In Reply Vora et al are correct that residual confounding is a possible explanation for our study findings, which we clearly acknowledged in our Original Investigation.1 Having said that, we believe that residual unmeasured confounding is unlikely given that we were able to measure and adjust for a large number of clinically relevant covariates, and that these covariates were highly balanced between drug cohorts before adjustment, suggesting that medication choice was not being driven by patient health status. Vora et al also raise the question that our use of claims data could lead to biased ascertainment of events. However, our study1 relied on well-validated hospitalized or fatal outcomes only and was conducted following a carefully designed protocol in which study investigators remained blinded to outcome data until all analytical and statistical decisions were finalized. In a previous study2 of dabigatran vs warfarin using the same database and outcome definitions, we obtained results for the standard dose of dabigatran closely similar to those from RE-LY, lending support to our study methods. Finally, while Vora et al worry that our analysis could be biased by the inclusion of patients taking rivaroxaban with a creatinine clearance of 30 to 50 mL/min, the subgroup analysis in our study including those without chronic kidney disease (which was also limited to standard-dose patients) replicated all our major findings.
Graham DJ, Wernecke M, Kelman JA. Dabigatran Compared With Rivaroxaban vs Warfarin—Reply. JAMA Intern Med. 2017;177(5):743–744. doi:https://doi.org/10.1001/jamainternmed.2017.0567
Customize your JAMA Network experience by selecting one or more topics from the list below.
Create a personal account or sign in to: