Introduction

Atrial fibrillation (AF) is the most common sustained cardiac arrhythmia with a rapidly increasing prevalence worldwide. Since patients with AF have a 4- to 5-fold increase in the risk of stroke, clinicians should identify individuals who may benefit from prophylaxis with oral anticoagulation (OA). Although vitamin K antagonists (VKAs) have been the pillar of stroke and thromboembolism prevention in AF, their use is associated with a number of clinical limitations. In this setting, a generation of new oral anticoagulants (NOACs) have emerged as an alternative to the use of VKAs. This manuscript reviews the evidence that supports the efficacy and safety of NOACs and will help clinicians to make evidence-based decisions when selecting the most appropriate form of OA for patients with nonvalvular AF. As defined by the American Heart Association/American College of Cardiology/Heart Rhythm Society, nonvalvular AF is defined as AF that occurs in the absence of rheumatic mitral stenosis, a mechanical or bioprosthetic heart valve, or mitral valve repair [1]. Acronyms and names of clinical trials discussed in this manuscript can be found in Table 1.

Table 1 Acronyms of clinical trials discussed in this review
Full size table

Oral anticoagulants

Vitamin K antagonists

Warfarin, the most commonly used VKA, works by preventing the carboxylation of coagulation proteins for factors II, VII, IX and X. The efficacy of warfarin to prevent stroke and systemic embolism (SE) in patients with valvular and nonvalvular AF has been well documented in clinical trials and meta-analyses [2] (Table 2).

Table 2 Main pharmacologic and clinical properties of oral anticoagulants used in atrial fibrillation
Full size table

Direct thrombin inhibitors

Dabigatran

The efficacy and safety of dabigatran (110 and 150 mg twice daily [bid]) relative to warfarin were assessed in the RE-LY trial [3], a randomized, open, blinded, endpoint study. While both doses of dabigatran were noinferior to warfarin in decreasing the risk of stroke and SE compared with VKA, only high-dose dabigatran was statistically superior to VKA (150 mg, RR 0.66; 95 % confidence interval [CI], 0.53–0.82; p < 0.001; 110 mg, RR 0.91; 95 % CI, 0.74–1.11, p = 0.34). The risk of intracranial hemorrhage (ICH) was higher in the VKA group compared to either dose of dabigatran. In the extension registry known as RELY-ABLE, the rate of major bleeding was higher in patients taking higher-dose dabigatran compared with lower-dose dabigatran (Tables 2, 3).

Table 3 Clinically relevant results from the main phase III trials and subanalyses
Full size table

Oral factor Xa inhibitors

Three oral factor Xa inhibitors (rivaroxaban, apixaban, edoxaban) have been compared to warfarin in large, multinational, randomized, double-blind, double-dummy, event-driven, non-inferiority trials. In addition, apixaban was compared to aspirin in a 4th trial.

Rivaroxaban

The efficacy and safety of rivaroxaban (20 mg once daily [qd] or 15 mg qd in patients with creatinine clearance [CrCl] 30–49 mL/min) as compared to warfarin were assessed in the ROCKET-AF trial [4]. Rivaroxaban was noninferior to warfarin for the prevention of stroke or SE (RR, 0.88; 95 %, CI, 0.74–1.03; p < 0.001 for noninferiority and p = 0.12 for superiority). The risk of ICH was lower in the rivaroxaban arm compared with the warfarin group (Tables 2, 3).

Apixaban

The efficacy and safety of apixaban 5 mg bid (or 2.5 bid in patients with ≥2 of the following characteristics: age ≥80 years, body weight ≤60 kg, serum creatinine ≥1.5 mg/dL) were assessed in the ARISTOTLE trial [5]. Apixaban was not only superior to VKA in preventing strokes or SE (RR 0.79; 95 % CI, 0.66–0.95; p < 0.001 for noninferiority and p = 0.01 for superiority), it also reduced the risk of major bleeding compared with warfarin (Tables 2, 3). The AVERROES trial [6] was a double-blind study that randomized patients with AF who either tried but failed VKA treatment or who were considered unsuitable for VKA therapy to apixaban 5 mg bid or aspirin 81–324 mg qd. The risk of stroke or SE was reduced by 55 % (RR 0.45; 95 % CI, 0.32–0.62; p < 0.001)) in the apixaban group without increasing the risk of major bleeding compared to the group maintained on aspirin monotherapy.

