Oral anticoagulation in patients with rheumatic heart disease for prevention of systemic thromboembolism is accepted clinical practice. The incidence of stroke in patients with nonrheumatic atrial fibrillation is about five times the rate of patients in sinus rhythm. However, contradictory findings in several small retrospective studies have precluded determination of a gold standard for patients with nonrheumatic atrial fibrillation so far. Recently, the results of five prospective, placebo-controlled studies in patients with nonrheumatic atrial fibrillation treated with anticoagulation have been published. A consistent risk reduction of thromboembolism ranging from 37 to 87% in patients treated with warfarin was reported. This risk reduction occurred in excess of a relatively low incidence of intracerebral and/or fatal bleeding complications. The efficacy of prevention of thromboembolism was comparable for high intensity anticoagulation (International Normalized Ratio (INR) 2.8-4.2) and low dose anticoagulation (INR 1.5-2.7). However, fatal and/or intracerebral bleedings only occurred with INR > or = 2.6. In subgroup analysis, recent congestive heart failure, arterial hypertension, and previous apoplex or arterial thromboembolism were independent clinical predictors of increased risk for thromboembolism, whereas results in patients with chronic and intermittent atrial fibrillation were comparable. In 69 patients with lone atrial fibrillation, no single event occurred in the follow-up period. Thus, lone atrial fibrillation does not seem to carry an increased risk for stroke when strict criteria for diagnosis of lone atrial fibrillation are applied. In two of the five studies, aspirin was additionally randomized. Since contradictory findings resulted, the role of aspirin for prophylaxis of stroke still needs to be determined.(ABSTRACT TRUNCATED AT 250 WORDS)