Occurrence and Characteristics of Stroke Events in the Atrial Fibrillation Follow-up Investigation of Sinus Rhythm Management (AFFIRM) Study FREE

An estimated 2 260 000 individuals in the United States in 2000 have atrial fibrillation (AF). The prevalence will be such that by 2020 there will be 3 330 000 individuals in the United States with AF.¹ The median age of patients with AF is 75 years,² and stroke attributable to AF increases with advancing age, from 1.5% for those aged between 50 and 59 years to 23.5% for those between 80 and 89 years. Atrial fibrillation is an important risk factor for stroke. The annual risk of stroke in patients with nonvalvular AF not taking warfarin sodium is 3% to 5%, and AF is thought to be responsible for 15% of thromboembolic strokes. Additionally, AF increases the risk of death, after adjustment for other risk factors.³

The role of warfarin anticoagulation therapy in patients with AF is established by randomized clinical trials.^4- 8 These trials showed a reproducible reduction in stroke risk associated with warfarin therapy compared with control populations. In aggregate, the control populations of the randomized clinical trials had an annual stroke rate of 4.5% reduced by 68% with warfarin therapy.⁹ The risk of stroke in patients with AF depends on the presence or absence of certain predictors of increased stroke risk. Consistent predictors of increased stroke risk include advancing age, hypertension, prior stroke or transient ischemic attack (TIA), and poor left ventricular function. Additional predictors in some of the studies include diabetes mellitus, female gender, age older than 75 years, hormone therapy, associated coronary artery disease, and certain transesophageal echocardiographic features.^9- 10 Most patients in those studies who had an ischemic stroke did not receive adequate anticoagulation therapy.⁹ Patients who did receive adequate anticoagulation therapy had an 80% reduction in their stroke risk. Some of the randomized trials also examined aspirin therapy compared with placebo and found a 22% to 24% risk reduction associated with aspirin therapy.^4- 5,9

Since the presence of AF carries an increased risk of stroke, the prevention of AF or its successful reversion to sinus rhythm might be expected to reduce the risk of stroke. The Atrial Fibrillation Follow-up Investigation of Rhythm Management (AFFIRM) Study was designed to explore the effects of maintaining sinus rhythm in patients with AF compared with controlling cardiac rate without attempting to maintain sinus rhythm.^11- 13 The AFFIRM Study found no difference in mortality for patients treated with the sinus rhythm control strategy compared with those treated with the rate control strategy.¹³ This article examines differences in the occurrence or characteristics of stroke in the 2 groups.

The AFFIRM Study was a randomized evaluation of treatment of AF by 1 of 2 strategies: (1) ventricular rate control and anticoagulation therapy, or (2) sinus rhythm control and anticoagulation therapy. The study was approved by each local institutional review board, and each patient gave written, informed consent. The primary end point was total mortality, analyzed by intention-to-treat. Secondary end points included composites of total mortality, disabling intracranial bleeding (subdural and/or subarachnoid hemorrhage; intraparenchymal hemorrhage), stroke (embolus, thrombosis, and hemorrhage), disabling anoxic encephalopathy, cardiac arrest, and major noncentral nervous system bleeding. Other secondary end points were cost, quality of life, and functional status. This article evaluates the incidence of stroke.

All patients were to receive anticoagulation therapy with warfarin at the outset of this study. Patients randomized to the rate control strategy were to receive anticoagulation therapy for the duration of the study. The sinus rhythm control group could have their warfarin therapy stopped after 4 (preferably a minimum of 12) weeks if they maintained sinus rhythm and continued taking an antiarrhythmic drug, although physicians were encouraged to continue anticoagulation therapy throughout the study. Stroke occurrence was assessed at follow-up visits. All stroke events reported to the AFFIRM Clinical Trial Center, classified by the local principal investigator and relevant clinical documents, were obtained. These documents were then blinded for the randomization group (rate control or sinus rhythm control) and reviewed by a member of the AFFIRM Stroke Events Subcommittee. If a disagreement in the classification of the event existed between the local principal investigator and the stroke events committee member, the entire subcommittee reviewed the event. A final determination was made by consensus as to whether a stroke or embolic event had occurred and, if so, its presumed mechanism. The designation of stroke mechanism was based on prespecified clinical and imaging criteria (available on the AFFIRM Web site at http://www.axioresearch.com/affirm/). The status of warfarin or aspirin therapy and sinus rhythm at the time of the event was also noted. When a definitive stroke mechanism could not be determined, a “probable” mechanism was assigned, wherever possible.