Edoxaban

The efficacy and safety of edoxaban 30 mg qd and edoxaban 60 mg qd were tested in the ENGAGE AF-TIMI 48 trial [7]. Importantly, the edoxaban dose was reduced 50 % in selected patients (CrCl 30–50 mL/min; body weight ≤60 kg; use of the potent permeability glycoprotein (P-gp) inhibitors verapamil, quinidine, and dronedarone), and dose modifications were permitted after randomization depending on whether patients developed one of the above criteria or changed one of the above concomitant medications. Both dose-regimens of edoxaban were noninferior to warfarin (median time-in-therapeutic range 68.4 %) in decreasing the risk of stroke or SE. While there was a trend favoring higher-dose superiority in terms of stroke and SE prevention when compared with warfarin while on-treatment (RR 0.79; 97.5 % CI, 0.63–0.99; p < 0.001 for noninferiority, p = 0.02 for superiority), there was an unfavorable trend with lower-dose edoxaban (RR 1.07; 97.5 % CI, 0.87–1.31; p = 0.005 for noninferiority, p = 0.44 for superiority). Major bleeding, intracranial and fatal bleeding, cardiovascular mortality, and net clinical outcomes (composite of cardiovascular events, death from any cause, or bleeding) were significantly lower with both doses of edoxaban compared with warfarin (Table 3).

Meta-analysis comparing the safety and efficacy of NOACs with warfarin

A comprehensive meta-analysis [8] of the main phase III warfarin-controlled clinical trials aforementioned showed that NOACs significantly decreased strokes and SE by 19 % (RR 0.81; 95 % CI, 0.73–0.91; p < 0.0001), hemorrhagic stroke by 51 % (RR 0.49; 95 % CI, 0.38–0.64; p < 0.0001) and mortality by 10 % (RR 0.90; 95 % CI, 0.85–0.95; p = 0.0003) as compared with warfarin. Interestingly, this benefit was driven predominantly by a halving in the hemorrhagic stroke rate. NOACs also markedly reduced ICH compared to warfarin (RR 0.48; 95 % CI, 0.39–0.59; p < 0.0001), while there was a borderline statistical reduction in major bleeding compared to the use of warfarin (RR 0.86; 95 % CI, 0.73–1.00; p = 0.06), with significant reductions in bleeding observed with apixaban and edoxaban.

Special clinical scenarios

Elderly patients

Elderly patients are usually undertreated with OACs due to concerns about increased bleeding. However, evidence shows in geriatric populations that the benefit-risk balance still favors VKA for the vast majority of elderly patients.

Although both doses of dabigatran are effective in preventing stroke and SE in elderly individuals, a post hoc analysis of elderly patients in the RE-LY trial showed that both higher- and lower-dose dabigatran were associated with higher risk of extracranial bleeding in patients ≥75 years (Table 3). In the ROCKET AF, ARISTOTLE and ENGAGE AF-TIMI 48 trials, the efficacy and safety of rivaroxaban, apixaban and edoxaban, respectively, were consistent in elderly and nonelderly individuals (Table 3).

Authors’ recommendation. FXa inhibitors have a good efficacy and safety profile for the prevention of stroke or SE in patients ≥75 years.

Patients at low and high risk of stroke and SE

Since rivaroxaban and edoxaban were tested in trials with a high proportion of patients with CHADS2 scores ≥3, when treating high-risk patients, physicians might prefer one of these two agents. On the other hand, patients with CHADS2 scores ≤1 represent approximately 1/3 of the populations studied in the RE-LY and ARISTOTLE trials (and were excluded from ROCKET AF and ENGAGE AF-TIMI 48), thus providing a much more robust experience with the use of dabigatran and apixaban in low-risk patients (Table 3).

Authors’ recommendations. All NOACs, but in particular rivaroxaban and edoxaban should be considered in patients with CHADS2 scores ≥3. Dabigatran and apixaban should be considered in individuals with CHADS2 scores ≤1.

Drug interactions

All OACs have significant drug interactions, although the NOACs have far fewer than VKAs. The most important drug–drug interactions with warfarin are summarized in Table 2. While all NOACs are substrates of the P-gp, only rivaroxaban and apixaban are substantially metabolized by the cytochrome (CYP) 3A4 system. Inducers of the P-gp (e.g. carbamazepine) decrease the plasma concentration of NOACs, while P-gp inhibitors (e.g., verapamil) have the opposite effect. For rivaroxaban and apixaban, inducers of the CYP3A4 (e.g., phenytoin) will decrease their plasma concentration, while CYP3A4 inhibitors (e.g., clarithromycin) will increase their plasma concentrations (Table 2). Importantly, the ENGAGE AF-TIMI 48 trial [7] was the only phase III study of a NOAC that reduced the dose of the study drug (edoxaban) before randomization or during the course of the study in patients taking potent P-gp inhibitors.

