Rivaroxaban versus warfarin in nonvalvular atrial fibrillation.

Abstract

The use of warfarin reduces the rate of ischemic stroke in patients with atrial fibrillation but requires frequent monitoring and dose adjustment. Rivaroxaban, an oral factor Xa inhibitor, may provide more consistent and predictable anticoagulation than warfarin.
In a double-blind trial, we randomly assigned 14,264 patients with nonvalvular atrial fibrillation who were at increased risk for stroke to receive either rivaroxaban (at a daily dose of 20 mg) or dose-adjusted warfarin. The per-protocol, as-treated primary analysis was designed to determine whether rivaroxaban was noninferior to warfarin for the primary end point of stroke or systemic embolism.
In the primary analysis, the primary end point occurred in 188 patients in the rivaroxaban group (1.7% per year) and in 241 in the warfarin group (2.2% per year) (hazard ratio in the rivaroxaban group, 0.79; 95% confidence interval [CI], 0.66 to 0.96; P<0.001 for noninferiority). In the intention-to-treat analysis, the primary end point occurred in 269 patients in the rivaroxaban group (2.1% per year) and in 306 patients in the warfarin group (2.4% per year) (hazard ratio, 0.88; 95% CI, 0.74 to 1.03; P<0.001 for noninferiority; P=0.12 for superiority). Major and nonmajor clinically relevant bleeding occurred in 1475 patients in the rivaroxaban group (14.9% per year) and in 1449 in the warfarin group (14.5% per year) (hazard ratio, 1.03; 95% CI, 0.96 to 1.11; P=0.44), with significant reductions in intracranial hemorrhage (0.5% vs. 0.7%, P=0.02) and fatal bleeding (0.2% vs. 0.5%, P=0.003) in the rivaroxaban group.
In patients with atrial fibrillation, rivaroxaban was noninferior to warfarin for the prevention of stroke or systemic embolism. There was no significant between-group difference in the risk of major bleeding, although intracranial and fatal bleeding occurred less frequently in the rivaroxaban group. (Funded by Johnson & Johnson and Bayer; ROCKET AF ClinicalTrials.gov number, NCT00403767.).

Full text

Page 1

n engl j med 365;10 nejm.org september 8, 2011
883
The new england
journal of medicine
established in 1812
september 8, 2011 vol. 365 no. 10
Rivaroxaban versus Warfarin in Nonvalvular Atrial Fibrillation
Manesh R. Patel, M.D., Kenneth W. Mahaffey, M.D., Jyotsna Garg, M.S., Guohua Pan, Ph.D., Daniel E. Singer, M.D.,
Werner Hacke, M.D., Ph.D., Günter Breithardt, M.D., Jonathan L. Halperin, M.D., Graeme J. Hankey, M.D.,
Jonathan P. Piccini, M.D., Richard C. Becker, M.D., Christopher C. Nessel, M.D., John F. Paolini, M.D., Ph.D.,
Scott D. Berkowitz, M.D., Keith A.A. Fox, M.B., Ch.B., Robert M. Califf, M.D.,
and the ROCKET AF Steering Committee, for the ROCKET AF Investigators*
Abstract
From the Duke Clinical Research Institute
(M.R.P., K.W.M., J.G., J.P.P., R.C.B.) and
Duke Translational Medicine Institute
(R.M.C.), Duke University Medical Center,
Durham, NC; Johnson & Johnson Phar-
maceutical Research and Development,
Raritan (G.P., C.C.N.), and Bayer HealthCare
Pharmaceuticals, Montville (J.F.P., S.D.B.)
— both in New Jersey; Massachusetts Gen-
eral Hospital and Harvard Medical School
— both in Boston (D.E.S.); Ruprecht-Karls-
University, Heidelberg (W.H.), and Hospi-
tal of the University of Münster, Münster
(G.B.) — both in Germany; the Cardiovas-
cular Institute, Mount Sinai Medical Cen-
ter, New York (J.L.H.); Royal Perth Hospi-
tal, Perth, WA, Australia (G.J.H.); and the
University of Edinburgh and Royal Infir-
mary of Edinburgh — both in Edinburgh
(K.A.A.F.). Address reprint requests to
Dr. Patel at Duke Clinical Research Institute,
Duke University Medical Center, Rm. 0311
Terrace Level, 2400 Pratt St., Durham, NC
27705, or at manesh.patel@duke.edu.
* A complete listing of the steering com-
mittee members and trial investigators
in the Rivaroxaban Once Daily Oral Di-
rect Factor Xa Inhibition Compared
with Vitamin K Antagonism for Preven-
tion of Stroke and Embolism Trial in
Atrial Fibrillation (ROCKET AF) is pro-
vided in the Supplementary Appendix,
available at NEJM.org.
This article (10.1056/NEJMoa1009638) was
published on August 10, 2011, at NEJM.org.
N Engl J Med 2011;365:883-91.
Copyright © 2011 Massachusetts Medical Society.
Background
The use of warfarin reduces the rate of ischemic stroke in patients with atrial fibril-
lation but requires frequent monitoring and dose adjustment. Rivaroxaban, an oral
factor Xa inhibitor, may provide more consistent and predictable anticoagulation than
warfarin.
Methods
In a double-blind trial, we randomly assigned 14,264 patients with nonvalvular atrial
fibrillation who were at increased risk for stroke to receive either rivaroxaban (at a
daily dose of 20 mg) or dose-adjusted warfarin. The per-protocol, as-treated primary
analysis was designed to determine whether rivaroxaban was noninferior to warfa-
rin for the primary end point of stroke or systemic embolism.
Results
In the primary analysis, the primary end point occurred in 188 patients in the riva-
roxaban group (1.7% per year) and in 241 in the warfarin group (2.2% per year)
(hazard ratio in the rivaroxaban group, 0.79; 95% confidence interval [CI], 0.66 to 0.96;
P<0.001 for noninferiority). In the intention-to-treat analysis, the primary end point
occurred in 269 patients in the rivaroxaban group (2.1% per year) and in 306 patients
in the warfarin group (2.4% per year) (hazard ratio, 0.88; 95% CI, 0.74 to 1.03;
P<0.001 for noninferiority; P = 0.12 for superiority). Major and nonmajor clinically rel-
evant bleeding occurred in 1475 patients in the rivaroxaban group (14.9% per year) and
in 1449 in the warfarin group (14.5% per year) (hazard ratio, 1.03; 95% CI, 0.96 to 1.11;
P = 0.44), with significant reductions in intracranial hemorrhage (0.5% vs. 0.7%, P = 0.02)
and fatal bleeding (0.2% vs. 0.5%, P = 0.003) in the rivaroxaban group.
Conclusions
In patients with atrial fibrillation, rivaroxaban was noninferior to warfarin for the
prevention of stroke or systemic embolism. There was no significant between-group
difference in the risk of major bleeding, although intracranial and fatal bleeding
occurred less frequently in the rivaroxaban group. (Funded by Johnson & Johnson
and Bayer; ROCKET AF ClinicalTrials.gov number, NCT00403767.)
The New England Journal of Medicine
Downloaded from nejm.org at COLUMBIA UNIV HEALTH SCIENCES LIB on September 14, 2011. For personal use only. No other uses without permission.
Copyright © 2011 Massachusetts Medical Society. All rights reserved.

