Author + information
- Received January 5, 2013
- Revision received March 28, 2013
- Accepted April 2, 2013
- Published online September 24, 2013.
- K. E. Juhani Airaksinen, MD, PhD∗∗ (, )
- Toni Grönberg, BM∗,
- Ilpo Nuotio, MD, PhD†,
- Marko Nikkinen, MD‡,
- Antti Ylitalo, MD, PhD§,
- Fausto Biancari, MD, PhD|| and
- Juha E.K. Hartikainen, MD, PhD‡
- ∗Heart Center, Turku University Hospital, Turku, Finland
- †Division of Medicine, Department of Acute Internal Medicine, Turku University Hospital, Turku, Finland
- ‡Heart Center, Kuopio University Hospital, Kuopio, Finland
- §Heart Center, Satakunta Central Hospital, Pori, Finland
- ||Department of Surgery, Oulu University Hospital, Oulu, Finland
- ↵∗Reprint requests and correspondence:
Dr. K. E. Juhani Airaksinen, Heart Center, Turku University Hospital, Kiinamyllynkatu 4-8, FIN-20520 Turku, Finland.
Objectives This study sought to explore the incidence and risk factors of thromboembolic complications after cardioversion of acute atrial fibrillation.
Background Anticoagulation therapy is currently recommended after cardioversion of acute atrial fibrillation in patients with risk factors for stroke, but the implementation of these new consensus-based guidelines has been slow.
Methods A total of 7,660 cardioversions were performed in 3,143 consecutive patients with atrial fibrillation lasting <48 h in 3 hospitals. For this analysis, embolic complications were evaluated during the 30 days after 5,116 successful cardioversions in 2,481 patients with neither oral anticoagulation nor peri-procedural heparin therapy.
Results There were 38 (0.7%; 95% confidence interval [CI]: 0.5% to 1.0%) definite thromboembolic events (31 strokes) within 30 days (median 2 days, mean 4.6 days) after cardioversion. In addition, 4 patients suffered transient ischemic attack after cardioversion. Age (odds ratio [OR]: 1.05; 95% CI: 1.02 to 1.08), female sex (OR: 2.1; 95% CI: 1.1 to 4.0), heart failure (OR: 2.9; 95% CI: 1.1 to 7.2), and diabetes (OR: 2.3; 95% CI: 1.1 to 4.9) were the independent predictors of definite embolic events. Classification tree analysis showed that the highest risk of thromboembolism (9.8%) was observed among patients with heart failure and diabetes, whereas patients with no heart failure and age <60 years had the lowest risk of thromboembolism (0.2%).
Conclusions The incidence of post-cardioversion thromboembolic complications is high in certain subgroups of patients when no anticoagulation is used after cardioversion of acute atrial fibrillation. (Safety of Cardioversion of Acute Atrial Fibrillation [FinCV]; NCT01380574)
Anticoagulation is recommended for 3 weeks before and 4 weeks after cardioversion for patients with atrial fibrillation of unknown duration or duration >48 h (1–3). Systemic embolism might also occur with shorter duration of arrhythmia, but the need for anticoagulation is less clear, and it has been a common practice to cardiovert acute (<48 h) atrial fibrillation without anticoagulation (4–6). This practice has been recently called into question, and recent guidelines recommend anticoagulation therapy during and after cardioversion in patients with acute atrial fibrillation and risk factors for stroke (1).
Because the new recommendations are based mainly on consensus, and no randomized trials are available, we decided to evaluate the incidence and risk factors of thromboembolic complications after successful cardioversion of acute atrial fibrillation when no anticoagulation is used.
All patients with a primary diagnosis of atrial fibrillation (ICD-10 code I48) were identified from the institutional discharge registries in the 3 participating hospitals. Emergency clinic admission records and databases were then used to review all patients (>18 years of age) with acute (<48 h) atrial fibrillation who underwent cardioversion during the study period. In addition, only patients living in the hospital catchment area were included to get the adequate follow-up data after the cardioversion. Each of these 3 hospitals is the only referral hospital responsible for the acute care of patients with cardiac and stroke events in their catchment areas.
A total of 7,660 consecutive cardioversions were identified in 3,143 patients with acute (<48 h) atrial fibrillation treated at the emergency clinics of 2 university hospitals from 2003 through 2010 and 1 central hospital during 2010 (see trial registry data note following abstract). Of the 7,237 successful cardioversions, 5,116 were performed without peri-procedural and post-cardioversion oral anticoagulation or heparin, and these cardioversions were included in the present analysis. Total number of the included patients was 2,481.
