Intended for healthcare professionals

Education And Debate


BMJ 1996; 312 doi: (Published 06 January 1996) Cite this as: BMJ 1996;312:45
  1. Gregory Y H Lip,
  2. Gordon D O Lowe



    • Antithrombotic prophylaxis with long term warfarin or aspirin reduces thromboembolic risk in atrial fibrillation

    • Identification, risk assessment, and regular review of all patients with atrial fibrillation should be routine in general and hospital practice

    • Risk stratification is easily performed on clinical grounds—echocardiography may refine it

    Although atrial fibrillation has long been recognised as a risk factor for thromboembolic events, preventive treatment has, until recently, been both empirical and controversial. Only in the past few years have antithrombotic drugs been proved by prospective clinical studies to be effective against strokes in atrial fibrillation.

    Atrial fibrillation is an important risk factor for stroke, and antithrombotic treatment should be considered in most patients with this arrhythmia

    Atrial fibrillation and thromboembolism

    The pathophysiological mechanism for thromboembolism seems to be the disturbed blood flow in the fibrillating left atrium, which predisposes to the formation of thrombi and arterial embolism, especially in the presence of underlying heart disease. About 15-20% of patients who have an acute stroke have atrial fibrillation at the time of their stroke; mortality is one and a half to three times higher in those patients than in patients with sinus rhythm at the time of stroke.

    Non-rheumatic atrial fibrillation and stroke

    • Non-rheumatic atrial fibrillation has been associated with a fivefold increase in the risk of ischaemic stroke

    • The yearly risk is 5-7%, increasing with age

    • Computed tomography studies have shown that silent ischaemic cerebral infarction is present in 26% of patients with non-rheumatic atrial fibrillation

    Underlying heart disease

    Underlying heart disease—including valvar and hypertensive heart disease, an enlarged left atrium, and poor left ventricular function—is a contributory risk factor for stroke and thromboembolism in atrial fibrillation. The risk of thromboembolism in atrial fibrillation is 18 times greater if rheumatic heart disease is present. In the stroke prevention in atrial fibrillation study, the presence of recent (within three months) congestive heart failure and left ventricular dysfunction on echocardiography also contributed to an increased risk of thromboembolism in atrial fibrillation. This is consistent with the observation that thromboembolism is a common cause of death in patients with congestive heart failure, occurring in up to 30% of patients, which is partly preventable by anticoagulation.


    Computed tomogram showing large thrombotic stroke.

    Left atrial enlargement and spontaneous echo contrast

    An enlarged left atrium may contribute to an increased risk of atrial thrombi and thromboembolism in patients with atrial fibrillation. An enlarged left atrium has also been associated with “spontaneous echo contrast” on transoesophageal echocardiography—a smoke-like appearance of blood in the atria, suggesting sluggish flow. It is associated with dilated left atria, intracardiac thrombi, thromboembolism, stroke, abnormal rheology, and coagulation.

    Previous cerebrovascular disease or thromboembolism

    A history of stroke, transient ischaemic attack, or other embolic events adds to the risk of stroke and mortality in atrial fibrillation. The pooled analysis by the Atrial Fibrillation Investigators showed that a previous stroke or transient ischaemic attack was an independent risk factor for further strokes. Using warfarin in such patients reduced the annual rate of stroke from 12% a year to 5%.

    Echocardiogram (with diagram) showing prosthetic mitral valve and thrombus in left atrium.

    Echocardiogram (with diagram) showing prosthetic mitral valve and thrombus in left atrium.

    In a Glasgow study, among patients admitted with acute severe stroke, those with atrial fibrillation (25%) had a significantly higher hospital mortality than those in sinus rhythm (67% v 44%). This was confirmed in the Oxfordshire community stroke project, in which the 30 day mortality among patients with acute stroke was three times higher among those with atrial fibrillation (17%) than among those in sinus rhythm.

    Paroxysmal and chronic atrial fibrillation and stroke risk

    • A third of patients with paroxysmal atrial fibrillation develop chronic atrial fibrillation over two to three years

    • The risk of stroke is highest during the first months after the initial diagnosis of atrial fibrillation or immediately after a transition from paroxysmal to chronic atrial fibrillation

    • The recent pooled analysis by the Atrial

    Fibrillation Investigators suggested that patients with paroxysmal atrial fibrillation and chronic atrial fibrillation had a similar risk of stroke, although the length of time a patient was in atrial fibrillation had no discernible effect on the risk

    Hypertension and diabetes

    A history of hypertension and diabetes add to the risk of stroke in atrial fibrillation. Using warfarin in patients with a history of hypertension or diabetes would reduce the annual rate of stroke respectively from 5-6% to 2% and from 9% to 3%.

