Assessment of Use vs Discontinuation of Oral Anticoagulation After Pulmonary Vein Isolation in Patients With Atrial Fibrillation

Key Points

Question  Do patients with atrial fibrillation and risk factors for stroke benefit from oral anticoagulation therapy after pulmonary vein isolation?

Findings  In this nationwide cohort study of 1585 patients with atrial fibrillation, those with a CHA₂DS₂-VASc score of 2 or more who discontinued oral anticoagulation treatment after pulmonary vein isolation experienced more ischemic strokes compared with those who continued their oral anticoagulation treatment.

Meaning  Patients with atrial fibrillation and a CHA₂DS₂-VASc score of 2 or more may benefit from continuous oral anticoagulation treatment after pulmonary vein isolation to reduce the risk of stroke.

Importance  Pulmonary vein isolation (PVI) is a recommended treatment for patients with atrial fibrillation, but it is unclear whether it results in a lower risk of stroke.

Objectives  To investigate the proportion of patients discontinuing anticoagulation treatment after PVI in association with the CHA₂DS₂-VASc (congestive heart failure, hypertension, age ≥75 years [doubled], diabetes, stroke [doubled], vascular disease, age 65-74 years, sex category [female]) score, identify factors predicting stroke after PVI, and explore the risk of cardiovascular events after PVI in patients with and without guideline-recommended anticoagulation treatment.

Design, Setting, and Participants  A retrospective cohort study was conducted using Swedish national health registries from January 1, 2006, to December 31, 2012, with a mean-follow up of 2.6 years. A total of 1585 patients with atrial fibrillation undergoing PVI from the Swedish Catheter Ablation Register were included, with information about exposure to warfarin in the national quality register Auricula. Data analysis was performed from January 1, 2015, to April 30, 2016.

Exposures  Warfarin treatment.

Main Outcomes and Measures  Ischemic stroke, intracranial hemorrhage, and death.

Results  In this cohort of 1585 patients, 73.0% were male, the mean (SD) age was 59.0 (9.4) years, and the mean (SD) CHA₂DS₂-VASc score was 1.5 (1.4). Of the 1585 patients, 1175 were followed up for more than 1 year after PVI. Of these, 360 (30.6%) discontinued warfarin treatment during the first year. In patients with a CHA₂DS₂-VASc score of 2 or more, patients discontinuing warfarin treatment had a higher rate of ischemic stroke (5 events in 312 years at risk [1.6% per year]) compared with those continuing warfarin treatment (4 events in 1192 years at risk [0.3% per year]) (P = .046). Patients with a CHA₂DS₂-VASc score of 2 or more or those who had previously experienced an ischemic stroke displayed a higher risk of stroke if warfarin treatment was discontinued (hazard ratio, 4.6; 95% CI, 1.2-17.2; P = .02 and hazard ratio, 13.7; 95% CI, 2.0-91.9; P = .007, respectively).

Conclusions and Relevance  These findings indicate that discontinuation of warfarin treatment after PVI is not safe in high-risk patients, especially those who have previously experienced an ischemic stroke.

Atrial fibrillation (AF) is the most common cardiac arrhythmia, with a prevalence in individuals aged between 50 and 89 years of 9.6% in men and 4.9% in women.¹ Atrial fibrillation increases the risk of mortality and stroke^2,3 and leads to impaired quality of life.^4,5 For symptomatic relief, pulmonary vein isolation (PVI) could be considered in patients with symptomatic, drug-resistant AF.⁶ Whether the risk of stroke is reduced after PVI remains unclear. According to current guidelines, anticoagulation is recommended for 2 to 3 months after PVI, and continuously thereafter in patients with risk factors for stroke, according to the CHA₂DS₂-VASc (congestive heart failure, hypertension, age ≥75 years [doubled], diabetes, stroke [doubled], vascular disease, age 65-74 years, sex category [female]) score.⁶ However, previous observational studies have shown very low rates of stroke in patients with AF after successful PVI despite cessation of oral anticoagulation (OAC) treatment,^7,8 possibly indicating that PVI may influence the risk of stroke. In these studies, about half of the patients had a CHADS₂ (congestive heart failure, hypertension, age ≥75 years, diabetes, stroke [doubled]) score of 0; that is, a substantial portion of the investigated patients had a very low risk of stroke at baseline. In a Danish study based on discharge registries, Karasoy et al⁹ found no difference in the rate of stroke between patients with AF who continued or discontinued anticoagulation treatment after PVI. In that study, about one-third of the patients had CHA₂DS₂-VASc scores of 0, 1, and 2 or more. In another study, however, discontinuation of OAC treatment after ablation was common and associated with an increased risk of cardioembolism for high-risk patients in the long term and for all patients within the first 3 months after the procedure.¹⁰

