The prognosis of patients with atrial fibrillation is improved when sinus rhythm is restored: report from the Stockholm Cohort of Atrial Fibrillation (SCAF)
- 1Karolinska Institute at South Hospital, Stockholm, Sweden
- 2Institute of Environmental Medicine, Karolinska Institute, Stockholm, Sweden
- 3AstraZeneca R&D, Mölndal, Sweden
- 4Division of Cardiology, Sahlgrenska University Hospital, Göteborg, Sweden
- 5Department of Cardiology, South Hospital, Karolinska Institute, Stockholm, Sweden
- Dr L Friberg, GlobenHeart Clinic, Arenavägen 33, Box 100 01, S-121 26 Stockholm, Sweden;
- Accepted 18 February 2009
- Published Online First 19 March 2009
Background: Clinical trials have indicated that an active rhythm control strategy aiming at restoration of sinus rhythm in patients with atrial fibrillation (AF) is no better than a rate-control strategy in terms of mortality and morbidity. To what extent restoration and maintenance of sinus rhythm per se affect long-term prognosis in AF patients is less clear.
Aim: To investigate if there are differences in mortality and morbidity between direct current (DC)-cardioverted AF patients who remain in sinus rhythm after cardioversion and those who relapse early.
Method: 361 cardioverted patients from the Stockholm Cohort Study on Atrial Fibrillation were studied by means of medical records and national registers. Patients were followed for a mean of 4.2 years from DC cardioversion regarding all-cause mortality and for a mean of 3.2 years for a composite endpoint of death, ischaemic stroke, myocardial infarction or hospitalisation for heart failure.
Results: All-cause mortality tended to be lower in patients who had been successfully cardioverted and had had no known relapse of AF within the first 3 months after cardioversion (hazard ratio (HR) 0.57, 95% CI 0.30 to 1.06, p = 0.076). They also had a significantly lower incidence of the composite endpoint than those who relapsed early (HR 0.51, 95% CI 0.32 to 0.82, p = 0.0058).
Conclusion: Restoration and 3 months maintenance of sinus rhythm was associated with improved long-term prognosis. The results imply that an active DC cardioversion approach is justified.
In 2002, the RACE1 2 and AFFIRM3 4 trials concluded that a rhythm control strategy offered no advantage regarding survival over a rate control strategy in patients with atrial fibrillation and risk factors for stroke or death. In an “on treatment” analysis5 of AFFIRM, sinus rhythm and anticoagulation were associated with improved survival. A similar result was reported from the Danish DIAMOND study.6
Relapses of atrial fibrillation (AF) are common, and persistent AF often progresses to permanent AF. While direct current (DC) cardioversion is an easily available and highly effective means to restore sinus rhythm, surprisingly little is known about the long-term prognosis in patients who have undergone a primarily successful DC cardioversion.
The aim of the study was to investigate if there are any differences in mortality and morbidity between DC-cardioverted AF patients who remain in sinus rhythm after cardioversion and those who relapse early.
The Stockholm Cohort Study on Atrial Fibrillation (SCAF)7 8 consists of 2912 patients with atrial fibrillation or flutter who were treated as inpatients or outpatients at the South Hospital in Stockholm or at the Gustavsberg Primary Care Centre during 2002. The complete digitally stored information was examined and validated according to a predefined protocol. Complementary information about previous and current disease was obtained from the National Hospital Discharge Register going back to 1987 nationally and for Stockholm County to 1972. The cohort was followed prospectively with annual re-examinations of medical records and through national registers of hospital discharges and mortality. The study was approved by the ethical committee of Karolinska Institutet.
During follow-up, 416 patients were cardioverted at least once at the South Hospital. Patients with predominantly paroxysmal AF during follow-up (n = 53), and two patients who died during the index-generating hospital period were excluded. The remaining 361 patients were followed from the date of the first DC cardioversion after inclusion in the cohort in 2002 until the end of 2006 regarding all-cause mortality and cardiovascular morbidity.
Information about deaths was obtained from the official Swedish population register in which all deaths, irrespective of where they occurred, are entered within approximately 14 days. We also used a combined endpoint consisting of all-cause mortality, ischaemic stroke, myocardial infarction and hospitalisation for heart failure (as a primary hospital discharge diagnosis). Information about such events was obtained from the National Register of Hospital Discharges using ICD codes I63-64, I21 and I50 respectively. The reporting from the National Hospital Discharge Register is subjected to a delay, which is why we used a shorter observation period for the combined endpoint (until the end of 2005) than for the endpoint death from any cause (until the end of 2006).
Patients were analysed with regard to known relapse of AF or not using all available information from medical records and registers. The time zero was set to be the date of the first cardioversion during the study period. Rhythm status (sinus rhythm or not) was determined during the first 3 months of follow-up when comparatively reliable information about rhythm was available for most patients. After 3 months, no further reclassifications were performed. Thus, subjects experiencing a first relapse after more than 3 months were classified as “no relapse within 3 months” and remained so throughout the study period. Subjects that did not survive 3 months after the cardioversion were classified according to the rhythm on the latest contact prior to death (figs 1, 2).
In the analyses of means and proportions, we used the Student t test (two-sided) and χ2 analysis. We used two outcome variables: all-cause mortality and a composite endpoint consisting of all-cause mortality, hospitalisation for acute myocardial infarction (I21), ischaemic stroke (I63) or heart failure (I50) as principal diagnosis. Survival analyses were performed using Cox regression for both univariable and multivariable analyses. In multivariable analyses, we used a forward stepwise Cox procedure to identify factors that were significantly associated with the dependent variable. We kept these factors for the final model and further added age (continuous), sex and most recent information about rhythm before 3 months after cardioversion irrespective of whether these variables showed a significant association with the outcome or not. All analyses were performed in SPSS 15.0 and 16.0.
During follow-up, 1140 DC cardioversions were performed on 416 patients with AF. Patients who were selected for cardioversion differed in important aspects from those who were not. Decisions in favour of cardioversion were more often taken in younger, healthier patients (table 1).
Half of the patients (49%) had only one cardioversion. One single patient was cardioverted 34 times. The majority had not been cardioverted earlier (72%). Most had AF which started within 48 h before admission to the hospital (87%). In 15 patients, DC cardioversion was attempted but did not result in sustained sinus rhythm at discharge. Thus, the success rate, defined as discharge from hospital while in sinus rhythm, was 96%.
Recurrences of AF were frequent, especially during the first months after cardioversion. The median time to recurrence was 9 months for patients having their first or second cardioversion, compared with 3 months for patients with two or more earlier DC cardioversions. Patients who remained in sinus rhythm where generally younger and healthier than those who relapsed (table 2).
Warfarin was used by 40% of all patients at the time of the first cardioversion and with a lower proportion in patients without relapse of AF within the first 3 months (32% vs 53%). Digoxin was also less commonly administered at baseline to patients without early relapse (6% vs 14%). Differences of this kind were also present regarding beta blocking agents, class 1 antiarrhythmics and amiodarone.
During a mean follow-up of 4.2 years from the first DC cardioversion, all-cause mortality tended to be higher in those who relapsed within the first 3 months than those who did not (40 vs 23 deaths/1000 patients year at risk, p = 0.08). There were almost half as many deaths in those who had been in sinus rhythm on the latest recorded contact during follow-up as those in AF at the latest contact (39 vs 20 deaths/1000 patients year at risk, p = 0.043). After adjustment for age and comorbidity, the mortality appeared to be almost half as high in patients without early relapse (HR 0.57; 95% CI 0.30 to 1.06).
Using a composite endpoint consisting of death, hospitalisation for ischaemic stroke, acute myocardial infarction or heart failure, we found that patients with relapse of AF within the first 3 months of cardioversion reached this endpoint more often than patients without early relapse (117 vs 52 events/1000 patients year at risk, p = 0.00073; table 3). This difference in favour of sinus rhythm remained significant in the multivariable analysis (HR 0.51; 95% CI 0.32 to 0.82, p = 0.0058).
In these analyses, we used the time of cardioversion as time zero. We also made the same analyses using 3 months after cardioversion as the time zero. The results were similar but with wider confidence intervals possibly due to a shortened time at risk (hazard ratio for death 0.57, 95% CI 02.29 to 1.13, p = 0.11 and for the composite endpoint 0.60, 95% CI 0.35 to 102, p = 0.060).