Page 2

The new england journal of medicine
n engl j med 365;10 nejm.org september 8, 2011
884
A
counts for up to 15% of strokes in persons of all
ages and 30% in persons over the age of 80 years.2
The use of vitamin K antagonists is highly effec-
tive for stroke prevention in patients with nonval-
vular atrial fibrillation and is recommended for
persons at increased risk.3-5 However, food and
drug interactions necessitate frequent coagulation
monitoring and dose adjustments, requirements
that make it difficult for many patients to use such
drugs in clinical practice.6-8
Rivaroxaban is a direct factor Xa inhibitor that
may provide more consistent and predictable anti-
coagulation than warfarin.9,10 It has been reported
to prevent venous thromboembolism more effec-
tively than enoxaparin in patients undergoing
orthopedic surgery11,12 and was noninferior to
enoxaparin followed by warfarin in a study involv-
ing patients with established venous thrombosis.13
This trial was designed to compare once-daily oral
rivaroxaban with dose-adjusted warfarin for the
prevention of stroke and systemic embolism in
patients with nonvalvular atrial fibrillation who
were at moderate-to-high risk for stroke.14
trial fibrillation is associated
with an increase in the risk of ischemic
stroke by a factor of four to five1 and ac-
Methods
Study Design and Oversight
The Rivaroxaban Once Daily Oral Direct Factor
Xa Inhibition Compared with Vitamin K Antago-
nism for Prevention of Stroke and Embolism Trial
in Atrial Fibrillation (ROCKET AF) was a multi-
center, randomized, double-blind, double-dummy,
event-driven trial that was conducted at 1178 par-
ticipating sites in 45 countries.14 The study was
supported by Johnson & Johnson Pharmaceutical
Research and Development and Bayer HealthCare.
The Duke Clinical Research Institute coordinated
the trial, managed the database, and performed
the primary analyses independently of the spon-
sors. Pertinent national regulatory authorities and
ethics committees at participating centers ap-
proved the protocol, which is available with the
full text of this article at NEJM.org. The members
of an international executive committee designed
the trial, were responsible for overseeing the study’s
conduct, retained the ability to independently ana-
lyze and present the data, made the decision to
submit the manuscript for publication, and take
responsibility for the accuracy and completeness
of the data and all analyses. The first academic
author wrote the initial draft of the manuscript.
Study Participants
We recruited patients with nonvalvular atrial fibril-
lation, as documented on electrocardiography, who
were at moderate-to-high risk for stroke. Elevated
risk was indicated by a history of stroke, transient
ischemic attack, or systemic embolism or at least
two of the following risk factors: heart failure or
a left ventricular ejection fraction of 35% or less,
hypertension, an age of 75 years or more, or the
presence of diabetes mellitus (i.e., a CHADS2 score
of 2 or more, on a scale ranging from 1 to 6, with
higher scores indicating a greater risk of stroke).
According to the protocol, the proportion of pa-
tients who had not had a previous ischemic stroke,
transient ischemic attack, or systemic embolism
and who had no more than two risk factors was
limited to 10% of the cohort for each region; the
remainder of patients were required to have had
either previous thromboembolism or three or
more risk factors. Complete inclusion and exclu-
sion criteria are provided in the Supplementary Ap-
pendix, available at NEJM.org. All patients provided
written informed consent.
Study Treatment
Patients were randomly assigned to receive fixed-
dose rivaroxaban (20 mg daily or 15 mg daily in
patients with a creatinine clearance of 30 to 49 ml
per minute) or adjusted-dose warfarin (target in-
ternational normalized ratio [INR], 2.0 to 3.0).
Patients in each group also received a placebo tab-
let in order to maintain blinding. Randomization
was performed with the use of a central 24-hour,
computerized, automated voice-response system.
A point-of-care device was used to generate en-
crypted values that were sent to an independent
study monitor, who provided sites with either real
INR values (for patients in the warfarin group in
order to adjust the dose) or sham values (for pa-
tients in the rivaroxaban group receiving placebo
warfarin) during the course of the trial. Sham INR
results were generated by means of a validated
algorithm reflecting the distribution of values in
warfarin-treated patients with characteristics sim-
ilar to those in the study population.15
It was intended that patients would continue
to take the assigned therapy throughout the
course of the trial, unless discontinuation was
considered to be clinically indicated. Follow-up
The New England Journal of Medicine
Downloaded from nejm.org at COLUMBIA UNIV HEALTH SCIENCES LIB on September 14, 2011. For personal use only. No other uses without permission.
Copyright © 2011 Massachusetts Medical Society. All rights reserved.