We used nationwide registries for catheter ablation and anticoagulation to retrospectively investigate 1585 patients with AF undergoing PVI between 2006 and 2012 at the 10 centers performing PVI in Sweden. Our main objectives were to investigate the proportion of patients discontinuing anticoagulation treatment after PVI in association with their CHA₂DS₂-VASc score, identify factors predicting stroke after PVI, and explore the risk of stroke, intracranial hemorrhage, and death after PVI in patients with and without guideline-recommended anticoagulation treatment.

This study is based on 6 nationwide Swedish national health quality registries; information from the registries was cross-matched by the National Board of Health and Welfare and Statistics Sweden. To safeguard personal integrity, all data were deidentified before access. The study was approved by the ethics committee at the Umeå University Faculty of Medicine, which did not require patient consent.

Information about all patients undergoing PVI in Sweden between January 1, 2006, and December 31, 2012, was collected from the Swedish Catheter Ablation Register. Only the first PVI was considered for each patient. The register was launched in 2004, and today, all 10 centers that perform catheter ablations are part of the register, which has a coverage of 94%.¹¹

The National Patient Register was used to collect information about concurrent illnesses, as well as electrical cardioversions. The register was launched in 1964 and complete coverage began in 1987. More than 99% of all somatic and psychiatric hospital discharges are registered in the Inpatient Register, which is a part of the National Patient Register. The validity of the Inpatient Register has been shown to be high for many, but not all, diagnoses, with a positive predictive value in general between 85% and 95%.¹² Since 2001, the National Patient Register also contains information on outpatient visits to specialized health care professionals. Primary care diagnoses are not included in the National Patient Register. The register contains information about dates of admission and discharge, primary and secondary diagnoses according to the International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10), and external causes of injury or poisoning, as well as codes for surgical procedures. Diagnostic definitions are presented in the eTable in the Supplement. Information from the National Patient Register was included since use of ICD-10 codes started January 1, 1997, and were used until December 31, 2012.

The Swedish Stroke Register has an active registration of patients with stroke and includes patients from all hospitals in Sweden admitting patients with stroke.¹³ Because the system uses active registration, there is a low risk for previous events counting as a new event. The coverage in the Swedish Stroke Register is estimated to be 97%¹³ and data from the register were therefore used to collect information about complications (diagnoses of ischemic stroke and intracranial hemorrhage) during the follow-up period.

Auricula is a national quality register for AF and anticoagulation. The register contains a web-based dosing system for warfarin sodium, with international normalized ratio values and warfarin doses for all patients in the register. Almost 50% of patients treated with anticoagulation in Sweden are registered in Auricula.¹⁴ Through Auricula, we collected information about the starting and ending dates for warfarin treatment for patients undergoing PVI. Auricula also contains information about treatment with new oral anticoagulants (ie, apixaban, dabigatran, and rivaroxaban).

Information about dates of death was collected from the Cause of Death Register, which, since 1961, contains information about date and cause of death for all persons who were registered in Sweden at time of death. From the Dispensed Drug Register, we collected information about exposure to antiplatelet agents, for calculation of the HAS-BLED (hypertension, abnormal renal or liver function, stroke, bleeding history or predisposition, labile international normalized ratio, elderly, drugs or alcohol concomitantly) score. The Dispensed Drug Register carries information about all prescribed and dispensed drugs from Swedish pharmacies since July 1, 2005.

