- Email: [email protected]
Overview of the randomized trials to prevent stroke in atrial fibrillation
Overview of the Randomized Trials To Prevent Stroke in Atria1 Fibrillation DANIEL E. SINGER,
MD
Epidemiologic studies have show that atria1 fibrillation (AF) raises the risk of stroke approximately fivefold, and that because AF is so common among the stroke-prone elderly, it accounts for about 15% of all strokes. Five recently completed, randurni~ed trials consistently found that the anticoagulant warfarin can prevent most of the additional stroke risk due to AF. This effect was seen at low doses. The trials have also demonstrated that warfarin therapy can be safe if careful patient selection and monitoring are implemented. Three of the trials prwided inconsistent, and currently inconclusive evidence about the e$acy of aspirin. The trials have not settled the anticoagukztion decision for all patients. Warfarin remains a demanding and risky therapy, which many patients and physicians do not find attractive. Future research should attempt to refine the risk ofstroke,and ofmajor hemorrhage during warfarin therapy among patients with AF, and should seek safer, less demanding, yet effective antithrombotic regimens. Ann Epidemiol 1993;3:563-567. KEY WORDS:
Randomized
trials, anticoagulation,
INTRODUCTION The past decade has witnessed a remarkable advance in our understanding of the stroke risk posed by atria1 fibrillation (AF), and of the prevention ofthis risk by warfarin anticoag ulation. For historic and clinical reasons, the relationship between AF and stroke has been categorized according to the presence or absence of rheumatic heart disease. AF in the setting of rheumatic heart disease, primarily mitral stenosis, has long been a widely accepted risk factor for stroke (1). The assumption has been that in patients with obstructed outflow of blood due to mitral stenosis, AF increases the stasis of blood in the atrium and promotes thrombus formation and subsequent emboli. Lifelong anticoagulation is standard therapy for such patients, although the clinical studies underlying this well-accepted management strategy are fairly weak (1). Presumably, the clear consensus in favor of anticoagulants in rheumatic AF results from the dramatic presentation of stroke or visceral embolus in these predominantly young patients. The vast majority of patients with AF have “nonrheumatic” AF associated with aging. These patients often have other conditions, such as hypertension and coronary artery disease, that predispose to stroke. The controversy about nonrheumatic AF focused on whether this rhythm disorder
From the General Internal Medicine Unit, Medical Services, Massachusetts General Hospital and the Departments of Medicine and of Health Care Policy, Harvard Medical School, Boston, MA. Address reprint requests to: Daniel E. Singer, MD, General Internal Medicine Unit, Bulfinch 1, Massachusetts General Hospital, Boston, MA 02114.
warfarin, atria1 fibrillation.
itself led to stroke, and even if it did, whether anticoagulants’ efficacy would be worth their risk in such an elderly group of patients. In the United States, the Framingham Heart Study (2) has provided a rich source of information on AF and stroke. The prevalence of AF is approximately 2% for individuals in their 6Os, and rises to 10% for those older than 80. Stroke risk among subjects without AF rises with age. The striking finding, however, is that the stroke risk among those with AF rises proportionately, with a relative risk of four to five persisting across all ages. The public health implication of these data is that AF accounts for 15% of all strokes, or about 70,000 strokes each year in the United States (2, 3).
THE RANDOMIZED PREVENT STROKE
TRIALS OF WARFARIN TO IN ATRIAL FIBRILLATION
Against this background, five randomized trials of the anticoagulant warfarin in nonrheumatic AF were begun independently in the mid 1980s (4-8). The trials include the Copenhagen Atria1 Fibrillation, Aspirin, Anticoagulation Trial (AFASAK), the Boston Area Anticoagulation Trial for Atria1 Fibrillation (BAATAF), the Stroke Prevention in Atria1 Fibrillation (SPAF) trial, the Canadian Atria1 Fibrillation Anticoagulation (CAFA) trial, and the Veterans’ Administration Stroke Prevention in Nonrheumatic Atria1 Fibrillation (SPINAF) trial. All were primary prevention trials, with a very small fraction of patients having had a prior stroke. All were stopped early because of warfarin’s efficacy.
564
Singer
AEP Vol. 3, No. 5 September 1993: 563-567
RCTS TO PREVENT STROKE IN ATRIAL FIBRILLATION
TABLE 1. Clinical characteristics of patients at entry to randomized trials” Feature
AFASAK
Male (%)
54 75h
Age (Y) AF > 1 y (%) PAF (%) Hx stroke (%) HBP (%) Diabetes (%) Angina (%) Hx MI (96) CHF (%)
0 4 31 10 16 8 51
BAATAF
SPAF
CAFA
70 66 72 34 8“ 55 19 10 6 19
73 67 a2 7 4“
70 68 68 16 3 51 16 25 16 28
34 10 20 12 20
SPINAF 100 67 86‘ 0 0 62 20 23 21 30
a These are the features of the control or placebo groups from each trial. b The mean age is provided except for AFASAK, which provided the median. of AF was not provided by AFASAK. For SPINAF, 86% had AF for longer than 6 ma ’ In SPAF and CAFA, percentages are for history of stroke and/or transient ischemic attack. AF = atrial fibrillation; PAF = paroxysmal or intermittent AF; Hx = history of; HBP = hypertension; MI = myocardial infarction; CHF = congestive heart failure.
