For patients with atrial fibrillation and a high risk of thrombosis, the standard prophylaxis is warfarin, an anticoagulant, at a dose adjusted to the INR. Warfarin and aspirin are both reasonable choices for patients with a moderate risk of thrombosis. Dabigatran, an oral anticoagulant that inhibits thrombin, has been authorised for patients with atrial fibrillation and a moderate or high risk of thrombosis, without associated valvular abnormalities. Clinical evaluation of dabigatran is based on a randomised "non-inferiority" trial comparing two doses of dabigatran (110 mg and 150 mg) taken twice daily, and adjusted-dose warfarin, in 18 113 patients treated for an average of 21 months. Overall mortality was about 4% per year and did not differ between the 3 groups. There was a greater reduction in the annual incidence of stroke or systemic embolism with the higher dose of dabigatran (1.1%) than with warfarin (1.7%). This is not the case for the lower dose (1.5%). In patients with good INR control, the difference in favour of the higher dose of dabigatran was no longer statistically significant. All-cause treatment discontinuation rates were higher with dabigatran than with warfarin (21% versus 17%, p<0.001). The incidence of serious bleeding did not differ statistically between warfarin and the higher dose of dabigatran (3.57% versus 3.32%), but was lower with the lower dose of dabigatran (2.87%). Compared with warfarin, dabigatran appears to be associated with about a 0.2% excess of myocardial infarction (0.73% versus 0.53%). Dyspepsia is also more frequent with dabigatran (6% versus 1.4%). Hepatic adverse effects appear to be mild but need to be monitored. The effects of dabigatran are potentiated by combination with P-glycoprotein inhibitors and drugs that impair renal function. Combination with other antithrombotic agents should also be avoided. Treatment with dabigatran does not require monitoring of haemostasis, whereas vitamin K inhibitors necessitate close INR monitoring. In contrast, renal function must be monitored, as renal impairment increases the risk of bleeding. Unlike vitamin K antagonists, there is no antidote for dabigatran overdose. In practice, warfarin remains the standard drug for patients with atrial fibrillation and a moderate or high risk of thrombosis. Aspirin is an alternative for moderate-risk patients. When the risk is significant and the INR cannot be maintained within the target range despite close monitoring, dabigatran is the alternative to warfarin, provided the patient is closely monitored, especially for changes in renal function.