OUP user menu

Safety and efficacy of interrupted dabigatran for peri-procedural anticoagulation in catheter ablation of atrial fibrillation: a systematic review and meta-analysis

Aref A. Bin Abdulhak, Abdur Rahman Khan, Imad M. Tleyjeh, John A. Spertus, Susan U. Sanders, Kristy E. Steigerwalt, Musa A. Garbati, Reem A. Bahmaid, Alan P. Wimmer
DOI: http://dx.doi.org/10.1093/europace/eut239 1412-1420 First published online: 16 August 2013


Aims To examine the safety (defined as bleeding risk) and efficacy (defined as prevention of thromboembolic events) of interrupted dabigatran for peri-procedural anticoagulation in catheter ablation (CA) of atrial fibrillation (AF) in comparison with warfarin.

Methods and results Reviewers independently searched literature databases from January 2010 through April 2013 for studies comparing the safety and efficacy of dabigatran and warfarin in CA of AF and extracted pre-defined data. The Mantel–Haenszel method was used to pool data of bleeding and thromboembolism outcomes into random and fixed effect model meta-analyses, respectively. Odds ratios (ORs), and risk difference (RD) analysis when studies reported no events in either arm, were used to generate an overall effect estimate of both outcomes. Publication bias and heterogeneity were assessed by contour funnel plot and the I2 test, respectively. Nine citations, including 3036 patients (1073 dabigatran), met the inclusion criteria. There was no significant difference between interrupted dabigatran and warfarin therapy in CA of AF in occurrence of bleeding [dabigatran 58 (5.4%), warfarin 103 (5.2%); OR 0.92 (95% confidence interval (CI) 0.55–1.45); χ2 = 13.03—P = 0.11; I2 = 39%] or thromboembolism [dabigatran 5 (0.4%), warfarin 2 (0.1%); OR 2.15 (95% CI—0.58–7.98); χ2 = 2.14, P = 0.54; I2 = 0%; RD 0.00 (95% CI—0.00 to 0.01); χ2 = 3.37, P = 0.81; I2 = 0%]. Analysis of pre-defined subgroups (published articles vs. abstracts), sensitivity analyses (interrupted warfarin, USA studies, and Japanese studies) and fixed effect model analyses showed similar results. Heterogeneity was mild in the bleeding outcome analysis and zero in thromboembolism. There was no evidence of publication bias in either meta-analysis.

Conclusion Meta-analysis of currently available studies showed no significant difference in bleeding and thromboembolism between interrupted dabigatran and warfarin therapy in CA of AF. Dabigatran appears to be safe and effective for peri-procedural anticoagulation in CA of AF.

  • Dabigatran
  • Warfarin
  • Catheter ablation
  • Atrial fibrillation

What's new?

  • Dabigatran is a novel oral anticoagulant approved for stroke prevention in non-valvular atrial fibrillation (AF).

  • Safety and efficacy of dabigatran in catheter ablation of AF is still not yet defined.

  • This first meta-analysis of the available observational studies provides evidence that dabigatran is safe and effective in catheter ablation of AF.


Catheter ablation (CA) for atrial fibrillation (AF) is an effective rhythm-control strategy for selected patients. However, there is risk of thromboembolism because the procedure involves ablation in the systemic circulation and often conversion of AF to sinus rhythm.1 The adoption of more aggressive peri-procedural anticoagulation has been associated with a decline in the rate of transient ischaemic attack (TIA) and stroke.2 Although the use of intra-procedural heparin is the standard of care, several different pre- and post-procedural anticoagulation strategies have been used.3 Pre-ablation uninterrupted warfarin therapy has been adopted by many institutions recently as it has proven safe and has been associated with very low thromboembolic risk.49 However, the international normalization ratio (INR) in as many as 50% or more of patients on warfarin at the time of ablation may fall outside the therapeutic range,10 with potential for increasing the risk of bleeding or thromboembolic complications, or resulting in a delay or cancellation of the procedure.

