Vous êtes sur la page 1sur 10

CLINICAL RESEARCH STUDY

Net Clinical Benet of Non-vitamin K Antagonist


Oral Anticoagulants Versus Warfarin in Phase III
Atrial Fibrillation Trials
Giulia Renda, MD, PhD,a Marta di Nicola, PhD,b Raffaele De Caterina, MD, PhDa,c
a
Institute of Cardiology, Department of Neurosciences, Imaging and Clinical Sciences-Center of Excellence on Aging, G. dAnnunzio
University, Chieti, Italy; bLaboratory of Biostatistics, Department of Experimental and Clinical Sciences, G. dAnnunzio University,
Chieti, Italy; cG. Monasterio Foundation, Pisa, Italy.

ABSTRACT
OBJECTIVES: The evaluation of the net clinical benet allows an integrated assessment of both the antiischemic and the prohemorrhagic effects of non-vitamin K antagonist oral anticoagulants compared with
warfarin, andin the absence of direct comparisonsmay inform clinical decisions. We estimated the net
clinical benet of non-vitamin K antagonist oral anticoagulants versus warfarin across the 4 phase III
clinical trials performed in patients with atrial brillation.
METHODS: We considered various composites of the main ischemic and hemorrhagic events, estimating the
rate ratio of all treatment groups versus warfarin for each composite outcome. Because the clinical relevance of the various ischemic or hemorrhagic events is not identical, we then attributed to each of them a
weight, according to its impact on death, as derived from previous studies. We evaluated a weighed net
clinical benet of each non-vitamin K antagonist oral anticoagulant compared with warfarin in the 4 trials.
RESULTS: The composite outcome of ischemic hemorrhagic stroke was reduced by dabigatran 150 mg
and apixaban. The composite of disabling stroke life-threatening bleeding was reduced by all none
vitamin K antagonist oral anticoagulants. The composite of ischemic stroke systemic embolism
myocardial infarction hemorrhagic stroke major bleeding was reduced by apixaban and edoxaban at
both doses. By attributing weights to these events according to their impact on mortality, all non-vitamin K
antagonist oral anticoagulants featured a favorable net clinical benet compared with warfarin, albeit to a
quantitatively different extent.
CONCLUSIONS: The choice of the proper antithrombotic treatment in patients with atrial brillation has to
consider the net clinical benet of each drug. However, all non-vitamin K antagonist oral anticoagulants
have a better efcacy/safety prole than warfarin in patients with atrial brillation.
2015 Elsevier Inc. All rights reserved.  The American Journal of Medicine (2015) 128, 1007-1014
KEYWORDS: Atrial brillation; Hemorrhagic risk; Net clinical benet; Oral anticoagulants; Thromboembolic risk

Funding: None.
Conict of Interest: GR received consultant and speaker fees from
Boehringer-Ingelheim, Daiichi-Sankyo, and Bayer. RDC received consultant and speaker fees from Bayer, Boehringer Ingelheim, Bristol-Myers
Squibb/Pzer, Daiichi Sankyo, Novartis, and Merck, and research grants
from Boehringer-Ingelheim.
Authorship: All authors had access to the data and played a role in
writing this manuscript.
Requests for reprints should be addressed to Raffaele De Caterina, MD,
PhD, Institute of Cardiology, G. dAnnunzio University e Chieti,
C/o Ospedale SS. Annunziata, Via dei Vestini - 66013 Chieti, Italy.
E-mail address: rdecater@unich.it
0002-9343/$ -see front matter 2015 Elsevier Inc. All rights reserved.
http://dx.doi.org/10.1016/j.amjmed.2015.03.034

Non-vitamin K antagonist oral anticoagulants,1 including


the direct thrombin inhibitor dabigatran etexilate and the
direct factor Xa inhibitors rivaroxaban, apixaban, and edoxaban, have emerged as alternatives to vitamin K antagonists
for thromboembolic prevention in patients with nonvalvular
atrial brillation. These agents have been compared with
warfarin, the most commonly used vitamin K antagonist, in
major phase III clinical trials.2-5 Individually taken, these
agents are all noninferior to warfarin in safety and efcacy,
and all signicantly reduced intracranial hemorrhage.6,7
All anticoagulants affect hemostatic function; therefore,
their benet must be balanced against risks. In patients with

1008

The American Journal of Medicine, Vol 128, No 9, September 2015

atrial brillation, the possible favorable change (reduction)


(ROCKET AF), Apixaban for Reduction In STroke and
of ischemic stroke, systemic embolism, and myocardial
Other ThromboemboLic Events in atrial brillation
infarction has to be balanced against an increased risk in
(ARISTOTLE), and Effective aNticoaGulation with factor
hemorrhagic events, including intracranial bleeding. Current
xA next GEneration in Atrial Fibrillation-Thrombolysis In
guidelines for the management of antithrombotic therapy in
Myocardial Infarction study 48 (ENGAGE AF-TIMI 48)
atrial brillation recommend both thromboembolic and
studies.2-5 For the RE-LY trial, we used the latest pub8,9
hemorrhagic risk stratication.
lished data update,12 containing
several primary efcacy and
However, the clinical relevance
CLINICAL SIGNIFICANCE
safety outcome events in addition
of outcome eventsaverted or
to the primary publication.2-5
causedis not equivalent, because
 We evaluated the net clinical benet for
such events differ in severity and
We rst used the simple algevarious non-vitamin K antagonist oral
may cause disabilities and affect
braic summation of ischemic and
anticoagulants in phase III clinical trials
mortality in different ways. As an
hemorrhagic events, alone or with
comparing them with warfarin in atrial
example, a hemorrhagic stroke
cardiovascular or total mortality.
brillation, weighing nonfatal efcacy
induced by an anticoagulant is
We then used the approach
and safety outcomes according to their
more likely to cause disability and
applied in the ACTIVE A13 and
prognostic impact on mortality.
death than an averted ischemic
RE-LY14 cohorts to gather a reli10
stroke. A previous analysis from
able estimate of the weighed net
 Although non-vitamin K antagonist oral
clinical benet, calculating the
the Atrial brillation Clopidogrel
anticoagulants have shown variable
number needed to treat in each
Trial with Irbesartan for prevenefcacy and safety relative to warfarin,
arm of warfarin comparison of the
tion of Vascular Events (ACTIVE)
according to this analysis all have a
4 major warfarin-controlled phase
W trial, a large trial examining the
better and strikingly similar net clinical
III trials of non-vitamin K antageffect of clopidogrel plus aspirin
benet in patients with atrial
onist oral anticoagulants in atrial
versus vitamin K antagonists for
brillation.
brillation, as detailed next.
the prevention of thromboembolism in atrial brillation,11 suggested
the
possibility
and
Evaluation of Composite Outcomes
importance of weighing the severity of events differentially
We rst considered unweighed different composites of main
occurring with different antithrombotic treatments using
events: (1) ischemic stroke and hemorrhagic stroke; (2)
their variable prognostic effect on mortality as a way to
disabling stroke and life-threatening bleeding; and (3)
adjust for the difference in severity of the nonfatal event.
ischemic stroke, systemic embolism, myocardial infarction,
Therefore, the modern concept of weighed net clinical
hemorrhagic stroke, and adjusted major bleeding (dened as
benet has been introduced to evaluate the overall
major bleeding minus hemorrhagic stroke, discussed later).
comparative benets of different antithrombotic treatments,
In a second analysis, each of these composites of events also
integrating information on anti-ischemic and prohemorwas combined with cardiovascular mortality and all-cause
rhagic effects.
mortality. The denition of each event was dened acOn this basis, we assessed the net clinical benet of
cording to what was reported in the original description of
various dose regimens of non-vitamin K antagonist oral
the trial design and rationale.15-18
anticoagulants compared with warfarin in the 4 warfarinThe frequency of ischemic stroke and unknown or
controlled trials so far performed on nonvalvular atrial
uncertain stroke was reported jointly in all trials with the
brillation. Such analyses, in the absence of direct comexception of the ROCKET-AF, in which unknown stroke
parisons of the different non-vitamin K antagonist oral
was reported separately5; therefore, for uniform data, we
anticoagulants, may inform choices of drugs and drug regadded the frequency of unknown stroke to that of ischemic
imens in the individual patient.
stroke throughout the collection of original data.
The frequency of disabling stroke was reported separately only in the ROCKET AF trial,5 whereas in the other
MATERIALS AND METHODS
trials, reports included the frequency of disabling plus fatal
Overall Strategy for the Analysis
stroke together.2-4 Because we conducted the unweighed
analysis both alone and combined with cardiovascular
In the current report, we estimated the net clinical benet of
mortality or all-cause mortality, we included only disabling
all non-vitamin K antagonist oral anticoagulants compared
strokes in the analysis (without fatal strokes) to avoid
with warfarin in phase III clinical trials in nonvalvular
counting the same fatal events twice. Therefore, to obtain
atrial brillation, namely, the Randomized Evaluation of
the frequency of disabling stroke in each trial, we subtracted
Long-Term Anticoagulation Therapy (RE-LY), Rivaroxthe frequency of fatal stroke from the composite of disabling
aban Once Daily, Oral, Direct Factor Xa Inhibition
plus fatal stroke. Because the frequency of fatal stroke was
Compared With Vitamin K Antagonism for Prevention
not reported in the original RE-LY trial report,2 this was
of Stroke and Embolism Trial in Atrial Fibrillation