’Duration
AFASAK AFASAK (4) was a three-armed trial comparing warfarin at a target international normalized ratio (INR) of 2.8 to 4.2 (roughly corresponding to a prothrombin time (PT) ratio of 1.7 to 2.0), aspirin at 75 mg/d, and an aspirin-placebo. AFASAK’s target INR was the highest of the trials. As with the BAATAF and SPAF trials, AFASAK did not blind warfarin therapy. AFASAK was probably the closest to a community trial of warfarin. Patients were identified at two outpatient electrocardiography laboratories used by general
TABLE
2. Results of the randomized
trials of warfarin for atria1 fibrillation
AFASAK” Warfarin Subjects (n) Emboli (n) Person-y
practitioners. Of 1842 eligible patients, 1007 entered the trial, probably the highest recruitment fraction of any of the trials. Slightly more than half of the patients were men (Table 1). The median age was in the mid 70s. A high proportion of patients had hypertension. Over 50% were reported as having heart failure, although the specificity of this diagnosis is not clear. Three hundred thirty-six subjects were assigned to receive placebo (i.e., aspirin-placebo); 336 to receive aspirin, and 335 to receive warfarin. Thirty-eight percent of patients assigned to the warfarin group withdrew from the study. This high percentage of withdrawals may have resulted from AFASAK’s high recruitment fraction; more highly selected samples of patients may be more compliant. AFASAK reported four embolic events in the warfarin group, 20 in the aspirin group, and 21 in the placebo group, corresponding to an annual rate of 1.6% in warfarin versus 5.6% in the placebo group (Table 2). There were two major hemorrhages in patients on warfarin, including one fatal intracerebral bleed. The marked benefit of warfarin coupled with the low rate of major hemorrhage led to the early stopping of the trial. Two additional points about AFASAK are worth noting. First, there was subsequent disagreement about the event rates reported for an intention-to-treat analysis. If patients who withdrew from therapy are included in the analysis, the numbers of thromboembolic events become 10,2 1, and 25 for warfarin, aspirin, and placebo, respectively (P - 0.05). In contrast, only three patients in the warfarin arm who sustained a thromboembolic event were actually
BAATAF
SPAF
CAFA
SPINAF
rate
335 4 250 1.6%
212 2 487 0.41%
210 6 260 2.3%
187 5 200 2.5%
260 4 456 0.88%
Control Subjects (n) Emboli (n) Person-y Annual rate
336 21 373 5.6%
208 13 435 3.0%
211 18 244 7.4%
191 11 212 5.2%
265 19
52%
79% 52-90%
Annual
Preventive Efficacy 95% CI
71% 23-90%
Intracranial hemorrhage rate due to warfarin Meta-analvtic
0.40%
estimate of the oreventive
efficacy of warfarin:
86% 51-969/o
0.21%
69% 27-85%
0%
(- 36)-87% 0.50%
440 4.3%
0.22%
73% (95% CI: 57-83%)b
a Person-y in the warfarin and control arms of AFASAK were inferred from data given in their report (4); p erson-y were given explicitly in the other trials (5-8). h Meta-analytic summary estimated using Rothman’s adaptation of the Mantel-Haensrel technique (19). CI = confidence interval.
565
Singer RCTS TO PREVENT STROKE IN ATRIAL FIBRILLATION
AEP Vol. 3, No. 5 September 1993:563-567
taking warfarin, and of these, two had briefly discontinued
trasted with 18 in the placebo group. In group 2, where
warfarin, resulting in an INR well below target at the time
patients were assigned to aspirin or aspirin-placebo
of their event.
there were 25 embolic events in those on aspirin versus 28
only,
in the placebo group (not statistically significant). Subgroup analyses indicated
BAATAF The BAATAF
trial (5) randomized
2 12 patients to receive
“low-dose” warfarin (target PT ratio of 1.2 to 1.5) and 208 to the control
group. Therapy
was not blinded.
Subjects
in the control group could take aspirin, and approximately half did. Nearly three-fourths Table
of the patients were male (see
1). The average age was 68. A small proportion
subjects had intermittent
of
AF. Most had had AF for several
years. Half the subjects had no clinical heart disease other than AF, although many of these had hypertension betes. Only 10% of BAATAF
or dia-
patients assigned to the war-
serious strokes. There was one fatal intracerebral
hemorrhage
in a pa-
hematoma
with full
tient on warfarin and one subdural
recovery. Two patients in the placebo group had subdural hematomas
(with full recovery).
There was no increase in
major bleeding from other anatomic
sites with warfarin.
CAFA The CAFA
trial (7) was a double-blind,
placebo-controlled
trial of low-dose warfarin (target INR of 2 to 3) in AF. After
farin group withdrew from therapy permanently. In the BAATAF
that aspirin did not confer benefit on
patients over age 75, and was not very effective at preventing
trial 13 strokes were observed in the
the AFASAK
report and a preliminary report of the SPAF
control arm versus two in the warfarin group, for a preven-
trial, the CAFA
tive efficacy of 86% (see Table 2). There was one presumed
allow all patients access to warfarin. CAFA
subdural hematoma
using an “efficacy” analysis where events occurring
Otherwise,
in a patient on warfarin that was fatal.
the rate of serious bleeding was essentially the
same in warfarin-treated
as in control
subjects.
Eight of
the 13 control patients who sustained a stroke were taking aspirin. Analyses to remove confounding domized assessment
of aspirin estimated
in this nonranthat the rate of
investigators
decided to stop their trial to
By this approach,
there were five primary embolic events
in the warfarin group versus 11 in the placebo group, for a risk reduction
of 52% (see Table 2). Of these five events
in the warfarin group, two occurred in patients who were
stroke persisted at severalfold higher on aspirin than on
not taking the drug, and in an additional were well below the anticoagulation occurred
trial (6) randomly
assigned patients to receive
warfarin at a low dose (target PT ratio of 1.3 to 1.8), aspirin at 325 mg/d, or aspirin-placebo. blinded.
Warfarin therapy was not
The SPAF trial also studied patients who would
hemorrhage,
two, the INRs
target. Major bleeding
in five patients on warfarin,
intracranial
The SPAF
more
than 28 days after stopping therapy would not be counted.
warfarin (9). SPAF
counted events
including one fatal
and in one patient
on placebo
(2.5% versus 0.5%/yr). SPINAF The SPINAF trial (8) was the only completed trial to provide
not or could not take warfarin. These latter patients were
a fully blinded assessment of warfarin. As a Department
separately randomized
Veterans
to aspirin at 325 mg/d or aspirin
of
Affairs study, it entered only men (see Table 1).
placebo. It was estimated that a relatively small fraction of
Patients with intermittent
eligible patients were actually recruited, a finding consistent
characteristics
with the other American studies. Patients in the SPAF war-
more cardiovascular
farin trial had a somewhat
younger average age at entry,
those in the other studies. The SPINAF trial used a 12-hour
with congestive heart failure than
cutoff to distinguish transient ischemic attacks from strokes,
and a smaller proportion in the BAATAF.
A third of patients had intermittent
and the overwhelming 1 year (see Table
AF,
majority had had AF for more than
1).
AF were not eligible. The entry
of subjects in SPINAF
suggested somewhat
disease, but were generally similar to
rather than the more usual 24-hour cutoff used by the other studies. In this study, warfarin at a low dose (target PTratio of 1.2 to 1.5) reduced the risk of stroke by 79% (see Table
SPAF showed that warfarin reduced the risk of embolic
2). There was one nonfatal intracerebral
hemorrhage
in the
events by 69% from 7.4 to 2.3%/yr (see Table 2). Of the
warfarin-treated
six patients in the warfarin arm who had strokes, four were
There
not taking warfarin at the time of their stroke.
group versus four in the placebo group (all gastrointestinal hemorrhages).
SPAF provided a mixed picture of aspirin’s effectiveness. It revealed an overall reduction
in embolic
risk of 42%;
however, this effect was markedly heterogeneous
in the two
group versus none in the placebo group.
were six other major hemorrhages
in the warfarin
This group of studies, as a whole, provides a consistency of evidence that is truly distinctive among prevention trials.
separately randomized studies of aspirin. In the study which
Warfarin can reduce the risk of stroke in AF by more than
included warfarin as a possible therapy (group l), there was
two-thirds (see Table 2, in particular the meta-analytic
only one embolic event observed in the aspirin group con-
mary estimate), which is almost all the additional risk con-
sum-
566
Singer RCTS
TO PREVENT
STROKE
IN ATRIAL
AEP Vol. 3, No. 5 September 1993: 563-567
FIBRILLATION
ferred by AF. This beneficial effect is conferred by low-dose anticoagulation, and at least in the trials, can be achieved with little additional risk of serious hemorrhage. In contrast, the current evidence bearing on aspirin’s efficacy is inconsistent and inconclusive. Nonetheless, warfarin is often an unattractive therapy for physicians and patients (10). Patients who are feeling well are reluctant to undertake a lifelong regimen of a risky therapy that demands frequent blood tests for monitoring, and ongoing concern about diet and other medications. It is also commonplace for patients who have begun warfarin to quit. Quitting may result from patients’ fatigue with the rigors of monitoring blood tests and medical visits, or from a “minor” bleeding event that undermines patient enthusiasm (11). More widespread use of warfarin and better sustained compliance would result from easier, but effective, antithrombotic regimens, and from better targeting of warfarin therapy to those likely to benefit most from therapy. Several of the trials identified risk factors for stroke among patients with AF. These factors have varied across studies. Age, a history of hypertension, and clinical heart disease such as coronary artery disease or congestive heart failure have increased the risk of stroke in at least one trial and in other observational studies. It also appears that patients who had a prior stroke while in AF are at higher risk for a second stroke. No echocardiographic features have been consistently found as risk factors. The lack of consistent support for the proposed risk factors, and their nonspecific nature (i.e., not specific for AF) suggest that they are not strong determinants of stroke. It appears that the only patients with AF who are at low enough risk to defer warfarin therapy are those who are young (e.g., < 60) and who have no other clinical heart disease.
risk in order to target warfarin therapy more rationally. Data on risk factors from the individual randomized trials have been sparse because the extreme efficacy of warfarin prompted the stopping of each trial before large numbers of strokes had been observed. There is now a collaborative working group pooling the results of the five trials, thereby increasing the number of analyzable end points (16). Further advances in risk assessment will need new clinical information and may need different study designs as well. New information would include imaging information from transesophageal echocardiography, and better hematologic measures such as the prothrombin activation fragment Fl+l. Different study designs would include case-control studies that can accumulate large numbers of patients with AF who have had a stroke more easily than randomized trials (17). Finally, more systematic study is needed of risk factors for serious hemorrhage while on warfarin. We remain fairly ignorant of determinants of hemorrhage other than elevated I’Ts (18). In particular, we need more precise estimates of the risk faced by the older patients who make up the bulk of those with AF.
