The long-term use of oral anticoagulants like warfarin in artery disease has long been controversial. Possible aims of treatment include the primary or secondary prevention of systemic embolism, preventing recurrence after myocardial infarction or the progression of transient cerebral ischemia to a complete stroke, and the prevention of artery graft occlusion. The value of long-term anticoagulation is generally accepted in the few situations where, as in patients with mechanical heart valve prostheses, mitral valve disease and atrial fibrillation, or idiopathic dilated cardiomyopathy, the risk of arterial thromboembolism without anticoagulation is known to be high and there is good evidence that anticoagulants are effective, so the benefit:risk balance strongly favours their use. In many settings, however, it is hard to justify long-term warfarin treatment as the benefit:risk balance remains unknown; either because the risk of thromboembolism without anticoagulation remains to be clearly defined (as in the case of patients with 'lone' atrial fibrillation), or because possible benefits have not been well documented (as after transient cerebral ischemia or peripheral vascular surgery). Finally, there is the difficult problem of estimating the benefit from long-term anticoagulation after myocardial infarction. It seems that warfarin can reduce the likelihood of non-fatal reinfarction with relative safety in highly selected patients, but whether it reduces mortality, and how its effect compares with that of other, more recent, therapies aimed at preventing reinfarction, remains unknown.