Renda et al

Net Clinical Benet of Non-Vitamin K Antagonist Oral Anticoagulants Versus Warfarin

acquired from the Advisory Committee Brieng Document


of dabigatran published by the US Food and Drug Administration.19 To avert the multiple counting of the same events
in the analysis, we dened adjusted major bleeding as the
number of major bleeding minus the number of hemorrhagic
strokes.
For life-threatening bleeding, such denition was given in
the RE-LY2,16 and ENGAGE AF-TIMI 483,18 trials, but not
in the other 2 trials; thus, for the ARISTOTLE4,17 and
ROCKET-AF5,15 trials, we used data reported under the
headings Global Strategies for Opening Occluded Coronary
Arteries severe bleeding (GUSTO) and critical bleeding,
respectively, because these coincided with life-threatening
bleeding in terms of clinical relevance. To determine the
frequency of an event not explicitly reported in a trial, we
calculated this as the event rate divided by the median
duration of the follow-up for each treatment arm of the trial.
For each of the 3 composite outcomes, alone and combined with cardiovascular mortality or all-cause mortality,
we estimated the rate ratio (RR) and the relative 95% condence intervals (CIs)20 of all treatment groups versus
warfarin.

Calculation of the Weighed Net Clinical Benet


We evaluated the weighed net clinical benet of each
non-vitamin K antagonist oral anticoagulant compared with
warfarin in the 4 trials, considering ischemic stroke, systemic embolism, myocardial infarction, hemorrhagic stroke,
and adjusted major bleeding as a composite event. We
evaluated the crude incidence rate (IR) per 100 patient-years
of each weighed event for patients receiving a none
vitamin K antagonist oral anticoagulant (1 or 2 doses) and
Table 1

1009

for patients receiving warfarin. The net clinical benet was


dened as the weighed sum of those IRs in the warfarin
groups minus the weighed sum of events in the nonvitamin K antagonist oral anticoagulant groups: net clinical
benet [IRischemic stroke_warfarin w1IRsystemic
embolism_warfarin w2IRmyocardial infarction_warfarin w3IRhemorrhagic stroke_warfarin w4IRadjusted major bleeding_warfarin] e
[IRischemic stroke_NOAC w1IRsystemic embolism_NOAC
w2IRmyocardial infarction_NOAC w3IRhemorrhagic stroke_NOAC
w4IRadjusted major bleeding_NOAC], where NOAC is nonvitamin K antagonist oral anticoagulant and w1, w2, w3,
and w4 are the weights associated with systemic embolism,
myocardial infarction, hemorrhagic stroke, and adjusted
major bleeding, respectively. Because we did not know the
mortality associated with each event considered during the 4
trials, we could not calculate the weights of these events in
the specic trial populations. Therefore, we assumed that the
impact of each event on mortality is similar in the 4 trial
populations and used the weights calculated in a recent
analysis of net clinical benet evaluating the combined data
from the ACTIVE and RE-LY trial databases.21 Therefore,
weights were 0.61 for systemic embolism, 0.89 for
myocardial infarction, 3.23 for hemorrhagic stroke, and 0.63
for adjusted major bleeding (this latter is, in practice, major
extracranial bleeding). These weights were expressed in
terms of ischemic stroke equivalents prevented per 100
persons/year, using ischemic stroke as the reference event
for each event (weight 1).
Finally, we estimated the number of patients needed to
treat to prevent all grouped events included in the net
clinical benet calculation per year. The number needed to
treat thus was calculated as 1/net clinical benet for all
non-vitamin K antagonist oral anticoagulants.

