Principal Findings:

A total of 21,105 patients were randomized, 7,035 to high-dose edoxaban, 7,034 to low-dose edoxaban, and 7,036 to warfarin. Baseline characteristics were fairly similar between the three arms. The mean CHADS₂ score was 2.8, with the majority having a score of ≤3 (77%). Nearly 28% had a history of prior cardiovascular disease. AF was paroxysmal in 25% of patients, and permanent in 51%. Target therapeutic range was 68.4% in the warfarin arm.

Noninferiority was demonstrated for the primary endpoint of stroke or systemic embolic event in the modified intent-to-treat population between warfarin and high-dose edoxaban arms (annualized rate 1.5% vs. 1.18%, hazard ratio [HR] 0.79, 95% confidence interval [CI] 0.63-0.99, p for noninferiority < 0.001), and warfarin and low-dose edoxaban (1.5% vs. 1.61%, HR 1.07, 95% CI 0.87-1.31, p for noninferiority = 0.005). In the intent-to-treat population, superiority testing was performed with borderline results (p = 0.08, p = 0.10, respectively). On subgroup analysis, patients who had never received warfarin had a significant reduction with high-dose edoxaban compared with warfarin (p < 0.05 for interaction). All strokes were similar (annualized rate 1.69% vs. 1.49% vs. 1.91%, p < 0.05), but hemorrhagic strokes were lower (annualized rate 0.47% vs. 0.26% vs. 0.16%, p < 0.001). Major adverse cardiac event rates were lower with high-dose edoxaban (annualized rate 4.98% vs. 4.41% vs. 4.90%; p = 0.01, p = 0.69, respectively).

The primary safety outcome of International Society on Thrombosis and Haemostasis (ISTH) major bleeding was higher in the warfarin arm (annualized rate 3.43% vs. 2.75% vs. 1.61%, p < 0.001 for both comparisons). Fatal bleeding (annualized rate 0.38% vs. 0.21% vs. 0.13%, p = 0.006, p < 0.001, respectively) and intracranial hemorrhage (0.85% vs. 0.47% vs. 0.26%, p < 0.001 for both) were also lower. Gastrointestinal (GI) bleeding was higher in the high-dose edoxaban arm, but lower in the low-dose edoxaban arm (1.23% vs. 1.51% vs. 0.82%, p < 0.05). Other adverse events including elevation of liver enzymes were similar.

On subgroup analysis, patients with paroxysmal AF overall had lower rates of stroke/systemic embolization compared with patients with persistent or permanent AF (HR 0.77, 95% CI 0.66-0.91). Mortality was also lower. However, bleeding risk was similar between the three subtypes. In all three subtypes, results for efficacy and safety of edoxaban (both doses) vs. warfarin were comparable to those noted in the overall population. In the subgroup of patients on amiodarone, edoxaban levels tended to be higher (due to P-glycoprotein inhibition). The primary efficacy endpoint was significantly lower in patients on low-dose edoxaban compared with warfarin in amiodarone-treated patients (HR 0.60, p for interaction = 0.009). No such interaction was observed in the high-dose edoxaban group. Major bleeding rates were also similar.

Genetic analysis: 14,348 patients were included in the genetic analysis, of which 4,833 were on warfarin. Patients were genotyped for polymorphisms in CYP2C9 (*2 and *3 alleles) and VKORC1 (–1639G→A), and classified as normal responders (61.7%), sensitive responders (35.4%), and highly sensitive responders (2.9%) to warfarin. Major or clinically relevant bleeding was lowest in normal responders and highest in highly sensitive responders (4.6% vs. 5.8% vs. 14.1%; HRs = 1.26 and 3.21, p < 0.05, respectively). In the edoxaban arm, there was no effect of genotype on risk of bleeding. Compared with warfarin, the risk of major or clinically relevant bleeding within 90 days was lowest in the low-dose edoxaban arm, but also reduced in the high-dose arm.

In patients in whom the dose of edoxaban was reduced for clinical reasons (creatinine clearance 30-50 ml/min, bodyweight 60 kg or less, or concomitant medication with potent P-glycoprotein interaction), antifactor Xa levels were reduced in a commensurate fashion, with no change in efficacy, but a greater reduction in bleeding compared with warfarin (major bleeding: higher dose p interaction = 0.02, lower dose p interaction = 0.002).

Efficacy of edoxaban based on renal function: There appeared to be lower efficacy of edoxaban 60 mg in patients with higher levels of creatinine clearance (CrCl) >95 ml/min (HR for edoxaban vs. placebo for primary endpoint: 1.36, 95% CI 0.88-2.10, p for interaction = 0.08). Bleeding was lower with edoxaban for all levels of CrCl.