Index date was defined as the date of PVI. A cardiovascular event was defined as a diagnosis of ischemic stroke or intracranial hemorrhage in the Stroke Register, or date of death in the Cause of Death Register that occurred after PVI. If a patient experienced more than 1 ischemic stroke or instance of intracranial bleeding during follow-up, only the first complication of a kind was considered. Time at risk was defined as time from PVI until the first complication; consequently, time after an ischemic stroke was not considered as time at risk for a new ischemic stroke. Time after an ischemic stroke was, however, considered as time at risk for intracranial bleeding and death.

Risk of ischemic stroke and hemorrhage was calculated using the risk score systems CHA₂DS₂-VASc¹⁵ and HAS-BLED.¹⁶ When calculating the CHA₂DS₂-VASc score, women with no other risk factors for stroke were considered to have a CHA₂DS₂-VASc score of zero. Regarding HAS-BLED, no score points were given for use of nonsteroidal anti-inflammatory drugs or labile international normalized ratio values since this information was missing. A patient was considered to be taking an antiplatelet agent if it was dispensed at a pharmacy within 3 months before PVI. Relapse of AF was defined as a new PVI registered in the Catheter Ablation Register or a diagnosis of electrical cardioversion in the National Patient Register that occurred beyond 3 months after the index PVI.

Data analysis was performed from January 1, 2015, to April 30, 2016, using SPSS statistics, version 23 (SPSS Inc). Descriptive information about the study population at baseline is presented as numbers with percentage and means (SDs).

Rate of complications was calculated as events per 100 years at risk. Rates of complications with and without warfarin treatment were calculated separately, as well as for patients with low or increased risk of stroke according to the CHA₂DS₂-VASc score. Risk of stroke for patients with a CHA₂DS₂-VASc score of 2 or more and for patients with a previous diagnosis of ischemic stroke was calculated with Cox proportional hazards regression analyses and risk of stroke was compared between time with and without warfarin treatment.

Factors associated with ischemic stroke during follow-up were calculated using univariable and multivariable logistic regression analyses. All factors included in the CHA₂DS₂-VASc score were analyzed, as well as age as a continuous variable. The number and proportion of patients discontinuing warfarin treatment within the first year after PVI were calculated considering all patients with at least 1 year of follow-up, divided into patients with CHA₂DS₂-VASc scores of 0, 1, or 2 or more. P < .05 was considered significant (all tests were 2-sided).

In total, 6551 PVIs were performed between 2006 and 2012; of these, 4364 PVIs were first-time ablations. Of all 4364 patients undergoing a first-time PVI, 3137 were male (71.9%), mean (SD) age was 58.6 (9.7) years, and the mean (SD) CHA₂DS₂-VASc score was 1.4 (1.4).

A total of 1585 of the 4364 patients undergoing first-time PVI were also registered in Auricula, all of whom were treated with warfarin; these 1585 patients were included in the present study. Of these 1585 patients, 73.0% were male, mean (SD) age was 59.0 (9.4) years, and the mean (SD) CHA₂DS₂-VASc score was 1.5 (1.4) (Table 1). For patients with a CHA₂DS₂-VASc score of less than 2, total time at risk was 2160 years; for patients with a CHA₂DS₂-VASc score of 2 or more, total time at risk was 1504 years.

In total, 1175 of the 1585 included patients had more than 1 year of follow-up after PVI. Of these, 360 (30.6%) discontinued warfarin treatment within the first year (Figure 1). In patients with a CHA₂DS₂-VASc score of 0, 1, and 2 or more, 158 of 373 (42.4%), 96 of 275 (34.9%), and 106 of 527 (20.1%), respectively, discontinued warfarin treatment within the first year after PVI. A total of 59 of 1480 patients (4.0%) with at least 3 months of follow-up discontinued warfarin treatment during the first 3 months after PVI.