We found an association between restored and maintained sinus rhythm and a favourable prognosis regarding all-cause mortality as well as with a composite outcomes parameter including death, ischaemic stroke, myocardial infarction and heart failure. Sinus rhythm appeared to be predictive, and not just a marker of an otherwise healthy patient with low cardiovascular risk.
Intention-to-treat or on-treatment analysis?
Patients selected for DC cardioversion are not representative of all patients with persistent AF. They are younger and healthier, and are thus expected to have a better prognosis irrespective of the rhythm. The best way to control for such differences in background characteristics is through a randomised, controlled trial. AF patients can be randomised to a treatment strategy, but they cannot be randomised to a rhythm. Relapses are frequent, and the efficacy of current prophylactic medications is generally poor. Randomised, controlled on-intention-to-treat studies like RACE or AFFIRM can thus only answer questions about which treatment strategy may be better, not about the how the prognosis is affected if normal sinus rhythm is restored and maintained. In fact, the rhythm control strategy was abandoned in 61% of the patients randomised to rhythm control in the RACE study.1 In the AFFIRM trial,3 the corresponding figure was 37.5%. Thus, it is important to separate the lack of success of the rhythm control strategy from the beneficial effect of remaining in sinus rhythm, of which the main reports from these studies have nothing to say.
In our prospective, observational study, we analysed patients “on-treatment”—for example, according to available information about rhythm. In order to avoid problems with large differences in background characteristics between cardioverted and non-cardioverted patients, we restricted the comparisons to the relatively homogeneous group of cardioverted patients. In multivariable analyses, we found that sinus rhythm was consistently associated with a 40–50% lower risk of death and of reaching the composite endpoint. This is similar to what was reported in the AFFIRM substudy where the database was reanalysed “on-treatment.”5 In this study, sinus rhythm on the most recent contact was indeed associated with a considerable reduction in the risk of death (HR 0.5, 99% CI 0.4 to 0.7). A post-hoc analysis was done in the RACE study, too9 but was not conclusive due to a small number of patients (n = 49) and a composite endpoint which, rather than including all-cause mortality, included pacemaker implantations and intolerance to antiarrhythmics. In this study, more patients in AF died from cardiovascular causes, although the investigators concluded that a rate-control strategy was as good as a rhythm-control strategy.
Use of warfarin
In our study, warfarin was used by just over 40% of cardioverted patients. This is probably less than it ought to be according to international guidelines10 but is consistent with warfarin use in clinical practice.7 11–14 In RACE and AFFIRM, oral anticoagulation was mandatory in the rate control arms but not in the rhythm control arms except in conjunction with DC cardioversion. In both these studies, there was an accumulation of thromboembolic strokes after cessation of warfarin treatment in the rhythm control arms. In the AFFIRM substudy,5 a strong association between survival and warfarin use was found (HR 0.47, 99% CI 0.36 to 0.61). A similar association between survival and warfarin use was found in the SCAF cohort (HR 0.49, 95% CI 0.30 to 0.79), as previously reported.8 Thus, unequal warfarin use between study arms in the AFFIRM and RACE studies may have affected results to the disadvantage of the rhythm control strategy, which in the present debate easily, but erroneously, may be interpreted to mean that rhythm does not matter.
Representativity of the study population
The study population, that is those selected for DC cardioversion, were younger and healthier than the total AF population. Thus, we studied the incidence of death and cardiovascular events in a low-risk population, and it is conceivable that the results would have been different in another set of patients. However, DC cardioversion is most often attempted where the prospects of success are good. Therefore, we believe that it is clinically relevant to study the prognosis of these low-risk patients in relation to whether sustained sinus rhythm was restored or not.
The duration of the ongoing AF episode is difficult to assess retrospectively, but among the DC cardioverted were both short episodes that could be treated at once and shorter or longer episodes in patients already on adequate warfarin treatment.