Study Design and Baseline Patient Characteristics in the Trials Examined in the Present Analysis

Trial Name

Treatment Arm

Dose

No. of
Patients

Median
Follow-up (y)

Mean
Age (y)

Male
Gender (%)

Mean
CHADS2

Median
TTR (%)

RE-LY (2)

Dose-adjusted warfarin
Dabigatran 150 mg
Dabigatran 110 mg
Dose-adjusted warfarin
Rivaroxaban
Dose-adjusted warfarin
Apixaban
Dose-adjusted warfarin

INR 2.0-3.0
150 mg BID
110 mg BID
INR 2.0-3.0
20 mg (or 15 mg*) OD
INR 2.0-3.0
5 mg (or 2.5 mg) BID
INR 2.0-3.0

6022
6076
6015
7090
7131
9081
9120
7036

71.6
71.5
71.4
71.2
71.2
64.5
69.1
72

63.3
63.2
64.3
60.3
60.3
65
64.4
62.5

2.1
2.2
2.1
3.46
3.48
2.1
2.1
2.8

67
NA
NA
58
NA
66
NA
68

Edoxaban 60 mg
Edoxaban 30 mg

60 mg (or 30 mg) OD
30 mg (or 15 mg) OD

7035
7034

72
72

62.1
61.2

2.8
2.8

NA
NA

ROCKET-AF (3)
ARISTOTLE (4)
ENGAGE
AF-TIMI 48 (5)

1.9
1.8
2.8

ARISTOTLE Apixaban for Reduction In STroke and Other ThromboemboLic Events in atrial brillation; BID twice daily; CHADS2 Congestive heart
failure, Hypertension, Age greater than 75, Diabetes mellitus, and prior Stroke or transient ischemic attack; ENGAGE AF-TIMI 48 Effective aNticoaGulation
with factor xA next GEneration in Atrial Fibrillation-Thrombolysis In Myocardial Infarction study 48; INR international normalized ratio; NA not
applicable; OD once daily; RE-LY Randomized Evaluation of Long-Term Anticoagulation Therapy; ROCKET-AF Rivaroxaban Once Daily, Oral, Direct
Factor Xa Inhibition Compared With Vitamin K Antagonism for Prevention of Stroke and Embolism Trial in Atrial Fibrillation; TTR time in therapeutic range.
*15 mg once daily with creatinine clearance 30-49 mL/min.
2.5 mg twice per day with 2 of the following criteria: age 80 y, body weight <60 kg, or serum creatinine of 1.5 mg/dL, 133 mmol/L.
30 mg or 15 mg with any 1 or multiple of the following: creatinine clearance 30-50 mL/min, body weight 60 kg, or concomitant administration of
verapamil, quinidine, or dronedarone strong P-glycoprotein inhibitors.

1010

The American Journal of Medicine, Vol 128, No 9, September 2015

All analyses were performed with the open-source statistical R software (version 3.0.0, The R Foundation for
Statistical Computing).22 To assess the effect of weight
variations on the stability of the results, we also conducted a
sensitivity analysis with increased and decreased weights by
25% and 50% of their original values.

and edoxaban at both doses (Table 2, Figure 1). The same


composite, in combination with cardiovascular mortality
and all-cause mortality, was reduced by apixaban, edoxaban at both doses, and dabigatran 150 mg (Table E1,
Figure E1, available online).

Weighed Net Clinical Benet

RESULTS
The study design and baseline characteristics of patients
who were enrolled in the trials considered are shown in
Table 1.