Among all patients with a CHA₂DS₂-VASc score of 0, 1, and 2 or more, 215 of 524 (41.0%), 179 of 358 (50.0%), and 421 of 703 (59.9%), respectively, continued to take warfarin beyond 1 year after PVI. Of those still taking warfarin after 1 year, 81 of 215 (37.7%), 76 of 179 (42.5%), and 136 of 421 (32.3%) patients with a CHA₂DS₂-VASc score of 0, 1, and 2 or more, respectively, had a relapse of AF within 1 year and 174 of 215 (80.9%), 144 of 179 (80.4%), and 272 of 421 (64.6%) patients with a CHA₂DS₂-VASc score of 0, 1, and 2 or more, respectively, had a relapse of AF within the entire follow-up period. Time at risk was 2515 years with warfarin treatment and 1167 years without warfarin treatment.

In patients with a CHA₂DS₂-VASc score of 2 or more and in patients who had previously experienced an ischemic stroke, a Cox proportional hazards regression analysis showed a higher risk of stroke in patients who discontinued warfarin treatment: for those with a CHA₂DS₂-VASc score of 2 or more, the hazard ratio was 4.6 (95% CI, 1.2-17.2; P = .02) (Figure 2) and for those with previous ischemic stroke, the hazard ratio was 13.7 (95% CI, 2.0-91.9; P = .007).

In the group with CHA₂DS₂-VASc scores of less than 2, no significant differences were seen between patients taking and not taking warfarin in rates of ischemic stroke (1 in 1310 years [0.1% per year] vs 1 in 843 years [0.1% per year]), intracranial hemorrhage (1 in 1310 years [0.1% per year] vs 0 in 847 years), or death (1 in 1312 years [0.1% per year] vs 1 in 847 years [0.1% per year]) were seen (Table 2). In patients with CHA₂DS₂-VASc scores of 2 or more who discontinued warfarin treatment, the rate of ischemic stroke was higher (5 in 311 years [1.6% per year]) compared with patients continuing warfarin treatment (4 in 1192 years [0.3% per year]; P = .046).

In total, 11 of 1585 patients (0.7%) experienced an ischemic stroke during follow-up (7 males [63.6%]; mean [SD] age, 61.3 [5.5] years; mean [SD] CHA₂DS₂-VASc score, 2.5 [1.7]). Previous ischemic stroke was the only predictor for a new ischemic stroke after PVI in a univariable logistic regression analysis (odds ratio, 4.1; 95% CI, 2.2-7.6; P < .001), and it was still the only significant predictor of a new ischemic stroke after multivariable analysis including the same factors (odds ratio, 12.1; 95% CI, 3.4-43.1; P < .001).

Within the first year after index PVI, excluding the first 3 months after the procedure, 211 patients (13.3%) underwent electrical cardioversion and 298 (18.8%) underwent a new PVI. In total, 408 patients (25.7%) underwent electrical cardioversion and/or a new PVI. During the follow-up period of on average 2.6 years, 693 patients (43.7%) underwent electrical cardioversion beyond 3 months after PVI, and 580 (36.6%) underwent another PVI. In total, at least 953 patients (60.1%) had a relapse of AF beyond 3 months after PVI, leading to electrical cardioversion and/or another PVI.

Of 11 patients with ischemic stroke during follow-up, 2 had a relapse of AF within the first year after PVI and 8 patients had a relapse of AF within the entire follow-up period.

In this retrospective register study of 1585 patients with AF undergoing PVI, we found an increased risk of ischemic stroke if warfarin treatment is discontinued after PVI in patients with a CHA₂DS₂-VASc score of 2 or more. This finding was driven primarily by a higher risk of stroke in patients with a previous ischemic stroke. In patients with a CHA₂DS₂-VASc score of 0 or 1, there was no difference in the rate of stroke between patients with or without warfarin treatment. Thus, our results support the current guidelines that treatment with OAC should be continued after PVI in high-risk patients. In addition, a previous ischemic stroke was associated with new ischemic stroke, a finding even more pronounced in patients who discontinued warfarin treatment.

