1.1. Background

Approximately 250,000 Canadians are affected by atrial fibrillation (AF).¹ Stroke is a complication of AF, and Canadians with AF are five times more likely to have a stroke and are twice as likely to die than individuals without AF.^1,2 AF and stroke are more common among the elderly.^3,4

Preventing thromboembolic events such as stroke is an important part of managing AF patients. Antithrombotic strategies for AF patients include anticoagulant drugs, notably the coumadin class of vitamin K antagonists (VKAs), such as warfarin, and antiplatelet agents, such as aspirin. VKAs reduce the risk of stroke in patients with AF by more than 60% when compared with no treatment, and by 30% to 40% when compared with low-dose aspirin.^5,6 However, VKA use is associated with some drawbacks, including a need for laboratory monitoring, an increased risk of bleeding complications, and several food and drug interactions. An improved understanding of how the blood clotting cascade works has led to the development of new oral anticoagulants (NOACs) that exhibit more predictable pharmacokinetics and pharmacodynamics, thereby obviating the need for laboratory monitoring. The NOACs that have either been approved, or are under review by regulators, for the prevention of thromboembolic events in AF patients include dabigatran, a direct thrombin inhibitor, and the direct Factor Xa (FXa) inhibitors, rivaroxaban, apixaban, and edoxaban. Ximelagatran, a direct thrombin inhibitor, was the first NOAC to be approved for use, but was withdrawn from the market in 2006 because of safety concerns.

While dabigatran, apixaban, and rivaroxaban have been demonstrated to be effective in preventing stroke/systemic embolism (SE) in AF patients, the relative effectiveness and associated bleeding risks of these NOACs, both among themselves and in comparison to warfarin, is not clear. Therefore, the aim of this project was to systematically review and analyze the safety and effectiveness of three NOACs – namely dabigatran, rivaroxaban, and apixaban – compared with warfarin in patients with non-valvular AF. In addition, the cost-effectiveness of the NOACs and warfarin was assessed using economic modelling.

1.2. Primary Research Questions

In patients with non-valvular AF:

• What is the clinical effectiveness and safety of new oral anticoagulants compared with warfarin?
• What is the cost-effectiveness of new oral anticoagulants compared to warfarin?
• How do the new oral anticoagulants compare to optimal warfarin therapy when considering the time spent in the time in therapeutic range (TTR)?
• How do the new oral anticoagulants compare to warfarin therapy in specific groups of patients with older age, other medical conditions, or who are taking other drug therapies?
• What are the costs associated with warfarin when patients are stratified according to TTR? How do these compare with estimates for the new oral anticoagulants?
• What is the cost-effectiveness of new oral anticoagulants compared to warfarin when stratified by age and CHADS₂ score (CHADS₂: C= congestive heart failure, H = hypertension, A = older than age 75 years, D = diabetes mellitus, S₂ = prior stroke or history of transient ischemic attack).?

1.3. Methods

Clinical evidence was selected systematically according to a predefined protocol following accepted guidelines.⁷ Trials were selected for inclusion in the systematic review and subsequent analyses if they were carried out in AF patients, included treatment with one or more NOACs and warfarin, and included the following outcomes: all-cause stroke/SE (stroke/SE), major bleeding, intracranial bleeding, major gastrointestinal (GI) bleeding, all-cause mortality, and myocardial infarction (MI).

Next, randomized controlled trials (RCTs) for the NOAC identified in the systematic review were used for a mixed treatment comparison (MTC) that comprised a network meta-analysis for each of the aforementioned outcomes.

Finally, the results of the MTC were used to evaluate the cost-effectiveness of warfarin and the NOACs following standard procedures.⁸ The price of apixaban was assumed because this drug was not approved at the time that this report was prepared.

1.4. Results

1.5. Strengths of Review

• Clinical results have been presented using relative and absolute effect measures.
• Attempts were made to adjust for heterogeneity by formally comparing results from subgroups as versus the alternative — an informal comparison.
• Results presented at the study level and using a Bayesian and frequentist network meta-analysis.
• Robustness of the cost-effectiveness results was demonstrated through extensive sensitivity analyses.

1.6. Limitations of Review

• The validity of indirect comparisons is determined by the extent of clinical and methodological trial similarity, so that differences in study populations, interventions, and outcome definitions are potential sources of incomparability.
• Heterogeneity of patient populations in the included RCTs is a key limitation, and the small number of trials available in the published literature limited the ability to adjust for this heterogeneity during analysis.
• The small number of trials in the published literature limited the modelling that could be considered primarily to fixed effects models and, as a result, study variability could not be fully incorporated.
• Not all outcomes were comprehensively or consistently reported across the three RCTs included in the network meta-analysis.
• Trial design is a potential source of bias. The open-label design RE-LY trial may have led to performance, ascertainment, or adjudication bias, when compared with the ROCKET-AF and ARISTOTLE trials, which both employed a double dummy design.
• ROCKET-AF did not report results using ITT data for many outcomes, whereas RE-LY and ARISTOTLE used ITT. As a result, only analyses for stroke/SE and all-cause mortality use for rivaroxaban use ITT data throughout.
• While outcome definitions for efficacy end points are similar across the included trials, definitions of bleeding events, especially minor bleeding, differed substantially.
• All studies included in the network meta-analysis were multinational, which could impact the generalizability to the Canadian health care system.
• The limited follow-up (maximum median of two yrs) from the three RCTs and the sensitivity of the results to the duration of treatment effect leads to uncertainty around whether the included NOACs will be cost-effective in the long term.

1.7. Conclusions

The results of this report highlight the paucity of clinical evidence available to definitively compare the efficacy and safety of the NOACs as a thromboprophylaxis in AF patients. There were three large, well-designed and properly conducted studies available for analysis, each comparing NOACs and warfarin. However, there was heterogeneity across the three RCTs and, with only three studies handling this, heterogeneity was limited. This should be borne in mind when interpreting the results.

Compared with adjusted-dose warfarin, dabigatran 150 mg and apixaban produced statistically significant reductions in stroke/SE, whereas rivaroxaban and dabigatran 110 mg did not. Apixaban and dabigatran 110 mg were associated with significantly less major bleeding versus adjusted-dose warfarin, whereas there was no association with major bleeding with dabigatran 150 mg and rivaroxaban. All treatments were associated with a significant reduction in intracranial bleeding relative to adjusted-dose warfarin, whereas no treatments were associated with a significant reduction in MI relative to adjusted-dose warfarin. No treatments were associated with a significant reduction in GI bleeding relative to adjusted-dose warfarin, but dabigatran 150 mg and rivaroxaban were associated with a significant increase. Apixaban was associated with a significant reduction in all-cause mortality relative to adjusted-dose warfarin. Even though some differences were statistically significant between treatments for some outcomes, the absolute differences for the NOACs versus warfarin where statistical significance was achieved ranged from two to a maximum of eight fewer events per 1,000 patients.

The results of the subgroup analyses were further limited by a paucity of data, but suggested that there may be subpopulations for which the use of NOACs may be more or less beneficial. In patients where warfarin treatment is well-controlled (TTR ≥ 66%), the use of NOAC may be rather less favourable — the absolute risk increase of a major bleed exceeded the absolute risk reduction of stroke/SE. In elderly patients (≥ 75 years old), the NOACs may be more favourable for preventing thromboembolic events than warfarin, but are associated with a greater risk of major bleeding. Indeed, for dabigatran 150 mg and rivaroxaban, the absolute risk of major bleeding exceeded or approached the absolute risk of stroke/SE, respectively, in this population. For low to moderate risk patients (CHADS₂ < 2), the NOACs were less favourable than warfarin for preventing stroke/SE, but were preferable to warfarin for reducing the absolute risk of major bleeding.

In addition to the limitations imposed by the clinical data, the results of the analysis of cost-effectiveness of the NOACs and warfarin were limited by the uncertainty regarding the pricing of the NOACs. Analysis of the base case suggested that either dabigatran 150 mg or apixaban would be the most cost-effective treatment option. Analysis of subpopulations suggested that dabigatran 150 mg was the most cost-effective treatment option in younger patients (< 80 years old), and in centres where the TTR is < 66%, whereas apixaban was the most cost-effective treatment option in older patients (80 years old) and in centres where the TTR was ≥ 66%. None of NOACs were likely to be considered cost-effective for patients with a previous major stroke. However, the results of the cost-effectiveness analysis were highly sensitive to the patient population under consideration, reinforcing the need for tailoring the treatment of individual patients according to individual characteristics that affect treatment outcomes(includes the degree of control of which warfarin therapy (TTR), age, risk of stroke, and history of thromboembolic events).

In conclusion, the limited number of RCTs available for analysis and heterogeneity in several important features of these RCTs means that there is uncertainty regarding the comparative clinical or cost-effectiveness of the NOACs and warfarin. To fully elucidate the comparative effectiveness of these agents and facilitate the identification of subpopulations within which the NOACs might be more beneficial than warfarin, rigorously conducted comparative RCTs or network meta-regression analyses of patient-level data are required.