Page 3

Rivaroxaban vs. Warfarin in Atrial Fibrillation
n engl j med 365;10 nejm.org september 8, 2011
885
procedures and restrictions on concomitant med-
i cations are summarized in the Supplementary
Appendix.
Outcomes
The primary efficacy end point was the compos-
ite of stroke (ischemic or hemorrhagic) and sys-
temic embolism. Brain imaging was recommend-
ed to distinguish hemorrhagic from ischemic
stroke. In the presence of atherosclerotic periph-
eral arterial disease, the diagnosis of embolism
required angiographic demonstration of abrupt
arterial occlusion.
Secondary efficacy end points included a com-
posite of stroke, systemic embolism, or death from
cardiovascular causes; a composite of stroke, sys-
temic embolism, death from cardiovascular causes,
or myocardial infarction; and individual compo-
nents of the composite end points. The principal
safety end point was a composite of major and
nonmajor clinically relevant bleeding events. Bleed-
ing events involving the central nervous system
that met the definition of stroke were adjudicated
as hemorrhagic strokes and included in both the
primary efficacy and safety end points. Other overt
bleeding episodes that did not meet the criteria
for major or clinically relevant nonmajor bleeding
were classified as minor episodes.
An independent clinical end-point committee
applied protocol definitions to adjudicate all sus-
pected cases of stroke, systemic embolism, myo-
cardial infarction, death, and bleeding events that
contributed to the prespecified end points. De-
tailed definitions of the end-point events are pro-
vided in the Supplementary Appendix.
Statistical Analysis
The primary hypothesis was that rivaroxaban
would be noninferior to warfarin for the preven-
tion of stroke or systemic embolism. The primary
analysis was prespecified to be performed in the
per-protocol population, which included all pa-
tients who received at least one dose of a study
drug, did not have a major protocol violation, and
were followed for events while receiving a study
drug or within 2 days after discontinuation (group
A in Fig. 1 in the Supplementary Appendix).16-19
For the primary analysis, we determined that
a minimum of 363 events would provide a power
of 95% to calculate a noninferiority margin of
1.46 with a one-sided alpha level of 0.025. How-
ever, 405 events were selected as the prespecified
target to ensure a robust statistical result. On the
basis of a projected event rate of 2.3% per 100
patient-years in the warfarin group and a projected
14% rate of annual attrition, it was estimated that
approximately 14,000 patients would need to be
randomly assigned to a study group.
If noninferiority was achieved in the primary
analysis, a closed testing procedure was to be con-
ducted for superiority in the safety population
during treatment, which included patients who
received at least one dose of a study drug and were
followed for events, regardless of adherence to the
protocol, while they were receiving the assigned
study drug or within 2 days after discontinuation
(group B in Fig. 1 in the Supplementary Appen-
dix). Key secondary efficacy end points were also
tested for superiority in the as-treated safety popu-
lation.20 Testing for noninferiority and superior-
ity was also performed in the intention-to-treat
population, which included all patients who un-
derwent randomization and were followed for
events during treatment or after premature dis-
continuation (group C in Fig. 1 in the Supplemen-
tary Appendix).
In addition, we performed post hoc analyses of
events in the intention-to-treat population and
events occurring during the end-of-study transi-
tion to open-label treatment with conventional
anticoagulant agents. In the warfarin group, we
used the method of Rosendaal et al.21 to calculate
the overall time that INR values fell within the
therapeutic range. Comparative analyses of treat-
ment efficacy were performed according to quar-
tiles of time that INR values fell within the thera-
peutic range at the participating clinical sites.
Event rates per 100 patient-years are presented
as proportions of patients per year. Hazard ratios,
confidence intervals, and P values were calculated
with the use of Cox proportional-hazards mod-
els with treatment as the only covariate. Testing
for noninferiority was based on a one-sided sig-
nificance level of 0.025; testing for superiority was
based on a two-sided significance level of 0.05.
Results
Recruitment and Follow-up
From December 18, 2006, through June 17, 2009,
a total of 14,264 patients underwent randomiza-
tion (Fig. 1 in the Supplementary Appendix). The
study was terminated on May 28, 2010. The pro-
portions of patients who permanently stopped
The New England Journal of Medicine
Downloaded from nejm.org at COLUMBIA UNIV HEALTH SCIENCES LIB on September 14, 2011. For personal use only. No other uses without permission.
Copyright © 2011 Massachusetts Medical Society. All rights reserved.