SUMMARY The past decade of research established that AF is a major risk factor for stroke and that most of this additional risk is reversible by warfarin therapy. The randomized trials studied selected patients and monitored anticoagulation very carefully. To realize the great potential benefit of antithrombotic therapy in AF, we will need easier, yet safe anticoagulant therapy regimens, better systems of monitoring anticoagulation, and more informed patient selection.
REFERENCES FUTURE
RESEARCH
Future research should focus on the two themes of easier, safer anticoagulation and more precise risk assessment for patients with AF. We need to define the lowest effective level of anticoagulation. Such low doses of warfarin should add safety and may allow less frequent monitoring. Newer ways of monitoring anticoagulation should be tested. These would include more mathematically sophisticated monitoring schedules to allow less frequent testing, as well as office-based or even home-based (via fingerstick capillary blood samples) measurement of INRs to add convenience (12, 13). More widespread use of dedicated clinics or personnel to manage warfarin would add effectiveness and safety. The PT test itself should be compared to better measures of in vivo effectiveness of anticoagulants (14, 15). One hundred patients need to take warfarin each year to prevent about three strokes. A major research goal should be to identify stronger and more precise indicators of stroke
1. Dunn M, Alexander J, De Silva R, Hildner F. Antithrombotic in atrial fibrillation,
Chest.
therapy
1989;95(suppl): 118S-27s.
2. Wolf PA, Abbott RD, Kannel WB. Atrial fibrillation: A major contributor to stroke in the elderly, Arch Intern Med. 1987;147:1561-4. 3. United States Bureau of the Census. Statistical States: 1992. 112th ed. Washington, DC:15.
Abstract
of the United
4. Petersen P, Godtfredsen J, Boysen G, Andersen ED, Andersen B. Placebo-controlled, randomised trial of warfarin and aspirin for prevention of thromboembolic complications in chronic atrial fibrillation: The Copenhagen AFASAK study, Lancet. 1989;1:175-9. 5
The Boston Area Anticoagulation Trial for Atrial Fibrillation Investigators. The effect of low-dose warfarin on the risk of stroke in patients with nonrheumatic atrial fibrillation, N Engl J Med. 1990;323:150511.
6
Stroke Prevention in Atrial Fibrillation Investigators. Stroke prevention in atria1 fibrillation study: Final results, Circulation. 1991;84:52739.
7. Connolly SJ, Laupacis A, Gent M, Roberts RS, Cairns JA, Joyner C. Canadian Atrial Fibrillation Anticoagulation (CAFA) study, J Am Coil Cardiol. 1991;18:349-55. 8. Ezekowitz MD, Bridgers SL, James KE, et al. Warfarin
in the preven-
AEP Vol. 3, No. 5 September 1993: 563-567
tion of stroke associated with nonrheumatic J Med. 1992;327:1406-12.
RCTS TO PREVENT
atria1 fibrillation,
N Engl
therapy:
STROKE
A randomized
IN ATRIAL
prospective
Singer FIBRILLATION
567
study. Ann Intern Med 1989;lll:
730-737.
9. Singer DE, Hughes RA, Gress DR, et al. The effect of aspirin on the risk of stroke in patients with atria1 fibrillation: The BAATAF study, Am Heart J. 1992;124:1567-73. 10. Kutner M, Nixon G, Silverstone F. Physicians’ attitudes toward oral anticoagulants and antiplatelet agents for stroke prevention in elderly patients with atrial fibrillation, Arch Intern Med. 1991;151:1950-3. 11
Lancaster TR, Singer DE, Sheehan MA, et al. The impact of long-term warfarin thrapy on quality of life: Evidence from a randomized trial, Arch Intern Med. 1991;151:1944-9.
12
Kent DL, Vermes D, McDonell, Henikoff planning optimal follow-up for outpatients tion. Med Decis Making 1992;12:132-141.
J, Fihn SD. A model for on warfarin anticoagula-
13. White RH, McCurdy SA, van Marensdorff H, Woodruff DE, Leftgoff L. Home prothrombin time monitoring after the initiation of warfarin
14. Furie B, Furie BC. Molecular and cellular biology of blood coagulation, N Engl J Med. 1992;326:800-6. 15. Bauer KA, Rosenberg RD. The pathophysiology of the prethrombotic state in humans: Insights gained from studies using markers of hemostatic system activation, Blood. 1987;70:343-50. 16. Walker MD. Atrial fibrillation and antithrombotic prospective meta-analysis, Lancet. 1989; 1:325-6.
prophylaxis:
A
17. Moulton AW. Singer DE, Haas JS. Risk factors for stroke in patients with nonrheumatic atria1 fibrillation: A case-control study, Am J Med. 1991;91:156-61. 18. Landefeld CS, Rosenblatt MW, Goldman L. Bleeding in outpatients treated with warfarin: Relationto the prothrombin time and important remedial lesions, Am J Med. 1989;87:153-9. 19. Rothman 213-4.