Diagnosis of atrial fibrillation was confirmed by 12-lead electrocardiography according to the standard criteria. All cardioversion procedures were performed with electrocardiographic monitoring and full resources for cardiopulmonary resuscitation. Electrocardiogram was recorded before and after the procedure to verify the changes in cardiac rhythm. Cardioversion was considered successful if sinus rhythm was restored and the patient was discharged from the cardioversion unit in sinus rhythm.
All case records were reviewed with standardized data collection protocol to get information on baseline characteristics and medication of the patients and management of the patients during the index cardioversion and during a 30-day follow-up after cardioversion. All possible thromboembolic and bleeding complications, death, and any conditions that caused the patient to consult a physician within 30 days after the cardioversion were recorded. The duration of the index arrhythmia was determined from the symptom history in the patient records and the exact time of cardioversion.
The primary study outcome measure was definite thromboembolic event within 30 days after index cardioversion. Definite thromboembolic event was defined as a stroke documented clinically and confirmed by computerized tomography or magnetic resonance imaging to be caused by cerebral infarction or a systemic embolism confirmed by imaging, surgery, or autopsy. Probable embolic complications included transient ischemic attacks that were not confirmed by imaging or other embolism suspected clinically but not confirmed by imaging. After completion of manual registration of data, a computer-based cross-checking of strokes was performed from discharge register data of the included patients to ensure the complete coverage of all events. Bleedings were defined according to the Thrombolysis In Myocardial Infarction criteria (7). All potential outcome measures were classified by at least 2 members of the study team.
The study protocol was approved by the Medical Ethics Committee of the Hospital District of Southwest Finland and the ethics committee of the National Institute for Health and Welfare. Informed consent was not required, because of the register-based nature of the study. The study complies with the Declaration of Helsinki.
Statistical analyses were performed with SPSS software (version 20.0, SPSS, Inc., Chicago, Illinois) and SAS software (version 9.2, SAS Institute, Inc., Cary, North Carolina). Continuous variables were tested for normal distribution by the Kolmogorov-Smirnov test. Data are presented as mean ± SD, median (interquartile range), or absolute number and percentages, as appropriate. Chi-square and Fisher exact tests were used to compare differences between proportions. Student t test and Mann-Whitney U test were used for analysis of continuous data. Differences were considered significant if the null hypothesis could be rejected at the 0.05 probability level. Rates of embolic events were calculated by dividing the number of events by the number of procedures. Because the primary study outcome was binary and repeated cardioversions of same individuals were included in analyses, the GENMOD procedure with repeated measures option was used in the univariate and multivariate analyses. After the univariate analyses, a logistic multivariable regression analysis was performed to identify independent predictors for definite thromboembolism. Because of multiple testing, only variables with a 2-sided p value <0.05 in the univariate analysis were accepted for the model. Classification tree analysis was used for classification of the risk for thromboembolism according to the independent risk factors. Validation of the classification tree procedure was assessed by cross-validation through 25 folds. The minimum number of patients for parent node was set to 50, and the minimum for child node was 10. The maximum classification tree depth was 5. We chose the chi-square automatic interaction detection method, and its best model on the basis of the obtained predicted probabilities. Minimum change in improvement was set at a significance level of 0.05. Receiver-operating characteristics (ROC) curve analysis was employed for estimating of the area under the curve of probabilities of regression models and to assess the ability of risk-scoring methods in predicting post-procedural thromboembolism. A p < 0.05 was considered statistically significant.
Cardioversion was successful in 5,116 (95.4%) of the 5,362 procedures in 2,481 patients who had no peri-procedural oral anticoagulation or heparin therapy. A great majority (88%) of cardioversions were electrical (Table 1). The mean age of patients averaged 61.0 ± 12.4 (median 62) years, and 1,580 (63.7%) were men. The baseline characteristics of the patients according to the incidence of definite thromboembolic complications are described in Table 1.
Thirty-eight definite embolic events occurred in 38 patients (0.7% of successful procedures; 95% confidence interval [CI]: 0.5% to 1.0%), and 31 of these were strokes. One patient experienced both stroke and systemic embolism. Four additional patients suffered from a transient ischemic attack after cardioversion (mean 2 days). The definite embolic events occurred between 1 and 27 days after cardioversion (median 2 days, mean 4.6 days).
