Published online before print August 2, 2007, doi: 10.1161/STROKEAHA.106.477133
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
(Stroke. 2007;38:2459.)
© 2007 American Heart Association, Inc.
Original Contributions |
Can Patients at Elevated Risk of Stroke Treated With Anticoagulants Be Further Risk Stratified?
From the Department of Medicine (L.B.), Bronx Veterans Affairs Medical Center, Bronx, and the Mt. Sinai School of Medicine, New York, NY; the Department of Medicine (B.F.G.), Washington University, St. Louis, Mo; AstraZeneca LP (J.H.), Wilmington, Del; Department of Cardiology (S.J.-M.), University Hospital Malmö, University of Lund, Malmö, Sweden; Department of Medicine (A.L.), St. Louis University, St. Louis, Mo; AstraZeneca R&D Mölndal (M.P.), Mölndal, Sweden; and the University of Texas Health Science Center (M.Z.), San Antonio, Tex.
Correspondence to Lawrence Baruch, MD, Bronx VA Hospital, 130 West Kingsbridge Rd, Bronx, NY 10468. E-mail baruchlarry{at}att.net
 |
Abstract |
|---|
 Top Abstract IntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Background and Purpose— Patients with atrial fibrillation have a varied risk of stroke, depending on age and comorbid conditions. The objective of this study was to assess the predictive value of stroke risk classification schemes and to identify patients with atrial fibrillation who are at substantial risk of stroke despite optimal anticoagulant therapy.
Methods— Seven recognized classification schemes—the American College of Chest Physicians 2001, American College of Chest Physicians 2004, Stroke Prevention in Atrial Fibrillation (SPAF), Atrial Fibrillation Investigators, Framingham, van Walraven, and CHADS2—were compared for their ability to predict ischemic stroke in patients receiving anticoagulant therapy. Data came from the Stroke Prevention using an ORal Thrombin Inhibitor in atrial Fibrillation III and V trials, which compared the efficacy of adjusted-dose warfarin and the direct thrombin inhibitor ximelagatran (36 mg twice daily) in preventing thromboembolic events in 7329 patients with chronic or paroxysmal nonvalvular atrial fibrillation who were at moderate or high risk of ischemic stroke. The main outcome measure was ischemic stroke, as determined by a central event adjudication committee.
Results— During 11 245 patient-years of follow-up, 159 patients had an ischemic stroke (1.4%/year). As indicated by c statistics and hazard ratios, 3 of the classification schemes predicted stroke significantly better than chance: Framingham (c=0.64), CHADS2 (c=0.65), and SPAF (c=0.61).
Conclusions— In a large cohort of atrial fibrillation patients at moderate or high risk of ischemic stroke treated with warfarin or ximelagatran, the CHADS2, SPAF, and Framingham schemes had greater predictive accuracy than chance. This predictive ability may allow clinicians to target high-risk patients for more aggressive intervention.
Key Words: anticoagulation • atrial fibrillation • direct thrombin inhibitors • risk prediction • stroke
|
Introduction |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Atrial fibrillation (AF), the most common cardiac arrhythmia, affects >2 million individuals in the United States.1 Much of AF-related morbidity resides in the 5- to 6-fold increased risk of ischemic stroke.2 Although anticoagulant and antiplatelet therapies reduce the incidence of stroke in AF patients,3 the risks of both stroke and bleeding vary.4 Accordingly, several schemes exist to risk-stratify the nonvalvular AF population to help select appropriate candidates for anticoagulant therapy.5–11 Existing risk stratification schemes use age and comorbid conditions, such as prior ischemic stroke, diabetes, heart failure, and hypertension, to classify patients as being at low, moderate, or high thromboembolic risk. Risk classification guides therapy with aspirin or warfarin.1,6,9
The American College of Chest Physicians (ACCP) at its Consensus Conference on Antithrombotic Therapy has promulgated the most recognized schemes.6,9 Other accepted schemes include the AF Investigators (AFI),5 Stroke Prevention in Atrial Fibrillation (SPAF),8 Framingham,11 CHADS2,7,12 and van Walraven.10 The stratification schemes vary in estimated stroke risk assigned to individual patients and perhaps in predictive ability.
The risk-stratification schemes arose from studies of patients not receiving anticoagulant therapy. Their predictive abilities in patients receiving anticoagulant therapy are unknown. In the Stroke Prevention using an ORal Thrombin Inhibitor in atrial Fibrillation (SPORTIF) III and SPORTIF V trials, all patients received anticoagulation. The trials compared the efficacy of warfarin with that of the oral direct thrombin inhibitor ximelagatran in preventing thromboembolic events in patients with nonvalvular AF at moderate or high risk of stroke.13–15 This substudy of the SPORTIF trials has compared the abilities of 7 recognized classification schemes—ACCP,6,9 AFI,5 SPAF,8 Framingham,11 CHADS2,7,12and van Walraven10—to risk-stratify patients and predict ischemic stroke in the anticoagulated SPORTIF participants. We hypothesized that the different classification schemes would vary in their predictive ability and their stratification of stroke risk.
|
Subjects and Methods |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Study Population
The rationale, design, and results of SPORTIF III and SPORTIF V have already been published.13–15 These randomized, multicenter, parallel-group trials compared fixed-dose oral ximelagatran with adjusted-dose warfarin for prevention of stroke and systemic embolism in nonvalvular chronic or paroxysmal AF patients at high risk of stroke based on the ACCP 2001 AF guideline recommendations.6
Every participant provided written, informed consent according to a protocol approved by local ethics committees and in accordance with the Declaration of Helsinki. Participants randomly received either fixed-dose ximelagatran, 36 mg twice daily, or dose-adjusted warfarin to maintain the international normalized ratio between 2.0 and 3.0. A masked, interactive, voice-response system allocated treatment according to an adaptive algorithm balanced by country, concomitant aspirin treatment at entry, and history of stroke or transient ischemic attack. Anticoagulants were administered open-label in SPORTIF III15 and double-blinded in SPORTIF V.13
Ascertainment of Outcomes
Periodic administration of a standard stroke-symptom questionnaire enhanced event detection; positive responses prompted additional evaluation. Local study-affiliated neurologists or stroke specialists, masked to treatment, assessed all possible primary events based on clinical findings and results of computed tomography or magnetic resonance imaging of the brain. A single, independent, masked central event adjudication committee reviewed the reports.
Classification Schemes
The SPAF,8 ACCP (2001, 2004),6,9 van Walraven,10 and AFI5 schemes estimate risk based on the presence of the following various factors, alone or in combination: age, female sex, diabetes, previous stroke or transient ischemic attack, hypertension or elevated systolic blood pressure, coronary artery disease, and left ventricular dysfunction. The Framingham scheme assigns point values to each of the following risk factors: age, gender, systolic blood pressure, diabetes, and prior stroke or transient ischemic attack.11 CHADS2 assigns 1 point each for congestive heart failure, hypertension, age 
75 years, and diabetes mellitus, with 2 points for a history of stroke or transient ischemic attack.7 The greater the number of CHADS2 or Framingham points, the greater the stroke risk.
Statistical Analyses
With the combined SPORTIF III and SPORTIF V data sets, the primary analysis for this substudy compared the predictive accuracy of the 7 classification schemes for the first occurrence of ischemic stroke. Secondary analyses assessed the predictive accuracy of the schemes for all strokes (ischemic and hemorrhagic) and separately for the composite of ischemic stroke and systemic embolic events. The intention-to-treat analyses included all randomized participants, even with incomplete follow-up and exposure truncated at the time of study withdrawal. For CHADS2 and Framingham schemes, a Cox proportional-hazards model (SAS version 8.2; SAS Institute Inc, Cary, NC) calculated stroke rates and quantified the hazard rate for stroke for each 1-point increase in score.
Time-to-event analyses determined the predictive validity of each of the classification schemes. The c statistic, a measure of the area under the receiver operating characteristic curve, quantified the predictive validity of the classification schemes and tested the hypothesis that these classification schemes performed significantly better than chance (indicated by a c statistic of 0.5).16,17 The c statistic quantifies discriminant ability, whereas the hazard ratio quantifies the increased relative risk of stroke across risk strata. The 95% CIs were calculated according to the Poisson approximation.18
To compare the CHADS2 classification scheme, which is based on point scores (0 to 6), with the schemes that categorize patients as being at low, medium or high risk, we also analyzed the predictive value of the CHADS2 scoring system after collapsing CHADS2 into 3 strata: low risk (CHADS2 0), moderate risk (CHADS2 1 to 2), and high risk (CHADS2 3 to 6). Likewise, collapsing Framingham scores of 0 to 7 as low-risk, 8 to 13 as moderate-risk, and 14 to 31 as high-risk strata permitted comparison of Framingham scores to other scoring systems.
|
Results |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
The baseline characteristics of the 7329 participants have been previously reported.14 The most common risk factors were hypertension, age, coronary artery disease, and left ventricular dysfunction (Table 1). The percentage of participants in the low-, moderate-, and high-risk cohorts varied, based on the specific risk stratification scheme used (the Figure). Because SPORTIF study-inclusion criteria required high-risk patients by ACCP 2001 criteria,6 few SPORTIF participants were classified as low risk according to the ACCP (2001, 2004), AFI, or CHADS2 classification schemes. Most participants were at high risk by ACCP 2001 (96%), ACCP 2004 (97.5%), van Walraven (99.2%), and AFI (85.1%) schemes.
|
|
In contrast, the Framingham scheme, with the ad hoc threshold of 14 points, characterized the fewest participants (21.2%) as being at high risk (Table 2). The Framingham and SPAF schemes classified a number of SPORTIF participants as being at low risk. CHADS2 characterized the largest cohort of SPORTIF participants as being at intermediate risk, whereas SPAF classified almost equal numbers of participants as being at moderate and high risk.
|
During the 11 245 patient-years of follow-up (mean, 1.5 years/patient), 159 participants had an ischemic stroke (1.4 per 100 patient-years; Table 2). The highest stroke rate occurred in patients identified as being at high risk by the Framingham scheme (Tables 2 and 3
). All schemes, other than van Walraven’s, which explicitly combines moderate- and high-risk patients, distinguished between moderate- and high-risk subjects. Only the Framingham and CHADS2 schemes distinguished low- from moderate-risk SPORTIF participants (Tables 2 and 3). No strokes occurred in the 238 patient-years categorized as low risk in CHADS2 and the smaller low-risk cohorts defined by the van Walraven (85 patient-years) and ACCP (29 patient-years) schemes.
|
The SPAF, Framingham, and CHADS2 c statistics had greater predictive accuracy for ischemic stroke than chance (Table 4). When Framingham and CHADS2 classification schemes were collapsed into 3 strata—low, moderate, and high—they maintained their predictive values, with c statistics of 0.61 and 0.64, respectively. CHADS2 had numerically the highest hazard ratio per increase in risk (Table 4).
|
Before schema collapse, the hazard ratio was 1.48 (1.31 to 1.66) for CHADS2, indicating a 48% increase in ischemic stroke rate per CHADS2 point (P<0.0001), and 1.10 (1.07 to 1.13) for Framingham, indicating a 10% increase per Framingham point (P<0.0001). This 5-fold effect ratio reflects the 5-fold greater scale for the Framingham (31-point system) compared with the CHADS2 (6-point system; Table 3) scheme.
The secondary analyses for the outcome of all strokes (ischemic and hemorrhagic; n=74 patients) and for the outcome combining ischemic stroke and systemic embolic events (n=169 patients) yielded results similar to those for ischemic stroke alone. Likewise, treatment with either warfarin or ximelagatran did not affect hazard ratios or c statistics for the various classification schemes. C statistics for ACCP 2001 (0.51), ACCP 2004 (0.51), and van Walraven (0.50) schemes reflected the lack of discriminant ability of these schemes in SPORTIF participants, arising from minimal exposure for low- and moderate-risk groups, based on the SPORTIF entry criteria.
|
Discussion |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
All 7 AF stroke prediction schemes were developed and evaluated in patients not receiving anticoagulant therapy.5–11 Clinical practice guidelines use the various risk-prediction schemes to stratify patients as being at low, moderate, or high risk; this leads to recommendations for antiplatelet or antithrombotic therapy. Although these stratification schemes help identify which patients are above a risk threshold that would lead to a benefit from anticoagulant therapy, they do not identify who remains at a higher level of risk despite the use of anticoagulant therapy. Identification of such patients might influence both clinical practice and the design of future clinical trials.
In this large cohort of AF patients at increased risk for stroke treated with warfarin or ximelagatran, the CHADS2, SPAF, and Framingham schemes had better accuracy and predictive value, as determined by their higher c statistics, than the AFI scheme in predicting ischemic stroke. CHADS2 and Framingham schemes more precisely classified the level of stroke risk. SPORTIF entry criteria created small samples of low- and moderate-risk patients for ACCP 2001, ACCP 2004, and van Walraven schemes, thus preventing definitive assessment of their abilities to predict treatment failure, as reflected in their c statistics. Of note, in the cohorts categorized as being at high risk by ACCP 2001 (1.4%), ACCP 2004 (1.5%), AFI (1.5%), and van Walraven (1.4%) schemes, stroke rates were similar and less than in high-risk cohorts identified by the CHADS2 (2.3%), SPAF (2.0%), and Framingham (2.7%) schemes.
The schemes differed significantly in classifying individual participant’s stroke risks. Many subjects classified as being at moderate or high risk by ACCP 2001, ACCP 2004, van Walraven, or AFI schemes were classified as being at low risk by Framingham and SPAF schemes. A number of patients classified as being at moderate risk according to the CHADS2 scheme were classified as being at high risk according to ACCP 2001, ACCP 2004, van Walraven, and AFI schemes and at low risk according to Framingham and SPAF schemes. The lowest stroke rates occurred in low-risk patients identified by the CHADS2, van Walraven, and ACCP 2004 schemes. However, for the van Walraven and ACCP 2004 schemes, the 95% CIs were wide.
Because many clinicians will consider newer strategies in "high-risk" patients, differences in stroke prediction may affect therapy. For example, dual antithrombotic therapy with warfarin and aspirin is recommended for patients with mechanical heart valves19 and may prevent ischemic events in selected high-risk patients with AF.20 Likewise, high-risk patients with AF may benefit from reducing, rather than holding, warfarin therapy before elective procedures.21 Likewise, they may more vigorously avoid subtherapeutic international normalized ratio values by using patient self-management, more frequent international normalized ratio monitoring, pharmacogenetic dosing,22 or anticoagulation clinics. Risk-prediction schemes also allow for more efficient clinical trials. Using the SPAF, Framingham, or CHADS2 scheme, investigators could selectively recruit high-risk patients, resulting in a clinical trial with smaller sample size, greater power, or shorter duration of follow-up.
Limitations
Individuals who enroll in clinical trials may not represent the general population. In particular, SPORTIF did not have an adequate sample size for low-risk cohorts as identified by several schemes, thereby impairing the performance of all schemes to discriminate risk. Evaluating all of the prediction schemes in a population with a broader spectrum of risk could validate these findings.
Conclusions
In this large, prospective cohort of AF patients with risk factors for stroke who were randomized to warfarin or ximelagatran therapy, CHADS2, SPAF, and Framingham schemes had a predictive accuracy significantly greater than expected by chance. Future trials should quantify the benefit of newer strategies or novel anticoagulants in high-risk patients identified by these schemes.
|
Acknowledgments |
|---|
Source of Funding
This study was sponsored by AstraZeneca, Mölndal, Sweden.
Disclosures
Other than serving as study investigators in the SPORTIF trials, Drs Baruch, Labovitz, and Juul-Möller have no funding or support to disclose relevant to this publication. Dr Gage has served as a consultant for Bristol-Myers Squibb. Dr Zabalgoitia has served as a consultant for AstraZeneca and on speakers’ bureaus for Merck and Pfizer. Ms Persson is an employee of AstraZeneca, the sponsor of the SPORTIF trials. Dr Horrow was an employee of AstraZeneca at the time this work was conducted.
Received November 2, 2006; revision received February 26, 2007; accepted February 28, 2007.
|
References |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
1. Fuster V, Ryden LE, Asinger RW, Cannom DS, Crijns HJ, Frye RL, Halperin JL, Kay GN, Klein WW, Levy S, McNamara RL, Prystowsky EN, Wann LS, Wyse DG, Gibbons RJ, Antman EM, Alpert JS, Faxon DP, Fuster V, Gregoratos G, Hiratzka LF, Jacobs AK, Russell RO, Smith SC, Klein WW, Alonso-Garcia A, Blomstrom-Lundqvist C, De Backer G, Flather M, Hradec J, Oto A, Parkhomenko A, Silber S, Torbicki A, American College of Cardiology/American Heart Association/European Society of Cardiology Board. ACC/AHA/ESC guidelines for the management of patients with atrial fibrillation: executive summary. J Am Coll Cardiol. 2001; 38: 1231–1266.
2. Wolf PA, Abbott RD, Kannel WB. Atrial fibrillation as an independent risk factor for stroke: the Framingham Study. Stroke. 1991; 22: 983–988.
3. Hart RG, Benavente O, McBride R, Pearce LA. Antithrombotic therapy to prevent stroke in patients with atrial fibrillation: a meta-analysis. Ann Intern Med. 1999; 131: 492–501.
4. Gage BF, Yan Y, Milligan PE, Waterman AD, Culverhouse R, Rich MW, Radford MJ. Clinical classification schemes for predicting hemorrhage: results from the National Registry of Atrial Fibrillation (NRAF). Am Heart J. 2006; 151: 713–719.[CrossRef][Medline] [Order article via Infotrieve]
5. Risk factors for stroke and efficacy of antithrombotic therapy in atrial fibrillation: analysis of pooled data from five randomized controlled trials. Arch Intern Med. 1994; 154: 1449–1457.
6. Albers GW, Dalen JE, Laupacis A, Manning WJ, Petersen P, Singer DE. Antithrombotic therapy in atrial fibrillation. Chest. 2001; 119: 194S–206S.
7. Gage BF, Waterman AD, Shannon W, Boechler M, Rich MW, Radford MJ. Validation of clinical classification schemes for predicting stroke: results from the National Registry of Atrial Fibrillation. JAMA. 2001; 285: 2864–2870.
8. Hart RG, Pearce LA, McBride R, Rothbart RM, Asinger RW. Factors associated with ischemic stroke during aspirin therapy in atrial fibrillation: analysis of 2012 participants in the SPAF I-III clinical trials: the Stroke Prevention in Atrial Fibrillation (SPAF) Investigators. Stroke. 1999; 30: 1223–1229.
9. Singer DE, Albers GW, Dalen JE, Go AS, Halperin JL, Manning WJ. Antithrombotic therapy in atrial fibrillation: the Seventh ACCP Conference on Antithrombotic and Thrombolytic Therapy. Chest. 2004; 126: 429S–456S.
10. van Walraven C, Hart RG, Wells GA, Petersen P, Koudstaal PJ, Gullov AL, Hellemons BS, Koefed BG, Laupacis A. A clinical prediction rule to identify patients with atrial fibrillation and a low risk for stroke while taking aspirin. Arch Intern Med. 2003; 163: 936–943.
11. Wang TJ, Massaro JM, Levy D, Vasan RS, Wolf PA, D’Agostino RB, Larson MG, Kannel WB, Benjamin EJ. A risk score for predicting stroke or death in individuals with new-onset atrial fibrillation in the community: the Framingham Heart Study. JAMA. 2003; 290: 1049–1056.
12. Gage BF, van Walraven C, Pearce L, Hart RG, Koudstaal PJ, Boode BSP, Petersen P. Selecting patients with atrial fibrillation for anticoagulation: stroke risk stratification in patients taking aspirin. Circulation. 2004; 110: 2287–2292.
13. SPORTIF Executive Steering Committee for the SPORTIF V Investigators. Ximelagatran vs warfarin for stroke prevention in patients with nonvalvular atrial fibrillation: a randomized trial. JAMA. 2005; 293: 690–698.
14. Halperin JL and the Executive Steering Committee, on behalf of the SPORTIF III, and V Study Investigators. Ximelagatran compared with warfarin for prevention of thromboembolism in patients with nonvalvular atrial fibrillation: rationale, objectives, and design of a pair of clinical studies and baseline patient characteristics (SPORTIF III and V). Am Heart J. 2003; 146: 431–438.[CrossRef][Medline] [Order article via Infotrieve]
15. Executive Steering Committee: Albers GW, Diener HC, Grind M, Halperin JL, Horrow J, Olsson SB, Petersen P, Vahanian A, on behalf of the SPORTIF III Investigators. Stroke prevention with the oral direct thrombin inhibitor ximelagatran compared with warfarin in patients with non-valvular atrial fibrillation (SPORTIF III): randomised controlled trial. Lancet. 2003; 362: 1691–1698.[CrossRef][Medline] [Order article via Infotrieve]
16. Harrell FE Jr, Lee KL, Mark DB. Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Stat Med. 1996; 15: 361–387.[CrossRef][Medline] [Order article via Infotrieve]
17. Harrell FE Jr, Califf RM, Pryor DB, Lee KL, Rosati RA. Evaluating the yield of medical tests. JAMA. 1982; 247: 2543–2546.
18. Cox D. Some simple approximate tests for Poisson variates. Biometrika. 1953; 40: 354–360.
19. Turpie AG, Gent M, Laupacis A, Latour Y, Gunstensen J, Basile F, Klimek M, Hirsh J. A comparison of aspirin with placebo in patients treated with warfarin after heart-valve replacement. N Engl J Med. 1993; 329: 524–529.
20. Stenestrand U, Lindback J, Wallentin L. Anticoagulation therapy in atrial fibrillation in combination with acute myocardial infarction influences long-term outcome: a prospective cohort study from the register of information and knowledge about Swedish heart intensive care admissions (RIKS-HIA). Circulation. 2005; 112: 3225–3231.
21. Gerson LB, Triadafilopoulos G, Gage BF. The management of anticoagulants in the periendoscopic period for patients with atrial fibrillation: a decision analysis. Am J Med. 2004; 116: 451–459.[CrossRef][Medline] [Order article via Infotrieve]
22. Voora D, Eby C, Linder MW, Milligan PE, Bukaveckas BL, McLeod HL, Maloney W, Clohisy J, Burnett RS, Grosso L, Gatchel SK, Gage BF. Prospective dosing of warfarin based on cytochrome P-450 2c9 genotype. Thromb Haemost. 2005; 93: 700–705.[Medline] [Order article via Infotrieve]
Related Article:
Stroke 2007 38: 2409.
This article has been cited by other articles:
 |
|
 |
|
  G. Y.H. Lip The Balance Between Stroke Prevention and Bleeding Risk in Atrial Fibrillation: A Delicate Balance Revisited Stroke, May 1, 2008; 39(5): 1406 - 1408. [Full Text] [PDF]
|
|
|
|
|
|
J. S. Taggar, T. Watson, and G. Y.H. Lip Can Patients With Atrial Fibrillation Be Optimally Risk Stratified for Stroke and Thromboembolism? Stroke, February 1, 2008; 39(2): e24 - e25. [Full Text] [PDF]
|
|
|
|
|
|
L. Baruch, B. Gage, and J. Horrow Response to Letter by Taggar et al Stroke, February 1, 2008; 39(2): e26 - e26. [Full Text] [PDF]
|
|
|
|
|
|
D. G. Hackam Predicting Stroke Risk in Patients With Atrial Fibrillation Stroke, September 1, 2007; 38(9): 2409 - 2409. [Full Text] [PDF]
|
|
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||