Until recently, warfarin was the only oral anticoagulant available for thromboembolism prevention in AF. In 2010, the US Food and Drug Administration approved the novel oral direct thrombin inhibitor dabigatran for prevention of stroke in non-valvular AF, with a large body of data demonstrating that the drug is as safe and at least as effective as warfarin in stroke prevention.11,12 Increasing numbers of patients are on dabigatran at the time a decision to proceed with CA of AF is made. How best to manage pre- and post-procedural anticoagulation for these patients is unknown.

In the absence of a large randomized clinical trial comparing the safety and efficacy of dabigatran for peri-procedural management at the time of AF ablation, several observational studies have been published with conflicting results.1318 A multicentre study by Lakkireddy et al.16 raised concern over the possibility of an increase in both bleeding and thromboembolic complications with the use of dabigatran. However, several recently published studies1315,18 have shown no difference between dabigatran and warfarin in rates of bleeding or thromboembolism.

Owing to the clinical importance of this issue for a widely performed therapeutic intervention, and the lack of definitive findings from individual published reports, we sought to systematically review the literature and perform a meta-analysis of the available studies to examine the safety and efficacy of interrupted peri-procedural dabigatran, as compared with warfarin, in CA of AF.


A systematic, pre-defined search strategy, study selection using specific criteria, data identification and extraction, and quality assessment were conducted according to the PRISMA19 and MOOSE20 guidelines for reporting a systematic review and meta-analysis of observational studies.

Search strategy

Multiple databases, including Embase, Medline/PubMed, and Scopus from January 2010 through April 2013, were queried. The following MeSH, Emtree, and text word search terms were used: CA, radiofrequency ablation, cryoablation, electrocoagulation, dabigatran, and warfarin. The search accounted for plurals and variations in spelling with the use of appropriate wildcards. Non-English publications were excluded. Meeting abstracts were searched in Embase. Efforts were made to identify research in progress, or in early stages, by examining clinical trial registries. The results were downloaded into EndNote®, a bibliographic database manager, and duplicates were removed. Three reviewers (A.A.B., S.U.S., and K.E.S.) independently extracted eligible studies from the search results. Disagreement was resolved by consensus.

Study selection

Any randomized control trial (RCT), case–control study, or cohort study which examined the safety and or efficacy of peri-procedural (pre- and post-procedural) dabigatran in comparison with warfarin in CA of AF was considered for analysis. Owing to the perceived importance of minimizing the duration of interruption of anticoagulation and in an effort to limit the degree of variation in the peri-procedural dabigatran regimen, the analysis was limited to studies in which dabigatran was held beginning the evening before or morning of the procedure and resumed on the same day as the procedure. Safety was assessed by occurrence of bleeding (defined as bleeding from any site), while efficacy was assessed by prevention of thromboembolic events (defined as ischaemic stroke, TIA, or systemic thromboembolism).

Data extraction and quality assessment

Two reviewers (A.A.B., A.R.K.) independently extracted data from eligible studies into a pre-defined data collection form. Data collected included country of the study, year of publication, study design, sample size, publication status, and bleeding and thromboembolic complications. In addition, data on the dabigatran and warfarin regimens, type of bridging therapy when used, duration of pre-procedural anticoagulation, ablation strategy, use of transoesophageal echocardiography, and patient information, including age, gender, and type of AF when available, were also recorded.

The Newcastle–Ottawa quality assessment scale (NOS) to evaluate the quality of observational studies was used.21 Quality assessment was limited to fully published articles. Newcastle–Ottawa quality assessment scale rates observational studies based on three parameters: selection, comparability between the exposed and non-exposed groups, and exposure/outcome assessment. It assigns a maximum of four stars for selection, two stars for comparability, and three stars for exposure/outcome assessment. Studies with less than five stars were considered of low quality, five to seven stars moderate quality, and more than seven stars high quality. The Jadad scale22 was used for quality assessment of the RCT which scores a maximum of two points for randomization, two points for blinding, and one point for description of withdrawal and drop out. Three reviewers independently (A.A.B., A.R.K., and M.G.) assessed the quality of the eligible studies. Disagreement was resolved by consensus.

Statistical analyses

We performed meta-analyses of bleeding and thromboembolism outcomes separately. Mantel–Haenszel method was used to pool data on the bleeding outcome into random effect model meta-analysis. The Mantel–Haenszel method with non-fixed zero cell correction was used to pool thromboembolism outcome data into fixed effect model meta-analysis. This method is recommended whenever sparse data are pooled and the sizes of the study arms are unequal.23 An odds ratio (OR), with its 95% confidence interval (CI), was used to calculate the overall effect estimate of both outcomes. Owing to several studies reporting no thromboembolic complications in either arm, a risk difference (RD) was also calculated for the thromboembolism outcome. Studies were analysed together as well as by subgroups according to the publication status. Statistical heterogeneity among included studies was assessed using χ2 and I2. The I2 statistic describes the proportion of variation in treatment estimate which is not related to sampling error.24 A value of zero indicates no heterogeneity, 25–49% low, 50–74% moderate, and >75% a high degree of heterogeneity. Publication bias was assessed by generating a contour funnel plot. Sensitivity analyses were carried out by repeating the meta-analysis on bleeding outcome using fixed effect model meta-analysis, reporting outcomes for USA and Japanese studies separately because Japanese studies included lower dabigatran dosing (110 mg). The statistical software Review Manager 5.2 (RevMan, Version 5.2) was used for all analyses.


Descriptive and qualitative review

The results of the search strategy are illustrated in Figure 1. A total of 290 citations were identified, with 9 citations, involving 3036 patients (1073 dabigatran), meeting the inclusion criteria.1318,2527 One was an RCT,27 one case–control,15 and the remaining were cohort studies.13,14,1618,2527 All citations except one16 were single-centre studies. Six citations were from the USA13,1517,25,26 and three were from Japan.14,18,27 All studies were fully published articles except two which were in abstract form.25,26 All studies were published in 2012–13. One study18 compared two different cohorts of patients on warfarin therapy with dabigatran; the uninterrupted warfarin group from this study was used in the primary analysis, and the interrupted warfarin group was used in the sensitivity analysis. All citations scored high on the NOS and the RCT scored high on the Jadad scale.

Figure 1

Flow diagram of the included studies.

Table 1 includes the baseline characteristics of the eligible studies. The mean age of patients in both the dabigatran and warfarin groups was ∼60, and the percentage of female patients ranged from 10–39%. Paroxysmal AF was more common than persistent. There was no significant difference between the dabigatran and the warfarin groups in mean age or gender within individual studies. The use of pre-procedural transoesophageal echocardiogram (TEE) varied among included studies; however, patients on dabigatran group underwent TEE in most of the studies. Irrigated radiofrequency CA was used in almost all the included studies. The target intra-procedural activated clotting time (ACT) ranged between 300 and 450.

