A two-sided evaluation of benefit and harm from antithrombotic treatment in atrial fibrillation: Balancing clinical application and statistical methodology