Factors affecting bleeding risk during anticoagulant
therapy in patients with atrial fibrillation:
Observations from the Atrial Fibrillation Follow-up Investigation of Rhythm Management (AFFIRM) Study John P. DiMarco, MD, PhD, a Gregory Flaker, MD, b Albert L. Waldo, MD, c Scott D. Corley, BS, d H. Leon Greene, MD, Robert E. Safford, MD, e Lynda E. Rosenfeld, MD, f Gladys Mitrani, PharmD, g Margit Nemeth, MD, a and The AFFIRM Investigators Charlottesville, Va, Columbia, Mo, Cleveland, Ohio, Seattle, Wash, Jacksonville, Fla, New Haven, Conn, and Los Angeles, Calif Background Stroke and systemic thromboembolism are serious problems for patients with atrial fibrillation (AF), but their incidence can be substantially reduced by appropriate anticoagulation. Bleeding is the major complication of anticoagulant treatment, and the relative risks for bleeding vs stroke must be considered when starting anticoagulation. Methods The AFFIRM trial included patients with AF and at least one risk factor for stroke, randomly assigning them to either a rate-control or rhythm-control strategy. All patients were initially treated with warfarin. The incidence of protocol- defined major and minor bleeding was documented during follow-up. Variables associated with bleeding were determined using a Cox proportional hazards model, using baseline and time- dependent covariates. Results The 4060 patients in the AFFIRM trial were followed for an average of 3.5 years. Major bleeding occurred in 260 patients, an annual incidence of approximately 2% per year, with no significant difference between the rate- control and rhythm-control groups. Increased age, heart failure, hepatic or renal disease, diabetes, first AF episode, warfarin use, and aspirin use were significantly associated with major bleeding. Minor bleeding was common in both treatment arms, with 738 patients reporting this problem in one or more visits. Conclusions Bleeding is a significant problem that complicates management of patients with AF. Risk factors for bleeding can be identified, and knowledge of these risk factors can be used to plan therapy. (Am Heart J 2005;149:650- 6.) Stroke and systemic thromboembolic events are the most common serious complications of atrial fibrillation (AF). 1-3 In patients with AF, the risk of stroke increases with age and other risk factors including hypertension, diabetes, atrial enlargement, and left ventricular dys- function. 4 Anticoagulant therapy with warfarin un- equivocally decreases the risk of stroke in AF. As a result, warfarin anticoagulation is now recommended for patients with AF and risk factors for stroke. 4-10 Bleeding is the major complication associated with warfarin therapy. 11,12 Up to one third of patients screened for participation in prior clinical trials on the use of warfarin in AF were deemed ineligible for chronic anticoagulation, primarily because of a high perceived bleeding risk. 5,11 After enrollment, the reported annual rates for major bleeding in the randomized clinical trials that evaluated anticoagulation in patients with AF were between 0.4% and 2.6% per year. 6,11-13 Rates of minor bleeding, which was not uniformly defined, were as high as 15.4%per year. 6 The risk of bleeding was related to the intensity of anticoagulation and to selected patient characteristics including hypertension, female gender, severe cardiac disease, and renal insufficiency. 11,12 Drug interactions with warfarin are also common. Pharmaco- kinetic interactions may contribute to unstable antico- agulation intensities that can increase risks for both bleeding and thromboembolism. Pharmacodynamic interactions with other agents commonly used in elderly patients with cardiovascular disease that may indepen- dently promote bleeding also occur. From the a Cardiovascular Division, University of Virginia Health System, Charlottesville, Va, b Columbia, Mo, c Cleveland, Ohio, d Seattle, Wash, e Jacksonville, Fla, f New Haven, Conn, and g Los Angeles, Calif. The AFFIRM investigators and their affiliations are listed in Ref. [13]. This study was supported by contract no. N01-HC-55139 with the National Heart, Lung, and Blood