Page 4

The new england journal of medicine
n engl j med 365;10 nejm.org september 8, 2011
886
their assigned therapy before an end-point event
and before the termination date were 23.7% in
the rivaroxaban group and 22.2% in the warfarin
group. The median duration of treatment expo-
sure was 590 days; the median follow-up period
was 707 days. Only 32 patients were lost to follow-
up. Because of violations in Good Clinical Prac-
tice guidelines at one site that made the data un-
reliable, 93 patients (50 in the rivaroxaban group
and 43 in the warfarin group) were excluded from
all efficacy analyses before unblinding. An addi-
tional issue with data quality was raised at an-
other trial site, but this issue was resolved with-
out the exclusion of the patients from the analysis
(for details, see the Supplementary Appendix).
Patient Characteristics and Treatments
Key clinical characteristics of the patients who
underwent randomization are shown in Table 1.
The median age was 73 years (a quarter of the
patients were 78 years of age or older), and 39.7%
of the patients were women. The patients had sub-
stantial rates of coexisting illnesses: 90.5% had
hypertension, 62.5% had heart failure, and 40.0%
had diabetes; 54.8% of the patients had had a
previous stroke, systemic embolism, or transient
ischemic attack. The mean and median CHADS2
scores were 3.5 and 3.0, respectively. Data on med-
ication use at baseline are provided in Table 1 in
the Supplementary Appendix. Previous use of vi-
tamin K antagonists was reported by 62.4% of pa-
tients. At some time during the study, 34.9% of
patients in the rivaroxaban group and 36.2% of
those in the warfarin group took aspirin concur-
rently with the assigned study drug. Among pa-
tients in the warfarin group, INR values were
within the therapeutic range (2.0 to 3.0) a mean
of 55% of the time (median, 58%; interquartile
range, 43 to 71).
Primary Outcome
In the per-protocol population (the patients in-
cluded in the primary efficacy analysis), stroke
or systemic embolism occurred in 188 patients
in the rivaroxaban group (1.7% per year) and in
241 patients in the warfarin group (2.2% per year)
(hazard ratio in the rivaroxaban group, 0.79; 95%
confidence interval [CI], 0.66 to 0.96; P<0.001
for noninferiority) (Table 2 and Fig. 1A). In the
as-treated safety population, primary events oc-
curred in 189 patients in the rivaroxaban group
(1.7% per year) and in 243 patients in the warfarin
Table 1. Characteristics of the Intention-to-Treat Population at Baseline.
Characteristic
Age — yr
Median
Interquartile range
Female sex — no. (%)
Body-mass index*
Median
Interquartile range
Blood pressure — mm Hg
Systolic
Median
Interquartile range
Diastolic
Median
Interquartile range
Type of atrial fibrillation — no. (%)
Persistent
Paroxysmal
Newly diagnosed or new onset
Previous medication use — no. (%)
Aspirin
Vitamin K antagonist
CHADS2 risk of stroke†
Mean score (±SD)
Score — no. (%)
2
3
4
5
6‡
Coexisting condition — no. (%)
Previous stroke, systemic em-
bolism, or transient
ischemic attack
Congestive heart failure
Hypertension
Diabetes mellitus
Previous myocardial infarction‡
Peripheral vascular disease
Chronic obstructive pulmonary
disease
Creatinine clearance — ml/min§
Median
Interquartile range
Rivaroxaban
(N = 7131)
Warfarin
(N = 7133)
73 73
65–78
2831 (39.7)
65–78
2832 (39.7)
28.328.1
25.2 –32.125.1–31.8
130130
120–140 120–140
8080
70–8570–85
5786 (81.1)
1245 (17.5)
100 (1.4)
5762 (80.8)
1269 (17.8)
102 (1.4)
2586 (36.3)
4443 (62.3)
2619 (36.7)
4461 (62.5)
3.48±0.943.46±0.95
925 (13.0)
3058 (42.9)
2092 (29.3)
932 (13.1)
123 (1.7)
934 (13.1)
3158 (44.3)
1999 (28.0)
881 (12.4)
159 (2.2)
3916 (54.9)3895 (54.6)
4467 (62.6)
6436 (90.3)
2878 (40.4)
1182 (16.6)
401 (5.6)
754 (10.6)
4441 (62.3)
6474 (90.8)
2817 (39.5)
1286 (18.0)
438 (6.1)
743 (10.4)
6767
52–8852–86
* The body-mass index is the weight in kilograms divided by the square of the
height in meters.
† The CHADS2 score for the risk of stroke ranges from 1 to 6, with higher scores
indicating an increased risk. Three patients (one in the rivaroxaban group and two
in the warfarin group) had a CHADS2 score of 1.
‡ P<0.05 for the between-group comparison.
§ Creatinine clearance was calculated with the use of the Cockcroft–Gault formula.
The New England Journal of Medicine
Downloaded from nejm.org at COLUMBIA UNIV HEALTH SCIENCES LIB on September 14, 2011. For personal use only. No other uses without permission.
Copyright © 2011 Massachusetts Medical Society. All rights reserved.

Page 5

Rivaroxaban vs. Warfarin in Atrial Fibrillation
n engl j med 365;10 nejm.org september 8, 2011
887
group (2.2% per year) (hazard ratio, 0.79; 95% CI,
0.65 to 0.95; P = 0.01 for superiority). Among all
randomized patients in the intention-to-treat anal-
ysis, primary events occurred in 269 patients in the
rivaroxaban group (2.1% per year) and in 306 pa-
tients in the warfarin group (2.4% per year) (haz-
ard ratio, 0.88; 95% CI, 0.74 to 1.03; P<0.001 for
noninferiority; P = 0.12 for superiority) (Fig. 1B).
During treatment in the intention-to-treat pop-
ulation, patients in the rivaroxaban group had a
lower rate of stroke or systemic embolism (188
events, 1.7% per year) than those in the warfarin
group (240 events, 2.2% per year) (P = 0.02) (Ta-
ble 2 and Fig. 2). Among patients who stopped
taking the assigned study drug before the end of
the study, during a median of 117 days of follow-
up after discontinuation, primary events occurred
in 81 patients in the rivaroxaban group (4.7% per
year) and in 66 patients in the warfarin group
(4.3% per year) (P = 0.58). (Details regarding the
time to events in patients who completed the study
and were switched to standard medical therapy are
provided in Fig. 2 in the Supplementary Appendix.)
Bleeding Outcomes
Major and clinically relevant nonmajor bleeding
occurred in 1475 patients in the rivaroxaban group
and in 1449 patients in the warfarin group (14.9%
and 14.5% per year, respectively; hazard ratio in
the rivaroxaban group, 1.03; 95% CI, 0.96 to 1.11;
P = 0.44) (Table 3). Rates of major bleeding were
similar in the rivaroxaban and warfarin groups
(3.6% and 3.4%, respectively; P = 0.58). Decreases
in hemoglobin levels of 2 g per deciliter or more
and transfusions were more common among pa-
tients in the rivaroxaban group, whereas fatal
bleeding and bleeding at critical anatomical sites
were less frequent. Rates of intracranial hemor-
rhage were significantly lower in the rivaroxaban
group than in the warfarin group (0.5% vs. 0.7%
per year; hazard ratio, 0.67; 95% CI, 0.47 to 0.93;
P = 0.02). Major bleeding from a gastrointestinal
site was more common in the rivaroxaban group,
with 224 bleeding events (3.2%), as compared with
154 events in the warfarin group (2.2%, P<0.001)
(Table 2 in the Supplementary Appendix). (Data
on nonhemorrhagic adverse events are provided
in Table 3 in the Supplementary Appendix.)
Secondary Efficacy Outcomes
The rates of secondary efficacy outcomes in the
as-treated safety population are presented in Ta-
ble 4 in the Supplementary Appendix. During
treatment, myocardial infarction occurred in 101
patients in the rivaroxaban group and in 126 pa-
tients in the warfarin group (0.9% and 1.1% per
year, respectively; hazard ratio in the rivaroxaban
group, 0.81; 95% CI, 0.63 to 1.06; P = 0.12). In the
same analysis population, there were 208 deaths
in the rivaroxaban group and 250 deaths in the
warfarin group (1.9% and 2.2% per year, respec-
tively; hazard ratio, 0.85; 95% CI, 0.70 to 1.02;
Table 2. Primary End Point of Stroke or Systemic Embolism.*
Study PopulationRivaroxaban Warfarin
Hazard Ratio
(95% CI)† P Value
No. of
Patients
No. of
Events
Event
Rate
No. of
Patients
No. of
Events
Event
RateNoninferiority Superiority
no./100
patient-yr
no./100
patient-yr
Per-protocol, as-treated
population‡
6958188 1.77004241 2.20.79 (0.66–0.96) <0.001
Safety, as-treated population7061 1891.77082243 2.20.79 (0.65–0.95)0.02
Intention-to-treat population§ 70812692.1 70903062.40.88 (0.75–1.03)<0.0010.12
During treatment188 1.72402.2 0.79 (0.66–0.96)0.02
After discontinuation 814.7 664.31.10 (0.79–1.52) 0.58
* The median follow-up period was 590 days for the per-protocol, as-treated population during treatment; 590 days for the safety, as-treated
population during treatment; and 707 days for the intention-to-treat population.
† Hazard ratios are for the rivaroxaban group as compared with the warfarin group.
‡ The primary analysis was performed in the as-treated, per-protocol population during treatment.
§ Follow-up in the intention-to-treat population continued until notification of study termination.
The New England Journal of Medicine
Downloaded from nejm.org at COLUMBIA UNIV HEALTH SCIENCES LIB on September 14, 2011. For personal use only. No other uses without permission.
Copyright © 2011 Massachusetts Medical Society. All rights reserved.