View this table:
Table 1

Characteristics of the included studies

StudyDesignSample Size (D)Country of studyMean age, year (D, W)F, % (D, W)PAF, % (D, W)Duration AC pre-ablationTEE useDabigatranDabigatran doseWarfarinACT, secondAblation catheterAblation strategy
Bassiouny et al.13Cohort967 (344)USA59, 6325, 2857, 50NRSome patients1–2 doses held before procedure; resumed at conclusion of the procedure150 mgUninterrupted350–450Irrigated RFPVI
Imamura et al.14Cohort227 (101)Japan61, 6225, 2956, 49At least 4 weeksAll patientsHeld 12–24 h before procedure; resumed 3 h after ablation110–150 mgInterrupted; +bridging heparin>300Irrigated RFPVI ± focal/linear ablation
Khan et al.25 (Abstract)Cohort116 (50)USA56, NR39, NRNRNRNRHeld 24 h before the procedure; resumed 6 h after the procedure150 mgUninterrupted; ±LMWH bridgingNRRF; cryoablationNR
Kim et al.15Case–control763 (191)USA61, 6120, 2653, 48At least 4 weeksAll Dab patients; some War patientsPM and AM held; resumed 4 h after sheath removal150 mgUninterrupted300–350Irrigated RFPVI ± focal/linear ablation
Lakkireddy et al.16Cohort290 (145)USA60, 6021, 2157, 5730 daysAll Dab patientsAM dose held; resumed 3 h post-procedure150 mgUninterrupted300–400Irrigated RFPVI ± focal/linear ablation
Mendoza et al.26 (Abstract)Cohort118 (60)USA63, 6410, 12NRNRNRAM dose held; Resumed after sheath removal150 mgUninterrupted300–350NRPVI
Nin et al.27RCT90 (45)Japan61, 6126, 2076, 71At least 3 weeksAll patientsAM dose held; resumed 4 h post-procedure110 mgInterrupted300–400Irrigated RFPVI + focal/linear ablation
Snipelisky et al.17Cohort225 (31)USA61, 6519, 2668, 47NRNRAM dose held; resumed at PM150 mgInterruptedAt least 350RFCANR
Yamaji et al.18Cohort309 (106)A
300 (106)B
Japan60, 62
60, 61
25, 25
25, 23
65, 61
65, 63
At least 30 daysAll patientsAM dose held; resumed 3 h post-procedure ± IV heparin used110–150 mgGroup A: uninterrupted
Group B: interrupted
300–350Irrigated RFVI ± focal/linear ablation
  • ACT, activated clotting time; D, Dabigatran; F, female; LMWH, low molecular weight heparin; NR, not reported; PAF, paroxysmal atrial fibrillation; PVI, pulmonary vein isolation; RCT, randomized control trial; RFCA, radiofrequency catheter ablation; RF, radiofrequency; TEE, transoesophageal echocardiogram; W, Warfarin.

Konduru et al.2 observed that the time required to reach target ACTs was prolonged and that target ACTs were often not reached using a standard heparin dosing protocol when patients had been on dabigatran prior to their procedure. In a larger cohort of patients, Bassiouny et al.13 reported a similar observation, as well as improved success in reaching target ACTs with adjustments to unfractionted heparin (UFH) dosing.

Dabigatran dosing was 150 mg twice daily for USA studies, with what appears to be uncommon reported use of 75 mg twice daily for patients with significant renal insufficiency. In contrast, dabigatran 110 mg twice daily was used exclusively in one Japanese study, and for patients with renal dysfunction or advanced age in the other two studies from Japan, with other patients receiving 150 mg twice daily dosing.

Meta-analysis of bleeding

Interrupted peri-procedural dabigatran in CA of AF was as safe as warfarin with no difference between the anticoagulants in occurrence of bleeding events [dabigatran 58 (5.4%), warfarin 103 (5.2%); OR 0.92 (95% CI 0.55–1.454); χ2 = 13.03—P = 0.11; I2 = 39%]. Subgroup analyses had similar results, published articles [OR 0.92 (95% CI 0.54–1.71); χ2 = 12.04—P = 0.06; I2 = 50%] and abstracts [OR 0.49 (95% CI 0.09–2.78); χ2 = 0.39—P = 0.54; I2 = 0%], as did fixed effect model analysis [OR 0.95 (95% CI 0.67–1.36); χ2 = 13.03—P = 0.11; I2 = 39%]. There was mild heterogeneity, I2 = 39% among the studies.

Sensitivity analysis showed no statistically significant difference in bleeding events between dabigatran and warfarin therapy in both Japan [OR 0.57 (95% CI 0.25–1.33); χ2 = 3.36—P = 0.19; I2 = 40%] and USA [OR 1.27 (95% CI 0.77–2.10); χ2 = 5.43—P = 0.37; I2 = 8%] studies. Heterogeneity dropped to a negligible level among USA studies, I2 decreased from 39 to 8%, while it remained around 40% in studies from Japan. Sensitivity analysis also showed no significant difference between interrupted dabigatran and interrupted warfarin therapy [OR 0.68 (95% CI 0.34–1.37); χ2 = 5.04—P = 0.17; I2 = 41%]. Figure 2 shows the total number of patients, contribution of each study, results of the pooled analyses by subgroups, and heterogeneity. Visual assessment of the contour funnel plot (Figure 3) revealed no asymmetry, indicating absence of publication bias.

Figure 2

Random effect model meta-analysis of bleeding risk in the dabigatran group compared with warfarin in CA of AF. Error bars indicate the CI.

Figure 3

Contour funnel plot of bleeding outcome.

Meta-analysis of thromboembolism

Meta-analysis of the thromboembolism outcome showed no significant difference between interrupted dabigatran as compared with warfarin [dabigatran 5 (0.4%), warfarin 2 (0.1%); OR 2.15 (95% CI—0.58–7.98); χ2 = 2.14; P = 0.54; I2 = 0%]. Only four studies were included in the OR calculation as the remaining studies reported zero events in both arms.15,17,18,25 Since multiple studies were necessarily excluded by this analytical method, an RD analysis was conducted and showed no difference between the dabigatran and the warfarin groups [RD 0.00 (95% CI—−0.00 to 0.01); χ2 = 3.37; P = 0.81; I2 = 0%] in prevention of thromboembolism. Heterogeneity was zero in both analyses as shown in Figures 4 and 5. A contour funnel plot (Figure 6) revealed no visual evidence of asymmetry.

Figure 4

Fixed effect model meta-analysis of thromboembolic risk in the dabigatran group compared with warfarin in CA of AF with OR as measure of overall effect estimate. Error bars indicate the CI.

Figure 5

Fixed effect model meta-analysis of thromboembolic risk in the dabigatran group compared with warfarin in CA of AF with RD measure of overall effect estimate. Error bars indicate the CI.

Figure 6

Contour funnel plot of thromboembolic outcome.

The thromboembolic manifestations in the dabigatran group consisted of one pulmonary embolism,13 one case of medial longitudinal fasciculus syndrome [magnetic resonance imaging (MRI) showed multiple cerebral infarcts] which resolved completely the next day,14 and three patients who developed unspecified neurological symptoms recovered completely at 3 months follow-up.16 In the warfarin group, there was one patient with right-sided weakness and expressive aphasia (MRI showed small left middle cerebral artery stroke) which resulted in minimal deficit at 3 months follow-up,13 and one unspecified cerebrovascular accident.26


In this first systematic review and meta-analysis assessing the safety and efficacy of dabigatran vs. warfarin for peri-procedural anticoagulation in CA of AF, we found that interrupted dabigatran did not significantly differ from warfarin. This may have important implications for management of patients undergoing this procedure, especially considering the increasing number of patients with AF on chronic dabigatran therapy. However, this evidence is largely based on meta-analysis of observational studies that are limited by inherent biases, difference in the study designs, and use of unadjusted data.

The likelihood of peri-procedural TIA or stroke in CA for AF has declined as anticoagulation strategies have evolved. Historically, with interrupted warfarin without bridging therapy and insufficient intra-procedural anticoagulation, thromboembolic event rates as high as 5% were reported.28 Subsequent strategies, as reflected in the first worldwide survey of CA of AF, were associated with lower thromboembolic event rates.29 The stroke/TIA rate in the survey was 0.76%30 and in a subsequent report from an experienced centre was 1%.30 More recent series including >3000 patients where warfarin was not interrupted for the procedure [and higher ACTs (>350 s) were targeted], report very low cerebral thromboembolic event rates (0–0.098%) while maintaining relatively low bleeding rates.4,6

Dabigatran, in contrast, because of the lack of a well-studied, readily available antidote, has generally been held before CA of AF, raising questions regarding the safety of the procedure given prior experience with interrupted anticoagulation without bridging.28 A report of initial experience with dabigatran in CA for AF by Lakkireddy et al.16 supported this concern, finding higher rates of thromboembolic as well as bleeding complications in comparison with warfarin. However, the medication, which is a renally cleared, reversible, direct thrombin inhibitor, has potential advantages over warfarin, including a rapid onset of therapeutic anticoagulation, a relatively abbreviated time to reversal of anticoagulation after the medication is held, and a predictable anticoagulant effect.3133 In addition, as many as 50% or more of patients on uninterrupted warfarin present for CA of AF with an INR outside the therapeutic range.9 These factors underlie the idea that dabigatran could be used peri-procedurally with a minimized window off of fully therapeutic anticoagulation with results similar to those achieved with uninterrupted but often sub- or supra-therapeutic warfarin.