Institute, Bethesda, MD. No other financial disclosures. Submitted April 14, 2004; accepted November 13, 2004. Reprint requests: John P. DiMarco, MD, PhD, Cardiovascular Division, University of Virginia Health System, Box 800158, Charlottesville, VA 22908. E-mail: jpd4h@virginia.edu 0002-8703/$ - see front matter n 2005, Elsevier Inc. All rights reserved. doi:10.1016/j.ahj.2004.11.015 The AFFIRM Study 14-16 compared rate-control and rhythm-control strategies in patients with AF who had risk factors for stroke. All patients were eligible for warfarin at baseline. Patients in the rate-control group were supposed to continue their warfarin regimen with a target internationalized normalized ratio ( INR) of 2 to 3 throughout the study. Warfarin anticoagulation was encouraged, but not mandated, for patients who remained in sinus rhythm in the rhythm-control group. Most, however, continued their warfarin regimen. Amiodarone and aspirin, 2 drugs with well-described pharmacokinetic or pharmacodynamic interactions with warfarin that might promote bleeding, were used in many patients in the AFFIRM trial. In this report, we describe the rates of and risk factors for bleeding in the AFFIRM trial. Methods Four thousand sixty patients were randomized in the AFFIRM trial and their clinical characteristics have been previously reported. 15,16 The average length of follow-up was 3.5 years (range 0 to 5.9 years). 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, intra- parenchymal hemorrhage), stroke (embolus, thrombosis, hemorrhage), disabling anoxic encephalopathy, cardiac arrest, and major noncentral nervous system (CNS) bleeding. Other secondary end points were cost, quality of life, and functional status. This paper evaluates bleeding. Patients were seen at the time of randomization and at follow-up visits at 2 months, 4 months, and every 4 months thereafter. At each visit, the study coordinator documented any interim bleeding event that occurred and whether warfarin, aspirin, and antiarrhythmic drugs were used at any point during the preceding time interval. A diagnosis of major bleeding outside the CNS was defined as bleeding that required transfusion of z2 units of blood, hospitalization in an intensive Table I. Baseline characteristics All patients No major bleeding Major bleeding P n (%) 4060 3800 260 Randomized to rhythm control 2033 (50) 1909 (50) 124 (48) .427 Age (y) (mean F SD) 69.7 F 9.0 69.6 F 9.0 72.3 F 8.2 b.001* Women 1594 (39) 1478 (39) 116 (45) .068 Minority 461 (11) 428 (11) 33 (13) .482 History of hypertension 2876 (71) 2686 (71) 190 (73) .412 History of CAD 1551 (38) 1439 (38) 112 (43) .094 History of CHF 939 (23) 857 (23) 82 (32) .001* History of diabetes 813 (20) 748 (20) 65 (25) .038* History of stroke or TIA 542 (13) 497 (13) 45 (17) .052 History of hepatic or renal disease 231 (6) 205 (5) 26 (10) .002* Recent history of smoking 496 (12) 471 (12) 25 (10) .186 Qualifying episode of AF is first episode documented 1391 (36) 1291 (35) 100 (40) .113 Left atrial enlargement (size N4.0 cm) 2023 (65) 1888 (65) 135 (66) .724 Left ventricular dysfunction (ejection fraction b50%) 788 (26) 724 (26) 64 (32) .063 Mitral regurgitation N2+ 647 (20) 603 (20) 44 (21) .799 Figures in parentheses are percent values. CAD, coronary artery disease; CHF, congestive heart failure; TIA, transient ischemic attack. *P b .05. Table II. Number of patients who had major bleeding events* Site of bleeding All patients Rate control patients Rhythm control patients P n (%) 4060 2027 2033 Ally 260 (9.2) 136 (9.6) 124 (8.9) .452 Non-CNS 203 (7.3) 107 (7.7) 96 (6.9) .436 CNSz 59 (2.1) 29 (1.9) 30 (2.2) .892 Primary intraparenchymal hemorrhage 34 (1.2) 18 (1.1) 16 (1.3) .732 Subdural or subarachnoid hemorrhage 24 (0.8) 11 (0.8) 13 (0.8) .677 Figures in parentheses are percent values. *Percentages were derived by Kaplan-Meier analysis. P values were based on the log-rank statistic. yThe sum of non-CNS and CNS bleeding events is greater than the number for ball Q types because there were 2 patients who had both non-CNS and CNS events. zThe sum of primary intraparenchymal hemorrhage and subdural or subarachnoid hemorrhage is less