KJ. Modern
Epidemiology.
Boston:
Little,
Brown;
1986:
MD
Epidemiologic studies have show that atria1 fibrillation (AF) raises the risk of stroke approximately fivefold, and that because AF is so common among the stroke-prone elderly, it accounts for about 15% of all strokes. Five recently completed, randurni~ed trials consistently found that the anticoagulant warfarin can prevent most of the additional stroke risk due to AF. This effect was seen at low doses. The trials have also demonstrated that warfarin therapy can be safe if careful patient selection and monitoring are implemented. Three of the trials prwided inconsistent, and currently inconclusive evidence about the e$acy of aspirin. The trials have not settled the anticoagukztion decision for all patients. Warfarin remains a demanding and risky therapy, which many patients and physicians do not find attractive. Future research should attempt to refine the risk ofstroke,and ofmajor hemorrhage during warfarin therapy among patients with AF, and should seek safer, less demanding, yet effective antithrombotic regimens. Ann Epidemiol 1993;3:563-567. KEY WORDS:
Randomized
trials, anticoagulation,
INTRODUCTION The past decade has witnessed a remarkable advance in our understanding of the stroke risk posed by atria1 fibrillation (AF), and of the prevention ofthis risk by warfarin anticoag ulation. For historic and clinical reasons, the relationship between AF and stroke has been categorized according to the presence or absence of rheumatic heart disease. AF in the setting of rheumatic heart disease, primarily mitral stenosis, has long been a widely accepted risk factor for stroke (1). The assumption has been that in patients with obstructed outflow of blood due to mitral stenosis, AF increases the stasis of blood in the atrium and promotes thrombus formation and subsequent emboli. Lifelong anticoagulation is standard therapy for such patients, although the clinical studies underlying this well-accepted management strategy are fairly weak (1). Presumably, the clear consensus in favor of anticoagulants in rheumatic AF results from the dramatic presentation of stroke or visceral embolus in these predominantly young patients. The vast majority of patients with AF have “nonrheumatic” AF associated with aging. These patients often have other conditions, such as hypertension and coronary artery disease, that predispose to stroke. The controversy about nonrheumatic AF focused on whether this rhythm disorder
From the General Internal Medicine Unit, Medical Services, Massachusetts General Hospital and the Departments of Medicine and of Health Care Policy, Harvard Medical School, Boston, MA. Address reprint requests to: Daniel E. Singer, MD, General Internal Medicine Unit, Bulfinch 1, Massachusetts General Hospital, Boston, MA 02114.
warfarin, atria1 fibrillation.
itself led to stroke, and even if it did, whether anticoagulants’ efficacy would be worth their risk in such an elderly group of patients. In the United States, the Framingham Heart Study (2) has provided a rich source of information on AF and stroke. The prevalence of AF is approximately 2% for individuals in their 6Os, and rises to 10% for those older than 80. Stroke risk among subjects without AF rises with age. The striking finding, however, is that the stroke risk among those with AF rises proportionately, with a relative risk of four to five persisting across all ages. The public health implication of these data is that AF accounts for 15% of all strokes, or about 70,000 strokes each year in the United States (2, 3).
THE RANDOMIZED PREVENT STROKE
TRIALS OF WARFARIN TO IN ATRIAL FIBRILLATION
Against this background, five randomized trials of the anticoagulant warfarin in nonrheumatic AF were begun independently in the mid 1980s (4-8). The trials include the Copenhagen Atria1 Fibrillation, Aspirin, Anticoagulation Trial (AFASAK), the Boston Area Anticoagulation Trial for Atria1 Fibrillation (BAATAF), the Stroke Prevention in Atria1 Fibrillation (SPAF) trial, the Canadian Atria1 Fibrillation Anticoagulation (CAFA) trial, and the Veterans’ Administration Stroke Prevention in Nonrheumatic Atria1 Fibrillation (SPINAF) trial. All were primary prevention trials, with a very small fraction of patients having had a prior stroke. All were stopped early because of warfarin’s efficacy.
564
Singer
AEP Vol. 3, No. 5 September 1993: 563-567
RCTS TO PREVENT STROKE IN ATRIAL FIBRILLATION
TABLE 1. Clinical characteristics of patients at entry to randomized trials” Feature
AFASAK
Male (%)
54 75h
Age (Y) AF > 1 y (%) PAF (%) Hx stroke (%) HBP (%) Diabetes (%) Angina (%) Hx MI (96) CHF (%)
0 4 31 10 16 8 51
BAATAF
SPAF
CAFA
70 66 72 34 8“ 55 19 10 6 19
73 67 a2 7 4“
70 68 68 16 3 51 16 25 16 28
34 10 20 12 20
SPINAF 100 67 86‘ 0 0 62 20 23 21 30
a These are the features of the control or placebo groups from each trial. b The mean age is provided except for AFASAK, which provided the median. of AF was not provided by AFASAK. For SPINAF, 86% had AF for longer than 6 ma ’ In SPAF and CAFA, percentages are for history of stroke and/or transient ischemic attack. AF = atrial fibrillation; PAF = paroxysmal or intermittent AF; Hx = history of; HBP = hypertension; MI = myocardial infarction; CHF = congestive heart failure.