    Duration and onset of atrial fibrillation

    The onset of atrial fibrillation may be related to the imminence of stroke. In the Framingham study, atrial fibrillation was present at the time of stroke in 24% of subjects, and about a third of the strokes associated with atrial fibrillation occurred within six months of onset of the arrhythmia. In addition, a further stroke within six months of the first may be more common in patients with continued atrial fibrillation.

    Many patients with paroxysmal atrial fibrillation go on to develop chronic atrial fibrillation, and thromboembolic complications often occur during the transition.

    Paroxysmal atrial fibrillation may be associated with the sick sinus syndrome, and patients with this condition are at particular risk of stroke and thromboembolism.

    Clinical trials of warfarin and aspirin prophylaxis Warfarin

    In the mid-1980s five randomised prospective clinical trials were independently started to define the value of anticoagulant prophylaxis with warfarin in patients with atrial fibrillation. These five primary prevention studies showed consistent results. Since publication of the results another primary prevention trial (the second stroke prevention in atrial fibrillation study) has been reported, as well as a secondary prevention trial in patients with non-rheumatic atrial fibrillation and a history of transient ischaemic attack or minor stroke (European atrial fibrillation trial).

    Important stroke prevention trials

    View this table:

    The pooled analysis of the five primary intervention trials showed that warfarin reduced the annual rate of stroke; led to a similar risk reduction in stroke with residual deficit; reduced mortality by a thrid; and reduced the rate of the combined adverse outcome (stroke, systemic embolism, or death) by half. The efficacy of warfarin prophylaxis was in fact underestimated because most strokes in patients allocated to warfarin occurred while the patient was not taking his or her warfarin.


    Efficacy of warfarin in atrial fibrillation trials—total risk reduction for all five trials combined is 68% (P<0.001).

    The five primary prevention trials used different target therapeutic ranges of the international normalised ratio of the prothrombin time; the target range in all studies was 1.5 to 4.5. The minimum risk of stroke seemed to occur in the range 2.0 to 3.0, which is now accepted as the optimum therapeutic range for warfarin prophylaxis of thromboembolism in all patients except those with older, mechanical prosthetic heart valves and those with recurrent thromboembolism despite warfarin prophylaxis at the range 2.0 to 3.0 (such patients merit higher intensity warfarin (range 3.0 to 4.5)). At present, therefore, a target range of 2.0 to 3.0 can be recommended in the absence of the above indications for high intensity warfarin.


    Ranges of international normalised ratios used in atrial fibrillation trials.

    Intracranial haemorrhage is the most feared complication of warfarin prophylaxis. The annual risks of intracranial haemorrhage increased from 0.1% in the controls to 0.3% in patients taking warfarin in the pooled analysis. This risk was associated with an international normalised ratio >3.0 and with uncontrolled hypertension; there was also a non-significant association with increasing age. Lower intensity warfarin (for example, target ratio 1.5 to 2.5) is currently being compared with higher intensity warfarin in the third stroke prevention in atrial fibrillation study; meanwhile some physicians use low intensity warfarin in patients for whom the usual target range of 2.0 to 3.0 is judged to carry an unacceptable risk of bleeding.

    Despite the convincing evidence from the pooled analysis of randomised trials that warfarin prophylaxis is highly effective and seems to outweigh substantially the risk of intracranial bleeding (as well as other types of major bleeding), the efficacy of warfarin in clinical trials may not be reproduced in practice. This is due to the highly selected patients in these studies.

    Risk of bleeding with anticoagulation treatment

    • Risk of bleeding in patients taking warfarin increases exponentially with an increase in anticoagulant effect, rising from an annual risk of 0.2% with an international normalised ratio of 2.0 to 3.0% with a ratio of 4.0

    • Bleeding is most common in patients with unstable anticoagulation control and those with a history of thromboembolism and ranges from 11% to 40%

    • Complications related to bleeding may occur often, however, even when the international normalised ratio is within the therapeutic range

    Does evidence from prevention studies apply to clinical practice?

    • The patient population was highly selected—for example, >90% of those screened in many of the studies were excluded from entry

    • The rate of patient withdrawal in the studies was often high (19% to 38%)

    • Patients in usual clinical practice may have different risks of thromboembolism and bleeding from selected patients participating in trials

    • The compliance, and therefore safety, of warfarin may be poorer when it is used in clinical practice than when it is used in carefully selected, well motivated, and closely monitored study participants

    Further questions on antithrombotic treatment in atrial fibrillation

    • What is the efficacy of aspirin (which does not need laboratory monitoring and carries a lower risk of bleeding) compared with warfarin as antithrombotic prophylaxis?

    • For which patients with the highest risk of thromboembolism may the risk of bleeding and the monitoring of warfarin prophylaxis be most justified?

    • Are there any patients whose risk of thromboembolism is so low that neither aspirin nor warfarin prophylaxis is indicated?