than the number for bCNSQ because information on type of CNS event was missing for one patient. American Heart Journal Volume 149, Number 4 DiMarco et al 651 care unit, and/or discontinuation of anticoagulant or antiplate- let therapy. Major bleeding included CNS (intraparenchymal, subdural, or subarachnoid) and non-CNS hemorrhage. Minor bleeding included hematuria, gastrointestinal bleeding, epi- staxis, and other reported bleedings that did not meet the major bleeding criteria listed above. The specific sites of non-CNS bleeding were not recorded for each episode. There was no difference between patients with and those without major bleeding in the number of forms collected. In the group without bleeding, 80.5%, 11%, and 8.6% had no, 1, or 2 or more forms missing, respectively. In the group with major bleeding, 82.9%, 10.5%, and 6.6% had no, 1, or 2 or more forms missing, respectively ( P = .6). All patients were eligible for long-term anticoagulation at enrollment. Physicians were encouraged to maintain anti- coagulation in both treatment arms unless contraindications developed. However, the protocol allowed discontinuation of warfarin use after a minimum of 4 (and preferably 12) weeks if sinus rhythm had apparently been maintained by antiar- rhythmic drugs. Statistics Baseline characteristics of patients were compared with m 2 tests and t tests as appropriate. The primary analysis was an intention-to-treat comparison of time to a major bleeding episode. Rates were estimated by the method of Kaplan and Meier and compared by the log-rank test. Patients were censored at last contact, at withdrawal from the study, or at death if the outcome variable did not include stroke. A secondary analysis used multivariate Cox proportional hazards regression to assess the relationship of certain variables (ie, bcovariatesQ) with risk of major bleeding after adjusting for the other covariates included in the analysis. Data from 3788 patients were available for this analysis. Two hundred seventy-two patients had incomplete data and were excluded from the analysis but the major bleeding rate among these patients was similar to that among those included (log-rank statistic 0.58, P = .45). Covariates included randomized treatment arm, age, sex, history of coronary artery disease, history of congestive heart failure, history of hypertension, history of diabetes, history of stroke or transient ischemic attack, history of hepatic or renal disease, recent history of smoking, and qualifying for enrollment with a first episode of AF. In those patients with echocardiographic data available (n = 3024), left atrial enlargement (ie, size N4.0 cm), left ventricular dysfunction (ie, ejection fraction b50%), and mitral regurgitation were also considered. Also included in the Cox analysis were 3 variables whose values changed over time (ie, time-dependent covariates): warfarin use, aspirin use, and amiodarone use. Each patients use of the 3 drugs was determined at various time points over the course of follow-up based on the responses recorded on the patients data forms. Specific start and stop dates for use of these and other drugs were not recorded. Instead, responses on the forms indicated whether the patient had used a particular drug since the previous follow- up. A patient was considered to have used a drug if it was recorded as such on the most recent available form before the time point of interest. Stepwise analytical procedures were performed. A stepwise procedure is similar to a forward-selection procedure except that the variables already in the model do not necessarily remain in the model, as each forward selection step is followed by a backward elimination step. Results Baseline characteristics for patients with and those without major bleeding events are summarized in Table I. Patients with major bleeding events were older and more likely to have a history of congestive heart failure, diabetes, or hepatic or renal disease. Numbers and actuarial rates of major bleeding events are shown in Table II. Two hundred sixty patients (9.2%) had at least one major bleeding event. Of these, 136 had been randomized to the rate-control group and 124 to the rhythm-control group, a difference that was not significant. Major bleeding outside the CNS was observed in 203 patients (7.3%), Figure 1 Time to major bleeding. One major bleeding event that occurred after 5 years of follow-up is not shown. Table III. Internationalized normalized ratio values and major bleeding events INR range No. of major bleeding events 0-1.2 13 1.3-1.9 34 2.0-3.0 85 3.1-4.0 27 4.1-8.0 32 N8.0* 20 Total 211 *Three outliers (INR values 73.6, 81.0, and 91.7) excluded. American Heart Journal April 2005 652 DiMarco et al with 107 of these randomized to the rate-control group and 96 to the rhythm-control group. Major bleeding within the CNS (including primary intra- parenchymal hemorrhage and subdural or subarach- noid hemorrhage) was observed in 59 patients (2.1%). Two of these patients also experienced a major non- CNS bleeding episode. Of the patients who had a CNS bleeding event, 29 were in the rate-control group and 30 were in the rhythm-control group. Thirty-four of the patients with CNS bleeding had a primary intra- parenchymal hemorrhage, and 24 patients had a subdural or subarachnoid hemorrhage (the type of CNS hemorrhage was unknown for one patient). The incidence of major bleeding through 5 years of follow-up by treatment group is shown in Figure 1. The incidence of bleeding was constant throughout the trial, approximately 2% per year in both randomized groups. There was also no difference by treatment group in the incidence of a combined end point of major bleeding and death (log-rank statistic 0.81, P = .37, data not shown). A measured INR at the time of major bleeding was available for 214 of the 260 patients. There were 3 outliers who had reported INR values (ie, 91.7, 81.0, 73.6) that were outside acceptable standards. These values were removed from all quantitative analyses. The median INR at the time of major bleeding was 2.6 (n = 211, interquartile range 2.0-4.0). Fifty-two (24.6%) of the values, excluding those of the 3 outliers, were above 4.0 (Table III). The median INR at the time of a CNS bleed was 2.9 (n = 54, interquartile range 1.7-3.2 ), with 9 (16.7%) of the values above 4.0. Warfarin was being used at the time of major bleeding in 217 (83.8%) of 259 patients. Information on warfarin use was missing for one patient. Minor bleeding was frequent in the AFFIRM Study (Figure 2), reported in one or more occasions by 738 patients. It was reported once by 559 patients, twice by 115, and more than twice by 64. Minor bleeding was more frequent in the initial months of study participa- tion then occurred at a relatively constant rate during the rest of follow-up. Because minor bleeding likely influenced subsequent use of anticoagulants, we did not include minor bleeding in our analysis of predictors for major bleeding. However, 45 of the 260 patients who experienced a major bleeding episode had previously reported a minor bleeding one. Warfarin discontinuation During the AFFIRM trial, 1365 (33.6%) of the trial participants discontinued warfarin use for some period. Apparent maintenance of sinus rhythm in a rhythm- control patient was the listed reason for 613 (44.9%) discontinuations. Bleeding, reported for 285 patients (20.9%), was the second most common reason for warfarin discontinuation. Physical frailty, physician or patient refusal, and surgery were reported by smaller numbers. Some patients started and stopped warfarin use multiple times during the trial. The numbers given above are for the first discontinuation only. Risk factors for bleeding Several clinical factors were associated with an increased risk of major bleeding. In the multivariate model, age, history of congestive heart failure, history of diabetes, history of hepatic or renal disease, qualifying for enrollment with a first episode of AF, warfarin use, and aspirin use were associated with an increased risk of major bleeding (Table IV). Other factors examined did not show a significant association. Discussion Randomized clinical trials have shown that antico- agulation with warfarin results in a 60% to 70% reduction in the risk of stroke and systemic thrombo- embolism during AF. 4-10 The efficacy of warfarin is even higher if an on-therapy, rather than an intention-to-treat analysis, is used. 16 Because chronic anticoagulation also raises the risk of bleeding complications, current guidelines recommend warfarin only for those patients with AF in whom the risk for thromboembolic events outweighs the risk for serious bleeding. 2,3,17 All patients in the AFFIRM trial had one or more risk factors for thromboembolism. The expected embolic event rate, if not on anticoagulation, based on individual patient characteristics, ranged from 2.5% to over 12% per year. 18,19 The data presented here confirm that the risks for major bleeding during anticoagulation in this population are considerably lower than the anticipated benefit in stroke reduction. In the AFFIRM trial, all patients were considered eligible at baseline for chronic anticoagulation. Anti- Figure 2 Time to minor bleeding. American Heart Journal Volume 149, Number 4 DiMarco et al 653 coagulation was mandated by the protocol for all patients in the rate-control group unless contraindica- tions developed. Over 85% of the rate-control group patients at each follow-up visit continued their warfarin regimen. In the rhythm-control group, chronic anti- coagulation was recommended, but, in accordance with then current guidelines, anticoagulants could be dis- continued at the local physicians discretion if sinus rhythm was apparently maintained for at least 4 (and preferably 12) weeks. Despite this provision, most patients in the rhythm-control group continued their use of anticoagulants, with N70% reporting current warfarin use at each follow-up visit. In both groups, most ischemic strokes occurred if warfarin had either been discontinued (49% overall) or if the INR at the time of the event was subtherapeutic (32% overall). 14 The rate of major bleeding in the AFFIRM trial was approximately 2% per year. This figure is well below the expected stroke rate in similar populations with AF in the absence of anticoagulation 18,19 and is similar to that seen in other trials of warfarin anticoagulation in AF. 4-13 Seven covariates were significantly associated with an increased risk of bleeding: increased age, history of congestive heart failure, diabetes, hepatic or renal disease, first episode of AF, warfarin use, and aspirin use. The first 3 of these covariates are also risk factors that increase the probability of stroke, such that a favorable risk/benefit ratio is likely to be maintained in patients with these characteristics. Patients with renal disease are at increased risk for both thromboembolic events and bleeding because of their underlying condition. Signifi- cant hepatic disease is associated with a high incidence of bleeding and may contribute to variable warfarin metabolism, resulting in unstable INR values, a known risk factor for bleeding. 20 We also noted an increase in bleeding risk associated with the first episode of AF. Others have reported that the risk of bleeding is increased during the first few months after anticoagula- tion is begun then becomes constant at a lower rate in those who continue therapy. 20-22 It is likely that many patients in the AFFIRM trial with recurrent episodes of AF had already received several months of anticoagula- tion before enrollment, whereas those with a first episode of AF were still in the higher-risk, early treatment phase. It is not surprising that warfarin use was a risk factor for bleeding. As in other studies, 2,16 patients with major bleeding in the AFFIRM trial commonly had INR values above the desired range of 2 to 3. Even after the 3 outliers with unreliable INR values are excluded, the observation that the mean INR at the time of bleeding was substantially elevated emphasizes the fact that antico- agulation therapy must be carefully managed to minimize the potential for bleeding. The increased risk of bleeding associated with aspirin use is also clinically important. Aspirin has many uses and is commonly recommended in patients at risk for cardiovascular disease. However, use of aspirin is a known risk factor for bleeding, especially gastrointestinal bleeding, via aspirins effects on platelets and the gastrointestinal mucosa. Concomitant use of warfarin and aspirin has been recommended for patients with mechanical heart valves who are at very high risk for thromboembolism 23 because it may permit use of a lower-target INR without compromising protection against stroke. However, in studies on this issue, bleeding has been increased when aspirin is added to warfarin. 23 A potentiation of risk with combined aspirin and warfarin therapy has also been reported in patients with ischemic heart disease. 24 It is likely that aspirin was most commonly used in the AFFIRM trial as part of a treatment program for patients with coronary, cerebral, or peripheral vascular disease. In some patients, the risks associated with those disorders may be high enough to justify combined therapy. However, physicians will need to be particu- larly cautious in their selection of patients for such treatment. The potential value of adding aspirin must be clearly justified. Careful INR monitoring, use of the Table IV. Covariates significantly associated with major bleeding Covariate P Hazard ratio HR: 95% confidence limits Increase/decrease Lower Upper in risk Age at entry b.0001 1.05 1.04 1.07 +5%* Congestive heart failure .0108 1.43 1.09 1.89 +43% Diabetes .0152 1.44 1.07 1.93 +44% Hepatic or renal disease .0021 1.93 1.27 2.93 +93% First episode of AF .0476 1.30 1.00 1.68 +30% Warfarin use .0073 1.78 1.17 2.70 +78% Aspirin use b.0001 2.01 1.45 2.77 +101% *Per year of age. American Heart Journal April 2005 654 DiMarco et al minimally effective aspirin dosage, and limitations on therapy with other drugs that increase bleeding risk are required. Amiodarone potentiates the anticoagulant effects of a fixed dose of warfarin. 25-27 Amiodarone inhibits the metabolism of warfarin by CYP 2C9 and CYP 1A2, thereby increasing the INR within the first days to weeks of combined therapy. Since this interaction was reported, it has been recommended that warfarin-treated patients starting on amiodarone regimen should reduce their warfarin dose by 25% to 50%. Amiodarone was the initial drug chosen for 33% of the patients in the rhythm- control arm in the AFFIRM trial and was eventually used by 62% of those patients. 14 The potential for an interaction between warfarin and amiodarone was publicized among the investigators during the trial. With this precaution, we did not observe an increased risk for bleeding associated with the use of amiodarone, sug- gesting that the potential for interaction is manageable clinically. Propafenone, another antiarrhythmic drug used to treat patients with AF, may also interact with warfarin. 28 We did not look for an increased risk associated with propafenone in the AFFIRM trial because there were only 339 patients who used this drug even for a short time. Nevertheless, physicians should be aware of this potential interaction. Warfarin therapy is often difficult because its antico- agulant effect is influenced by dietary factors, drug interactions that may affect absorption and metabolism, and genetic factors. 29 The risk of bleeding while on warfarin regimen may be increased when other drugs that independently potentiate bleeding are also pre- scribed. We focused our analysis on aspirin because it was commonly used by patients in the AFFIRM trial, but other drugs commonly used by the elderly patients (eg, nonsteroidal anti-inflammatory agents, corticosteroids, other platelet inhibitors) must also be prescribed cautiously in patients receiving a warfarin regimen. The Stroke Prevention in Atrial Fibrillation II investigators 11 and Wehinger et al 30 have reported that use of more concomitant drugs was associated with bleeding but did not specifically identify individual agents. Limitations This study has several limitations. It has been shown that hemorrhagic risks of warfarin are lowered if patients are followed in specialized anticoagulation clinics. 20,21 It has also been shown that unstable INR values are potent predictors of hemorrhagic complications. 20 In the AFFIRM trial, anticoagulation was variably managed as directed by the patients primary physicians. Interna- tionalized normalized ratio values were only recorded by the study at the regular AFFIRM protocol visits and at the time of certain events. We therefore cannot comment on the details of anticoagulation management nor on the stability of INR values as a risk factor. However, we did observe that 62% of the INR values recorded at study visits were within the desired range, a proportion similar to that seen in previous anticoagulation trials. 4-10,13 In the time-dependent analysis presented in this paper, patients were considered to be taking the drug examined as a covariate if they had received the drug at any time during the interval since their prior study visit. Start and stop dates for use of the various drugs were not recorded. These limitations in data recording potentially introduced error because a patients drug use at the time of an event was 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, major bleeding), making it more difficult to detect significant associations rather than falsely strengthening the effect of one type of drug over another. Also, the analysis was not planned at the outset of the AFFIRM Study and is, therefore, both retrospective and non- randomized. Accordingly, it may suffer from hidden biases and other unidentified confounders. Because of these data collection limitations and potential biases, we concluded that analysis of interactions and sub- groups involving the time-dependent covariates could not be reliably performed. Conclusions In patients with nonvalvular AF and one or more risk factors for stroke who are eligible for warfarin therapy at baseline, major bleeding during anticoagulant therapy occurs at a rate of approximately 2% per year. Increasing age, congestive heart failure, diabetes, hepatic or renal disease, recent development of AF, the use of warfarin, and concurrent use of aspirin increase this risk. In this population, the risk of major bleeding during antico- agulation is lower than the expected rate of ischemic stroke or systemic thromboembolism. Knowledge of the risk factors associated with bleeding should help physicians decide for or against chronic anticoagulation in patients with AF at low or moderate embolic risk 31,32 and for or against addition of aspirin to patients receiving warfarin. References 1. Wolf PA, Abbott RD, Kannel WB. Atrial fibrillation as an inde- pendent risk factor for stroke: the Framingham study. Stroke 1991;22:983- 8. 2. Albers GW, Dalen JE, Laupacis A, et al. Antithrombotic therapy in atrial fibrillation. Chest 2001;119:194S- 206S. 3. Fuster V, Ryden LE, Asinger RW, et al. Management of patients with atrial fibrillation. A report of the American College of Cardiology/ American Heart Association Task Force on Practice Guidelines and the European Society of Cardiology Committee on Practice Guide- lines and Policy Conferences (Committee to develop guidelines for the management of atrial fibrillation). J Am Coll Cardiol 2001;104:2118- 50. American Heart Journal Volume 149, Number 4 DiMarco et al 655 4. Atrial Fibrillation Investigators. Risk factors for stroke and efficacy of anti-thrombotic therapy in atrial fibrillation: analysis of pooled data from five randomized controlled trials. Arch Intern Med 1994;154: 1449- 57. 5. Stroke Prevention in Atrial Fibrillation Investigators. Stroke pre- vention in atrial fibrillation study: final results. Circulation 1991;84:527- 39. 6. Connolly SJ, Laupacis A, Gent M, et al. Canadian Atrial Fibrillation Anticoagulation (CAFA) Study. J Am Coll Cardiol 1991;18:349- 55. 7. Stroke Prevention in Atrial Fibrillation Investigators. Warfarin versus aspirin for prevention of thromboembolism in atrial fibrillation: Stroke Prevention in Atrial Fibrillation II Study. Lancet 1994;343: 687- 91. 8. Petersen P, Boysen G, Godtfredsen J, et al. 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- 8. 9. EAFT (European Atrial Fibrillation Trial) Study Group. Secondary prevention in non-rheumatic atrial fibrillation after transient ischae- mic attack or minor stroke: EAFT (European Atrial Fibrillation Trial) Study Group. Lancet 1993;342:1255- 62. 10. Boston Area Anticoagulation Trial for Atrial Fibrillation Investiga- tors. The effect of low-dose warfarin on the risk of stroke in patients with nonrheumatic atrial fibrillation. N Engl J Med 1990;322: 1505- 11. 11. Stroke Prevention in Atrial Fibrillation Investigators. Bleeding during antithrombotic therapy in patients with atrial fibrillation. Arch Intern Med 1996;156:409- 16. 12. Levine MN, Raskob G, Landefeld S, et al. Hemorrhagic complica- tions of anticoagulant treatment. Chest 2001;119:108S- 21S. 13. Executive Steering Committee on behalf of the Sportif Investigators. Stroke prevention with the oral direct thrombin inhibitor ximelaga- tran compared with warfarin in patients with non-valvular atrial fibrillation (SPORTIF III): randomised controlled trial. Lancet 2003;362:1691- 8. 14. The Atrial Fibrillation Follow-up Investigation of Rhythm Manage- ment (AFFIRM) Investigators. A comparison of rate control and rhythm control in patients with atrial fibrillation. N Engl J Med 2002;347:1825- 33. 15. The AFFIRM Investigators. Baseline characteristics of patients with atrial fibrillation: the AFFIRM study. Am Heart J 2002;143: 991- 1001. 16. Albers GW. Atrial fibrillation and stroke. Three new studies, three remaining questions. Arch Intern Med 1994;154: 1443- 8. 17. Hylek EM, Singer DE. Risk factors for intracranial hemorrhage in outpatients taking warfarin. Ann Intern Med 1994;120: 897- 902. 18. Go AS, Hylek EM, Phillips KA, et al. Implications of stroke risk criteria on the anticoagulation decision in nonvalvular atrial fibrillation: the Anticoagulation and Risk Factors in Atrial Fibrillation (ATRIA) study. Circulation 2000;102:11- 3. 19. Go AS, Hylek EM, Borowsky LH. Warfarin use among ambulatory patients with nonvalvular atrial fibrillation: the Anticoagulation and Risk Factors in Atrial Fibrillation (ATRIA) study. Ann Intern Med 1999;131:927- 34. 20. Fihn SD, Callahan CM, Martin DC, et al. The risk and severity of bleeding complications in elderly patients treated with warfarin. Ann Intern Med 1996;124:970- 9. 21. McMahan DA, Smith DM, Carey MA, et al. Risk of major hemorrhage for outpatients treated with warfarin. J Gen Intern Med 1998;13:311- 6. 22. Copland M, Walker I, Tait RC. Oral anticoagulation and hemor- rhagic complications in an elderly population with atrial fibrillation. Arch Intern Med 2001;161:2125- 8. 23. Stein PD, Bussey HI, Dalen JE, et al. Antithrombotic therapy in patients with mechanical and prosthetic heart valve. Chest 2001;119:220S- 7S. 24. Fiore LD, Ezekowitz MD, Brophy MT, et al. Department of Veteran Affairs cooperative studies program clinical trial comparing com- bined warfarin and aspirin with aspirin alone in survivors of acute myocardial infarction. Primary results of the CHAMP study. Circulation 2002;105:557- 63. 25. Heimark LD, Wienkers L, Kunze K, et al. The mechanism of the interaction between amiodarone and warfarin in humans. Clin Pharmacol Ther 1992;51:398- 407. 26. Kerin NZ, Blevins RD, Goldman L, et al. The incidence, magnitude, and time course of the amiodarone-warfarin interaction. Arch Intern Med 1988;148:1779- 81. 27. Sanoski CA, Bauman JL. Clinical observations with the warfarin/ amiodarone interaction: dosing relationships with long-term therapy. Chest 2002;121:19- 23. 28. Kates RE, Yee YG, Kirsten EB. Interaction between warfarin and propafenone in healthy volunteer subjects. Clin Pharmacol Ther 1987;42:305- 11. 29. Hirsh J, Fuster V, Ansell J, et al. American Heart Association/ American College of Cardiology Foundation guide to warfarin therapy. J Am Coll Cardiol 2003;41:1633- 55. 30. Wehinger C, Stolberger C, Langer T, et al. Evaluation of risk factors for stroke/embolism and of complications due to anticoagulant therapy in atrial fibrillation. Stroke 2001;32:2246- 52. 31. van Walraven C, Hart RG, Wells GA, et al. A clinical prediction rule to identify patients with atrial fibrillation and a low risk for stroke while taking aspirin. Arch Intern Med 2003;163:936- 43. 32. van Walraven C, Hart RG, Singer DE, et al. Oral anticoagulants vs. aspirin in nonvalvular atrial fibrillation. An individual patient meta- analysis. JAMA 2002;288:2441- 8. American Heart Journal April 2005 656 DiMarco et al