Patients documented were at least 18 years old and treated according to the ESC Guidelines for the management of AF [1]. The OAC had to have been initiated at least 3 weeks before the intervention. Patients had undergone left atrial (LA) ablation by transeptal puncture according to the HRS/EHRA/ECAS expert consensus statement on catheter and surgical ablation of AF [8].

Given the observational, retrospective nature of the study, patient grouping was based on the historically documented OAC regimen according to a patient’s medical records. We defined three groups of patients: 1) in the first group (bridging group), patients had received phenprocoumon prior to the intervention. VKAs were paused and patients bridged with low molecular weight heparin (LMWH) in patients with an international normalised ratio (INR) ≤ 2.0 on the day of hospitalisation. Phenprocoumon was paused at least 2–3 days before the procedure. Bridging with LMWH was started when INR was < 2. After the procedure unfractionated heparin (UFH) was perfused with a target activated partial thromboplastin time (aPTT) of 60–80 s until removal of the pressure bandage. If no bleeding complications were seen, LMWH was administered until an INR < 2.0 was reached. 2) The second group (VKA group, VKA) consisted of patients treated with phenprocoumon before the procedure with an INR > 2 and < 3.5 on the day of the procedure. In these patients, phenprocoumon was administered uninterruptedly. 3) The DOAC group consisted of all patients on DOACs before the procedure. The drugs used were dabigatran (110 or 150 mg twice daily [BID]), rivaroxaban (15 or 20 mg once daily [OD]) and apixaban (2,5 or 5 mg BID). All DOACs were administered without discontinuation. Last dosages of dabigatran and apixaban were administered the morning of the procedure. Patients on rivaroxaban the evening before or the morning of the procedure as per the patient’s scheduled dose. No LMWH or antidote were administered. The next dose was given the evening of the procedure.

A bolus of 5,000 IU UFH was administered before or after the transseptal puncture and partial thromboplastin time monitored. During the procedure, the ACT was kept > 300 s using UFH. Maximal UFH dosages, mean ACT time, time to ACT > 300 were recorded for each patient.

No protamine nor other antidotes were given in any of the procedures.

The primary study aim was to determine whether different peri-procedural anticoagulation strategies have different safety outcomes within 48 h after the ablation. To determine the incidence of thromboembolic events during 48 h after ablation was a further objective of this study.

Bleeding complications were classified as major (requiring intervention) and minor (not requiring intervention). Major bleedings and vascular complications were defined as any bleeding (hemoglobin [Hb] decrease > 2 mg/dl) or vascular complication requiring prolonged hospital stay or additional therapy (ie, surgery, transfusion). Minor bleedings and vascular complications were defined as any bleeding (Hb decrease ≤2 mg/dl) or vascular complication (hematoma, pseudoaneurysm, atriovenous fistula, groin hematoma) that could be managed conservatively without further treatment and without blood transfusion. Major pericardial effusion was defined as cardiac tamponade with a level of fluid > 10 mm, or that needed further treatment such as pericardial puncture or surgery. Minor pericardial effusion was defined as a level of fluid < 10 mm that could be managed without further treatment. Thromboembolic events were defined as stroke, transient ischemic attack (TIA), or systemic thromboembolism. The CHA₂DS₂-VASc score was used to estimate the risk of thromboembolism, and a HAS-Bled score to indicate the risk of bleeding [9].

The data coordinating centre was responsible for maintenance of the study database, data validation, and analyses. Categorical variables were compared using the χ² test, while continuous variables were compared using the student’s t-test or Wilcoxon rank sum test, as appropriate. Pairwise results were corrected using the Bonferroni–Holm–Shaffer procedure for multiple comparisons. Statistical significance was tested two-sided with the alpha level of 5%.

Patients had a mean age of 59.5 years with 31.4% being female. Hypertension (71.2%) and coronary artery disease (15.0%) were frequent comorbidities. The mean ejection fraction was 62.6% and patients had a mean left atrial diameter of 42.6 mm. There were only minor differences in patient characteristics between groups (Table 1) with hypertension being less frequent in the DOAC group compared to the VKA (p = 0.009) and Bridging group (p = 0.001).

