Prevention of cerebral thromboembolism by oral anticoagulation with dabigatran after pulmonary vein isolation for atrial fibrillation: the ODIn-AF trial

Atrial fibrillation (AF) is the most common heart rhythm disorder. It affects approximately 4.5 million people in Europe [1]. AF strongly increases mortality, morbidity, and hospitalizations. Cardioembolic strokes represent the most severe complications related to AF [2], with AF being responsible for overall at least 20% of all ischemic strokes.

Oral anticoagulation (OAC) in AF patients significantly reduces the risk of stroke and death, and is the only therapeutic intervention that has repetitively shown to reduce mortality in patients with AF. Yet, OAC increases the risk of severe bleeding events [3]. The current guidelines recommend OAC in AF patients with relevant risk factors to reduce cardioembolic ischemic strokes [4]. For decades, vitamin K-antagonists (VKA) were the only substances available for OAC and have been shown to be effective in reducing the stroke risk in patients with AF [5]. VKA have major limitations, including multiple drug and dietary interactions and a very narrow therapeutic window.

Alternative anticoagulant treatments include recently developed oral thrombin inhibitors and oral factor Xa inhibitors. Among others, the phase III clinical trials investigating dabigatran proved the efficacy of non-vitamin K antagonist oral anticoagulants (NOAC) [6]. NOAC nowadays represent the established alternative for VKA in patients with non-valvular AF at embolic risk, and are recommended for initiation of OAC as first-line therapy in current guidelines [4].

Pulmonary vein isolation (PVI) is the cornerstone of catheter ablation for AF [7] representing the most successful treatment option for rhythm control [8]. Ablation of AF provides an improvement in quality of life [9] and reverses left ventricular dysfunction [10]. Recent data show that strict and early rhythm control reduces stroke risk and mortality [11]. Early PVI is the best treatment to achieve rhythm control [12]. Previous retrospective data show that the cardioembolic risk is relevantly reduced after successful PVI [13].

OAC should mandatorily be performed for 2–3 months after AF ablation procedures [4]. Continuation of OAC treatment after that period is controversial [13], even in the case of clinically successful catheter ablation; yet, current guidelines recommend the lifelong continuation of OAC in all these patients with a CHA₂DS₂VASc score > 1 (men) and > 2 (women), irrespective of evidence of recurrent AF (class of recommendation I, expert consensus) [4]. The net clinical benefit of OAC after successful AF ablation remains unclear, in particular in view of the risk of severe bleeding. The potential benefit for the individual patient who continues OAC after PVI is a risk reduction of clinically apparent strokes. The patients might additionally profit from an analogously reduced incidence of silent cardioembolic events. On the other hand, patients who discontinue OACs might benefit from the lower risk of potentially devastating bleeding events. There is insufficient prospective data supporting both assumptions and an urgent need for prospective data to clarify the best therapeutic strategy for these patients.

The ODIn-AF study is the first large-scale prospective study to systematically evaluate the necessity to continue OAC for prevention of silent cerebral lesions and clinical events in patients with paroxysmal or persistent AF followed-up for 12 months after successful PVI and with a relevant stroke risk, as assessed by a CHA₂DS₂VASc score ≥ 2.

A total of 448 patients were screened for the study across 22 centers in Germany between September 9th 2015 and September 15th 2021. Screening included medical history taking and verification of inclusion and exclusion criteria. Main reasons for screening failure were refusal of study participation by the patient, concomitant diseases such as hyperthyroidism or severe heart failure, need for extensive ablation strategies due to concomitant atrial flutter or ectopic tachycardia, and CHA₂DS₂VASc score lower than 2. No selection bias was expected by these screening exclusion and a total of 200 patients underwent PVI for AF (66% paroxysmal; 34% persistent) followed by the run in phase and were randomized to one of the two study arms (with dabigatran: OAC n = 99; without OAC: off OAC n = 101) (Fig. 1). The number of subjects completing the study, providing assessments 12 months after randomization, was comparable in both arms [OAC n = 87 (87.9%), off OAC n = 91 (90.1%)]. Demographics and baseline characteristics were well balanced between the study arms (Table 1).

The mean CHA₂DS₂VASc score in all patients was 2.6 ± 0.8, ranging between 2 (54.5%) and 5 (1.5%); the mean HasBled score was 1.4 ± 0.7, defining a population at high stroke risk and moderate bleeding risk. No differences were found regarding CHA₂DS₂VASc and HasBled scores between the treatment groups (Table 1).

