Fifteen per cent of cerebrovascular accidents have a cardiac origin, two thirds of which are due to atrial fibrillation (AF). The Framingham study showed the risk of an ischaemic cerebral event to be increased by 5.6 in AF unrelated to rheumatic heart disease and by 17.5 when AF is associated with valvular heart disease. The risk of embolism is higher in elderly subjects and in those with underlying cardiac disease. Other high risk conditions include hypertension, diabetes, hyperthyroidism and cases with echocardiographic changes: left atrial dilatation, pre-thrombotic state or intra-atrial thrombus, atheroma of the ascending aorta. This stratification of risk should be taken into account when deciding on treatment. Conscious of the importance of the risk of embolism in AF, several authors have undertaken, over the last few years, randomised studies of the prevention of thromboembolic complications of AF: the AFASAK, BAATAF, SPAF and SPINAF trials. All showed the unquestionable efficacy of warfarin, even at low doses, at the price of a haemorrhagic risk of less than 2% per year for severe haemorrhages. A more recent study (SPAF II) confirmed the value of aspirin at the dosage of 325 mg/day which would seem to be a good alternative to anticoagulant therapy when this is contraindicated, although aspirin is less effective. The indications for anticoagulant therapy have become clearer since the publication of these results. Anticoagulant therapy is essential in permanent AF whether or not associated with rheumatic heart disease.(ABSTRACT TRUNCATED AT 250 WORDS)