Page 6

The new england journal of medicine
n engl j med 365;10 nejm.org september 8, 2011
888
P = 0.07). In addition, in the intention-to-treat analy-
sis throughout the trial, there were 582 deaths in
the rivaroxaban group and 632 deaths in the war-
farin group (4.5% and 4.9% per year, respectively;
hazard ratio, 0.92; 95% CI, 0.82 to 1.03; P = 0.15).
Selected Subgroup Analyses
The effect of rivaroxaban, as compared with war-
farin, in both efficacy and safety analyses was con-
sistent across all prespecified subgroups (Fig. 3,
4, and 5 in the Supplementary Appendix). Fur-
thermore, the effect of rivaroxaban did not differ
across quartiles of the duration of time that INR
values were within the therapeutic range accord-
ing to study center (P = 0.74 for interaction) (Ta-
ble 5 in the Supplementary Appendix). Within
the highest quartile according to center, the haz-
ard ratio with rivaroxaban versus warfarin was
0.74 (95% CI, 0.49 to 1.12).
Discussion
In this randomized trial, we compared rivaroxa-
ban with warfarin for the prevention of stroke or
systemic embolism among patients with nonval-
vular atrial fibrillation who were at moderate-to-
high risk for stroke. In both the primary analy-
sis, which included patients in the per-protocol
population, and in the intention-to-treat analysis,
we found that rivaroxaban was noninferior to war-
farin. In the primary safety analysis, there was no
significant difference between rivaroxaban and
warfarin with respect to rates of major or nonma-
jor clinically relevant bleeding.
As prespecified in the statistical-analysis plan,
we analyzed the trial data in a variety of ways be-
cause we anticipated that some patients would
discontinue the study treatment and we wished to
evaluate both noninferiority and superiority. Al-
though an intention-to-treat analysis is the stan-
dard method for assessing superiority in a ran-
domized trial, noninferiority is best established
when patients are actually taking the randomized
treatment.16-19 Thus, the primary analysis was per-
formed in the per-protocol population during re-
ceipt of the randomly assigned therapy. In the
intention-to-treat population, we found no signifi-
cant between-group difference in a conventional
superiority analysis. In contrast, in the analyses
of patients receiving at least one dose of a study
drug who were followed for events during treat-
ment, we found that rivaroxaban was superior to
warfarin. The difference between these results
reflects the fact that among patients who discon-
tinued therapy before the conclusion of the trial,
no significant difference in outcomes would have
been anticipated, and none was seen.
The most worrisome complication of antico-
agulation is bleeding. Rates of major and nonma-
jor clinically relevant bleeding, the main measure
of treatment safety, were similar in the rivaroxa-
ban and warfarin groups. Bleeding that proved
fatal or involved a critical anatomical site occurred
less frequently in the rivaroxaban group, mainly
Cumulative Event Rate (%)
100
80
90
70
60
40
30
10
50
20
0
0 120240360 480 600720
840
6
4
3
1
5
2
0
0 120 240 360 480 600 720 840
Days since Randomization
B Events in Intention-to-Treat Population
100
A Events in Per-Protocol Population
No. at Risk
Rivaroxaban
Warfarin
6958
7004
6211
6327
5786
5911
5468
5542
4406
4461
2472
2539
3407
3478
1496
1538
Warfarin
Rivaroxaban
Cumulative Event Rate (%)
80
90
70
60
40
30
10
50
20
0
0120 240360480 600720
840
Days since Randomization
No. at Risk
Rivaroxaban
Warfarin
7081
7090
6879
6871
6683
6656
6470
6440
5264
5225
2951
2944
4105
4087
1785
1783
6
4
3
1
5
2
0
0 120 240 360 480 600 720 840
Warfarin
Rivaroxaban
Figure 1. Cumulative Rates of the Primary End Point (Stroke or Systemic
Embolism) in the Per-Protocol Population and in the Intention-to-Treat
Population.
The New England Journal of Medicine
Downloaded from nejm.org at COLUMBIA UNIV HEALTH SCIENCES LIB on September 14, 2011. For personal use only. No other uses without permission.
Copyright © 2011 Massachusetts Medical Society. All rights reserved.

