Principal Findings:

A total of 18,201 patients were randomized, 9,120 to apixaban and 9,081 to warfarin. Preliminary results indicate that apixaban is noninferior to warfarin for the primary endpoint of stroke or systemic embolism. Key secondary endpoints for efficacy and ISTH (International Society on Thrombosis and Haemostasis) major bleeding demonstrated superiority, as compared with warfarin.

Baseline characteristics were fairly similar between the two arms. About a quarter of the patients were recruited in North America. Hypertension was noted in 88% of the patients, prior myocardial infarction (MI) in 14%, diabetes in 25%, congestive heart failure (CHF) in 35%, 19% had prior stroke, and about 31% of the patients were at least 75 years of age. The mean CHADS₂ score was 2.1, with about 34% with a score of ≤1. About 57% had been on a vitamin K antagonist prior to randomization. In the patients on warfarin, the median time in therapeutic range was 66%.

The primary efficacy outcome of stroke or systemic embolism was significantly reduced in the apixaban arm as compared with warfarin (1.27%/year vs. 1.6%/year, hazard ratio 0.79, 95% confidence interval 0.66-0.95, p < 0.001 for noninferiority, p = 0.01 for superiority). All strokes (1.19%/year vs. 1.51%/year, p = 0.01), hemorrhagic strokes (0.24%/year vs. 0.47%/year, p < 0.001), and all-cause mortality (3.52%/year vs. 3.94%/year, p = 0.047) were similarly reduced in the apixaban arm. The rates of systemic embolism (0.09%/year vs. 0.10%/year, p = 0.7), ischemic strokes (0.97%/year vs. 1.05%/year, p = 0.42), and MI (0.53%/year vs. 0.61%/year, p = 0.37) were similar between the two arms. No evidence of heterogeneity for the primary endpoint was noted for any of the subgroups tested.

The primary safety outcome of ISTH major bleeding was significantly reduced in the apixaban arm (2.13%/year vs. 3.09%/year, p < 0.001). Other forms of bleeding such as intracranial bleeding (0.33%/year vs. 0.80%/year, p < 0.001), major or clinically relevant bleeding (4.07%/year vs. 6.01%/year, p < 0.001), GUSTO (Global Use of Strategies to Open Occluded Coronary Arteries) severe bleeding (0.52%/year vs. 1.13%/year, p < 0.001), TIMI (Thrombolysis In Myocardial Infarction) major bleeding (0.96%/year vs. 1.69%/year, p < 0.001), and any bleeding (18.1%/year vs. 25.8%/year, p < 0.001) were all reduced with apixaban, as compared with warfarin. On subgroup analysis, there were two significant interactions: patients with diabetes and normal renal function did not demonstrate a significant difference in ISTH major bleeding with apixaban, as compared with warfarin.

Drug discontinuation was lower with apixaban: 25.3% vs. 27.5% (p = 0.001). Rates of liver function abnormalities were similar between the two arms.

Results based on renal function: Stroke, systemic embolism, and bleeding rates all trended higher in both arms as renal function declined. The efficacy of apixaban as compared with warfarin appeared to be similar irrespective of renal function. However, a significant reduction in major bleeding was noted in patients with impaired renal function with apixaban as compared with warfarin. Annual bleeding rates were 1.46% vs. 1.84% for glomerular filtration rate (GFR) ≥80, 2.45% vs. 3.21% for GFR 50-80, and 3.21% vs. 6.44% for GFR ≤50, respectively (p for interaction = 0.03).

In patients with worsening of renal function over time during 12 months, apixaban, compared  with warfarin, had numerically lower rates of stroke or systemic embolism (HR 0.80, 95% CI 0.51-1.24; p = 0.86), ischemic or unspecified stroke (HR 0.88, 95% CI 0.52-1.48, p = 0.94), and major bleeding (HR 0.76, 95% CI 0.54-1.07, p = 0.73).