Results from previous studies support our results that high-risk patients benefit from anticoagulation after PVI.^17,18 In a Swiss study by Tran et al,¹⁷ mean (SD) CHA₂DS₂-VASc score in patients experiencing a thromboembolic event during follow-up was 3.1 (1.3) compared with 2.5 (1.7) in our study. In a German study by Nührich et al,¹⁸ 460 patients with AF undergoing PVI were included, of whom 83 had experienced a previous stroke. The results showed that patients with a previous stroke more often experienced a new thromboembolic event during follow-up than did patients without a previous stroke. All patients with a thromboembolic event during follow-up had recurrence of AF, and all patients were treated with OAC.

In a study by Noseworthy et al,¹⁰ an increased risk of cardioembolism was seen in the first 3 months after ablation in all patients and beyond 3 months in high-risk patients among those with any time off OAC therapy. In that study, however, time taking OACs was deducted from prescription data, with uncertainties regarding actual time periods with and without OAC. Several studies have shown a low risk of stroke after PVI, suggesting that the risk of serious bleeding would outweigh the benefit of anticoagulation.^9,19,20 In the study by Karasoy et al,⁹ there was no information about exact date of initiation and termination of anticoagulation treatment; instead, anticoagulation treatment duration was estimated with an algorithm by quantification of average daily dose, which generates substantially less precise information. Also, definitions of concurrent illnesses diverged. In our study, a clinical diagnosis of hypertension was counted as 1 point in the CHA₂DS₂-VASc score, while Karasoy et al⁹ required the use of 2 different antihypertensive drugs for the same diagnosis. Also, in contrast with our study, Karasoy et al⁹ did not consider cardiovascular events during the first 3 months after PVI. In our study, a diagnosis of ischemic stroke or intracranial bleeding in the Stroke Register or a date of death in the Cause of Death Register were considered complications, while Karasoy et al⁹ used a wider definition of complications, including ischemic stroke, transient ischemic attack, and systemic embolism, as well as hospitalization for intracranial bleeding and bleeding from the respiratory, gastrointestinal, or urinary tract.

In a newly presented Swedish register-based study by Friberg et al,²¹ the risk of stroke was decreased in patients with AF undergoing PVI compared with a propensity score–matched population with AF not undergoing PVI. This study did not, however, support discontinuation of OAC treatment after PVI. A PVI may decrease the risk of stroke, but according to our study, also supported by the results of the study by Noseworthy et al,¹⁰ the possible risk reduction conferred by a PVI is not enough for cessation of OAC therapy in patients with risk factors for stroke.

Patients undergoing PVI in our study were in general young, with a mean age of 59.0 years compared with a mean age of 76.2 years for the average patient with AF in Sweden.²² Patients undergoing PVI also had few concurrent illnesses, with a mean CHA₂DS₂-VASc score of only 1.5. In a study from northern Sweden, 85.1% of all patients with AF had a CHA₂DS₂-VASc score of 2 or more,²³ while the corresponding rate in our study was 44.4%. Almost three-quarters of the patients in our study who underwent PVI were male; the low proportion of women undergoing PVI has been previously described.^9,24 Men have a higher incidence of AF than do women (2.8% vs 2.1%).²³ Atrial fibrillation also affects men at a younger age than women,²³ and as the mean age of 59.0 years in our study indicates, PVI is most often performed in younger patients. It is unclear if these epidemiologic sex differences explain the entire numerical difference between men and women undergoing PVI, especially since women have more symptoms of AF than do men.^25,26

Most patients in our study (1421 of 1480 [96.0%]) continued warfarin treatment during the first 3 months after PVI, as current guidelines recommend,⁶ and 815 of 1175 (69.4%) continued warfarin treatment beyond 1 year after PVI. As expected, the proportion of patients continuing warfarin treatment long-term was higher in patients with a CHA₂DS₂-VASc score of 2 or more (421 of 527 [79.9%]) than in patients with a CHA₂DS₂-VASc score of zero (215 of 373 [57.6%]). The latter proportion may be explained by the fact that relapse of AF was common; 953 (60.1%) of 1585 patients underwent either electrical cardioversion and/or a new PVI beyond 3 months after the index PVI until the end of follow-up. Therefore, 174 (80.9%) of the 215 patients with a CHA₂DS₂-VASc score of zero who continued taking warfarin had an indication for OAC treatment (electrical cardioversion or PVI), at least intermittently. Consequently, most patients with an indication for continuous anticoagulation continued to receive warfarin treatment.