The most frequent type of AF was paroxysmal (53.2%), followed by persistent AF (36.8%) and atypical AFL (12.2%) with more patients in the DOAC group having paroxysmal AF compared to the VKA group (p < 0.001). Further, the left atrial diameter was slightly smaller in both the VKA and DOAC groups (p < 0.001 for both comparisons) than in the Bridging group. In addition, there was a significant difference of reported coronary artery disease between the Bridging and DOAC group (p = 0.026).

Patients had a median CHA₂DS₂-VASc of 2 (range 0 to 6), with the majority being classed as 1 or 2 across groups (Table 2) and 26.0% having a Score > 2. There were no differences between groups with respect to their thromboembolic risk (CHA₂DS₂-VASc). The median HAS-BLED on the other hand was 1 (range 0 to 5) with the bleeding risk being lower in the DOAC group compared to groups Bridging (p = 0.006) and VKA (p = 0.001).

Patients were being treated with a variety of concomitant drugs (Table 3). Significant differences were observed in the rate of betablocker, angiotensin converting enzyme (ACE) inhibitor and statin use. Noteworthy was that more patients in the Bridging group (11.7%) received aspirin compared to patients in the VKA (6.6%; p = 0.014) and DOAC (2.6%; p < 0.001) groups. The same was true for other antiplatelet drugs (e.g. clopidogrel, ticagrelor etc.).

The mean procedure time was 209.6 min with a longer duration in the Bridging (241.5 min) and VKA groups (225.4 min) compared to DOAC (185.1 min; both p < 0.001 vs. Bridging) (Table 4).

The intra-procedural total heparin requirement was higher in the DOAC group compared to the Bridging and VKA groups, irrespective of whether the dose overall or adjusted by hour or hours and bodyweight was considered. On the other hand, the mean ACT was significant lower in the DOAC group (315.7 s) compared to groups Bridging (337.3 s; p = 0.001) and VKA (335.6 s; p < 0.001). Further, the time to ACT> 300 was the shortest in the DOAC group (26.8 ± 17.2 min) and significantly longer than in the Bridging group (70.1 ± 66.7 min) and VKA (36.6 ± 34.5 min) (both p < 0.001).

Out of 780 patients documented 27 suffered from major complications (3.5%) and 59 from minor complications (7.6%). Furthermore, 8 patients experienced a thromboembolic event (1.0%). No patient died within 48 h.

Patients on VKA being bridged with heparin (bridging) had a higher risk of suffering from major complications both in comparison to the VKA (odds ratio [OR] 3.42; 95% confidence interval [CI] 1.29 to 9.10; p = 0.019) and DOAC groups (OR 3.01; 95%CI 1.19 to 7.61; p = 0.025) (Table 5). This was mostly on the account of major complications without cerebrovascular events, with ORs of 4.86 and 4.47, respectively. The risk of minor complications as well as groin haemorrhage was not different between groups.

Patients with thromboembolic events are displayed in Table 6. All 4 patients receiving uninterrupted DOACs (1.1% of all; 2 males, age range 45 to 73 years) had no signs of stroke upon computed tomography (CT) scan and were considered to have suffered from TIA. In one of those patient puncture related paraesthesia may have resulted in the clinical appearance of temporary paraesthesia of the right leg. The patient receiving VKA being bridged with heparin reported visual impairment, but no signs of stroke were evident on CT scan. Three patients in the VKA group (age range 47 to 72 years, 2 males) reported complications within 48 h, two of them were confirmed to have stroke and one patient TIA. Overall, in the DOAC and the Bridging group there have been only TIAs, whereas in the VKA group strokes occurred.

The incidence of periprocedural thromboembolic events reported in the literature in patients undergoing AF ablation ranges from 0.1% to 1.1% [2, 10], and bleeding complications were reported to occur within a range of 12% to 20% [11]. In our study the overall rate of thromboembolic complications was 1.0% (n = 8), the rate of major bleeding complication 2.4% and the rate of minor bleeding complications 7.6%. This is comparable, albeit lower than in previously reported studies.