Baseline characteristics were compared with χ² and t tests, as appropriate. The primary analysis was an intention-to-treat comparison of time to ischemic stroke. For all time-to-event analyses, rates were estimated using Kaplan-Meier techniques and compared using the log-rank test. Patients were censored at last contact, at study withdrawal, or at death.

A secondary analysis used multivariate Cox proportional hazards regression to assess the relationship of certain variables (ie, covariates) with risk of ischemic stroke after adjusting for other covariates included in the analysis. Prespecified covariates were age, sex, sinus rhythm at randomization, duration of the qualifying AF episode, history of coronary artery disease, history of hypertension, history of congestive heart failure, and left ventricular ejection fraction. The following baseline variables were also included in the analysis: history of stroke or TIA, history of diabetes mellitus, recent history of smoking, and in the 3311 patients who had an echocardiogram, left atrial dimension and mitral regurgitation.

Also included in the Cox proportional hazards regression analysis were 3 variables whose values changed over time (ie, time-dependent covariates): presence of AF, warfarin therapy, and aspirin therapy. Time-dependent covariates were obtained from baseline and follow-up forms. Follow-up occurred at 2 and 4 months after randomization and at 4-month intervals thereafter until the patient died, was lost to follow-up, or completed the AFFIRM Study. Specific start and stop dates for the presence of AF and for use of the various drugs were not recorded. Instead, responses on the follow-up form indicated whether the patient was in AF or taking warfarin at the time of the follow-up (or, for aspirin therapy and other drug therapy, at any time since the previous follow-up). At the time of a stroke event, the status of the time-dependent covariates was determined. A patient was considered to have been in AF or to have used a drug if it was recorded as such on the patient’s most recent form prior to the stroke event.

Following a stepwise approach, each covariate was evaluated for inclusion in the Cox proportional hazards regression model. Covariates that met the statistically significance criterion of P<.05 were included in the final model.

The AFFIRM Study randomized 4060 patients at 213 sites in the United States and Canada, and the mean follow-up was 3½ years (range, 2-6 years).¹² Baseline characteristics did not differ between the rate control and sinus rhythm control groups (Table 1). The mean ± SD age was 69.7 ± 9.0 years, with 39.3% being women and 11.4% being minorities. Hypertension was the most common cardiovascular condition, reported in 70.8% of patients; 38.2% had coronary artery disease. A history of TIA or stroke was present in 13% of the patients, congestive heart failure in 23.1%, and diabetes mellitus in 20.0%.

Initially, 96% of patients received anticoagulation therapy with warfarin. In the rate control arm of the study, during the first 3 years of follow-up, more than 90% of patients were taking warfarin (Figure 1). Eighty-four percent of the rate control patients received warfarin therapy throughout the study. In the sinus rhythm control arm of the study warfarin therapy declined, but the overall proportion of patients receiving warfarin therapy at any point in time remained approximately 70%. Fifty-two percent of the sinus rhythm control patients never stopped treatment with warfarin. Sixty-two percent of international normalization ratios (INRs) were within the recommended range of 2.0 to 3.0.