Continuous variables are summarized by mean ± standard deviations. The randomized treatment groups did not show differences and were well balanced.

AA antiarrhythmic agent, AF atrial fibrillation, ACE angiotensin-converting enzyme, AT angiotensin, BMI body mass index, CAD coronary artery disease, CKD chronic kidney disease, Cryo cryoballoon ablation, CTI cavotricuspid isthmus, LVEF left ventricular ejection fraction, NOAC non-vitamin K antagonist oral anticoagulant, OAC oral anticoagulation, PVI pulmonary vein isolation, RF radiofrequency ablation, VKA vitamin K antagonist.

^aStable heart failure is defined as New York Heart Association (NYHA) stage II or a LVEF of less than 50%

^bCHA₂DS₂VASc score (an assessment of the risk of stroke among patients with atrial fibrillation) ranges between 0 and 9, with higher scores indicating a higher risk of stroke.

^cThe absence of symptoms is defined as a European Heart Rhythm Association (EHRA) score of I. The EHRA score groups symptoms related to AF into four classes from I (asymptomatic) to IV (severe symptoms at rest).

After inclusion, first antral PVI was conducted in all patients. One hundred fifty patients (75%) were treated with cryoballoon ablation, fifty patients (25%) were treated with point-by-point RF ablation, 13% received additional RF ablation of the cavotricuspid isthmus for typical atrial flutter. In case of AF recurrence in the 3-month observation period of the run in phase, second ablation was conducted in a total of six patients (2 patients received cryoballoon ablation, four patients received RF), followed by a second run in phase. No differences were seen in ablation type or need for secondary ablation in the patients randomized to OAC versus off OAC arm. Use of antiarrhythmic drugs (Class I and Class III) prior to was 20.5% in the study population and was reduced to 4.3% at final visit 3. At 12 months, the drop-out rates were comparable (12.1% in the OAC and 9.9% in the off OAC group; Fig. 1).

In the patients treated with dabigatran, the compliance to therapy was rated good in 70.7% of the patients receiving OAC at visit 3 (Suppl. Table 2). Two patients at close out visit refused 72-h Holter ECG. All other patients received 72-h Holter ECGs protocol conform at 3-, 9-, and 12-month FU visits with mean recording durations of > 69 h for all visits in both groups. During the 12-month follow-up, ten patients from off OAC group started treatment with OAC, mainly due to documented AF recurrences, that triggered protocol compliant allocation to OAC. Two patients from OAC group stopped their treatment because of relevant gastrointestinal side effects of the OAC. No differences were present in these cross-overs compared to these patients adherent to randomized treatments (non-cross-overs), regarding demographical or clinical characteristics. Clinical reasons for switching treatment are depicted in Suppl. Table 3. AF recurrences were equally distributed across the groups and occurred in 9% in the OAC arm versus 7.9% in the off OAC arm.

On baseline MRI, just one patient randomized to OAC showed cerebral microlesions. Regarding the ITT analysis, the primary endpoint (occurrence of new micro- and macro-embolic lesions up to 12 months after randomization), two patients (2%) on OAC showed new cerebral microlesions and no patient (0%) off OAC, ****P = 0.1517; 95%-CI:-0.860%;5.5711% (Table 2; Fig. 2). The two patients with microembolisms were male, 65 and 67 years of age and both had a CHA₂DS₂VASc score of 2. Thirteen patients in the OAC group and eleven in the off OAC group did not receive cerebral MRI at close out, i.e., due lost to FU, withdrawal of consent or declined MR examination. As predefined, analysis for the primary endpoint was also performed with these missing data considered as failure, and thus counted as occurrence of lesions. Again, there was no significant difference between the OAC and off OAC arm (15.2% versus 10.9%; P = 0.4545, 95% CI:-5.056%;13.577%). No macro-embolic lesion appeared on MRI in the treatment groups.

DVT deep vein thrombosis, FU follow-up, MI myocardial infarction, MRI magnetic resonance imaging, OAC oral anticoagulation, PE pulmonary embolism, TIA transient ischemic attack.

^aAs predefined for the primary endpoint, missing values were counted as occurrence of cerebral embolism.

^bAll observed bleedings that occurred were graded minor.

Secondary outcomes were analyzed by ITT. No clinically apparent cardioembolic event or stroke occurred in the groups in the 12 months of follow-up. No patient died during the trial. No differences were found between the OAC and the off OAC group in any of the other secondary endpoints (Table 2). Relevant bleedings did not differ among the groups (Table 2). The use of antiplatelet therapy was equal between the groups during FU (OAC: 0%; off OAC: 4.3%). The changes from baseline of the neuropsychological test (MOCA) and quality of life (QoL EQ-5D score) were not found to differ between the groups (Fig. 3).

