Randomized controlled trials (RCTs) examining warfarin use for stroke prevention in atrial fibrillation (AF) may not accurately reflect real-world populations. We aimed to determine the representativeness of the RCT populations to real-world patients and to describe differences in the characteristics of trial populations from trial eligible patients in a real-world setting. We hypothesized that a significant fraction of real-world patients would not qualify for the RE-LY, ROCKET-AF, and ARISTOTLE trials and that real-world patients qualifying for the studies may have more strokes and bleeding events. We compared the inclusion and exclusion criteria, patient characteristics, and clinical outcomes from RE-LY, ROCKET-AF, and ARISTOTLE against data from the Michigan Anticoagulation Quality Improvement Initiative (MAQI2), a regional network of six community- and academic-based anticoagulation clinics. Of the 1446 non-valvular AF patients in the MAQI2 registry taking warfarin, approximately 40-60% would meet the selection criteria used in RE-LY (788, 54.5%), ROCKET-AF (566, 39.1%), and ARISTOTLE (866, 59.9%). The most common reasons for exclusion from one or more trial were anemia (15.1%), other concurrent medications (11.2%), and chronic kidney disease (9.4%). Trial-eligible MAQI2 patients were older, more frequently female, with a higher rate of paroxysmal AF, and lower rates of congestive heart failure, previous stroke, and previous myocardial infarction than the trial populations. MAQI2 patients eligible for each trial had a lower rate of stroke and similar rate of major bleeding than was observed in the trials. A sizable proportion of real-world AF patients managed in anticoagulation clinics would not have been eligible for the RE-LY, ROCKET-AF, and ARISOTLE trials. The expected stroke risk reduction and bleeding risk among real-world AF patients on warfarin may not be congruent with published clinical trial data.
Keywords: Anticoagulation; Atrial fibrillation; Outcomes; Real world; Warfarin.