    Aspirin has been evaluated as primary prophylaxis of systemic thromboembolism—in two randomised trials compared with no antithrombotic treatment and in one compared with warfarin: aspirin was only about half as effective as warfarin, with a 36% decrease in risk of stroke. Aspirin prophylaxis had no significant effect on stroke with residual deficit or mortality; but there was a 28% decrease in the rate of the combined outcome of stroke, systemic embolism, or death.

    The annual risk of stroke seems low, however, in patients at moderate risk treated with aspirin, and aspirin is simpler to monitor and has a lower risk of bleeding. Aspirin was less effective than warfarin, however, when given as secondary prophylaxis in patients with previous stroke or transient ischaemic attack, who have a higher risk of thromboembolism.


    Efficacy of aspirin in atrial fibrillation trials—total risk reduction for all three trials combined is 21% (P=0.04).

    Risk stratification for warfarin, aspirin, or no antithrombotic prophylaxis

    The relative risk of stroke in people with non-rheumatic atrial fibrillation (and hence the absolute benefits of antithrombotic prophylaxis with warfarin or aspirin) varies greatly with the presence of risk factors for thromboembolism. The risk of stroke in non-rheumatic atrial fibrillation seems similar in men and women, in people with continuous atrial fibrillation, and in those with paroxysmal atrial fibrillation.

    Risk of stroke in patients with atrial fibrillation, stratified by risk, with and without antithrombotic prophylaxis. Values are percentages per patient per year

    View this table:

    A moderate risk of stroke (annual risk 4%) exists in people with non-rheumatic atrial fibrillation, without previous embolism, who are (a) aged under 65 years with a history of diabetes or hypertension; (b) aged 65-74 (with or without a history of diabetes or hypertension); or (c) aged 75 years or more with no history of diabetes or hypertension. In such people aspirin seems as effective as warfarin (reducing the annual risk of stroke to 1% to 2%).

    A low risk of stroke (annual risk 1%) exists in people with non-rheumatic atrial fibrillation who are aged under 65 years with no history of embolism, hypertension, or diabetes. In such people antithrombotic prophylaxis with either aspirin or warfarin does not seem warranted unless some other indication coexists.

    Echocardiography may complement the clinical risk stratification of risk of stroke in atrial fibrillation, although the precise contribution in non-rheumatic atrial fibrillation is uncertain (echocardiographic data were not included in the pooled analysis of the Atrial Fibrillation Investigators). There is an increase in risk of stroke, however, with a large left atrium, impaired left ventricular function, or calcification of the mitral valve. An increased risk of systemic embolism has also been observed in patients with atrial fibrillation associated with heart valve disease (for example, rheumatic heart disease), heart valve prosthesis, heart failure, intracardiac thrombus, or thyrotoxicosis. One possible algorithm for risk stratification and selection of prophylaxis in atrial fibrillation is shown.


    Algorithm for risk stratification and selection of prophylaxis in atrial fibrillation.

    Practical considerations in antithrombotic prophylaxis

    Before starting antithrombotic prophylaxis with warfarin or aspirin it is important to balance the risks (especially bleeding) and benefits in a patient. To minimise the risk of intracranial bleeding with prophylactic warfarin, hypertension should be adequately controlled, and the risks and benefits of warfarin reviewed annually, especially in patients aged over 80.

    Factors that may increase risk of bleeding with warfarin

    • Age

    • Uncontrolled hypertension (systolic blood pressure > 180 mm Hg or diastolic blood pressure >100 mm Hg)

    • Alcohol excess

    • Liver disease

    • Poor drug or clinic compliance

    • Bleeding lesions (especially gastrointestinal blood loss—for example, peptic ulcer disease—and previous cerebral haemorrhage)

    • Tendency to bleeding (including coagulation defects, thrombocytopenia)

    • Concomitant use of aspirin with oral anticoagulants

    In patients with acute stroke and atrial fibrillation, intracranial haemorrhage should therefore be excluded—for example, with computed tomography—before starting warfarin. This is important as 11% of patients with haemorrhagic stroke have atrial fibrillation, compared with 18% with stroke due to cerebral infarction. It is important to start anticoagulation at the right time after an acute cardioembolic stroke as there is the risk of haemorrhagic transformation of the cerebral infarct, by bleeding into the infarcted, softened brain.

    Recommendations by American College of Chest

    • Physicians on starting anticoagulation after acute thromboembolic stroke

    • Patients with small or moderate thromboembolic strokes in whom an intracranial bleed is excluded by computed tomography at >/=48 hours: Heparin followed by warfarin (international normalised ratio 2.0-3.0)

    • Patients with large embolic strokes or uncontrolled hypertension: No anticoagulation for 5-14 days because of the increased risk of haemorrhagic transformation

    • Patients with non-valvar atrial fibrillation as the presumed cause of thromboembolism: Warfarin only (after computed tomography at 48 hours) owing to the low risk of early recurrent thromboembolism

    Key references