Safety results were analyzed in the predefined safety population (Table 3) different to ITT populations, the ten patients from off OAC who switched their treatment to OAC were considered to be allocated to the OAC group; the two patients that switched from OAC that discontinued treatment were considered to be allocated to the off OAC group. Adverse events were reported by a total of 108 patients [OAC n = 72 (67.3%), off OAC n = 46 (49.5%); P = 0.011]. Nine patients [OAC n = 6 (5.6%), off OAC n = 3 (3.2%); P = 0.5080] reported drug-related AEs. The percentage of patients with SAEs (52 in total) was higher in the OAC arm [OAC: n = 34 (31.8%); off OAC: n = 18 (19.4%); P = 0.046]. Cardiac disorders represented the most common AEs and were significantly elevated in the patient group with OAC (OAC: 35.5%, off OAC: 10.8%; P < 0.0001), mainly triggered by arrhythmia recurrences.

AF is an independent risk factor for cardioembolic events and increases the stroke risk by almost fivefold [18]. Cerebral microembolisms represent an independent risk factor for the development of later clinically apparent cerebral embolic events [19, 20] and were, therefore, used as subclinical surrogate for the risk of clinical apparent cerebral embolic events. Medication with OAC (VKA and NOAC) according to the CHADS₂ and CHA₂DS₂VASc scores reduces stroke risk and improves survival in AF patients [11]. Current guidelines [4] suggest long-term OAC after successful catheter ablation for AF, although prospective data are lacking and even in the era of NOAC, bleeding complications remain a main cause of morbidity and mortality.

ODIn-AF is the first randomized, prospective study to evaluate the effect of OAC on the 12 months incidence of silent cerebral embolic infarcts in patients with an elevated of cardioembolic events, but free from AF after successful PVI. There was no significant difference found in the patient groups on and off OAC. Incidences of cerebral micro-insults were low in both groups but no excess microembolism was present in the patients off OAC. Despite a CHA₂DS₂VASc score ≥ 2 in both groups, clinically significant embolisms and strokes were not found in secondary endpoint and safety analyses. Bleeding events under OAC with dabigatran were very low, and lower than the bleeding risk estimable from the RE-LY study. Here, a major bleeding event rate of 3.32% per year was found under OAC (150 mg dabigatran bid) [8] and life-threatening bleeding occurred in up to 1.45%. Cessation of OAC was not associated to better quality of life in the patient group off OAC. OAC therapy seems, thus, to be generally well accepted in the investigated on OAC group.

The predictive values of CHADS₂ and CHA₂DS₂VASc scores were previously investigated in patients after AF catheter ablation [21]. Both scores remained valid predictors of thromboembolic events after ablation in particular in patients with and without AF recurrences after ablation, with recurrence of AF representing an independent predictor of embolism. Elimination of AF, thus, seems the most important factor for freedom from embolic events. In ODIn-AF, only patients proven to be free from AF by routine clinical assessment 6 months after catheter ablation of AF were randomized. ODIn-AF followed a feasible and generally conductible follow-up protocol for AF recurrences, with regular 72-h Holter ECG and patient contact in case of clinically apparent AF relapse or palpitations. The aim of this screening protocol (rather than continuous monitoring by implantable loop recorders) for AF recurrences was to establish an easily executable follow-up in general practice.

The expert consensus of current guidelines advises to continue OAC in patients with elevated stroke risk independent from detected AF recurrence, as stroke rate and rate of silent embolisms may persistently be elevated due to asymptomatic AF relapses. Current recommendations, therefore, put patients on potential bleeding risk while hoping to prevent embolic events, yet in the absence of prospective data supporting this strategy. The results of a retrospective cohort study showed that discontinuation of warfarin treatment after PVI was associated to elevated risk of cardioembolic events, especially those who have previously experienced an ischemic stroke [22]. In ODIn-AF, history of previous stroke was, therefore, an exclusion criteria.

Two large-scale, but retrospective, studies showed that up to 25% of patients with a CHA₂DS₂VASc score ≥ 2 discontinued OAC after AF ablation [23, 24]. The clinical impact of this frequent violation of recommended OAC therapy by patients and/or physicians remained unclear. Liu and co-workers showed in a large-scale meta-analysis of prospective observational studies that discontinuation of OAC after successful AF ablation is safe and observed an increased risk of major bleeding in patients remaining on OAC [25]. In accordance to that, the ODIn-AF study suggests that cessation of OAC after AF ablation in the absence of AF might be a safe strategy in such patients.

