We have entered an exciting new era of anticoagulation—with multiple new oral anticoagulants becoming available for use as potential replacements for warfarin, the vitamin K antagonist that has been the only oral agent available for 50 years! The most prominent of the indications for anticoagulation is for stroke prevention in patients with atrial fibrillation, where long-term, often life-long therapy is needed in millions of patients. Two agents have already been approved by the Food and Drug Administration (FDA), and a third is anticipated later this year. There are 2 main classes of drugs that are undergoing testing or have been tested in large randomized trials: direct thrombin inhibitors (dabigatran [1]) and factor Xa inhibitors (rivaroxaban [(2)], apixaban [(3)], edoxaban [4]); in these trials, these drugs are being compared with warfarin.

Three large randomized trials (totaling over 50,000 patients with atrial fibrillation) comparing one of the newer oral anticoagulants to warfarin have already been published. Each new drug has shown equal or superior efficacy to warfarin, suggesting that these agents offer excellent alternatives to warfarin for stroke prevention. But faced with these growing new options, how do physicians choose between them? Are there differences between the drugs in their efficacy and/or safety that would help inform our clinical decision to prescribe one over the other? Unfortunately, no trials have been conducted (yet) that directly compare one new agent versus another.

In this issue of the Journal, Lip et al. (5) have tried to answer these questions by applying statistical methods to make indirect comparisons between the new agents. Because each trial had a common comparator arm, one could assume similarity in outcomes of that arm and then look at the relative benefit (or harm) that each drug displays versus the standard arm. They have used the Bucher method to compare hazard ratios of the new agent to the comparator to try to determine the relative efficacy between the 2 newer agents.

For such comparisons, however, we need to proceed with extreme caution: attempts at indirect comparisons were made among fibrinolytic agents—tissue plasminogen activator, streptokinase, and anistreplase—where relative reductions in mortality in placebo-controlled trials were 25%, 25%, and 50%, respectively (6). On the basis of these results, one might be tempted to conclude that anistreplase is superior, but in a direct comparison of these drugs in ISIS-3 (Third International Study of Infarct Survival), no difference in endpoints was seen (9). As such, indirect comparisons can be misleading, and extreme caution should be exercised when using such methods to draw definitive conclusions.

After evaluating the efficacy of each new drug individually, Lip et al. (5) then went on to calculate a weighted average of all the new agents combined to estimate the relative benefit of “any new oral anticoagulant” versus warfarin. This latter approach is akin to pooling or meta-analysis of all the studies, and provides a reasonable overview of the “class effect.” For all the new oral anticoagulants, they found that this class as a whole significantly reduced the risk of: 1) stroke or systemic embolism by 21% (p < 0.001); 2) stroke by 23% (p < 0.001); 3) hemorrhagic stroke by 53% (p < 0.001); and 4) all-cause mortality by 12% (p < 0.001). In addition, major bleeding was lower for any new oral anticoagulant by 13% (p < 0.001) compared with warfarin. This straightforward and statistically sound analysis of the data strongly supports the use of the newer agents over warfarin, as a class.

For the indirect comparisons between individual agents, however, the picture remains a bit confusing. Lip et al. report a “significantly lower risk of stroke or systemic embolism (by 26%) for dabigatran 150 mg twice daily compared with rivaroxaban.” On the other hand, they conclude “we found no profound significant differences in efficacy between apixaban, and dabigatran (both doses) or rivaroxaban.” They also report no significant difference in myocardial infarction event rates among the 3 agents. Yet, within the individual trials, the rate of myocardial infarction was definitely higher with dabigatran (1), but not with either of the factor Xa inhibitors (2). These conflicting results lend support to the conclusion that such methods for indirect comparisons may not be the most accurate due to several sources of confounding.

The authors acknowledge that there are multiple limitations in comparing different trials, despite the common comparator arm—notably the patient population differs. Variations in the baseline risk of patients in 1 study could influence the absolute rates of the endpoints in that study, confounding the drug–drug comparison being studied. For example, the ROCKET trial (Rivaroxaban-once daily, oral, direct factor Xa inhibition compared with vitamin K antagonism for prevention of stroke and Embolism Trial in Atrial Fibrillation) trial enrolled older patients with higher CHADS₂ (Congestive heart failure, Hypertension, Age, Diabetes, Prior Stroke) risk scores, and as such, would be expected to have higher rates of bleeding and stroke—which might influence the relative drug effect seen in this trial. Therefore, 2 variables would differ when comparing a similar endpoint between these 2 trials—the drug being tested and the patient population—and it would be unclear whether any apparent difference observed in the relative efficacy could be attributed to the drug alone. Similarly, the optimal use of warfarin (as measured by the “time in therapeutic range”) also differed between these trials, resulting in another possible source of confounding that could influence the relative efficacy of the drugs when comparing the trials.

Traditional metrics for determining efficacy derive directly from the results of the randomized trial. Indeed, this is exactly what the FDA does in the prescribing insert: the FDA approves and lists only those findings and indications that were directly reported in the randomized trials. As such, we would find that dabigatran 150 mg twice daily significantly reduced stroke or systemic embolism (and within that endpoint, ischemic stroke alone), but with a similar rate of major bleeding to warfarin. The 110-mg dose (unfortunately not approved by the FDA) had similar efficacy for stroke or systemic embolism but a lower risk of major bleeding compared with warfarin (1). For apixaban, significant reductions in stroke or systemic embolism, major bleeding, and mortality were seen (3). Rivaroxaban showed similar rates of stroke or systemic embolism and major bleeding versus warfarin (2). But whether apixaban and dabigatran 150 mg are superior to rivaroxaban cannot readily be concluded by comparing the relative efficacy of each agent versus comparator within its own respective trial. We can say though that these 2 agents have been shown to be superior for stroke prevention, and apixaban for mortality reduction, whereas neither can be said for rivaroxaban.

So what are we to do? Should we use the indirect comparisons put forth by Lip et al. (5) since that provides the only comparative data we have? In general, the authors appear to be saying that there are more similarities between these agents than differences, as has also been previously noted by Mega (10). However, because of the statistical limitations of such comparisons, although of some interest, we feel the differences they report on some endpoints are not robust enough to be relied upon for the clinical care of patients. Instead, we would turn to direct evidence from trials and the indications put forth by the FDA to select the appropriate agent, at the dose tested, for use in the patient population studied within the trial.

⁎Editorials published in the Journal of the American College of Cardiology reflect the views of the authors and do not necessarily represent the views of JACC or the American College of Cardiology.