Table 2 summarizes the stroke outcomes in the AFFIRM Study; the occurrence of ischemic stroke over 5 years is shown in Figure 2. No differences were detected between treatment arms for the various types of stroke events and for all types combined Figure 3. Two hundred eleven patients (8.2%) had at least 1 stroke event. Ischemic stroke occurred in 157 patients (6.3%), primary intraparenchymal hemorrhage in 34 (1.2%), and subdural or subarachnoid hemorrhage in 24 (0.8%). Ischemic stroke was fatal for 25 patients, in 16 with primary intraparenchymal hemorrhage, and in 7 with subdural or subarachnoid hemorrhage. Disabling (but nonfatal) ischemic stroke occurred in 48 patients, in 6 patients with primary intraparenchymal hemorrhage, and in 9 patients with subdural or subarachnoid hemorrhage. The mechanism of ischemic stroke could be determined for 71 (45%) of the 157 patients who had an ischemic stroke. The mechanism was most frequently cardioembolic, with 35 patients (49%) having a cardioembolic stroke. Twelve patients (17%) had an atherothrombotic stroke, 19 (27%) a lacunar stroke, and 5 (7%) had a stroke of another mechanism. Eighty-six patients were classified as having a stroke of an unknown mechanism, typically because of an incomplete diagnostic workup. Of these 60% were suspected to be cardioembolic.

Most patients in both treatment arms who had an ischemic stroke were either not receiving warfarin or had a subtherapeutic INR (Table 3). At the time of the ischemic stroke, most patients in the rate control group were in AF, whereas most patients in the sinus rhythm control group were in sinus rhythm.

The relationship of ischemic stroke to AF and warfarin use and other variables was further investigated using Cox proportional hazards regression analysis. Seven variables were found to be significantly associated with risk of stroke (Table 4), excluding the treatment arm. Increasing age, female gender, the qualifying episode lasting 2 or more days, a history of stroke or TIA, and a history of diabetes mellitus were significantly associated with an increased risk of stroke. Among the time-dependent covariates, the presence of AF (associated with a 60% increase in risk) and warfarin therapy (associated with a 69% decrease in risk) were significantly related to stroke. Aspirin therapy was not significantly associated with a risk of stroke.

A total of 1894 patients in the sinus rhythm control group had a period of sinus rhythm that on average lasted 548 days. Patients were censored at the onset of AF. Ischemic strokes occurred in 25 of these patients; warfarin therapy associated with a reduction in stroke risk (hazard ratio = 0.33, 95% confidence limits = 0.14, 0.77, P = .01).

The reasons for stopping warfarin therapy are outlined in Table 5. Older patients more commonly stopped warfarin therapy because of frailty and risk of falls, surgery, and patient refusal compared with younger patients. Surgery was the reason for stopping warfarin therapy in 13% of patients who discontinued its use. Data are insufficient to determine the number who had a stroke following discontinuation of warfarin therapy for that reason.

The results of the AFFIRM Study showed no difference in mortality or stroke occurrence between rate control and sinus rhythm control strategies. More patients assigned to the sinus rhythm control group had sinus rhythm on follow-up visits; however, mortality and stroke occurrence were not significantly influenced. Patients recorded as having AF on their most recent follow-up form had a 60% greater chance of having an ischemic stroke than those recorded as not having AF, even after adjusting for warfarin therapy. The established high-risk features of age, hypertension, and prior TIA or stroke also were associated with occurrence of stroke in the AFFIRM Study.

While the use of warfarin therapy in the AFFIRM Study population was excellent compared with studies in the literature,^15- 16 enough patients stopped warfarin therapy that analyses showed that continuation of warfarin therapy is crucial to preventing strokes, even if the patient has presumed sinus rhythm. Warfarin therapy should be continued whenever possible in a patient older than 65 years or having another risk factor for stroke, even if the patient is presumed to be in continuous sinus rhythm.

Furthermore, sinus rhythm was better maintained in the sinus rhythm control arm than in most studies of antiarrhythmic drugs.^17- 19 Therefore, we had a greater opportunity to see beneficial effects of sinus rhythm on stroke occurrence than previous studies. It is unclear why patients in sinus rhythm should continue to have cardioembolic strokes. One speculation is that these patients continue to have episodes of AF, unrecognized both to themselves and to their physicians.^20- 21 Another would be that cardiac sources unrelated to AF (eg, left ventricular thrombus or patent foramen ovale) are responsible for the continued risk of cardioembolic strokes. It might seem paradoxical that treatment arm (rate or sinus rhythm) was unrelated to stroke risk while the presence of AF was strongly associated. These observations probably reflect the fact that available antiarrhythmic drugs are still not highly effective in preventing AF. It is well known that for every symptomatic episode of AF, there are many that are asymptomatic.²¹ Perhaps patients taking antiarrhythmic therapy are more likely to be asymptomatic with bouts of AF.