Efficacy in patients at high risk for falls (n = 900, 4.3%): These patients were older, and had a higher prevalence of comorbidities including prior stroke/transient ischemic attack, diabetes, and coronary artery disease. These patients had a higher risk of bone fractures, major and life-threatening bleeding, and all-cause death; ischemic events including stroke/systemic embolic events were similar. The primary endpoint of stroke/systemic embolic event was similar between the high-dose edoxaban and warfarin arms (2.81% vs. 2.85%, p for interaction = 0.74). Major bleeding was also similar (5.43% vs. 5.55%, p for interaction = 0.45).

Interaction with body mass index (BMI): In an adjusted analysis, higher BMI (continuous, per 5 kg/m² increase) was significantly and independently associated with lower risks of stroke/systemic embolic event (hazard ratio [HR] 0.88, p = 0.0001), ischemic stroke/systemic embolic event (HR 0.87, p < 0.0001), and death (HR 0.91, p < 0.0001), but with increased risks of major bleeding (HR 1.06, p = 0.025) and major or clinically relevant nonmajor bleeding (HR 1.05, p = 0.0007). Trough edoxaban concentration and anti-Factor Xa activity were similar across BMI groups >18.5 kg/m² regardless of whether a dose adjustment in edoxaban was performed, while time in therapeutic range for warfarin improved significantly as BMI increased (p < 0.0001). The efficacy and safety profiles of edoxaban relative to warfarin were similar across BMI groups.

Subanalysis among Asian patients (n = 2,909): Asians on average weighed about 20 kg (44 lbs) less than non-Asians, and had a higher prevalence of chronic kidney disease requiring dose adjustments. The trough edoxaban concentration and anti-Factor Xa activity were 20–25% lower for Asians compared with non-Asians, particularly when the lower dose was used. Even after adjusting for lower dose, levels were lower (possibly due to younger age and lower use of concomitant amiodarone). Despite a lower time in therapeutic range and a greater percentage of time with a subtherapeutic INR, Asian patients treated with warfarin had a higher adjusted risk of intracranial hemorrhage (ICH) (adjusted HR 1.71 [1.05–2.77], p = 0.03), and clinically relevant nonmajor bleeding (adjusted HR 1.56 [1.34–1.82], p < 0.001) compared with non-Asians. Compared to warfarin, higher-dose edoxaban significantly reduced ICH, while preserving the efficacy of stroke prevention in both Asians and non-Asians.

Low-dose vs. high-dose edoxaban: Composite of stroke/systemic embolism, major bleeding, and death was less frequent with low-dose edoxaban compared with high-dose edoxaban (7.3% vs. 8.0%, HR 0.90, 95% CI 0.84-0.98, p = 0.014). However, stroke/systemic embolism was higher (2.0% vs. 1.6%, HR 1.31, 95% CI 1.12-1.52, p < 0.001), including a higher risk of stroke (1.9% vs. 1.5%, p = 0.001); MI (0.9% vs. 0.7%, p = 0.03). Major bleeding was lower (1.8% vs. 2.9%, p < 0.0001), including intracranial hemorrhage (0.3% vs. 0.4%, p = 0.035) and GI major bleeding (0.9% vs. 1.5%, p < 0.0001).

Dose reduction among patients ≥80 years of age (n = 2,966): Dose reduction was warranted among patients with creatinine clearance of ≤50 mL/min, body weight ≤60 kg, or taking strong P-glycoprotein (permeability glycoprotein) inhibitors at the time of randomization or during the treatment phase. 1,700 patients met dose reduction criteria: edoxaban 30 mg (n = 578), edoxaban 60 mg (n = 560), or warfarin (n = 562). Patients ≥80 years who received edoxaban 30 mg, regardless of dose-reduction criteria, had a median trough concentration that approximated that of patients <80 years who received edoxaban 60 mg. In patients ≥80 years receiving edoxaban 60 mg, the median (IQR) trough edoxaban concentration was 35% higher than in patients <80 years receiving edoxaban 60 mg. Among patients ≥80 years, irrespective of dose-reduction criteria, the net clinical outcome was 22% lower (HR 0.78, 95% CI 0.68-0.91, p = 0.001) with edoxaban 30 mg compared with warfarin. Rates of stroke or systemic embolic event were similar between groups (HR 0.93, 95% CI 0.69-1.27, p = 0.66), but major bleeding events overall were lower with edoxaban 30 mg (HR 0.59, 95% CI 0.45-0.77, p < 0.001). All-cause death was 17% lower (HR 0.83, 95% CI 0.70-1.00, p = 0.046) with edoxaban 30 mg vs. warfarin in patients ≥80 years. Although rates of ischemic stroke were not different between treatments, intracranial hemorrhage was 51% lower (HR 0.49, 95% CI 0.28-0.87, p = 0.02) in patients randomized to edoxaban 30 mg.