Two patients suffered from pulmonary embolism, and there were 11 deaths during the 30-day follow-up (median 3 days, mean 7 days after cardioversion), including 3 patients with fatal stroke. One patient died of aortic dissection, which occurred within 24 h of cardioversion but was unlikely to be related to the cardioversion. There were no major bleeding complications during the follow-up, and the only clinically relevant bleeding event was related to surgical embolectomy.
Logistic regression analyses showed that age (odds ratio [OR]: 1.05; 95% CI: 1.02 to 1.08, p < 0.001), female sex (OR: 2.1; 95% CI: 1.1 to 4.0, p = 0.03), heart failure (OR: 2.9; 95% CI: 1.1 to 7.2, p = 0.03), and diabetes (OR: 2.3; 95% CI: 1.1 to 4.9, p = 0.03) were the only independent predictors of embolic events (Table 2).
Chi-square automatic interaction detection analysis showed that heart failure, diabetes, and age >60 years were predictive of post-procedural definite thromboembolism (area under the ROC curve: 0.68, 95% CI: 0.60 to 0.76) (Fig. 1). A cutoff of 60 years was chosen on the basis of ROC curve analysis and the results of classification tree analysis. The highest risk of thromboembolism (9.8%) was observed in patients with heart failure and diabetes, whereas patients without heart failure and <60 years of age had the lowest risk of thromboembolism (0.2%).
There were 246 failed cardioversions in patients with no peri- or post-procedural anticoagulation. None of these patients suffered from thromboembolic complications during the 30-day follow-up.
Our results show that, in general, embolic events are quite rare (<1%) within 30 days after cardioversion of acute atrial fibrillation, even without perioperative anticoagulation. The incidence of thromboembolism is within the same range as in unselected patients undergoing elective or acute cardioversion during therapeutic anticoagulation (3,8). However, increasing age, female sex, heart failure, and diabetes increase the risk of thromboembolic complications substantially, and in the presence of multiple risk factors, the risk becomes unacceptably high (approximately 10%), significantly higher than after elective cardioversion of atrial fibrillation when using conventional pre- and post-cardioversion anticoagulation (Fig. 1). Interestingly, both of the recommended stroke risk scores (Congestive heart failure, Hypertension, Age ≥75, Diabetes mellitus, and prior Stroke or transient ischemic attack [doubled]; and Congestive heart failure, Hypertension, Age ≥75 [doubled], Diabetes mellitus, and prior Stroke, transient ischemic attack, or thromboembolism [doubled], Vascular disease, Age 65 to 74, Sex category [female]) were highly predictive for thromboembolism also in this acute setting (Table 1) (9,10).
Most embolic events occurred shortly after successful cardioversion, which supports the view that the conversion of atrial arrhythmia to sinus rhythm is responsible for the thromboembolism also after short attacks (<48 h) of arrhythmia (11). Indeed, transesophageal echocardiography has shown that left atrial thrombi—a clear contraindication to cardioversion—are found in 4% of the patients already within the first 48 h of atrial fibrillation when no anticoagulation is used (12). In the aforementioned study, low ejection fraction turned out to be a significant predictor for the presence of thrombi, which is in line with our present observation on heart failure as a major risk factor for post-cardioversion embolism. Another study by Stoddard et al. (13) reported a prevalence of 14% for left atrial thrombi when the duration of arrhythmia was <72 h. It is also noteworthy that the absence of thrombus before cardioversion does not guarantee safe cardioversion, because restoration of sinus rhythm results in a fall in blood flow velocity of the left atrial appendage (14), and it is now generally accepted that this accentuation of atrial stasis promotes new thrombus formation and predisposes to embolization. Our observations are in agreement with these data and suggest that atrial stunning might also occur in the setting of short attacks of atrial fibrillation and render to embolism especially when the patient has clinical features favoring thrombus formation. Most of the embolic events in this setting occurred within 3 to 4 days after cardioversion. According to the case records, spontaneous conversion of later recurrences of atrial fibrillation seemed to be responsible for at least 1 of the late (Day 27) thromboembolic events.