Information about rhythm
Reliable information about time spent in atrial fibrillation or in sinus rhythm is difficult to obtain without continuous monitoring of ECG of all patients at all times. For this study, we had to rely on what was documented on spontaneous hospital contacts during follow-up. We used available information about relapses from medical records and from diagnostic codes in the National Hospital Discharge registry covering the whole country. We believe that the information about symptomatic relapses within the first 3 months after cardioversion is relatively reliable, since cardioverted patients are offered one or more return visits for a control ECG and possibly for a change of prophylactic regime. Yet, we are aware that the relapse rate may have been underestimated, since patients may have been treated at a primary care centre without contacting a hospital, or they may have had asymptomatic relapses. It is less likely that patients in sinus rhythm have been misclassified as being in AF. In other words, if sinus rhythm really is of importance for the prognosis, the way we classified exposure to the different rhythms would tend to give a conservative estimate of this effect.
Information about medication
The information about the patients medication at various times during follow-up is fragmentary for the same reasons as the knowledge about rhythm. Information only became available when, and if, patients sought medical attention at the hospital or the primary care centre. If we were not to disregard the impact of warfarin all together, which we thought would hardly be justifiable, we had a choice between using information about warfarin exposure at baseline or at the latest contact preceding an event or censoring. In our opinion, the latter was the better option because of the closer time relationship to occurring endpoint events during follow-up. However, we believe that any resulting misclassifications of true exposure of warfarin would most probably serve to underestimate, rather than overestimate, the effect.
In an observational study such as the present one, it is vital to take into account and make adjustments for differences in baseline characteristics between the groups compared. We used information from various sources, including registers going back several decades in order to minimise the impact of differences in baseline characteristics. Still there may be differences that we have failed to adjust for. Randomisation is the preferred method to neutralise differences in background characteristics. However, randomisation is not an applicable technique to control for confounding for the purpose of the present study, since it is not possible to randomise to sinus rhythm.
Differences that we have failed to adjust for include lifestyle factors. The documentation in the medical records is often unreliable regarding factors like smoking and alcohol consumption, although these factors may be of great importance for the health of the patient. Information is often fragmentary regarding factors like physical activity, obesity and impaired glucose tolerance, which also may have a profound impact on the prognosis.
In addition, it is likely that a patient who looks fit and healthy is treated differently than one who appears to be in poor general health, even if there is no diagnosis. Someone who looks ill is more likely to be ill than someone who appears to be in good health, for example, having a cancer disease that was not known at the time.
When we compared patients who remained in sinus rhythm with those who relapsed, we found that patients who relapsed were using amiodarone or class 1 antiarrhythmic agents at baseline more often than those who remained in sinus rhythm (table 1). It is unlikely that this was because antiarrhythmic agents make patients relapse more easily, or that their toxicity causes a higher mortality and morbidity in relapsing patients. A more likely explanation is that doctors had recognised that these patients were at higher risk of relapse and therefore chose more potent drugs. The total number of patients treated with specific class I or class III antiarrhythmics was too small to affect the outcome of the multivariable analyses, but nevertheless it is likely that the use of various drugs may have affected the results. Thus, undocumented differences in health, medication and lifestyle factors may have contributed to residual confounding in the analyses.
Validity of endpoints
The endpoint all-cause mortality must be regarded as highly reliable, since the information we used was obtained from the official Swedish population register in which all deaths are entered within a few weeks. Unreported deaths are very unusual in Sweden.
For the combined endpoint, we also used clinical diagnoses entered in the National Hospital Discharge Register. It is conceivable that doctors occasionally may have used wrong diagnostic codes at hospital discharge. It is also probable that some patients may have had silent myocardial infarctions or ischaemic stroke events without seeking medical care. However, such unrecorded events are not likely to be very frequent and are furthermore not likely to have affected any of the compared groups more than others.
Our results suggest that restoration and maintenance of sinus rhythm may improve the prognosis in patients with atrial fibrillation. Therefore, we propose that all eligible patients should be given at least one opportunity to be restored to sinus rhythm before atrial fibrillation is accepted as permanent.
We want to express our gratitude towards AL Fernström, RN, for helping us with the medical records.
Funding: This study was supported by unrestricted grants from the Stockholm County Council and AstraZeneca R&D, Mölndal, Sweden.
Competing interests: NH and NE are part-time employees of AstraZeneca R&D, Mölndal, Sweden. MR has been lecturing and acted as a national coordinator in clinical trials sponsored by Astra Zeneca. He is also a member of the Boehringer Ingelheim Swedish advisory committee regarding haemostasis.
Ethics approval: Ethics approval was provided by the Regional Ethics Committee of Stockholm.