Continuation of oral anticoagulation therapy with VKA during catheter ablation is the recommended periprocedural strategy in the recent HRS/EHRA/APHRS (Heart Rhythm Society/European HeartRhythm Association/Asia Pacific Heart Rhythm Society) consensus statement [8]. For DOACs, the European Heart Rhythm Association’s practical guide on the use of non-VKA anticoagulants in patients with non-valvular AF recommends a cessation [12], although large trials showed that uninterrupted DOACs were also safe during ablation procedure [6, 7]. Nonetheless ‘bridging’ VKA-therapy is still a common strategy using different heparins and regimes [5, 13]. Recently data were presented in VKA-treated patients, that bridging with LMWH has no benefit regarding thromboembolic events but is inferior concerning bleeding complications [14, 15]. For catheter ablation an increase of groin hematoma and other bleeding complications has been described [4].

We found a three times higher rate of major complications in the Bridging group (8.1%) than in patients using uninterrupted regimens (2.5% and 2.9%, respectively). This was mostly due to major vascular events and pericardial effusion requiring puncture, although the single endpoint missed statistical significance. There was also a trend for more groin hemorrhages in the Bridging group, which was significant for minor events only. Regarding particularly pericardial effusion and major bleedings that needed further intervention, it might be of special clinical interest that the risk of occurrence is even higher in the Bridging group. This was on the background that patient characteristics were fairly similar albeit with a higher HAS-BLED score for Bridging and VKA patients compared to DOAC patients (p = 0.006) and more patients using aspirin and other anti-platelet drugs in the Bridging group compared to both other groups. Previous studies have shown excessive anticoagulation to be associated with a higher incidence of large effusion and tamponade [16]; the higher level of anticoagulation in the bridging group may thus explain their nominally higher rate of major pericardial effusion.

Overall, patient-related bleeding risk was very low (median HAS-BLED Score 1). Therefore, its influence on the complications is rather negligible and it can be assumed that the complications tend to be due to the intervention and the periprocedural management. Upon a thorough literature review, we found a recent meta-analysis including 13 studies with over 17.000 patients that summarized major bleeding complication with either uninterrupted warfarin compared to interrupted warfarin or heparin bridging [14]. Patients on uninterrupted warfarin had a lower risk of combined stroke/TIA compared with the bridging group (OR 0.25, 95% CI 0.10 to 0.62; p = 0.003). The results further confirm our finding of a reduced rate of major bleeding (OR 0.72, 95% CI 0.54 to 0.95; p = 0.02) as well as minor bleeding (OR 0.33, 95% CI 0.21 to 0.52; p < 0.0001) with uninterrupted warfarin compared to interrupted warfarin. This correspond to a recent published single-centre trial from Germany which was comparable to our study with respect to patient characteristics [13]. A notable difference between the data and our results was that DOACs were paused 48 h prior to ablation. Gunawardene et al. showed bleeding rates of 5.22% in the DOAC group, 6.97% in VKA group, and 10.8% in the Bridging group. The combined complication risk (thromboembolic events and bleeding) was nearly 2-fold higher in the bridging group (OR 1.9, 95% CI 1.0 to 3.7, p = 0.049) compared to the others.

One potential reason for the high bleeding rate seen in the Bridging group may be borne from the difficulty to quantify hemostasis exactly when restarting VKA after the intervention [13]. Recent guidelines recommend to give the first LMWH in procedures with high bleeding risk 48 – 72 h after the intervention [17]. In our patient population, UFH was perfused until removal of the pressure bandage and LMWH was started as soon as possible if no bleeding complications were seen (until INR < 2.0).

Prior literature recommending bridging VKA was potentially based on a medium to high risk for thromboembolic events population where, at the same time the risk of bleeding was underestimated. In our particular population, the majority of enrolled patients (approx. 74%) was documented to have a low thromboembolic risk with a CHA₂DS₂-VASc of 2 or less (Table 2) which was comparable to other study reports [13, 18, 19]. Further, we observed only a few thromboembolic events with a rate of about 1%, which is in agreement with other data [6]. In agreement with our data, the recent guidelines do not recommend bridging low-risk patients (ACC: CHA₂DS₂-VASc score 0–4 with no prior stroke/TIA/SE; CHEST: CHADS₂-score less than two and no prior stroke/TIA) [5, 20].