Composite Outcomes
We rst analyzed the composite outcomes of main thrombotic and bleeding events in terms of RR of all treatment
arms versus warfarin. When considering the composite of
ischemic stroke hemorrhagic stroke, we found that
dabigatran 150 mg and apixaban signicantly reduced this
composite outcome (Table 2), and such reduction also was
evident when the composite included cardiovascular
mortality or all-cause mortality (Table E1, available
online). Rivaroxaban and edoxaban 60 mg reduced only
the composite outcome of ischemic stroke hemorrhagic
stroke cardiovascular /all-cause mortality (Table 2,
Table E1, available online). Dabigatran 110 mg and
edoxaban 30 mg had no effect on this composite
(Table 2, Table E1, available online).
The composite outcome of disabling stroke lifethreatening bleeding, alone or combined with cardiovascular or all-cause mortality, was reduced by all non-vitamin K
antagonist oral anticoagulants (Table 2, Table E1, available
online).
The overall composite outcome of ischemic stroke
systemic embolism myocardial infarction hemorrhagic
stroke adjusted major bleeding was reduced by apixaban

We evaluated the weighed net clinical benet of the 4


non-vitamin K antagonist oral anticoagulants compared
with warfarin by a weighed estimate of a composite of
ischemic stroke systemic embolism myocardial
infarction hemorrhagic stroke adjusted major bleeding.
The crude IRs per 100 patient-years of each weighed event
for all treatment groups are reported in Table 3. All
non-vitamin K antagonist oral anticoagulants featured a
directionally favorable net clinical benet compared with
warfarin in terms of reduction of ischemic stroke
equivalents and the related number needed to treat to
prevent all grouped events included in the net clinical
benet per year, with some quantitative differences
between the non-vitamin K antagonist oral anticoagulants
(Figure 2). Sensitivity analysis using weights increased
and decreased by 25% and 50% compared with those
considered for the main analysis produced similar results,
showing the same favorable prole of non-vitamin K
antagonist oral anticoagulants compared with warfarin
(Table E2, available online).

DISCUSSION
This analysis shows that all non-vitamin K antagonist oral
anticoagulants in nonvalvular atrial brillation feature a
favorable net clinical benet compared with warfarin, with
some quantitative differences among them in the warfarin
comparison.

Table 2 Rate Ratio and Corresponding 95% Condence Interval for Each Treatment Arm of the Various Trials Versus Warfarin for the
Various Composite Outcomes Considered

Treatment
Dabigatran 150 mg
Dabigatran 110 mg
Rivaroxaban
Apixaban
Edoxaban 60 mg
Edoxaban 30 mg

Ischemic Stroke
Hemorrhagic Stroke

Disabling Stroke
Life-threatening Bleeding

Ischemic Stroke Hemorrhagic


Stroke Myocardial Infarction Systemic
Embolism Adjusted Major Bleeding

0.65 (0.51-0.81)
<.001
0.91 (0.74-1.12)
.382
0.83 (0.68-1.00)
.058
0.79 (0.66-0.96)
.015
0.88 (0.75-1.02)
.106
1.12 (0.96-1.30)
.146

0.8 (0.67-0.95)
.009
0.78 (0.66-0.93)
.005
0.70 (0.56-0.87)
<.001
0.55 (0.44-0.68)
<.001
0.67 (0.53-0.84)
<.001
0.69 (0.55-0.88)
.002

0.93 (0.83-1.03)
.201
0.93 (0.83-1.03)
.205
0.92 (0.83-1.03)
.151
0.78 (0.70-0.87)
<.001
0.87 (0.79-0.95)
.004
0.85 (0.77-0.93)
<.001

RR and 95% CI of each treatment arm versus warfarin are reported in each case, with P value immediately below, for the various composite outcomes;
statistically signicant P values are in bold. Adjusted major bleeding major bleeding  hemorrhagic stroke.

Renda et al

Net Clinical Benet of Non-Vitamin K Antagonist Oral Anticoagulants Versus Warfarin

Figure 1 RR and 95% CI of all treatment arms in the phase III


trials comparing a non-vitamin K antagonist oral anticoagulants
with warfarin for the overall composite outcome including
unweighed ischemic stroke systemic embolism myocardial
infarction hemorrhagic stroke adjusted major bleeding
(major bleeding minus hemorrhagic stroke). NOAC nonvitamin K antagonist oral anticoagulant.