The main limitation of our study lies in the retrospective, nonrandomized design, leading to difficulties balancing differences between patient groups at baseline. As in other register-based studies, missing data are a problem, possibly leading to lower score points in CHA₂DS₂-VASc and HAS-BLED than were actually present. Also, since information about international normalized ratio values and use of NSAIDs were missing, the calculated HAS-BLED score could be falsely low. There was no information on patients’ rhythm status after PVI, which could affect the risk of stroke. We were, however, presumably able to detect most patients with relapse of symptomatic AF through the available information about electrical cardioversion and reablation. We also had no information about type of AF (ie, paroxysmal, persistent, long-standing persistent, or permanent), which could affect the PVI results and thereby perhaps also the risk of stroke. Our material includes data from 2006 through 2012; at this time, both CHADS₂ and CHA₂DS₂-VASc were used for stroke risk stratification in patients with AF. Among patients with a CHADS₂ score of zero, a significant proportion still have risk factors for stroke according to the CHA₂DS₂-VASc score, which may have led to discontinuation of warfarin therapy in patients we today would define as high risk and therefore have continued OAC therapy. In Auricula, treatment with new oral anticoagulants has been registered since their introduction. It is possible that some patients switched from warfarin to treatment with new oral anticoagulants during follow-up without being registered in Auricula. Given the slow introduction of new oral anticoagulants in Sweden since November 2011, this possibility is unlikely to have influenced our results. Only about one-third of the patients undergoing PVI were registered in Auricula and are included in our study. There is, however, no selection of patients included in Auricula; in most regions, Auricula includes all or none of the patients receiving OAC.

Overall, the majority of patients undergoing PVI in Sweden continue their treatment with warfarin, according to current guidelines. Nevertheless, in a substantial proportion of patients with high risk of stroke, warfarin therapy was withdrawn, and a higher rate of stroke was observed among these patients. These findings support the notion that it is unsafe to stop OAC treatment after PVI in patients at high risk of stroke.

Accepted for Publication: September 14, 2016.

Corresponding Author: Sara Själander, MD, PhD, Department of Public Health and Clinical Medicine, Umeå University, 901 87 Umeå, Sweden (sara.sjalander@lvn.se).

Published Online: November 23, 2016. doi:10.1001/jamacardio.2016.4179

Author Contributions: Dr S. Själander had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: S. Själander, Holmqvist, Smith, Platonov, Kesek, A. Själander.

Acquisition, analysis, or interpretation of data: S. Själander, Holmqvist, Smith, Platonov, Kesek, Svensson, Blomström-Lundqvist, Tabrizi, Tapanainen, Poci, Jönsson, A. Själander.

Drafting of the manuscript: S. Själander.

Critical revision of the manuscript for important intellectual content: Smith, Platonov, Kesek, Svensson, Blomström-Lundqvist, Tabrizi, Tapanainen, Poci, Jönsson, A. Själander.

Statistical analysis: S. Själander, A. Själander.

Obtained funding: S. Själander, A. Själander.

Administrative, technical, or material support: Smith, Platonov, Kesek, Blomström-Lundqvist, Tabrizi, Tapanainen, Poci.

Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Dr Tabrizi reported being a minority shareholder of Arrhythmia Center Stockholm. No other disclosures were reported.

Funding/Support: This study was supported by grants LVNFOU415691, LVNFOU534001 (Dr S. Själander) and LVNFOU481841 (Dr A. Själander) from the Department of Research and Development, County Council of Västernorrland, grant VISARENORR363951 from the Northern County Councils (Dr A. Själander), and The Heart Foundation of Northern Sweden (Dr S. Själander).

Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Additional Contributions: Aigars Rubulis, MD, PhD, Swedish Catheter Ablation Register, coordinated local patient data. He was not compensated for his contribution. We thank all who contributed with patient data in the national quality registries Auricula, the Swedish Catheter Ablation Register, and the Swedish Stroke Register (Riks-Stroke).