Similar thromboembolism and bleeding rates for dabigatran and warfarin were reported in three recently published larger studies; two comparing interrupted dabigatran to uninterrupted warfarin,13,15 and one comparing interrupted dabigatran with interrupted warfarin with UFH bridging.14 These findings are supported by the present meta-analysis which showed no difference in bleeding or thromboembolism risk for those treated with interrupted dabigatran in comparison with warfarin. Similar results were observed in secondary analyses comparing interrupted dabigatran with interrupted warfarin combined with bridging anticoagulation [either intravenous UFH or subcutaneous low molecular weight heparin (LMWH)]. The lack of difference between dabigatran and warfarin was consistent in all subgroup and sensitivity analyses.

Clinical implications

Presently available data, including recent larger than previously reported individual studies, and the present meta-analysis, support the use of interrupted dabigatran as an alternative to warfarin for peri-procedural anticoagulation for CA for AF. Holding dabigatran for 1–2 doses before the procedure with resumption after the procedure on the same day has shown similar safety and efficacy as uninterrupted warfarin or interrupted warfarin with bridging anticoagulation with UFH or LMWH. A longer hold for dabigatran for patients with renal dysfunction is advisable, although supporting data in CA of AF are lacking. With anticoagulation being interrupted and the majority of studies using pre-procedural TEE for patients on dabigatran, routine use of pre-procedural TEE for patients on dabigatran may be warranted.

Particular attention to intra-procedural anticoagulation with UFH may be advised as larger doses of UFH appear to be necessary to achieve target ACTs for patients treated with dabigatran, although a mechanism for this difference has not been identified and its clinical significance is not currently known.


An extensive search strategy was conducted to help ensure representation of all currently available data from studies meeting the pre-defined inclusion criteria. The findings were robust and consistent among different subgroup and sensitivity analyses. There was no significant in between studies heterogeneity and no evidence of publication bias.


The study has several limitations. It involves meta-analysis of unadjusted data from mainly observational studies with all the inherent biases of such study design. Our meta-analysis could not control for variation of practices among different centres. Although no study reported dabigatran outcomes in comparison with historical warfarin use, it is possible that dabigatran use was concentrated in more recent cases and may have been favourably affected by safer procedural practices and greater clinical experience. The availability of data on the use of dabigatran during CA for AF is limited, as the drug was only recently approved and marketed, and while our efforts captured all the available studies as of 2013, ongoing studies will further improve the ability to estimate the risks and benefits of dabigatran as compared with warfarin.

Although the results of the meta-analyses are consistent, there is variability among individual study results. Different results may be attributable to differences in patient risk profiles (CHADS2 score), ablation strategies, ablation catheter, ACT ranges, TEE practices, physician experience, target INRs, and dabigatran dosing. Transoesophageal echocardiogram practices not only differed among centres, but often differed for dabigatran in contrast to warfarin patients at the same centre. How often intracardiac thrombus was identified during pre-procedural TEE and the ablation procedure postponed was not reported. Thrombembolic events were not analysed based on whether patients were in sinus rhythm or required cardioversion at the time of the ablation procedure, which may have differed among centres and between warfarin and dabigatran groups.


Meta-analysis of currently available studies shows no significant difference in bleeding or thromboembolism between interrupted dabigatran and warfarin therapy in CA of AF. Dabigatran appears to be safe and effective for peri-procedural anticoagulation in CA of AF, although small differences in comparison with warfarin cannot be excluded. Given limitations to the available studies, additional data from a large multicentre randomized study are needed.

Conflict of interest: none declared.


View Abstract