Page 7

Rivaroxaban vs. Warfarin in Atrial Fibrillation
n engl j med 365;10 nejm.org september 8, 2011
889
because of lower rates of hemorrhagic stroke and
other intracranial bleeding. In contrast, bleeding
from gastrointestinal sites, including upper, lower,
and rectal sites, occurred more frequently in the
rivaroxaban group, as did bleeding that led to a
drop in the hemoglobin level or bleeding that re-
quired transfusion. Even though patients in our
trial were at increased risk for bleeding events,
rates of major bleeding were similar to those in
other recent studies involving patients with atrial
fibrillation.4,15,22,23
Among patients in our study who survived and
did not reach the primary end point, the rate of
premature, permanent cessation of randomized
treatment (14.3% in year 1) was slightly higher
than in other studies (average, 11%).15,23 This may
have been a consequence of the trial’s double-blind
design or the inclusion of patients with more co-
existing illnesses. Among patients who perma-
nently discontinued their assigned treatment be-
fore the end of the study, only about half were
treated thereafter with a vitamin K antagonist.
This observation suggests that for at least some
of the patients who participated in the trial, the
risks of open-label therapy with currently available
anticoagulants were ultimately judged to outweigh
the risk of stroke or systemic embolism. Event
rates were similar at 30 days and 1 year after with-
drawal, suggesting that the mechanism of events
did not involve hypercoagulability early after with-
drawal of rivaroxaban. Events occurring at the end
of the study were probably related to increased
difficulty in achieving the transition from blinded
trial therapy to the open-label use of a vitamin K
antagonist when the patient had previously been
assigned to the rivaroxaban group, since presum-
ably many patients who had previously been as-
signed to the warfarin group would have already
had a therapeutic INR.
Among patients in the warfarin group, the pro-
portion of time in which the intensity of anti-
coagulation was in the therapeutic range (mean,
55%), which was calculated from all INR values
during the study and for 7 days after warfarin
interruptions, was lower than in previous studies
of other new anticoagulants in patients with atrial
fibrillation (range, 64 to 68%). Among these trials,
the only study of blinded treatment was limited to
North American sites, which may have facilitated
trial compliance.15 Most earlier trials of warfarin
included fewer high-risk patients,3 and no previous
studies addressed patient populations with overall
levels of coexisting illnesses and geographic diver-
sity that were similar to those of the patients in
our study.24 Significant variations in the duration
of time in the therapeutic range may reflect re-
gional differences and differential skill in manag-
ing warfarin.25 In a recent analysis of anticoagu-
lation management involving more than 120,000
patients in the Veterans Affairs health care system,
the mean proportion of time in the therapeutic
range was 58%, with significant variation across
sites.24 The efficacy of rivaroxaban, as compared
with warfarin, was as favorable in centers with the
best INR control as in those with poorer control.
Cumulative Event Rate (%)
100
80
90
70
60
40
30
10
50
20
0
0 120240360480600720
840
10
6
4
8
2
0
0 120 240 360 480 600 720 840
Days since Randomization
B Events after Discontinuation
100
A Events during Treatment
No. at Risk
Rivaroxaban
Warfarin
7081
7090
6309
6397
5874
5976
5543
5602
4394
4432
2372
2408
3354
3401
1392
1407
Warfarin
Rivaroxaban
Cumulative Event Rate (%)
80
90
70
60
40
30
10
50
20
0
0 120240360 480 600720
840
Days since Drug Discontinuation
No. at Risk
Rivaroxaban
Warfarin
2088
1962
1270
1193
986
880
775
681
543
470
211
196
364
326
101
96
10
6
4
8
2
0
0120 240 360 480 600 720 840
Warfarin
Rivaroxaban
Figure 2. Cumulative Rates of the Primary End Point during Treatment
and after Discontinuation in the Intention-to-Treat Population.
The New England Journal of Medicine
Downloaded from nejm.org at COLUMBIA UNIV HEALTH SCIENCES LIB on September 14, 2011. For personal use only. No other uses without permission.
Copyright © 2011 Massachusetts Medical Society. All rights reserved.

Page 8

The new england journal of medicine
n engl j med 365;10 nejm.org september 8, 2011
890
In conclusion, in this trial comparing a once-
daily, fixed dose of rivaroxaban with adjusted-
dose warfarin in patients with nonvalvular atrial
fibrillation who were at moderate-to-high risk for
stroke, rivaroxaban was noninferior to warfarin
in the prevention of subsequent stroke or sys-
temic embolism. There were no significant dif-
ferences in rates of major and clinically relevant
nonmajor bleeding between the two study groups,
although intracranial and fatal bleeding occurred
less frequently in the rivaroxaban group.
Supported by Johnson & Johnson Pharmaceutical Research
and Development and Bayer HealthCare.
Dr. Patel reports receiving consulting fees from Ortho McNeil
Janssen and Bayer HealthCare and serving on an advisory board
for Genzyme; Dr. Mahaffey, receiving consulting fees from Adolor,
Alexion, Amgen, Argolyn Bioscience, AstraZeneca, Bayer, Bristol-
Myers Squibb, Daiichi Sankyo, Eli Lilly, Elsevier, Forest Labs,
Genentech, GlaxoSmithKline, Guidant, Ikarja, Johnson & Johnson,
Merck, Novartis, Pfizer, Proctor & Gamble, Sanofi Aventis, Schering
Plough, Scios, and WebMD, grant support from Abbott Vascular,
Amgen, Amylin, AstraZeneca, Bayer, Boehringer Ingelheim, Bristol-
Myers Squibb, CardioKinetix, Cierra, Cordis, Edwards LifeSciences,
Eli Lilly, Genentech, GlaxoSmithKline, Guidant, Innocoll Pharma-
ceuticals, Johnson & Johnson, KCI Medical, Luitpold Pharmaceuti-
cal, Medtronic, Merck, Momenta Pharmaceuticals, Novartis, Portola
Pharmaceutical, Pozen, Regado Biotechnologies, Sanofi Aventis,
Schering Plough, and the Medicines Company, and lecture fees
from Johnson & Johnson; Dr. Pan, being an employee of Johnson
& Johnson; Dr. Singer, receiving consulting fees from Johnson &
Johnson, Bayer HealthCare, Boehringer Ingelheim, Daiichi Sankyo,
Merck, Sanofi Aventis, Medtronic, and St. Jude Medical, grant sup-
port from Daiichi Sankyo, and lecture fees from Bristol-Myers
Squibb and Pfizer; Dr. Hacke, serving as a board member for Syg-
nis Pharma Germany and receiving consulting fees, grant support,
lecture fees, and payments for the development of educational
presentations from Boehringer Ingelheim, consulting fees from
Photothera USA and Codman USA, and lecture fees from Bayer
and Photothera USA; Dr. Breithardt, receiving consulting fees
from Bayer HealthCare, Sanofi Aventis, Bristol-Myers Squibb,
Boehringer Ingelheim, Boston Scientific, and Otsuka Pharmaceu-
tical, grant support from Sanofi Aventis, St. Jude, and Meda Phar-
ma, and lecture fees from Sanofi Aventis, Boehringer Ingelheim,
Bayer HealthCare, and Boston Scientific; Dr. Halperin, receiving
consulting fees from Astellas Pharma, Boehringer Ingelheim,
Bristol-Myers Squibb, Daiichi Sankyo, Pfizer, and Sanofi Aventis;
Dr. Hankey, receiving consulting fees from Sanofi Aventis, Schering
Plough, and Boehringer Ingelheim and lecture fees from Sanofi
Aventis and Pfizer; Dr. Becker, receiving grant support from
Johnson & Johnson, Bayer, Regado Biosciences, and Bristol-Myers
Squibb, consulting fees from Boehringer Ingelheim, and lecture
fees from Merck and AstraZeneca; Dr. Nessel, being an employee
of and having an equity interest in Johnson & Johnson; Drs.
Paolini and Berkowitz, being employees of Bayer HealthCare, and
Dr. Paolini, having an equity interest in Bayer HealthCare; Dr. Fox,
receiving grant support and lecture fees from Eli Lilly and lecture
fees from Sanofi Aventis and AstraZeneca; Dr. Califf, receiving con-
sulting fees from Kowa, Nile, Orexigen, Sanofi Aventis, Novartis,
and Xoma and grant support from Novartis, Merck, and Amilyn/
Lilly and having an equity interest in Nitrox. No other potential
conflict of interest relevant to this article was reported. Disclosure
forms provided by the authors are available with the full text of this
article at NEJM.org.
We thank Lisa G. Berdan, Karen Hannan, and Kimberly Schwabe
for their operational oversight of the trial and Elizabeth Cook for
providing editorial support in the preparation of the manuscript.
Table 3. Rates of Bleeding Events.*
Variable
Rivaroxaban
(N = 7111)
Events
Warfarin
(N = 7125)
Hazard Ratio
(95% CI)†P Value‡
Event Rate
no./100
patient-yr
Events Event Rate
no./100
patient-yr no. (%)no. (%)
Principal safety end point: major and nonmajor
clinically relevant bleeding§
1475 (20.7)14.91449 (20.3)14.51.03 (0.96–1.11)0.44
Major bleeding
Any395 (5.6)3.6386 (5.4)3.4 1.04 (0.90–1.20) 0.58
Decrease in hemoglobin ≥2 g/dl305 (4.3) 2.8254 (3.6) 2.31.22 (1.03–1.44)0.02
Transfusion183 (2.6) 1.6149 (2.1) 1.31.25 (1.01–1.55)0.04
Critical bleeding¶91 (1.3) 0.8133 (1.9) 1.20.69 (0.53–0.91)0.007
Fatal bleeding27 (0.4) 0.255 (0.8)0.50.50 (0.31–0.79)0.003
Intracranial hemorrhage55 (0.8)0.5 84 (1.2) 0.70.67 (0.47–0.93) 0.02
Nonmajor clinically relevant bleeding1185 (16.7)11.8 1151 (16.2)11.41.04 (0.96–1.13) 0.35
* All analyses of rates of bleeding are based on the first event in the safety population during treatment.
† Hazard ratios are for the rivaroxaban group as compared with the warfarin group and were calculated with the use of Cox proportional-hazards
models with the study group as a covariate.
‡ Two-sided P values are for superiority in the rivaroxaban group as compared with the warfarin group.
§ Minimal bleeding events were not included in the principal safety end point.
¶ Bleeding events were considered to be critical if they occurred in intracranial, intraspinal, intraocular, pericardial, intraarticular, intramuscular
(with compartment syndrome), or retroperitoneal sites.
The New England Journal of Medicine
Downloaded from nejm.org at COLUMBIA UNIV HEALTH SCIENCES LIB on September 14, 2011. For personal use only. No other uses without permission.
Copyright © 2011 Massachusetts Medical Society. All rights reserved.

