In reply
Bellelli et al raise 2 points. First, in our study, patients with prosthetic cardiac valves were excluded, thus introducing possible bias. However, although this is true, these patients make up only a small percentage of patients attending the anticoagulant clinic. They were excluded because they have a different target international normalized ratio of 3.0 to 4.5 compared with all other patients with atrial fibrillation (target international normalized ratio of 2.5), and are being treated with warfarin primarily because of their valve prosthesis and not atrial fibrillation.