The efcacy of antithrombotic drugs in avoiding


ischemic events in various settings, including atrial brillation, can be partially or totally offset by the risk of adverse
hemorrhagic events they confer because of their impact on
hemostasis. Therefore, the evaluation of the overall benet
of oral anticoagulants in the treatment of patients with atrial
brillation needs to consider both sides of the coin. There is
no unanimous denition of what exactly the net clinical
benet should be in atrial brillation trials with anticoagulation and how to estimate it. A composite of ischemic
and bleeding events has been used as a measure of net
clinical benet in phase III clinical trials comparing nonvitamin K antagonist oral anticoagulants with warfarin.
However, the exercise of adding the number of ischemic and
hemorrhagic events to obtain an overall outcome is not
accurate and can be misleading because of the different
clinical impact of the different events in terms of mortality,
disability, and costs. On this basis, attempts to weigh the
relevance of each event have been conducted.
The rst quantitative, empirically based description of the
weighed net clinical benet of warfarin treatment for patients
with atrial brillation was derived from the large Anticoagulation and Risk Factors in Atrial Fibrillation (ATRIA)
cohort, in which the net clinical benet was dened as the
annual rate of ischemic stroke and systemic embolism prevented by warfarin minus intracranial hemorrhages due to
warfarin multiplied by an arbitrary weight of 1.5.23 This
weight was chosen by giving an approximate quantitative
estimate to the observations that the outcome of an intracranial hemorrhage is generally worse than that of an ischemic
stroke, with the weighing factor of 1.5 reecting the impact
on mortality of an intracranial hemorrhage while receiving
warfarin versus having an ischemic stroke while not receiving
warfarin. A similar model was proposed recently to quantify
the balance between the risk of ischemic stroke and the risk of

1011

intracranial hemorrhage with the use of non-vitamin K


antagonist oral anticoagulants (dabigatran, rivaroxaban, and
apixaban) from a Danish National Patient Registry on patients with atrial brillation, considering 1.5 the weight of an
intracranial hemorrhage compared with an ischemic stroke,
which is assigned the normalized weight of 1.24
In an alternative approach, because ischemic stroke is a
common primary end point of efcacy for antithrombotic
treatments in atrial brillation, a weighed net clinical benet
has been expressed in units of ischemic stroke equivalents
prevented, as the reference weight for each event, and made
equal to 1. This approach was rst used for patients enrolled
in the ACTIVE A trial,13 who, as an entry criterion, were
unsuitable for vitamin K antagonists. The net clinical benet
of dual antiplatelet therapy in the ACTIVE A trial participants was dened as the weighed event incidence with dual
antiplatelet therapy subtracted from the sum of weighed
event incidence on control treatment, expressed as ischemic
stroke equivalents prevented per 100 patient-years.14 On the
basis of the same model, a more recent report explored the
net clinical benet of each dose of dabigatran and warfarin
in the RE-LY trial.21 We have applied the same criteria to
assess and compare the net clinical benet of all 4 nonvitamin K antagonist oral anticoagulants tested so far in
atrial brillation.

Composite Outcomes
For the rst analysis, we chose the main ischemic and
hemorrhagic events evaluated in the 4 clinical trials
designed to compare the efcacy and safety of non-vitamin
K antagonist oral anticoagulants versus warfarin, and we
grouped these events in various composite outcomes: a
composite of ischemic stroke hemorrhagic stroke, a
composite of disabling stroke life-threatening bleeding,
and an overall composite outcome of ischemic stroke
hemorrhagic stroke myocardial infarction systemic
embolism adjusted major bleeding. Cardiovascular mortality and all-cause mortality also were combined with all
composite outcomes in a secondary analysis. This latter
combination of nonfatal events with cardiovascular mortality
or total mortality (presented in Table E1 and Figure E1,
available online) is to be considered exploratory, because
it may include nonmutually exclusive events. We calculated
the RR of non-vitamin K antagonist oral anticoagulants
versus warfarin for each composite outcome. Our ndings
indicate that apixaban always demonstrates better efcacy
than warfarin in reducing all these composite outcomes.
In this analysis, dabigatran 150 mg demonstrates a better
efcacy than warfarin in reducing all the composite outcomes with the exception of the most comprehensive one of
ischemic stroke, systemic embolism, myocardial infarction,
hemorrhagic stroke, and adjusted major bleeding, probably
because this regimen was associated, in the RE-LY trial,
with a slightly higher number of myocardial infarctions and
an equivalent number of major bleeding compared with
warfarin. However, when combined with cardiovascular or

