In July 2005, leaders from academia, government, and industry convened in Washington, DC, to discuss key issues in the development of antithrombotic treatments for atrial fibrillation (AF). In addition to summarizing available data on the relative benefits and risks of currently available therapies in diverse clinical practice settings, we reviewed designs of ongoing trials and registries, focusing on areas of methodological controversy and uncertainty. Participants in this meeting described the growing burden of AF, summarized the data showing effectiveness of warfarin for prevention of stroke in AF, and noted that warfarin is both underused and poorly monitored and adjusted in general practice. There was consensus that there is an important unmet clinical need for better treatment of patients with AF at risk of stroke, including alternatives to warfarin that address its limitations. Comparative noninferiority trials to develop alternatives to warfarin must include warfarin management that is at least as good as that provided in historical trials. There was agreement that noninferiority trials can be done based on historical warfarin trials, and that placebo-controlled trials focused on patients not receiving warfarin in general practice can provide important information as well. Statistical principles for noninferiority in this setting were discussed, and a standard approach was proposed. A majority of clinical trial representatives suggested that large, simple, open-label trials would provide the most meaningful information relevant to future practice, but regulators cautioned that, in such a simple trial, one needs to ensure that the control group does at least as well as the historical controls for the noninferiority design to be interpretable. With this summary document, we hope to provide a helpful resource for future drug development for AF.