ArticlesDecision analysis and guidelines for anticoagulant therapy to prevent stroke in patients with atrial fibrillation
Introduction
Non-valvular atrial fibrillation increases the risk of stroke,1 but anticoagulant therapy decreases this risk by about two thirds.2, 3, 4 However, few clinicians have acted on this evidence5—a finding that is reflected in the underuse of anticoagulant therapy in patients with atrial fibrillation.6, 7, 8, 9
Clinical guidelines can change medical practice and improve health.10 Previous studies have shown not only demand for guidelines in this area,11 but also variation in the content and implications of those available.12 Guidelines have been classified into a hierarchy ranging from informal and formal consensus guidelines, through evidence-based guidelines, to evaluative guidelines.13 The evaluative method may offer the most comprehensive approach since it incorporates data on quality of life and allows explicit quantitative comparison of the benefits and risks of different therapies. However, it has not been widely used.14
Decision analysis permits explicit quantitative comparison of the benefits and risks of different therapies. Previous use of decision analysis has taken a rigorous approach to the synthesis of data from randomised controlled trials, but has not used similarly careful techniques in the assessment of epidemiological evidence,15 nor has it made the critical step of translating the results of this synthesis into practical tools for making clinical decisions.16, 17
We used a Markov decision analysis to model decision-making about warfarin treatment in patients with atrial fibrillation. The analysis included a systematic literature review followed by the development of evaluative guidelines. We then applied the guidelines to a cohort of patients with atrial fibrillation.8
Section snippets
Data acquisition
We did a systematic search of published studies on: effectiveness of anticoagulant and antiplatelet therapies; natural history and risk of stroke in patients with atrial fibrillation; adverse effects of warfarin; utility of relevant health states; and costs of treatment. Titles and abstracts were read, and potentially relevant articles retrieved for full appraisal. Our basic tools were MEDLINE and BIDS, the Cochrane Search Strategies,18, 19 and, for appraisal, the Evidence-Based Medicine
Data acquisition
There were six randomised, controlled trials of warfarin anticoagulation,3, 31, 32, 33, 34, 35 and a pooled analysis of five of them.2 One trial33 reported only an on-treatment analysis and was excluded. Four were not double-blind.3, 31, 33, 34 The pooled study2 analysed the data by intention-to-treat, which overcame the problem of on-treatment results, but included trials that were not double-blind. The two double-blind trials of warfarin and placebo32, 35 were terminated early, and only one
Discussion
We have shown that a Markov decision-analysis model can clarify the factors that affect clinical decision on anticoagulation. Decision analysis explicitly quantifies uncertainty—eg, we showed that the decision on whether to anticoagulate is sensitive to the estimate of the effectiveness of warfarin, which itself is available as a point measure with attendant uncertainty in the form of confidence intervals around this estimate. The incorporation of a patient's preference also makes explicit the
References (46)
- et al.
Prevalence of atrial fibrillation and eligibility for anticoagulants in the community.
Lancet
(1998) - et al.
Making decisions about antithrombotic therapy in heart disease.
Chest
(1995) - et al.
The role of decision analysis in informed consent: choosing between intuition and systematicity.
Soc Sci Med
(1997) Measurement of health state utilities for economic appraisal—a review.
J Health Econ
(1986)- et al.
Canadian Atrial Fibrillation Anticoagulation (CAFA) study.
J Am Coll Cardiol
(1991) - 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) - et al.
Bleeding complications of oral anticoagulant treatment: an inception-cohort, prospective collaborative study (ISCOAT).
Lancet
(1996) - et al.
Atrial fibrillation as an independent risk factor for stroke: the Framingham study.
Stroke
(1991) 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)Secondary prevention in non-ischaemic atrial fibrillation after transient ischaemic attack or minor stroke
Lancet
(1993)