’Duration
AFASAK AFASAK (4) was a three-armed trial comparing warfarin at a target international normalized ratio (INR) of 2.8 to 4.2 (roughly corresponding to a prothrombin time (PT) ratio of 1.7 to 2.0), aspirin at 75 mg/d, and an aspirin-placebo. AFASAK’s target INR was the highest of the trials. As with the BAATAF and SPAF trials, AFASAK did not blind warfarin therapy. AFASAK was probably the closest to a community trial of warfarin. Patients were identified at two outpatient electrocardiography laboratories used by general
TABLE
2. Results of the randomized
trials of warfarin for atria1 fibrillation
AFASAK” Warfarin Subjects (n) Emboli (n) Person-y
practitioners. Of 1842 eligible patients, 1007 entered the trial, probably the highest recruitment fraction of any of the trials. Slightly more than half of the patients were men (Table 1). The median age was in the mid 70s. A high proportion of patients had hypertension. Over 50% were reported as having heart failure, although the specificity of this diagnosis is not clear. Three hundred thirty-six subjects were assigned to receive placebo (i.e., aspirin-placebo); 336 to receive aspirin, and 335 to receive warfarin. Thirty-eight percent of patients assigned to the warfarin group withdrew from the study. This high percentage of withdrawals may have resulted from AFASAK’s high recruitment fraction; more highly selected samples of patients may be more compliant. AFASAK reported four embolic events in the warfarin group, 20 in the aspirin group, and 21 in the placebo group, corresponding to an annual rate of 1.6% in warfarin versus 5.6% in the placebo group (Table 2). There were two major hemorrhages in patients on warfarin, including one fatal intracerebral bleed. The marked benefit of warfarin coupled with the low rate of major hemorrhage led to the early stopping of the trial. Two additional points about AFASAK are worth noting. First, there was subsequent disagreement about the event rates reported for an intention-to-treat analysis. If patients who withdrew from therapy are included in the analysis, the numbers of thromboembolic events become 10,2 1, and 25 for warfarin, aspirin, and placebo, respectively (P - 0.05). In contrast, only three patients in the warfarin arm who sustained a thromboembolic event were actually
BAATAF
SPAF
CAFA
SPINAF
rate
335 4 250 1.6%
212 2 487 0.41%
210 6 260 2.3%
187 5 200 2.5%
260 4 456 0.88%
Control Subjects (n) Emboli (n) Person-y Annual rate
336 21 373 5.6%
208 13 435 3.0%
211 18 244 7.4%
191 11 212 5.2%
265 19
52%
79% 52-90%
Annual
Preventive Efficacy 95% CI
71% 23-90%
Intracranial hemorrhage rate due to warfarin Meta-analvtic
0.40%
estimate of the oreventive
efficacy of warfarin:
86% 51-969/o
0.21%
69% 27-85%
0%
(- 36)-87% 0.50%
440 4.3%
0.22%
73% (95% CI: 57-83%)b
a Person-y in the warfarin and control arms of AFASAK were inferred from data given in their report (4); p erson-y were given explicitly in the other trials (5-8). h Meta-analytic summary estimated using Rothman’s adaptation of the Mantel-Haensrel technique (19). CI = confidence interval.
565
Singer RCTS TO PREVENT STROKE IN ATRIAL FIBRILLATION
AEP Vol. 3, No. 5 September 1993:563-567
taking warfarin, and of these, two had briefly discontinued
trasted with 18 in the placebo group. In group 2, where
warfarin, resulting in an INR well below target at the time
patients were assigned to aspirin or aspirin-placebo
of their event.
there were 25 embolic events in those on aspirin versus 28
only,
in the placebo group (not statistically significant). Subgroup analyses indicated
BAATAF The BAATAF
trial (5) randomized
2 12 patients to receive
“low-dose” warfarin (target PT ratio of 1.2 to 1.5) and 208 to the control
group. Therapy
was not blinded.
Subjects
in the control group could take aspirin, and approximately half did. Nearly three-fourths Table
of the patients were male (see
1). The average age was 68. A small proportion
subjects had intermittent
of
AF. Most had had AF for several
years. Half the subjects had no clinical heart disease other than AF, although many of these had hypertension betes. Only 10% of BAATAF
or dia-
patients assigned to the war-
serious strokes. There was one fatal intracerebral
hemorrhage
in a pa-
hematoma
with full
tient on warfarin and one subdural
recovery. Two patients in the placebo group had subdural hematomas
(with full recovery).
There was no increase in
major bleeding from other anatomic
sites with warfarin.
CAFA The CAFA
trial (7) was a double-blind,
placebo-controlled
trial of low-dose warfarin (target INR of 2 to 3) in AF. After
farin group withdrew from therapy permanently. In the BAATAF
that aspirin did not confer benefit on
patients over age 75, and was not very effective at preventing
trial 13 strokes were observed in the
the AFASAK
report and a preliminary report of the SPAF
control arm versus two in the warfarin group, for a preven-
trial, the CAFA
tive efficacy of 86% (see Table 2). There was one presumed
allow all patients access to warfarin. CAFA
subdural hematoma
using an “efficacy” analysis where events occurring
Otherwise,
in a patient on warfarin that was fatal.
the rate of serious bleeding was essentially the
same in warfarin-treated
as in control
subjects.
