Sequential Monitoring of the Comparative Effectiveness and Safety of Dabigatran in Routine Care

Background: The increasing availability of electronic healthcare data enables ongoing monitoring of the effectiveness and safety of newly marketed medications. We sought to demonstrate a 5-year prospective monitoring system of dabigatran for stroke prevention that may expedite discovery and allow ongoing evidence development. Methods and Results: Between 2011 and 2015, we conducted 9 sequential analyses of dabigatran versus warfarin users in a sequential cohort design in 2 US claims databases. Analyses 4 through 9 were prespecified, and analyses 1 through 3 were added subsequently using the same methodology. New users of anticoagulants with nonvalvular atrial fibrillation were followed until a study outcome of hospitalization for stroke (hemorrhagic and ischemic) or hospitalization for major hemorrhage (intracranial and extracranial). Hazard ratios and 95% CIs were estimated after 1:1 propensity score matching. Sequential analyses 1 through 3 on stroke prevention using data through June 2012 were limited by few events leading to wide CIs. As data accumulated the effect estimate in analysis 4 visually stabilized at a 25% risk reduction with increasingly narrower CIs (−46% to +9% in December 2012 and −42% to −2% in September 2015). Improved data-adaptive confounding adjustment with high-dimensional propensity score reached a stable state already at analysis 3 and was slightly closer to the randomized clinical trial finding (−39%). The risk of major hemorrhage was 28% lower in dabigatran initiators (−35% to −20%) a finding that was stable throughout analyses 2 to 9. Conclusions: Prospectively monitoring the effectiveness and safety of dabigatran for stroke prevention allowed for early insights with increasing precision over time.