Page 9

Rivaroxaban vs. Warfarin in Atrial Fibrillation
n engl j med 365;10 nejm.org september 8, 2011
891
1. Wolf PA, Abbott RD, Kannel WB.
Atrial fibrillation as an independent risk
factor for stroke: the Framingham Study.
Stroke 1991;22:983-8.
2. Idem. Atrial fibrillation: a major con-
tributor to stroke in the elderly: the Fram-
ingham Study. Arch Intern Med 1987;147:
1561-4.
3. Hart RG, Benavente O, McBride R,
Pearce LA. Antithrombotic therapy to pre-
vent stroke in patients with atrial fibrilla-
tion: a meta-analysis. Ann Intern Med
1999;131:492-501.
4. Hart RG, Pearce LA, Aguilar MI. Meta-
analysis: antithrombotic therapy to pre-
vent stroke in patients who have nonval-
vular atrial fibrillation. Ann Intern Med
2007;146:857-67.
5. Singer DE, Albers GW, Dalen JE, Go
AS, Halperin JL, Manning WJ. Antithrom-
botic therapy in atrial fibrillation: the
Seventh ACCP Conference on Antithrom-
botic and Thrombolytic Therapy. Chest
2004;126:Suppl:429S-456S.
6. Albers GW, Yim JM, Belew KM, et al.
Status of antithrombotic therapy for pa-
tients with atrial fibrillation in university
hospitals. Arch Intern Med 1996;156:2311-
6.
7. Go AS, Hylek EM, Borowsky LH, Phil-
lips KA, Selby JV, Singer DE. Warfarin use
among ambulatory patients with nonval-
vular atrial fibrillation: the Anticoagula-
tion and Risk Factors in Atrial Fibrillation
(ATRIA) study. Ann Intern Med 1999;
131:927-34.
8. Piccini JP, Hernandez AF, Zhao X, et
al. Quality of care for atrial fibrillation
among patients hospitalized for heart
failure. J Am Coll Cardiol 2009;54:1280-9.
9. Kubitza D, Becka M, Wensing G,
Voith B, Zuehlsdorf M. Safety, pharmaco-
dynamics, and pharmacokinetics of BAY
59-7939 — an oral, direct Factor Xa in-
hibitor — after multiple dosing in healthy
male subjects. Eur J Clin Pharmacol 2005;
61:873-80.
10. Kubitza D, Becka M, Roth A, Mueck
W. Dose-escalation study of the pharma-
cokinetics and pharmacodynamics of riv-
a roxaban in healthy elderly subjects. Curr
Med Res Opin 2008;24:2757-65.
11. Eriksson BI, Borris LC, Friedman RJ,
et al. Rivaroxaban versus enoxaparin for
thromboprophylaxis after hip arthroplas-
ty. N Engl J Med 2008;358:2765-75.
12. Lassen MR, Ageno W, Borris LC, et
al. Rivaroxaban versus enoxaparin for
thromboprophylaxis after total knee ar-
throplasty. N Engl J Med 2008;358:2776-86.
13. The EINSTEIN Investigators. Oral riva-
roxaban for symptomatic venous throm-
boembolism. N Engl J Med 2010;363:2499-
510.
14. ROCKET AF Study Investigators. Riv-
aroxaban — once daily, oral, direct factor
Xa inhibition compared with vitamin K
antagonism for prevention of stroke and
Embolism Trial in Atrial Fibrillation: ratio-
nale and design of the ROCKET AF study.
Am Heart J 2010;159(3):340.e1-347.e1.
15. Albers GW, Diener HC, Frison L, et al.
Ximelagatran vs warfarin for stroke pre-
vention in patients with nonvalvular atrial
fibrillation: a randomized trial. JAMA
2005;293:690-8.
16. Matilde Sanchez M, Chen X. Choosing
the analysis population in non-inferiority
studies: per protocol or intent-to-treat.
Stat Med 2006;25:1169-81.
17. Brittain E, Lin D. A comparison of
intent-to-treat and per-protocol results in
antibiotic non-inferiority trials. Stat Med
2005;24:1-10.
18. Draft: Guidance for Industry: non-
inferiority clinical trials. Silver Spring,
MD: Food and Drug Administration, 2010.
(http://www.fda.gov/Drugs/Guidance
ComplianceRegulatoryInformation/
Guidances/default.htm.)
19. Piaggio G, Elbourne DR, Altman DG,
Pocock SJ, Evans SJ. Reporting of nonin-
feriority and equivalence randomized tri-
als: an extension of the CONSORT state-
ment. JAMA 2006;295:1152-60. [Erratum,
JAMA 2006;296:1842.]
20. Soulakova JN. Comparison of several
testing strategies for combination drug
efficacy trials based on the closure prin-
ciple. Stat Med 2009;28:260-73.
21. Rosendaal FR, Cannegieter SC, van
der Meer FJ, Briët E. A method to deter-
mine the optimal intensity of oral antico-
agulant therapy. Thromb Haemost 1993;
69:236-9.
22. Connolly S, Pogue J, Hart R, et al.
Clopidogrel plus aspirin versus oral anti-
coagulation for atrial fibrillation in the
Atrial fibrillation Clopidogrel Trial with
Irbesartan for prevention of Vascular Events
(ACTIVE W): a randomised controlled
trial. Lancet 2006;367:1903-12.
23. Connolly SJ, Ezekowitz MD, Yusuf S,
et al. Dabigatran versus warfarin in pa-
tients with atrial fibrillation. N Engl J
Med 2009;361:1139-51. [Erratum, N Engl
J Med 2010;363:1877.]
24. Rose AJ, Hylek EM, Ozonoff A, Ash
AS, Reisman JI, Berlowitz DR. Risk-
adjusted percent time in therapeutic
range as a quality indicator for outpatient
oral anticoagulation: results of the Veter-
ans Affairs Study to Improve Anticoagu-
lation (VARIA). Circ Cardiovasc Qual Out-
comes 2011;4:22-9.
25. Wallentin L, Yusuf S, Ezekowitz MD,
et al. Efficacy and safety of dabigatran
compared with warfarin at different levels
of international normalised ratio control
for stroke prevention in atrial fibrillation:
an analysis of the RE-LY trial. Lancet
2010;376:975-83.
Copyright © 2011 Massachusetts Medical Society.
References
receive immediate notification when an article
is published online first
To be notified by e-mail when Journal articles
are published Online First, sign up at NEJM.org.
The New England Journal of Medicine
Downloaded from nejm.org at COLUMBIA UNIV HEALTH SCIENCES LIB on September 14, 2011. For personal use only. No other uses without permission.
Copyright © 2011 Massachusetts Medical Society. All rights reserved.