bleeding events were all reported as bleeding events that did not fulfill this criterion. TEs were defined as stroke, transient ischemic attack, systemic embolism, or development of an intracardiac thrombus post-procedure. Finally, SCI was defined as a clinically silent new brain lesion detected by DWI brain MRI post-procedurally.

Data extractions and quality appraisal

Three investigators (J.R., R.C.R., and L.D.B.) independently screened all titles and abstracts and manually searched the full text versions of all relevant studies that fulfilled the inclusion criteria. References of the retrieved articles were independently reviewed for further identification of potentially relevant studies. Disagreements were resolved by consensus after discussion (J.R. and R.C.R.). We extracted characteristics of each study, including methodology and baseline patient characteristics, MBEs, minor bleeding events, and TEs. SCI events were also extracted from the relevant studies. If the previously mentioned information was not readily available in the written article, the principal investigator of that particular study was contacted to supply pertinent information.

Quality assessment

The quality and reporting of the included RCTs were assessed using the Cochrane Risk of Bias Tool (15). Six categories were included in the analysis: 1) selection bias: systematic differences between baseline characteristics of the groups that were compared; 2) performance bias: systematic differences between groups in the care that was provided, or in exposure to factors other than the interventions of interest. Blinding was not always possible, as was the case of the RCTs included in our study, because it would have been unethical not to monitor the international normalized ratio in the VKA arm; 3) detection bias: systematic differences between groups in how outcomes were determined. Blinding of outcome assessors might reduce the risk that knowledge of which intervention was received, rather than the intervention itself, because this affected outcome measurement; 4) attrition bias: systematic differences between groups due to withdrawals from a study. Withdrawals from the study led to incomplete outcome data; 5) reporting bias: systematic differences between reported and unreported findings; and 6) other biases: other sources of bias that were relevant only in certain circumstances. The quality of the included RCTs was summarized visually.

Statistical analysis

Descriptive statistics are presented as number of cases for dichotomous and categorical variables. Statistical analysis was performed in line with recommendations from the Cochrane Collaboration and PRISMA guidelines, using Review Manager (RevMan version 5.3, Cochrane Collaboration, 2014, Oxford, United Kingdom) (14). Heterogeneity was assessed using the I² statistics, which is the proportion of total variation observed among the studies attributable to differences between studies, rather than a sampling error (chance). Data was summarized across groups using the Mantel-Haenszel risk ratio (RR) fixed-effect model if I² was <25% (16). We considered I² <25% as low and I² ≥25% as high. The random effects model was used if I² was ≥25%. Publication bias was estimated visually by funnel plots (17).

Results

A total of 741 studies were identified using the specified search criteria (Figure 1). After evaluation of these studies, based on titles and abstracts, 11 RCTs were further analyzed in their full-text versions, 5 of which were discarded, leaving another 6 RCTs that fulfilled the inclusion criteria. These 6 RCTs had a total of 2,256 participants (72.7% male; average age 61.3 ± 2.6 years). Other RCTs were excluded due to a lack of information relevant to our study questions, because they did not follow an uninterrupted DOAC strategy, or their statistical power was poor due to the low number of enrolled patients (18). The summary of the primary and secondary outcomes can be found in the Central Illustration.

Characteristics of included studies

The baseline characteristics of the included trials are summarized in Table 1. Uninterrupted dabigatran was used in 317 (14.1%) patients (8), rivaroxaban in 187 (8.3%) patients (4,13), apixaban in 418 (18.5%) patients (19,20), and edoxaban in 316 (14.0%) patients (12). The remaining patients (n = 1,018; 45.1%) were on uninterrupted VKA. For detailed information regarding duration of oral anticoagulation before CA and the follow-up periods for each study, please refer to Online Table 1. The target international normalized ratio for patients receiving VKA was between 2.0 and 3.0.

Quality assessment and publication bias

Funnel plots did not suggest publication bias for the selected outcomes of minor bleeding events, TE, and SCI (Figure 2). All the RCTs included in this meta-analysis had good methodological quality that indicated a low risk of bias (Figure 3).

Impact on MBEs

We found a statistically significant reduction in MBEs in the uninterrupted DOAC strategy (RR: 0.45; 95% confidence interval [CI]: 0.20 to 0.99; p = 0.05), with a relative risk reduction (RRR) of 55%. The incidence of MBEs in the uninterrupted DOAC strategy was 2.3%, and the incidence of MBEs in the uninterrupted VKA strategy was 5.2%. (Central Illustration, Figure 4).

Impact on minor bleeding events

The minor bleeding rates in the DOAC and VKA groups were 13.9% and 13.7%, respectively. There were no significant differences between groups (RR: 1.12; 95% CI: 0.87 to 1.43; p = 0.39) (Central Illustration, Figure 5).

Impact on TEs

There were no significant differences between groups regarding TEs. The stroke rates in the DOAC and VKA groups were 0.4% and 0.7%, respectively (RR: 0.75; 95% CI: 0.26 to 2.14; p = 0.59) (Central Illustration, Figure 5).

Impact on SCI

Four of the 6 analyzed RCTs included a component to assess for post-CA SCI using DWI brain MRI (12,13,19,20). In our analysis, we could not find any statistically significant difference between the uninterrupted DOAC group (16.3%) and the uninterrupted VKA group (15.4%) regarding SCI as detected by brain MRI (RR: 1.09; 95% CI: 0.80 to 1.49; p = 0.58) (Central Illustration, Figure 6).

Discussion

CA of AF is an important rhythm control strategy for improvement of quality of life in patients with AF. A recent meta-analysis by Turagam et al. (21) showed a mortality benefit of CA of AF in patients with congestive heart failure. Although this procedure carries the risk of embolization, an uninterrupted VKA strategy has been shown to decrease this risk to <1% with no associated increase in the rate of bleeding complications (3).

The use of uninterrupted DOACs during CA of AF has been steadily increasing, mainly due to the published data showing that DOACs have a favorable safety and efficacy profile (5,6). A meta-analysis by Zhao et al. (22), which used observational data in >7,900 patients, showed no difference between groups in preventing TEs; DOAC use was associated with a lower risk of bleeding. The concern of not having readily available reversal agents in case of life-threatening bleeding (e.g., a pericardial tamponade) may be allayed by the Food and Drug Administration approval of andexanet alfa as a reversal agent for rivaroxaban and apixaban, as well as idarucizumab for dabigatran (23,24).

Our meta-analysis used all published RCT data that compared the outcomes of uninterrupted DOACs versus uninterrupted VKAs on MBEs, minor bleeding events, TEs, and SCI. We included a total of 2,256 participants who underwent CA of AF. The pertinent findings of this study were the following:

Given our findings, we believe it is reasonable and beneficial to offer patients who need to undergo CA of AF uninterrupted anticoagulation with DOACs as first-line therapy. These conclusions fall in line with the recent consensus statement on the use of uninterrupted DOACs for CA of AF that gave a Class I recommendation for the use of uninterrupted dabigatran or rivaroxaban (1). DOACs are more convenient for both the patient and the physician, have fewer interactions with medications and food, and do not require frequent blood testing to monitor the international normalized ratio.

The results of our meta-analysis proved that there was a significant decrease in the risk of MBEs between uninterrupted DOACs and uninterrupted VKAs during CA of NVAF. A previous study by our research group failed to reach statistical significance, which was likely due to a lack of statistical power because of its smaller sample size (11). This lack of power was solved with the inclusion of the data published by both the ASCERTAIN and the ELIMINATE-AF trials (12,13).

Two previous meta-analyses found results similar to ours using different methodologies. A meta-analysis by Zhao et al. (25), which analyzed 6 RCTs, found a significantly decreased risk of MBEs with DOACs. However, 1 of the included papers, by Zhu et al. (18), whose main goal was to analyze the metabolic effects of rivaroxaban (measuring outcomes such as serum total protein, albumin, and globulin levels), had poor statistical power due to the extremely small number of enrolled patients (30 patients each in the rivaroxaban and the VKA arms). Zhao et al. (25) also used inaccurate MBE rates in their analysis when referring to the ASCERTAIN trial, because they used preliminary data published in the Circulation Supplement for Abstracts presented at the American Heart Association 2016 meeting (26). The final, definitive data of the ASCERTAIN trial were published as a full paper in JACC: Clinical Electrophysiology in December 2018 (13).

A second meta-analysis by Cardoso et al. (27) found significantly decreased MBEs using uninterrupted DOACs. The authors included a total of 12 studies in its analysis, 3 of which were RCTs (4,8,19). The investigators found significantly decreased MBEs in a dedicated quantitative synthesis of studies with a lower risk of selection bias, which included the previously mentioned 3 RCTs and 3 additional studies: a multicenter prospective registry that assessed uninterrupted apixaban versus uninterrupted VKAs by Di Biase et al. (28); a study by Dillier et al. (10) that matched patients in the intervention arm (uninterrupted rivaroxaban) with control subjects (uninterrupted VKA) by age, sex, and type of AF (10); and a multicenter prospective registry that matched patients on uninterrupted rivaraxoban with patients on uninterrupted VKAs by Lakireddy et al. (9).

We want to highlight that our meta-analysis was the only one to include only RCTs in its analysis and to include all DOACs currently on the market, and to reclassify all bleeding events using the BARC classification.

Study limitations

Most of the RCTs included in our meta-analysis were exploratory trials with administratively determined trial sizes, because the sample required to provide sufficient power to establish formal noninferiority would have made them unfeasible. The AXAFA (Apixaban in patients at risk of stroke undergoing atrial fibrillation ablation) trial was the exception, because it was designed from the beginning to accumulate sufficient events for a formal noninferiority analysis by selecting patients with at least 1 risk factor for stroke (20).

Another limitation of our study was that the frequency of TEs was quite low; it was not specifically powered to measure this outcome, which might predispose to a type II error.

The results of our study should generate interest in the design of larger, well-designed RCTs comparing the safety and efficacy, not only of uninterrupted DOACs versus uninterrupted VKAs in CA of AF, but also comparing the differences in outcomes when using the different DOACs now available on the market. As in our previous study, we believe the number of patients assessed for SCI was still insufficient to determine if there was a difference between uninterrupted DOACs and uninterrupted VKAs. Finally, we believe that the main limitation of our analysis lies in the low number of events.

Conclusions

An uninterrupted DOACs strategy for CA of NVAF carries a lower risk of MBEs compared with uninterrupted VKAs. There were no significant differences among the other outcomes (minor bleeding events, TE, SCI) between groups. We thus believe DOACs should be considered a first option to offer to patients undergoing CA of AF, over traditional VKAs.

Footnotes