Individual time in therapeutic range (TTR) was calculated for each warfarin-treated patient by the Rosendaal method, excluding the first 7 days after randomization and treatment interruptions until 2 days after the last dose of warfarin. The center’s TTR was calculated as the median of the individual TTRs during the whole study in its warfarin patients. The center’s TTR was assigned as a proxy for centers’ quality of INR control for all of its patients. Centers in the lowest quartile (TTR <58: 32.6%) had more patients with CHADS₂ scores of 3-6 versus those in the other quartiles (58-65.7: 31.1%, 65.7-72.2: 30%, ≥72.2: 27%; p < 0.0001). They also had more patients with prior strokes, CHF, and were less likely to be on statin therapy. There was no difference for the primary efficacy outcomes between the four quartiles of TTR (p for interaction = 0.29). The primary safety endpoint was also similar (p for interaction). Major and clinically relevant bleeding was highest in the subgroup with TTR ≥72.2/highest quartile (p for interaction = 0.005). These results suggest that the benefits of apixaban over warfarin in preventing strokes and reducing bleeding were independent of the centers' TTR (i.e., quality of INR control).

Outcomes in patients with valvular heart disease: The trial included 4,808 patients with valvular heart disease. The benefits of apixaban compared with warfarin in reducing stroke or systemic embolism were consistent in patients with (HR 0.70, 95% CI 0.51-0.97) and without (HR 0.84, 95% CI 0.67-1.04) valvular heart disease (interaction p value = 0.38). Other efficacy outcomes, including stroke, ischemic stroke, hemorrhagic stroke, death from any cause, and several composite outcomes were consistently and similarly reduced with apixaban compared with warfarin. Bleeding rates in patients with (HR 0.79, 95% CI 0.61-1.04) and without (HR 0.65, 95% CI 0.55-0.77) valvular heart disease (interaction p value = 0.23) were similar.

Bleeding score: A bleeding score was developed using the ARISTOTLE cohort (ABC-bleeding score: age, biomarkers [GDF-15, high-sensitivity cardiac troponin, and hemoglobin], clinical history [previous bleeding]). This yielded a c-index of 0.68, compared with 0.61 and 0.65 for the HAS-BLED and ORBIT scores, respectively (p < 0.001 for both). This was then validated in the RE-LY cohort. The c-indices for the three scores were 0.71, 0.62, and 0.68, respectively (p < 0.01 for ABC-bleeding vs. the other two).

Methylated arginine derivatives asymmetric and symmetric dimethylarginine (ADMA/SDMA) levels: On multivariable analysis adjusting or established ris factors and treatment, tertile groups of ADMA concentrations were significantly associated with stroke/systemic embolism (p = 0.034) and death (p < 0.0001), whereas tertile groups of SDMA were associated with major bleeding and death (p < 0.001 for both). Neither ADMA nor SDMA predicted differential response to warfarin or apixaban. Disturbances of nitric oxide function may modulate both thrombotic and hemorrhagic risk in anticoagulated patients with AF.

Use of apixaban in patients at extremes of body weight: 10.9% had weight ≤60 kg (≤132 lbs) and 5.4% >120 kg (>264.5 lbs). The treatment effect of apixaban versus warfarin for the efficacy outcomes of stroke/systemic embolism, all-cause death, or myocardial infarction was consistent across the weight spectrum (interaction p value > 0.05). For major bleeding, apixaban had a better safety profile than warfarin in all weight categories and even showed a greater relative risk reduction in patients in the low (≤60 kg; HR 0.55, 95% CI 0.36-0.82) and midrange (>60-120 kg) weight groups (HR 0.71, 95% CI 0.61-0.83, interaction p value = 0.016).

Concomitant NSAID use: Baseline NSAID use (n = 832 [4.8%]); incident NSAID use (n = 2,185 [13.2%]). During a median follow-up of 1.8 years and when excluding those taking NSAID at baseline, incident NSAID use was associated with an increased risk of major bleeding (HR 1.61, 95% CI 1.11-2.33) and clinically relevant nonmajor bleeding (HR 1.70, 95% CI 1.16-2.48), but not gastrointestinal bleeding. No significant interaction was observed between NSAID use and randomized treatment for any outcome.

Dose adjustment: Dose-adjustment criteria in the trial were age ≥80 years, weight ≤60 kg (≤132 lbs), and creatinine ≥1.5 mg/dl. Patients who received the lower dose of apixaban (2.5 mg BID) compared with the standard dose had lower apixaban exposure. In patients with ≥2 dose-adjustment criteria, reductions in D-dimers and prothrombin fragment 1 + 2 (PF1+2) were consistent with those observed in the standard-dose population. Efficacy and safety were similarly preserved.