Author’s recommendation. In patients who must continue on a drug with a potent effect on the CYP3A4 system, clinicians should carefully review the most recent prescribing information for rivaroxaban and apixaban, or select a NOAC with little-no CYP metabolism (dabigatran, edoxaban). In patients on drugs that are potent inhibitors of the P-gp transporter and/or CYP3A4 system, edoxaban should be considered as it has no CYP3A4 metabolism and the ENGAGE AF-TIMI 48 trial [7] tested dose reductions of edoxaban in patients taking strong P-gp inhibitors before and after randomization.

Concomitant antiplatelet therapy

In the four phase III trials, the use of aspirin was associated with higher absolute rates of bleeding in both the NOAC and warfarin groups [3–5, 7]. However the benefits in prevention of stroke or SE were comparable irrespective of the use of aspirin in the four trials, except in the ENGAGE AF-TIMI 48 trial [7], wherein the relative efficacy of lower-dose edoxaban vs. warfarin was significantly improved in patients taking aspirin (interaction p = 0.02). Importantly, patients on dual antiplatelet therapy (DAPT) were excluded from the ARISTOTLE [5], ROCKET-AF [4] and ENGAGE AF-TIMI 48 [7] trials. A small number (n = 812) received DAPT in combination with dabigatran in the RE-LY trial (Table 3).

Authors’ recommendations. Current AF guidelines [1, 9, 10] recommend monotherapy with anticoagulation and omission of antiplatelet therapy in patients with stable CAD. In patients treated with DAPT, warfarin is preferred over NOACs. Lower-dose dabigatran or edoxaban may also be considered in patients on aspirin monotherapy.

Patients at high risk of acute coronary syndromes (ACS)

Analyses of DTIs compared to warfarin have provide conflicting information regarding whether DTIs such as dabigatran are as effective as warfarin to prevent myocardial infarction (MI). The initial data published in RE-LY reported a 35 and 38 % increased rate of MI with dabigatran 110 and 150 mg, respectively, compared to warfarin [3], while subsequent update from the RE-LY trial showed no significant increase in the risk of MI. Meanwhile, a large-scale Danish cohort found a 3-fold increase in the rate of MI in patients who switched from warfarin to dabigatran compared to patients who stayed on warfarin. Furthermore, a meta-analysis that included 7 trials of dabigatran (in patients with venous thromboembolism or AF) also found significant association between the risk of MI and the use of dabigatran compared to control groups (OR 1.27; 95 % CI, 1.00–1.61; p = 0.05). In the other three trials with FXa inhibitors, no significant difference in rates of MI was observed (Table 3).

Authors’ recommendations. In patients with increased risk of an ACS (e.g., prior), a FXa inhibitor is preferred over dabigatran.

Gastrointestinal adverse events

In the RE-LY trial [3], dyspepsia was significantly more common in both groups of patients taking dabigatran compared to the warfarin group, which led to high rates of drug discontinuation. This adverse effect could be related to the use of tartaric acid in dabigatran capsules to lower GI pH and improve bioavailability.

Data from their respective trials and a meta-analysis of the four warfarin-controlled trials concluded that patients treated with dabigatran, rivaroxaban, and higher-dose edoxaban had significantly higher rates of GI bleeding than patients on warfarin. In contrast, corresponding analyses with apixaban and lower-dose edoxaban showed a lower rate of GI bleeding compared with warfarin (Table 3).

Authors’ recommendation. In patients with dyspepsia or who developed dyspepsia on dabigatran, a FXa inhibitor is preferred. In patients with a history of recent GI bleeding, apixaban or lower-dose edoxaban appear to be safer alternatives than warfarin or the other NOACs.

Renal impairment

Since all oral anticoagulants, including warfarin, are renally excreted, dose adjustments and close monitoring is prudent in individuals with renal dysfunction. In the case of warfarin, patients with CrCl < 60 mL/min might require a dose reduction and close INR monitoring targeting an INR of 2.5. While there is strong evidence that the benefits of the use of warfarin outweighs the bleeding risk in non-dialysis patients with chronic kidney disease (CKD), the proof is conflicting in patients with CKD maintained on dialysis.

Secondary efficacy analyses of dabigatran, rivaroxaban and apixaban in patients with renal dysfunction demonstrated no evidence of heterogeneity in the rates of stroke or SE in subgroups defined by renal function in the RELY, ROCKET-AF and ARISTOTLE trials. Similarly, no treatment-subgroup interaction in major bleeding was observed in the RELY and ROCKET-AF trials. However, in the ARISTOTLE trial, the reduction in major bleeding in the apixaban group compared with warfarin was greater in patients with a CrCl of ≤50 mL/min (Cockroft-Gault, p value for interaction = 0.005). The lower rate (25 %) of renal elimination of apixaban compared to the other NOACs may partially explain these findings (Table 3). Based on pharmacokinetic and pharmacodynamics modeling of dabigatran, the U.S. FDA (United States Food and Drug Administration), but not the EMA (European Medicines Agency), approved the use of dabigatran 75 mg bid in subjects with a CrCl of 15–30 mL/min.