Eight of
the 13 control patients who sustained a stroke were taking aspirin. Analyses to remove confounding domized assessment
of aspirin estimated
in this nonranthat the rate of
investigators
decided to stop their trial to
By this approach,
there were five primary embolic events
in the warfarin group versus 11 in the placebo group, for a risk reduction
of 52% (see Table 2). Of these five events
in the warfarin group, two occurred in patients who were
stroke persisted at severalfold higher on aspirin than on
not taking the drug, and in an additional were well below the anticoagulation occurred
trial (6) randomly
assigned patients to receive
warfarin at a low dose (target PT ratio of 1.3 to 1.8), aspirin at 325 mg/d, or aspirin-placebo. blinded.
Warfarin therapy was not
The SPAF trial also studied patients who would
hemorrhage,
two, the INRs
target. Major bleeding
in five patients on warfarin,
intracranial
The SPAF
more
than 28 days after stopping therapy would not be counted.
warfarin (9). SPAF
counted events
including one fatal
and in one patient
on placebo
(2.5% versus 0.5%/yr). SPINAF The SPINAF trial (8) was the only completed trial to provide
not or could not take warfarin. These latter patients were
a fully blinded assessment of warfarin. As a Department
separately randomized
Veterans
to aspirin at 325 mg/d or aspirin
of
Affairs study, it entered only men (see Table 1).
placebo. It was estimated that a relatively small fraction of
Patients with intermittent
eligible patients were actually recruited, a finding consistent
characteristics
with the other American studies. Patients in the SPAF war-
more cardiovascular
farin trial had a somewhat
younger average age at entry,
those in the other studies. The SPINAF trial used a 12-hour
with congestive heart failure than
cutoff to distinguish transient ischemic attacks from strokes,
and a smaller proportion in the BAATAF.
A third of patients had intermittent
and the overwhelming 1 year (see Table
AF,
majority had had AF for more than
1).
AF were not eligible. The entry
of subjects in SPINAF
suggested somewhat
disease, but were generally similar to
rather than the more usual 24-hour cutoff used by the other studies. In this study, warfarin at a low dose (target PTratio of 1.2 to 1.5) reduced the risk of stroke by 79% (see Table
SPAF showed that warfarin reduced the risk of embolic
2). There was one nonfatal intracerebral
hemorrhage
in the
events by 69% from 7.4 to 2.3%/yr (see Table 2). Of the
warfarin-treated
six patients in the warfarin arm who had strokes, four were
There
not taking warfarin at the time of their stroke.
group versus four in the placebo group (all gastrointestinal hemorrhages).
SPAF provided a mixed picture of aspirin’s effectiveness. It revealed an overall reduction
in embolic
risk of 42%;
however, this effect was markedly heterogeneous
in the two
group versus none in the placebo group.
were six other major hemorrhages
in the warfarin
This group of studies, as a whole, provides a consistency of evidence that is truly distinctive among prevention trials.
separately randomized studies of aspirin. In the study which
Warfarin can reduce the risk of stroke in AF by more than
included warfarin as a possible therapy (group l), there was
two-thirds (see Table 2, in particular the meta-analytic
only one embolic event observed in the aspirin group con-
mary estimate), which is almost all the additional risk con-
sum-
566
Singer RCTS
TO PREVENT
STROKE
IN ATRIAL
AEP Vol. 3, No. 5 September 1993: 563-567
FIBRILLATION
ferred by AF. This beneficial effect is conferred by low-dose anticoagulation, and at least in the trials, can be achieved with little additional risk of serious hemorrhage. In contrast, the current evidence bearing on aspirin’s efficacy is inconsistent and inconclusive. Nonetheless, warfarin is often an unattractive therapy for physicians and patients (10). Patients who are feeling well are reluctant to undertake a lifelong regimen of a risky therapy that demands frequent blood tests for monitoring, and ongoing concern about diet and other medications. It is also commonplace for patients who have begun warfarin to quit. Quitting may result from patients’ fatigue with the rigors of monitoring blood tests and medical visits, or from a “minor” bleeding event that undermines patient enthusiasm (11). More widespread use of warfarin and better sustained compliance would result from easier, but effective, antithrombotic regimens, and from better targeting of warfarin therapy to those likely to benefit most from therapy. Several of the trials identified risk factors for stroke among patients with AF. These factors have varied across studies. Age, a history of hypertension, and clinical heart disease such as coronary artery disease or congestive heart failure have increased the risk of stroke in at least one trial and in other observational studies. It also appears that patients who had a prior stroke while in AF are at higher risk for a second stroke. No echocardiographic features have been consistently found as risk factors. The lack of consistent support for the proposed risk factors, and their nonspecific nature (i.e., not specific for AF) suggest that they are not strong determinants of stroke. It appears that the only patients with AF who are at low enough risk to defer warfarin therapy are those who are young (e.g., < 60) and who have no other clinical heart disease.
risk in order to target warfarin therapy more rationally. Data on risk factors from the individual randomized trials have been sparse because the extreme efficacy of warfarin prompted the stopping of each trial before large numbers of strokes had been observed. There is now a collaborative working group pooling the results of the five trials, thereby increasing the number of analyzable end points (16). Further advances in risk assessment will need new clinical information and may need different study designs as well. New information would include imaging information from transesophageal echocardiography, and better hematologic measures such as the prothrombin activation fragment Fl+l. Different study designs would include case-control studies that can accumulate large numbers of patients with AF who have had a stroke more easily than randomized trials (17). Finally, more systematic study is needed of risk factors for serious hemorrhage while on warfarin. We remain fairly ignorant of determinants of hemorrhage other than elevated I’Ts (18). In particular, we need more precise estimates of the risk faced by the older patients who make up the bulk of those with AF.
SUMMARY The past decade of research established that AF is a major risk factor for stroke and that most of this additional risk is reversible by warfarin therapy. The randomized trials studied selected patients and monitored anticoagulation very carefully. To realize the great potential benefit of antithrombotic therapy in AF, we will need easier, yet safe anticoagulant therapy regimens, better systems of monitoring anticoagulation, and more informed patient selection.
REFERENCES FUTURE
RESEARCH
Future research should focus on the two themes of easier, safer anticoagulation and more precise risk assessment for patients with AF. We need to define the lowest effective level of anticoagulation. Such low doses of warfarin should add safety and may allow less frequent monitoring. Newer ways of monitoring anticoagulation should be tested. These would include more mathematically sophisticated monitoring schedules to allow less frequent testing, as well as office-based or even home-based (via fingerstick capillary blood samples) measurement of INRs to add convenience (12, 13). More widespread use of dedicated clinics or personnel to manage warfarin would add effectiveness and safety. The PT test itself should be compared to better measures of in vivo effectiveness of anticoagulants (14, 15). One hundred patients need to take warfarin each year to prevent about three strokes. A major research goal should be to identify stronger and more precise indicators of stroke
1. Dunn M, Alexander J, De Silva R, Hildner F. Antithrombotic in atrial fibrillation,
Chest.
therapy
1989;95(suppl): 118S-27s.
2. Wolf PA, Abbott RD, Kannel WB. Atrial fibrillation: A major contributor to stroke in the elderly, Arch Intern Med. 1987;147:1561-4. 3. United States Bureau of the Census. Statistical States: 1992. 112th ed. Washington, DC:15.
Abstract
of the United
4. Petersen P, Godtfredsen J, Boysen G, Andersen ED, Andersen B. Placebo-controlled, randomised trial of warfarin and aspirin for prevention of thromboembolic complications in chronic atrial fibrillation: The Copenhagen AFASAK study, Lancet. 1989;1:175-9. 5
The Boston Area Anticoagulation Trial for Atrial Fibrillation Investigators. The effect of low-dose warfarin on the risk of stroke in patients with nonrheumatic atrial fibrillation, N Engl J Med. 1990;323:150511.
6
Stroke Prevention in Atrial Fibrillation Investigators. Stroke prevention in atria1 fibrillation study: Final results, Circulation. 1991;84:52739.
7. Connolly SJ, Laupacis A, Gent M, Roberts RS, Cairns JA, Joyner C. Canadian Atrial Fibrillation Anticoagulation (CAFA) study, J Am Coil Cardiol. 1991;18:349-55. 8. Ezekowitz MD, Bridgers SL, James KE, et al. Warfarin
in the preven-
AEP Vol. 3, No. 5 September 1993: 563-567
tion of stroke associated with nonrheumatic J Med. 1992;327:1406-12.
RCTS TO PREVENT
atria1 fibrillation,
N Engl
therapy:
STROKE
A randomized
IN ATRIAL
prospective
Singer FIBRILLATION
567
study. Ann Intern Med 1989;lll:
730-737.
9. Singer DE, Hughes RA, Gress DR, et al. The effect of aspirin on the risk of stroke in patients with atria1 fibrillation: The BAATAF study, Am Heart J. 1992;124:1567-73. 10. Kutner M, Nixon G, Silverstone F. Physicians’ attitudes toward oral anticoagulants and antiplatelet agents for stroke prevention in elderly patients with atrial fibrillation, Arch Intern Med. 1991;151:1950-3. 11
Lancaster TR, Singer DE, Sheehan MA, et al. The impact of long-term warfarin thrapy on quality of life: Evidence from a randomized trial, Arch Intern Med. 1991;151:1944-9.
12
Kent DL, Vermes D, McDonell, Henikoff planning optimal follow-up for outpatients tion. Med Decis Making 1992;12:132-141.
J, Fihn SD. A model for on warfarin anticoagula-
13. White RH, McCurdy SA, van Marensdorff H, Woodruff DE, Leftgoff L. Home prothrombin time monitoring after the initiation of warfarin
14. Furie B, Furie BC. Molecular and cellular biology of blood coagulation, N Engl J Med. 1992;326:800-6. 15. Bauer KA, Rosenberg RD. The pathophysiology of the prethrombotic state in humans: Insights gained from studies using markers of hemostatic system activation, Blood. 1987;70:343-50. 16. Walker MD. Atrial fibrillation and antithrombotic prospective meta-analysis, Lancet. 1989; 1:325-6.
prophylaxis:
A
17. Moulton AW. Singer DE, Haas JS. Risk factors for stroke in patients with nonrheumatic atria1 fibrillation: A case-control study, Am J Med. 1991;91:156-61. 18. Landefeld CS, Rosenblatt MW, Goldman L. Bleeding in outpatients treated with warfarin: Relationto the prothrombin time and important remedial lesions, Am J Med. 1989;87:153-9. 19. Rothman 213-4.
KJ. Modern
Epidemiology.
Boston:
Little,
Brown;
1986: