American Heart Journal
Volume 156, Issue 1 , Pages 1-3, July 2008

Risk factors and stroke risk stratification for atrial fibrillation: Limitations and new possibilities

  • Gregory Y.H. Lip, MD, FRCP

      Affiliations

    • Corresponding Author InformationReprint requests: Gregory Y.H. Lip, MD, FRCP, University Department of Medicine, City Hospital, Birmingham, England.

University Department of Medicine, City Hospital, Birmingham, England

Received 18 February 2008; accepted 29 February 2008. published online 06 May 2008.

Article Outline

 

Given the high risk of stroke and thromboembolism associated with atrial fibrillation (AF), great efforts have been directed toward risk stratification to identify those at highest risk, who would benefit most from anticoagulation therapy as thromboprophylaxis.1 Cohort data as well as nonwarfarin arms of clinical trials have identified clinical and echocardiographic risk factors that confer a significant risk of stroke.2

A wide range of risk factors have been proposed. A recent systematic review, conducted as part of the United Kingdom National Institute for Health and Clinical Excellence (NICE) national clinical guidelines for AF management, identified history of stroke or TIA, increasing age, hypertension, and structural heart disease (left ventricular dysfunction or hypertrophy) to be good predictors of stroke risk in AF patients.2 In this review, the evidence for stroke risk regarding diabetes mellitus, sex, and other characteristics were found to be less consistent in the AF population per se, although it was accepted that diabetes was an important risk for stroke generally. In another systematic review of stroke risk factors in AF by the Stroke Risk in Atrial Fibrillation Working Group,3 prior stroke/TIA (relative risk [RR] 2.5), increasing age (RR 1.5 per decade), a history of hypertension (RR 2.0), and diabetes mellitus (RR 1.7) were the most consistent independent risk factors. Prior stroke/TIA was the most powerful stroke risk factor and conferred a high stroke risk (>5% per year, averaging 10% per year). Of note, female sex was found to be inconsistently associated with stroke risk, whereas the evidence was considered “inconclusive” that either heart failure or coronary artery disease is independently predictive of stroke.3

Appreciation of stroke risk factors has informed the development of stroke risk stratification schema, but despite these efforts, cardiologists are generally not good at stroke risk assessments and applying treatment guidelines. In the EuroHeart survey of AF,4 for example, many low risk subjects were prescribed warfarin, whereas many high-risk subjects were not given anticoagulation.

Equally, all the published stroke risk stratification schema have many limitations, being predominantly derived and validated in nonwarfarin arms of clinical trial cohorts, although the Framingham stroke risk score is perhaps an exception.2 In the NICE systematic review, only 3 stroke risk stratification models were identified that were able to discriminate between different categories of stroke risk to at least 95% accuracy.2 Also, the identified stroke risk factors represent those that have been looked for, and (say) atherosclerotic vascular disease—such as coronary or peripheral artery disease—does not feature in most published stroke risk stratification schema as trial datasets did not systematically assess, record, or validate this risk factor.5 This is despite the recognition that many stroke patients with AF have significant carotid disease, and the presence of peripheral artery disease confers a high mortality in AF patients.6 Also, complex aortic plaque on the descending aorta is an independent risk factor for ischemic stroke in AF, reemphasizing the importance of atherosclerotic vascular disease.5

The CHADS2 schema is a popular and well-validated stroke risk stratification schema and is an acronym derived from the individual stroke risk factors as follows: Congestive heart failure, Hypertension, Age ≥75 years, Diabetes mellitus, and prior Stroke/TIA (1 point is assigned for each of the factors, apart from prior stroke, or TIA that is given 2 points). In a recent validation exercise comparing CHADS2 with other schema, a score of 0 was “low risk,” 1 to 2 was “moderate risk,” and 3 to 6 was “high risk.”7 This immediately causes a problem if a subject with AF has stroke/TIA as the only risk factor, as he/she is given a CHADS2 score of 2, which is moderate risk despite such patients being at very high risk of recurrent stroke.2, 8 Also, the risk factor of “age >75” does not confer a “single” uniform stroke risk, as illustrated by the Atrial Fibrillation Working Group,3 where increasing age increases stroke rate with an RR of 1.5 per decade. Well-controlled blood pressure (or a history of hypertension, now well controlled) may also be less of a major risk factor, given that stroke and systemic embolic events only rises markedly at mean systolic blood pressure of >140 mm Hg, at least in anticoagulated patients.9 Nonetheless, few published risk models have addressed the cumulative nature of risk factors where a combination of risk factors would confer a greater risk than individual risk factors alone, but the CHADS2 schema at least allows for this.2, 7

Stroke risk stratification schema that results in classification of a large proportion of AF subjects into the moderate risk category could potentially be less useful in everyday clinical practice because current treatment guidelines recommend the use of either warfarin or aspirin in such patients, which can sometimes cause confusion over which therapy should really be prescribed. Indeed, many guidelines imply that aspirin is an alternative to warfarin in “moderate risk” subjects. In a recent analysis, Baruch et al10 reported that the CHADS2 schema classified the largest cohort of participants as moderate risk (approximately 60% vs 35% who were classed as “high risk”), compared to most participants being classed at high risk by the American College of Chest Physicians (96%-97%), the van Walraven (99.2%), and AF Investigators (85.1%) schemes; the C-statistics for the various schema in these subjects ranged from to 0.53 to 0.65. Similarly, an analysis from the large ATRIA cohort study, found that current risk schemes had comparable, but only limited, overall ability to predict thromboembolism in AF, and of the schemes tested, the CHADS2 schema classified 61.2% of the study cohort as moderate risk, compared to other schemae, such as the AF Investigators (24.7%) and the American College of Chest Physicians schema (7.9%).11 Clearly, recommendations for antithrombotic therapy would vary widely depending on which scheme is applied for individual patients.

Can things improve? In the current issue of the Journal, Reitbrock et al12 provide an interesting report from the UK General Practice Research Database on 51 807 AF patients. Of this large cohort, 13% were low risk (CHADS2 score = 0) and, as with the recent article by Baruch et al10 and Fang et al,11 approximately 60% were classed as moderate risk (CHADS2 1-2). Although the excess 5-year risk for stroke in AF patients correlated well with CHADS2, adding sex, the extension of age categories, and reweighting of established risk factors did slightly improve the CHADS2 accuracy for stroke prediction (C-statistic 0.68-0.72). Of note, applying this reclassification resulted in a substantial number of patients actually changing their stroke risk category. The authors conclude that the CHADS2 score is a good predictor of the stroke risk, but this “could be improved.”

The downside here is that the reclassified CHADS2 score loses its well-liked simplicity and the revised score ranges from 0 to 14. This has some parallels with the stroke risk classification scheme based on 868 Framingham study subjects, using a complicated point system based on age (0-10 points), sex (6 points for female, 0 for male), blood pressure (0-4 points), diabetes mellitus (4 points), and prior stroke or TIA (6 points) to predict the combination of ischemic plus hemorrhagic stroke.13 Nonetheless, with portable handhelds and smartphones that can easily run calculation programs, this may be less of a problem. Another argument is that the improvement in C-statistic is modest and is quite a way off from being >0.90. For example, the analysis from the ATRIA cohort reported C-statistic values that ranged from 0.56 to 0.62, for the different schema.11

In a recent comprehensive analysis, the Stroke Risk in Atrial Fibrillation Working Group published a comparison of 12-risk stratification schemes to predict stroke in patients with nonvalvular AF.14 The most frequently included risk factors in various schema were previous stroke/TIA (in all schemes), age (83%), hypertension (83%), and diabetes (83%). Based on published test cohorts, all tested schemes stratified stroke risk, but (as expected) the absolute stroke rates varied widely. Also, the fractions of patients categorized by the different risk stratification schemes as low-risk varied from 9% to 49% and as high-risk from 11% to 77%.

Clearly, the article by Reitbrock et al12 illustrates that the accuracy for predicting stroke in the CHADS2 schema, such as many other published schema, can be improved upon. None of the validation studies have achieved a C-statistic of >0.75, and the sad fact is that we are still a long way from having the ideal stroke risk stratification schema. Other novel possibilities should perhaps be considered. For example, the addition of categorized plasma vWf levels (an index of endothelial damage/dysfunction) to both the NICE/Birmingham and CHADS2 clinical risk stratification schemes further refined risk stratification for stroke and vascular events.15 Thus, biomarkers such as vWf levels may eventually aid decisions about thromboprophylaxis, particularly among AF patients at moderate risk where uncertainty often arises. Other potential biomarkers include fibrin d-dimer, a fibrin degradation product that is an index of thrombogenesis, given that high d-dimer levels (irrespective of anticoagulation) are predictive of adverse events in AF patients.16 Wider application of echocardiography (both transthoracic and where necessary, transoesophageal) may help our efforts to improve stroke risk stratification.

Another improvement would be to abandon the artificial categorization of low risk, moderate risk, and high risk strata used in exiting schema, and describe risk in terms of the annual rate (without antithrombotic therapy) of stroke and thromboembolism. However, this approach would quote event rates largely dependent upon historical data, given that most AF cohorts are treated with some form of antithrombotic agent.

New anticoagulant drugs that avoid the inconvenience of warfarin may circumvent this discussion over risk stratification anyway, and because the role of aspirin for stroke prevention has been debated, perhaps the time may come that all AF patients be simply treated with a safe, convenient oral anticoagulant that does not need monitoring, has low bleeding risk, and has no drug, food, or alcohol interactions rather than warfarin and aspirin, as currently happens. Further refinement of stroke risk stratification and application of thromboprophylaxis in AF may yet be possible.

Back to Article Outline

References 

  1. Lip GY, Lim HS. Atrial fibrillation and stroke prevention. Lancet Neurol. 2007;6:981–993
  2. Hughes M, Lip GYH. Stroke and thromboembolism in atrial fibrillation: a systematic review of stroke risk factors, risk stratification schema and cost effectiveness data. Thromb Haemostat. 2008;
  3. Stroke Risk in Atrial Fibrillation Working Group. Independent predictors of stroke in patients with atrial fibrillation: a systematic review. Neurology. 2007;69:546–554
  4. Nieuwlaat R, Capucci A, Lip GY, et al. Antithrombotic treatment in real-life atrial fibrillation patients: a report from the Euro Heart Survey on Atrial Fibrillation. Eur Heart J. 2006;27:3018–3026
  5. Lip GYH. Coronary artery disease and ischaemic stroke in atrial fibrillation. Chest. 2007;132:8–10
  6. Conway DS, Lip GY. Comparison of outcomes of patients with symptomatic peripheral artery disease with and without atrial fibrillation (the West Birmingham Atrial Fibrillation Project). Am J Cardiol. 2004;93(11):1422–1425
  7. Gage BF, Walraven CV, Pearce L, et al. Selecting patients with atrial fibrillation for anticoagulation. Stroke risk stratification in patients taking aspirin. Circulation. 2004;110:2287–2292
  8. Secondary prevention in non-rheumatic atrial fibrillation after transient ischaemic attack or minor stroke. EAFT (European Atrial Fibrillation Trial) Study Group. Lancet. 1993;342:1255–1262
  9. Lip GY, Frison L, Grind M. Effect of hypertension on anticoagulated patients with atrial fibrillation. Eur Heart J. 2007;28:752–759
  10. Baruch L, Gage BF, Horrow J, et al. Can patients at elevated risk of stroke treated with anticoagulants be further risk stratified. Stroke. 2007;38:
  11. Fang MC, Go AS, Chang Y, et al. Comparison of risk stratification schemes to predict thromboembolism in people with nonvalvular atrial fibrillation. J Am Coll Cardiol. 2008;51:810–815
  12. Reitbrock S, Heeley E, Plumb J, et al. Chronic atrial fibrillation: incidence, prevalence and prediction of stroke using the Congestive heart failure, Hypertension, Age > 75, Diabetes mellitus, and prior Stroke or transient ischemic attack (CHADS2) risk stratification scheme. Am Heart J. 2008;156:57–64
  13. Wang TJ, Massaro JM, Levy D, et al. 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
  14. Stroke Risk in Atrial Fibrillation Working Group. Comparison of 12 risk stratification schemes to predict stroke in patients with nonvalvular atrial fibrillation. Stroke. 2008;[in press]
  15. Lip GY, Lane D, Van Walraven C, et al. Additive role of plasma von Willebrand factor levels to clinical factors for risk stratification of patients with atrial fibrillation. Stroke. 2006;37:2294–2300
  16. Nozawa T, Inoue H, Hirai T, et al. d-dimer level influences thromboembolic events in patients with atrial fibrillation. Int J Cardiol. 2006;109:59–65

PII: S0002-8703(08)00195-6

doi:10.1016/j.ahj.2008.02.020

American Heart Journal
Volume 156, Issue 1 , Pages 1-3, July 2008

Article Tools