Patients with atrial fibrillation (AF) and atrial thrombi have an increased risk for cerebral embolism. However, there is little knowledge about the long-term fate of atrial thrombi and the incidence of cerebral embolism in patients receiving continued oral anticoagulation. Forty-three consecutive patients with AF and atrial thrombi were enrolled in the study. Serial and prospective transesophageal echocardiographic studies, cranial magnetic resonance imaging (MRI), and clinical examinations were performed during a period of 12 months. Oral anticoagulation was continued or initiated in all patients. An international normalized ratio of 2.0 to 3.0 was regarded as effective. During follow-up, 56% of the thrombi disappeared (7 [16%] at 1 month, 18 [42%] at 3 months, 21 [49%] at 6 months, and 24 [56%] at 12 months). Patients with the disappearance of thrombi had significantly smaller thrombi compared with patients with persistent thrombi (1.5 +/- 0.8 cm in length and 0.8 +/- 0.5 cm in width vs 1.9 +/- 0.6 cm in length and 1.3 +/- 0.4 cm in width, p = 0.04), reduced echogenicity of thrombi (46% vs 89%, p <0.01), and smaller left atrial (LA) volume (83 +/- 27 vs 116 +/- 55 cm(3)). Seven patients (16%) had embolic lesions during follow-up MRI. Six of these patients (86%) had clinically apparent embolisms, and 1 died from stroke. The only independent predictors of cerebral embolism were an elevated peak emptying velocity of the LA appendage (p <0.01) and previous thromboembolic events (p = 0.02). Patients with AF and atrial thrombi have a large likelihood of cerebral embolism (16%) and/or death despite oral anticoagulation therapy. Thrombus size may predict thrombus resolution under continued anticoagulation.