In their article on the management of chronic asymptomatic atrial fibrillation, Disch and colleagues [1] used a decision model to conclude that low-dose amiodarone or quinidine would be preferred over warfarin, primarily through a reduction in stroke risk [1]. They concluded this in the context of several clinical trials showing the effectiveness of warfarin in reducing stroke risk in patients with atrial fibrillation [2] but had no clinical trial data supporting the effectiveness of quinidine, amiodarone, or any other antiarrhythmic agent for this purpose.

Their conclusion appears to hinge on two subtle assumptions embedded in the structure of the model. The first is that the presence of normal sinus rhythm on periodic electrocardiograms recorded during office visits indicates that asymptomatic atrial fibrillation is not occurring. In fact, asymptomatic atrial fibrillation occurs frequently in patients who are being followed for symptomatic atrial fibrillation [3]. For the patients addressed in this analysis—those whose atrial fibrillation is asymptomatic—this phenomenon may be more common and more difficult to detect. The second assumption is that persons with atrial fibrillation who revert to normal sinus rhythm have a risk for stroke similar to that of persons who have never had atrial fibrillation. The risk for stroke is nearly as high in patients with intermittent atrial fibrillation as it is in patients with chronic atrial fibrillation [4]. In addition, the risk factors common to stroke and atrial fibrillation (for example, atherosclerotic heart disease) do not disappear with successful cardioversion or with antiarrhythmic therapy. The authors acknowledge this risk and attempt to explain it by varying the underlying annual probability of stroke from a baseline of 0.1% to 1%. However, the upper bound of this range is probably too low to account for the many stroke risk factors among persons with atrial fibrillation who had successful cardioversion.

Despite the tenuous nature of the underlying assumptions, the conclusions were stated rather strongly. As clinicians concerned with the practical implications this analysis presented, we wonder whether modifying the model to account for the questions raised here would result in the disappearance of the apparent preference for antiarrhythmics.