It is questionable if the patient population after successful AF ablation and without clinical or ECG-documented AF recurrences exhibits the same risk for cardioembolic events as non-ablation patients. An analysis including 20 studies and > 20,000 patients showed elevated embolic risk after AF ablation for patients off OAC. The study designs and endpoints of these mainly retrospective analyses were heterogenous [26]. In a large retrospective analysis involving 37.908 patients [20], catheter ablation reduced the rate of cerebrovascular incidents by 53% at 12 months and by 41% during long-term follow-up. The stroke rates in AF patients with ablation were similar to a healthy, matched control population without documented AF. These results pointed toward a different stroke risk after successful AF ablation in patient populations free from AF recurrences.

Supporting this, data from an international study group compared the outcome of a large cohort of 1273 patients after catheter ablation to medically treated patients from the EURO Heart survey [27]. Freedom from AF was the strongest independent predictor of stroke-free survival. In this study cohort, OAC was stopped in 809 patients (64%) after ablation. Despite a CHADS₂ score of 0, 1, 2, and 3 or higher, the annual stroke rate was only 0.3%, 0%, 0.7%, and 0%, respectively, in these patients. Numerous retrospective, non-randomized trials further investigated OAC management and discontinuation of OAC after AF catheter ablation [28, 29]. Several meta-analyses were conducted evaluating cessation of OAC after AF ablation [24, 30]. The largest systematic review of 16 cohort studies (> 25,000 patients) compared embolic events in patients with and without OAC after AF ablation. Even after stratification for CHADS₂ and CHA₂DS₂VASc score, no differences were found regarding cardioembolic events between the groups [13]. Heterogeneity of the analyzed studies, low NOAC use, and the fact that patients with cessation of OAC had tendentially lower CHADS₂ and CHA₂DS₂VASc scores in some studies were potential confounders.

In conclusion, these inhomogeneous, retrospective analyses suggested that cessation of OAC after AF ablation could be safe and may lower bleeding risk. On the other hand, other studies, yet with obvious limitations, suggested potentially elevated risk in such patients [28]. In the light of these inconclusive results, randomized clinical trials were strongly encouraged [31], and ODIn-AF is the first study to show feasibility of cessation in a defined population of patients with CHA₂DS₂VASc score ≥ 2 and AF-free after PVI. Increasing number of risk factors with time may elevate stroke risk after PVI. In a study of 14,606 AF patients not on OAC with a CHA₂DS₂VASc score <  = 1, 49% of patients acquired at least one new risk factor over 4 years. It is, thus, mandatory to reassess stroke risk in patients off OAC on a regular base [32]. Kaplan et al. evaluated cardioembolic risk depending on AF duration and CHA₂DS₂VASc score in > 21,000 patients with implantable devices off OAC. The investigators showed that the stroke risk with a CHA₂DS₂VASc score 0 and 1 is low, and that with increasing CHA₂DS₂VASc score ≥ 2 cardioembolic risk rises inversely proportional to documented AF-episode durations. The higher the CHA₂DS₂VASc score, the shorter AF episodes seem to provoke cardioembolic events. AF in patients with CHA₂DS₂VASc score ≥ 2 may, thus, just be a coexistent factor for other vascular morbidities associated to AF, leading to additionally elevated stroke risk [33]. We did not show elevated thromboembolic events in patients off OAC with a median CHA₂DS₂VASc score of 2 after AF ablation. In ODIn-AF, regular Holter ECGs and clinical presentation/symptoms and validation by CEC were utilized for screening for AF recurrences. Yet, 10% of patients were identified in the off OAC group with AF recurrence and switched to OAC therapy OAC arm. None of the cross-overs showed relevant events regarding primary and secondary endpoints, pointing toward the fact that the in ambulatory setting, easy-to-apply Holter ECGs rather than continuous monitoring and clinical evaluation seem sufficient to monitor these patients and to safely change therapy to OAC, when AF relapses are present.

ODIn-AF is, thus, the first randomized study that evaluated the benefit of long-term NOAC therapy after effective AF ablation compared to cessation of NOAC 6 months after. No difference in regard to asymptomatic cerebral infarcts were detected. The results of ODIn-AF encourage larger scale randomized trials to further confirm the safety of withdrawal of OAC after effective ablation of AF.