Among the principles for selection of endpoints for trials of arrhythmia management, the most important is that the endpoint be clinically relevant. According to this principle, the endpoint must be closely related to the objective of therapy. In the case of atrial fibrillation, for too long it seems, it has been believed that the primary objective of therapy has been maintenance of sinus rhythm. However, it is important to recognize that maintenance of sinus rhythm is, in fact, a surrogate endpoint and, as the lesson of the Cardiac Arrhythmia Suppression Trial has taught us, arrhythmia suppression is not necessarily a good surrogate for clinical events of importance. The use of clinical events as endpoints is to be encouraged, but the problem is that these events occur with a low frequency in many patients with atrial fibrillation. Symptoms are a primary target of therapy and potentially a good endpoint, but good methodology for quantification of symptoms remains to be developed. Methodology for assessment of patient-perceived quality of life is more highly developed. Surrogate endpoints, such as measurement of sinus rhythm or heart rate control, have the dual problems that they have not been well calibrated with clinical events and they sometimes are overly precise. Other surrogates have not been investigated. Composite endpoints are potentially useful, and cost is an important endpoint. More research on the validity of some of these endpoints is needed if we are to advance our knowledge of the management of atrial fibrillation in a meaningful way.