At the time of our study, European and American guidelines did not have solid recommendations about anticoagulation in recent-onset (<48 h) atrial fibrillation (2). Thus, a general approach in our country was to convert atrial fibrillation without post-cardioversion oral anticoagulants as long as there was a clear history of arrhythmia onset within 48 h from the scheduled cardioversion. This practice was called into question in 2010 when the European guidelines recommended effective perioperative and long-term anticoagulation in patients with risk factors for stroke (1). The implementation of the guideline has been slow, because the evidence behind the recommendation is circumstantial and supported only by small retrospective cardioversion studies. Furthermore, prolonged use of low-molecular weight heparin to cover the start of warfarin treatment causes significant logistic problems in busy emergency rooms. To the best of our knowledge, there are only 6 small retrospective studies (n = >50) involving, in total, <1,500 patients shedding light on this issue (Table 3) (4,5,15–18). The incidence of thromboembolic events in these studies has been low, and all definite events occurred in elderly (age >75 years) women and after spontaneous conversion of arrhythmia to normal sinus rhythm. Our present data strongly support the new European recommendations.
Although our study is the first to include a meaningful number of consecutive patients for the decision making, it has certain limitations. A retrospective study does not allow characterization of the study cohort as accurately as in a well-executed prospective trial. We were dependent on the recording of data by the physicians who performed the cardioversion and were responsible for the follow-up. However, because of the good coverage of electronic patient records and stability of the population, we could reliably review the outcome from all included patients at the subsequent outpatient and hospital visits. An important limitation was that the onset of atrial fibrillation was based on onset of symptoms. It is well-known that symptoms might not be reliable for accurate marking of onset of atrial fibrillation. As expected in some of the cases, there was uncertainty of the exact time of arrhythmia onset especially during the night time, but according to the recorded patient history it was <48 h also in these cases. The multivariate model in the classification and regression trees analysis might be somewhat overfitted, although only a small number of variables were included in the models, and the results were cross-validated 25 times. All these limitations must be weighed against the advantage that the retrospective design avoids filtering and yields results that reflect clinical practice more accurately than those of a prospective study.
Our results in this large multicenter, retrospective patient cohort show that, although in general the risk of definite thromboembolic events after cardioversion of acute atrial fibrillation is quite low, it becomes unacceptably high in patients with conventional risk factors for thromboembolism. Our results support the current recommendation that these patients need effective peri-procedural anticoagulation followed by long-term oral anticoagulation. At present, it seems unethical to execute a randomized trial on anticoagulation in the high-risk patients in this setting, and in patients with a lower event rate the sample size would be several thousands of cardioversions, making the study unattractive.
The authors thank our study coordinator Tuija Vasankari, RN, for her input in data and study management. Acknowledgments of clinical investigators for the collection of the data by center: Turku University Hospital, Turku: I. Nuotio, T. Grönberg, A. Karmi; Satakunta Central Hospital, Pori: M. Ampio, K. Ruuhijärvi, A. Ylitalo; Kuopio University Hospital, Kuopio: M. Nikkinen, P. Autere, E. Parikka, T. Rautiainen, S. Rissanen, M-L. Sutinen, M. Tuhkalainen.
This work received support from the Finnish Foundation for Cardiovascular Research, Helsinki, Finland; and the Clinical Research Fund (EVO) of Turku University Hospital, Turku, Finland. The authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- confidence interval
- odds ratio
- receiver-operating characteristics
- Received January 5, 2013.
- Revision received March 28, 2013.
- Accepted April 2, 2013.
- American College of Cardiology Foundation
- Camm A.J.,
- Kirchhof P.,
- Lip G.Y.,
- et al.
- Fuster V.,
- Rydén L.,
- Cannom D.,
- et al.
- Gallagher M.M.,
- Hennessy B.J.,
- Edvardsson N.,
- et al.
- Rao A.K.,
- Pratt C.,
- Berke A.,
- et al.
- Nagarakanti R.,
- Ezekowitz M.D.,
- Oldgren J.,
- et al.
- Coppens M.,
- Eikelboom J.W.,
- Hart R.G.,
- et al.
- Kleemann T.,
- Becker T.,
- Strauss M.,
- Schneider S.,
- Seidl K.
- Stoddard M.F.,
- Dawkins P.R.,
- Prince C.R.,
- Ammash N.M.
- Grimm R.A.,
- Stewart W.J.,
- Maloney J.D.,
- et al.