None of the NOACs have been extensively studied in patients with end-stage renal disease (ESRD). However, based on pharmacokinetic studies, the U.S. FDA changed the prescribing information for apixaban in February 2014, recommending apixaban 5 mg bid (or 2.5 mg bid in patients ≥80 years or body weight ≤60 kg) in patients with AF maintained with hemodialysis. Similar analyses with edoxaban showed that its plasma concentrations are not significantly changed with hemodialysis suggesting that further dose adjustment of edoxaban in patients with CrCl < 30 mL/min may not be needed in these patients, although additional prospective data are needed.

Authors’ recommendation. In individuals with ESRD maintained on HD, apixaban or warfarin are recommended. Dabigatran 75 mg bid is a reasonable choice in patients with CrCl 15–30 mL/min, as are apixaban 5 mg bid (or 2.5 mg bid in patients with a body weight ≤ 60 kg and/or age ≥ 80 years), rivaroxaban 15 mg qd, and edoxaban 30 mg qd.

Hepatic dysfunction

Since all OACs undergo hepatic metabolism, liver function tests should be performed prior initiating anticoagulation and at least annually thereafter. In patients with mild-to-moderate liver disease with preserved synthetic function who are receiving a VKA, close monitoring for bleeding and frequent (at least once every 4 weeks) INR monitoring is recommended. Physicians may consider dabigatran (no CYP metabolism) or edoxaban (<4 % CYP metabolism) as reasonable alternatives to warfarin (Table 2) in patients with mild hepatic insufficiency or in combination with strong CYP inhibitors (see also next section below).

Authors’ recommendation. In patients with liver impairment, warfarin remains the first line option, although dabigatran and edoxaban also may be considered in patients with mild insufficiency.

Clinical circumstances in which NOACs are not recommended

Since NOACs have substantially shorter half-lives (5–14 h) compared with warfarin (20–60 h), strict adherence to therapy is essential to guarantee adequate anticoagulation effect. Therefore, clinicians may prefer to use warfarin in patients whom daily compliance is not ensured.

Dabigatran is the only NOAC that has been tested in patients with mechanical valve replacement. However, higher-dose dabigatran regimens (dose range between 150 and 300 mg bid, adjusted based on CrCl and trough concentration) were less effective and caused more bleeding than warfarin. The reasons for these findings are unclear, but some have speculated that inhibition of multiple clotting proteins including those affecting the contact pathway are needed to prevent thromboembolism in patients with intracardiac mechanical prostheses, and that high levels of thrombin inhibition with dabigatran could not achieve a similar benefit without increasing bleeding. Thus, VKAs remain the preferred anticoagulants for patients with mechanical heart valves. However, it should be noted that patients with valvular heart disease (other than rheumatic mitral stenosis), bioprosthetic cardiac valves and/or valve repair were included in several of the large NOAC trials without evidence of increased risk of thromboembolism (supplemental Table 1).

Although NOACs have been approved for adults, data in pediatric populations are limited. Currently, ongoing clinical trials are testing the use, pharmacokinetics, safety and tolerability of dabigatran (NCT01083732, NCT01773174, NCT01895777), rivaroxaban (NCT01684423, NCT01145859) and apixaban (NCT01707394, NCT01195727) in children and adolescents. Similarly, the safety of NOACs in pregnant women has not been well studied. While apixaban is allocated to the B group of the U.S. FDA pregnancy category, rivaroxaban and dabigatran are allocated to the C group and warfarin to the D (for women with mechanical heart valves) and X (for other pregnant populations) groups.

Authors’ recommendations. NOACs should be avoided in patients with poor compliance with medications, mechanical heart valves, children, adolescent and in women during pregnancy.

A summary of the author recommendations can be found in Fig. 1.

Fig. 1
figure 1

Clinical scenarios that clinicians should consider when selecting an oral anticoagulant for patients with nonvalvular atrial fibrillation. CAD coronary artery disease, CHADS 2 congestive heart failure, hypertension, age ≥75 years, diabetes mellitus, prior stroke or transient ischemic attack or thromboembolism, CYP cytochrome, DAPT dual antiplatelet therapy, GI gastrointestinal, HD hemodialysis, NOAC new oral anticoagulant; P-gp P-glycoprotein; U.S. FDA United States Food and Drug Administration, VKA vitamin-K antagonist

Full size image

Conclusions

Since head-to-head controlled comparison of NOACs are currently not available and large scale randomized trials are unlikely to occur, clinicians should consider available data from the individual warfarin-controlled trials, observational studies and knowledge of the pharmacokinetic profiles of each NOAC to make evidence-based decisions regarding selection of the most appropriate OAC in patients with atrial fibrillation.