The HAS-BLED bleeding risk assessment we performed was based on the guidelines at the time of study initiation [21]. Most patients were on low risk if categorized by HAS-BLED. This was comparable to other peri-ablation studies [6, 14]. The ACC Guideline for periprocedural management now focus on an expanded HAS-BLED Score with regard on the patient’s individual bleeding risk (including prior bleed event within 3 months, quantitative or qualitative platelet abnormality, INR above the therapeutic range at the time of the procedure, bleeding history from previous bridging and bleed history with similar procedure) [17]. This could be also a reason we missed to exclude patients from bridging.

A recent meta-analysis has summarized major bleeding complications of uninterrupted DOAC vs. uninterrupted VKA [6, 14]. They included 8 datasets covering 3.544 patients and found no difference in the risk of stroke/TIA (OR 0.65, 95% CI 0.14 to 2.96; p = 0.58) or major bleeding (OR 0.94, 95% CI 0.48 to 1.87; p = 0.87). In addition, no differences were observed in minor bleeding (OR 0.93, 95% CI 0.67 to 1.28; p = 0.66), cardiac tamponade (OR 1.00, 95% CI 0.43 to 2.31; p = 1.00) and all bleeding complications (OR 0.93, 95% CI 0.67 to 1.29; p = 0.65). They also reported no differences between DOACs [6].

Our study shows similar rates for cerebrovascular events (OR 1.21, 95% CI (0.27 to 5.45; p = 1.0), major complications (1.14, 95% CI 0.44 to 2.92; p = 0.82) and minor complications (OR 0.98, 95% CI 0.55 to 1.75; p = 1.0) in the DOAC compared to the VKA group, confirming that DOACs can be used safely during catheter ablation. It is noteworthy that in the DOAC and the Bridging group there have been only TIAs, whereas in the VKA group strokes occurred (Table 6). The reason could be that bridged patients are at higher level of anticoagulation than the uninterrupted VKA patients and DOACs might be highly effective in preventing stroke, especially when not paused. The “simply-passed-on” strategy used in the study seems to be a safe regime not resulting in more complications and preventing major cerebrovascular events. In contrast to our study most DOAC-protocols use delayed (e.g., start after procedure), shortly interrupted (e.g., morning dose paused) or otherwise changed periprocedural DOAC regimes [6, 13]. Our anticoagulation regime of giving the uninterrupted DOAC like prescribed is very simple and feasible for clinical practice. Further, in the DOAC group the rate of major pericardial effusion requiring further intervention was similar to the other groups with no statistical difference. As these left atrial procedures are classified as high-risk interventions due to the necessary transseptal puncture this result implicates that DOACs can be used unpaused without increasing the risk for this adverse event. In the recently presented RE-CIRCUIT study with 635 patients the uninterrupted DOAC was superior to VKA regarding the bleeding complications (1.6% in the dabigatran group vs 6.9% in the warfarin group) [7].

The “simply-passed-on” strategy is currently under further examination in some large studies such as AXAFA (Apixaban; clinicaltrials.​gov: NCT02227550) and VENTURE-AF (Rivaroxaban). The latter with a patient number of 250 patient randomly assigned to either uninterrupted rivaroxaban vs. uninterrupted VKA found the use of uninterrupted oral rivaroxaban to be feasible and event rates being similar to those for uninterrupted VKA therapy [22]. In contrast to our study these studies don’t compare to a VKA-Bridging group.

We have analyzed the amount of heparin administered during the study, which is quite different between the individual drugs. The highest heparin dosage was needed by the DOAC group. This has been already described [13]. In contrast to these publications the times that are required to achieve goal ACT was the lowest in the DOAC group compared to VKA or Bridging in our study.

This is an observational study with its known biases and limitations. There were no long-term information about hemostasis of the patients because the strategy and monitoring of anticoagulation management was based on the treating physician’s preferences. The post-procedure follow up was very short and only during the hospital stay. Additional follow up examinations might have brought up the long-term differences in the safety profile of the different regimes. Additionally, given that grouping was retrospective based on medical records, a modest proportional shift away from VKA use/bridging and towards DOAC use over time was seen, reflective of the latter’s increasing availability and uptake. As such, we cannot unequivocally rule out the potential influence of biases associated with temporal trends, such as movement towards use of cryobaloon ablation. The latter may have affected major PE rates, given that it most commonly occurs due to mechanical perforation (or steam pops) in patients undergoing radiofrequency ablation; nevertheless, there was substantial overlap in procedural timing between groups throughout the study.