1Santovito et al are concerned about the unequal distribution of baseline risk factors between patients treated with rhythm control compared with patients treated with rate control strategies. In the current study, great efforts were made to account for potential confounding by stroke risk, and therefore we used multivariable model risk adjustment and propensity-based matching. Nevertheless, these methods require that the confounding variables are known and measured. As mentioned in the Discussion in our article, the main limitation of our study was the possible residual confounding from unknown, unavailable, or unmeasurable factors. Although we were able to establish a large and well-balanced cohort of patients treated with rhythm and rate control therapies matched on the basis of the propensity score, there may still be residual confounding that may contribute to our findings, such as quality of anticoagulation.We agree with Santovito and colleagues that the main strategy to prevent stroke in patients with atrial fibrillation (AF) is the appropriate prescription of anticoagulant therapy. Sadly, the suboptimal proportion of anticoagulated patients in our study population is not that different from various other reports from real-world practice.
2-5We agree with Gasparovic and Kopjar's comment that rhythm therapy cannot be equated with sinus rhythm. Unfortunately, we did not have ECG information to determine sinus rhythm in patients. Similarly, we also did not have information on AF duration and subtypes.In their first concern, Parikh and Rashba suggested that it would be better to adjust our models for warfarin use instead of "any antithrombotic medication." Because the point estimate and the statistical significance of the association between rhythm control and the risk of stroke were barely modified whether the model was adjusted for just warfarin or for warfarin, or aspirin, or clopidogrel, we decided to report the latter. As for the second concern of Parikh and Rashba, we want to clarify that the groups of rhythm and rate control patients were defined according to prescriptions dispensed postdischarge. We believe that both variables in question can partially predict the clinician's choice of either the rhythm or rate control strategy, because they are both associated with the burden of AF. For example, patients with a long stay in the hospital at the index AF hospitalization probably have a higher burden of AF. Similarly, patients selected for treatment with warfarin are probably older and have additional comorbidities, such as diabetes mellitus, hypertension, congestive heart failure, or a history of stroke. Although we appreciate Parikh and Rashba's concern regarding the inclusion of outcome variables in our model, we are confident that these variables are actually markers for decisions made during the index encounter. Therefore, we believe that our choice of variables to be included in the propensity score was well considered and justified.Despite the limitations raised by Santovito et al, Gasparovic and...