Head-to-head or indirect comparisons of the novel oral anticoagulants in atrial fibrillation: Whatâs next?

doi:10.1160/TH12-07-0463 Thromb Haemost 2012; 108: 407–409 Atrial fibrillation (AF) as an age-related disease represents an increasing challenge to Western societies due to its increasing prevalence, and the associated burden of stroke and thromboembolism associated with this common arrhythmia. Patients with non-valvular AF (NVAF) are exposed to an increased risk of ischaemic stroke and systemic embolism, as well as mortality – which is reduced by antithrombotic therapy (1), (2). However, severe bleeding complications, including intracranial haemorrhage have limited the optimal use of vitamin K antagonists (VKAs), particularly in the elderly (3–5). This is despite the net clinical benefit balancing stroke against serious bleeding, largely being in favour of oral anticoagulation (rather than non-anticoagulation) (6). The VKAs require frequent laboratory guided dose adjustments to increase the time in therapeutic range (TTR) of an international normalised ratio (INR) between 2.0–3.0, to achieve optimal thromboprophylaxis (7). The considerable variability of INR values may be genetically determined as VKA metabolism depends on polymorphic cytochrome P450 and vitamin K-epoxide reductase systems; VKA pharmacokinetics and -dynamics are further complicated by multiple interactions with drugs and food, and by intercurrent diseases. As these ramifications are aggravated in elderly patients with multiple disease (multimorbidity) often receiving 5+ drugs (polypharmacy), they may be undertreated with VKA and receive less effective therapies such as aspirin or no prophylaxis (5). Oral direct factor Xa inhibitors and oral direct thrombin inhibitors specifically targeting only one coagulation factor have been developed (8). Their efficacy and safety have been investigated for prevention of ischaemic stroke and systemic embolism in patients with NVAF with dabigatran in the RE-LY trial (150 mg and 110 mg bid) (9), (10), rivaroxaban in the ROCKET AF trial (20 mg od) (11), and apixaban in the ARISTOTLE trial (5 mg bid) (12) versus INR-adjusted warfarin. All trials showed non-inferiority or superiority for the efficacy outcome of combined ischaemic stroke and systemic embolism. Rates of major bleeding complications were similar or even reduced in comparison with warfarin for all three novel oral anticoagulants (NOAC). The current scientific discussion aims at defining differentiating efficacy/safety profiles for the NOACs in patients suffering from NVAF. A head to head comparison of these NOACs would require a large number of patients, thus be expensive and therefore unlikely to be performed. In this issue of Thrombosis and Haemostasis, Mantha and Ansell (13) indirectly compared the three large studies based on analytical state-of-the-art techniques. They used the appropriate statistical tools for this indirect comparison, also termed “network meta-analysis” (NMA) (14, 15). As their key message, Mantha and Ansell show that dabigatran 150 mg bid seems to be more effective than dabigatran 110 mg bid and rivaroxaban 20 mg od to prevent ischaemic stroke and systemic embolism in patients with NVAF. Apixaban 5 mg bid (13, 19-21) and dabigatran 110 mg bid (20, 21) seem to be afflicted with less major complications compared to dabigatran 150 mg bid and rivaroxaban 20 mg od (13, 19-21). Indeed, dabigatran 110 mg bid was associated with less intracerebral bleeding compared to rivaroxaban 20 mg od. In this NMA, some of the following confounding factors may limit the validity of the results, as follows: (i) the open (RE-LY) or double blind design (ROCKET AF, ARISTOTLE) with warfarin, (ii) differences in reporting age as mean and standard deviation versus median and range, (iii) the time in therapeutic range (TTR) of INR values, (iv) numbers of risk factors according the CHADS2 score at entry, including history of previous stroke (16), (iv) concomitant use of aspirin, and (v) creatinine clearance. Indeed, there are ‘danger periods’ for increased bleeding following initiation of any anticoagulant, particularly relevant for inception cohorts (17) . Therefore, the paper Mantha and Ansell concludes that further to their NMA, a headto-head randomised controlled trial comparing the NOACs in patients with AF is still mandatory to assess their relative efficacy and safety (13). As a historical note on the method used in the analysis by Mantha and Ansell, the first NMA was published in 2003 (18); in 2007, five NMAs were published and even 43 NMAs in the first six months of 2012. It is noteworthy that a total of 4 NMAs on the efficacy and safety of NOACs in patients with NVAF have now been published within only three months (13, 19–21). All NMAs included the same three studies employing identical or similar statistical methods. NMAs typically encounter a number of statistical problems. Heterogeneity is a relevant aspect when several studies were performed for one indication. For each clinical endpoint (such as thromboembolism, 407 © Schattauer 2012