Complex questions command complex analyses: comparative effectiveness of drug treatment strategies in atrial fibrillation

Abstract

Listen

Translate article

Among the randomized clinical trials (RCTs) comparing rhythm- with ratecontrol trials conducted in the past 10 years [3–7], the AFFIRM trial is regarded as the most important, owing to its large sample size (4060 participants) and long follow-up (up to 6 years) [7]. Overall, the AFFIRM study showed that the strategy of restoring and maintaining sinus rhythm had no clear advantage in reducing overall mortality over the strategy of controlling the ventricular rate and allowing the AF to persist [7]. This conclusion held after combining the results of AFFIRM [7], and four other smaller trials [3–6], in a single meta-ana lysis estimate (OR: 0.87; 95% CI: 0.74–1.02 vs 1.12; 95% CI: 0.64–1.94 in the pooled ana lysis that included/excluded the AFFIRM trial) [8]. However, many clinicians feel uncomfortable with the rate-control strategy as the main therapeutic approach for AF. For some, the trial results do not mean that the rhythm control is not beneficial, but merely highlight the limitations of the current treatments to achieve and maintain sinus rhythm [9], while others believe the trial results are not generalizable to the full population of AF patients [8]. Such a conundrum demands more evidence through comparative effectiveness studies. To our knowledge, our group has published the first observational comparative effectiveness study of rhythm- versus rate-control drug strategies, published in July 2012 [10]. In this study, we used population-based administrative databases from Quebec (Canada) from 1999 to 2007 to select patients aged ≥66 years, hospitalized with an AF diagnosis who did not have AF-related drug prescriptions in the year before the admission, but received a prescription within 7 days of discharge. The study included 30,664 hospitalized AF patients who were followed-up for up to 8 years and, similar to the trials [7,8], found no difference in mortality between the rhythm- and rate-control drug treatment strategies in the first 3 - years post-treatment initiation. However, our long-term results differed from those of the clinical trials [7,8]; while the AFFIRM trial found no mortality difference at 5 years between the rhythm- and rate-control strategies [7], we found a significant 12% mortality reduction in the rhythm-control group found after 5 years of follow-up and a significant 23% reduc tion after 8 years. In the face of contradictory results between observational studies and RCTs trials, the observational design is oftentimes dismissed as being flawed due to poor control of confounding. We argue here, as others have done before us [11,12], that RCTs and observational studies are complementary designs and that much can be learned by examining results from different study designs and analyses [13,14]. Bias, due to unobserved confounding, is the Achilles’ heel of observational studies of drug effects [15–20], with lack of randomization being a particular concern in studies