Estimating Oral Anticoagulant Comparative Effectiveness in the Setting of Effect Heterogeneity: Comparing Clinical Trial Transport and Observational Epidemiologic Methods

Estimating Oral Anticoagulant Comparative Effectiveness in the Setting of Effect Heterogeneity: Comparing Clinical Trial Transport and Observational Epidemiologic Methods

Atrial fibrillation, a type of irregular heart-rate, affects 33 million adults worldwide and up to 9% of adults over 65. Even if patients with atrial fibrillation don't show any symptoms, they are at elevated risk of strokes from blood clots and those strokes are more frequently associated with death, hospitalization, and long-term disability than strokes in adults without atrial fibrillation. Warfarin is the usual drug used to prevent strokes for patients with atrial fibrillation. Warfarin is complicated for patients to take because each patient needs to have a specific dose to make sure they don't have strokes but also don't have harmful bleeding events. Novel oral anticoagulants (NOACs) are easier to manage with simple dose change for poor kidney functions and have been shown to be about as good as or better than warfarin in several drug trials. One of the first NOACs to be approved in the United States, dabigatran, was shown to be more effective than warfarin at preventing blood clots and strokes (HR 0.66, 95% C.I. 0.53-0.82) with no increase in bleeding (HR 0.93, 95% C.I. 1.07) in the RE-LY trial. However, these numbers may not be the same as the real-world treatment effects for dabigatran. Patients selected into trials are usually younger with fewer other diseases; this may change how well they respond to a drug. Some studies have used observational claims data to directly estimate effects of NOACs compared to warfarin in clinical care and seen different results from the trial. Unfortunately, these studies might be biased by things related to both treatment and the outcomes since they were not randomized. Complicating things further, ordinary patients and patients in clinical trials may not have their warfarin doses modified in the same ways. Finally, we don't know how much inaccurate observational data could change how well we can use trial data in general to get a better estimate of real-world treatment effects.To fill these gaps in the research, we propose to:1: "Weight" the RE-LY patients so that they look like older adults with atrial fibrillation collected from real-world Medicare claims data when it comes to a variety of risk factors and assess the difference between the RE-LY trial and Medicare warfarin patients (and, separately, RE-LY trial and Medicare dabigatran patients) in stroke, bleeding, and mortality risk after taking these risk factors into account. 2: Use these weighted trial populations to see how effective dabigatran is relative to warfarin in a general population while still retaining randomization of exposure and compare these effectiveness results with those of an observational study conducted in the same Medicare population at the same time to see how much the results agree with one another. 3 and 4: Assess how much the results in our weighted population would change if risk factors for stroke, bleeding, and mortality had been 3) missing or 4) recorded inaccurately in the RE-LY trial and Medicare cohorts.