Our findings on the mechanisms of stroke and the effects of warfarin therapy in the AFFIRM Study are consistent with reports from prior studies. Warfarin therapy was effective in preventing cardioembolic stroke^22- 25 whereas aspirin therapy reduced noncardioembolic strokes. In the AFFIRM Study, a high proportion of patients continued to receive warfarin therapy throughout the trial. Most ischemic strokes occurred in patients who were not taking warfarin or who had a subtherapeutic INR. Perhaps greater success in maintaining a therapeutic INR would reduce ischemic stroke events even further. Lacunar and atherothrombotic strokes accounted for about half of all ischemic strokes, probably owing to the beneficial effects of warfarin therapy on the risk for cardioembolic strokes in most patients.

The results found in this study are applicable to patients comparable with those included in the AFFIRM Study (>65 years or 1 or more risk factors for stroke). Specific start and stop dates for both the presence of AF and for drug use were not recorded. For the time-dependent analyses, a patient was considered to have been in AF or to have used a drug if it was recorded as such on the most recent form prior to a particular event. With determination of the presence or absence of AF only at follow-up visits, it is impossible to identify accurately all patients with paroxysmal AF. These limitations introduced potential error because a patient’s sinus rhythm and drug use at the time of an event were not known with certainty. The added error, however, would be expected to weaken the association between the time-dependent variables and the end point of interest (ie, ischemic stroke), likely making it more difficult to detect significant associations. Detailed evaluation of interactions among the various factors was precluded by the lack of precise dates and the increased risk of bias.

Few AFFIRM-enrolled patients experienced strokes. Among the 4060 enrolled patients, only 211 had at least 1 stroke event of any kind, and only 157 had an ischemic stroke. The stroke subtype classification was occasionally based on incomplete diagnostic studies. Even when all appropriate diagnostic studies are available, assigning a stroke mechanism is often uncertain.

The analyses described in this article suggest that several factors influence the occurrence of stroke in patients with AF (Table 4). Because they do not involve randomized comparisons, however, the results cannot conclusively demonstrate the effect these factors have on the risk of stroke in patients with AF.

Although this analysis detected no relationship between randomized treatment arm and incidence of stroke, these results suggest that several clinical and therapeutic variables are associated with stroke risk in patients with AF. Increasing age, female gender, a longer length of the qualifying episode, a history of stroke or TIA, and a history of diabetes mellitus are significantly associated with stroke.

In patients with AF at high risk for stroke or death, the presence of AF is associated with an increased risk of stroke and warfarin therapy with a reduced risk of stroke. These data suggest that the beneficial effect of warfarin therapy exists not only for patients experiencing AF but also for patients who have a history of AF but who are presumably in sinus rhythm. Anticoagulation therapy should be maintained in patients who have a history of AF and risk factors for stroke, even when recurrent AF has not been documented.

Correspondence: David G. Sherman, MD, Department of Medicine, Division of Neurology, University of Texas Health Science Center, 7703 Floyd Curl Dr, San Antonio, TX 78229-3900 (Sherman@uthscsa.edu).

Accepted for Publication: January 18, 2005.

Financial Disclosure: None.

Funding/Support: This study was supported by contract N01-HC-55139 with the National Heart, Lung, and Blood Institute, National Institutes of Health, Bethesda, Md.

Previous Presentation: This study was presented in part at the 55th Scientific Sessions of the American Academy of Neurology; April 2, 2003; Honolulu, Hawaii.

Group Information: A complete listing of the members of the AFFIRM Study was previously published (Am Heart J. 2002;143:991-1001).