Are the newer anticoagulants safer or more effective than warfarin in patients with atrial fibrillation?
In this meta-analysis, newer anticoagulants appear to be slightly more effective than warfarin in the short term (2 years) in preventing strokes of all kinds in patients with atrial fibrillation. However, they are no more effective than warfarin in preventing ischemic strokes, and they cause more gastrointestinal hemorrhage. The short-term nature of the included studies and a significant concern about publication bias suggests that the newer agents are by no means a slam dunk over warfarin. Since the patients taking warfarin only spent two thirds of their time in therapeutic range, perhaps efforts to improve performance may be a wiser use of resources. (LOE = 1a-)
Ruff CT, Giugliano RP, Braunwald E, et al. Comparison of the efficacy and safety of new oral anticoagulants with warfarin in patients with atrial fibrillation: a meta-analysis of randomised trials. Lancet 2014;383(9921):955-962.
Study Design: meta analysis of RCT’s
Funding: self-funded or unfunded
The authors of this study and the editors of Lancet did a fairly abysmal job in communicating how this study was conducted. The abstract claims the authors searched a single database to identify phase randomized trials comparing newer anticoagulants with warfarin in patients with atrial fibrillation. However, the methods section reports the authors conducted a prespecified analysis of 4 studies and doesn’t describe a thing about the search strategy. Additionally, the authors don’t include data on ximelagatran, which was pulled because of safety concerns. The authors pooled the data for nearly 72,000 patients with atrial fibrillation who received one of the newer anticoagulants (dabigatran, rivaroxaban, apixaban, or edoxaban; n = 42,411) or warfarin (n = 29,272). The authors evaluated the outcomes by intention to treat “when possible,” but made no adjustments for performing multiple analyses. Two of the four trials included nearly one third of the patients with CHADS2 scores of 0 to 1, which is a group that may not need anticoagulation. Overall, approximately one fourth of the patients had paroxysmal atrial fibrillation. The median duration of follow-up was 2.2 years and the warfarin-treated patients were in therapeutic range only two thirds of the time. Patients receiving the newer agents had a lower risk of stroke or systemic embolic events than patients taking warfarin (3.1% vs 3.8%; relative risk = 0.81; number needed to treat (NNT) =148; 95% CI, 103 – 261). There was no statistically significant differences in the rate of ischemic stroke or myocardial infarction. Patients taking the newer agents had fewer hemorrhagic strokes (0.4% vs 0.9%; NNT = 220; 170 – 308) and intracranial hemorrhages (0.7% vs 1.4%; NNT = 132; 108 – 169). Furthermore, all-cause mortality was slightly lower in patients receiving the newer agents (6.9% vs 7.7%; NNT = 129; 84 – 279). However, patients taking newer agents had more gastrointestinal bleeding (2.6% vs 2%; number needed to treat to harm = 185; 128 – 335). Although the study was unfunded, the authors had heavy ties to industry, which may explain their sloppiness in describing their methods. For example, they don’t even try to evaluate the potential for publication bias, an issue especially important since industry-sponsored studies have a long track record of not finding their way to publication.