1012
Table 3

The American Journal of Medicine, Vol 128, No 9, September 2015


Crude Incidence Rate per 100 Patient-years of Each Weighed Event for All Treatment Groups

Weight
RE-LY
Dabigatran 150 mg
Dabigatran 110 mg
Warfarin
ROCKET-AF
Rivaroxaban
Warfarin
ARISTOTLE
Apixaban
Warfarin
ENGAGE AF-TIMI 48
Edoxaban 60 mg
Edoxaban 30 mg
Warfarin

Ischemic Stroke

Systemic Embolism

Myocardial Infarction

Hemorrhagic Stroke

Adjusted
Major Bleeding

1.00

0.61

0.89

3.23

0.63

0.92
1.34
1.21

0.07
0.07
0.10

0.72
0.73
0.57

0.32
0.39
1.23

2.02
1.72
1.97

1.40
1.52

0.02
0.12

0.81
1.00

0.84
1.42

2.10
1.86

0.97
1.05

0.05
0.06

0.47
0.54

0.78
1.52

1.19
1.65

1.25
1.77
1.25

0.05
0.09
0.07

0.62
0.79
0.67

0.84
0.52
1.52

1.57
0.91
1.86

Adjusted major bleeding major bleeding e hemorrhagic stroke.


ARISTOTLE Apixaban for Reduction In STroke and Other ThromboemboLic Events in atrial brillation; ENGAGE AF-TIMI 48 Effective aNticoaGulation
with factor xA next GEneration in Atrial Fibrillation-Thrombolysis In Myocardial Infarction study 48; RE-LY Randomized Evaluation of Long-Term
Anticoagulation Therapy; ROCKET-AF Rivaroxaban Once Daily, Oral, Direct Factor Xa Inhibition Compared With Vitamin K Antagonism for Prevention of
Stroke and Embolism Trial in Atrial Fibrillation.

all-cause mortality, the overall composite outcome also was


reduced by dabigatran 150 mg.
Dabigatran 110 mg shows better efcacy than warfarin
only for the composite of disabling stroke life-threatening
bleeding, both alone and combined with cardiovascular or
all-cause mortality, probably because of the good performance of this drug regimen on bleeding. However, it does
not show advantages in reductions of the other outcomes,
possibly because of the higher number of the combination of
ischemic strokes and myocardial infarctions versus warfarin.
Edoxaban 60 mg shows better efcacy than warfarin in
reducing most of the outcomes, with the exception of the
composite of ischemic hemorrhagic stroke, probably
driven by the similar number of ischemic strokes versus
warfarin observed in the ENGAGE AF-TIMI 48 trial,
despite a reduction in hemorrhagic strokes. However, when
combined with cardiovascular or all-cause mortality, the
ischemic stroke hemorrhagic stroke outcome is reduced
by edoxaban 60 mg.
The lower-dose regimen with edoxaban 30 mg is disadvantaged compared with warfarin in the composite of
ischemic stroke hemorrhagic stroke. However, when this
composite is combined with cardiovascular and all-cause
mortality, edoxaban 30 mg performs better than warfarin.
The other composite outcomes are all reduced by edoxaban
30 mg, mainly in the lower risk of hemorrhagic events of
this treatment compared with warfarin.
Rivaroxaban demonstrates better efcacy than warfarin
for the composite of disabling stroke life-threatening
bleeding alone and combined with cardiovascular and allcause mortality. On the other hand, rivaroxaban does not
seem to be superior to warfarin in preventing the outcome of
ischemic stroke hemorrhagic stroke, probably because of

the similar number of ischemic strokes observed in the


ROCKET AF trial in the rivaroxaban and warfarin arms, but
it seems to perform better than warfarin when this outcome
is combined with cardiovascular and all-cause mortality. In
the analysis of the overall composite outcome, rivaroxaban
does not seem to be better than warfarin, possibly because of
the similar number of ischemic strokes and major bleeds,
unless such an overall outcome is combined with all-cause
mortality, favorably affected by rivaroxaban.
Overall, our results agree broadly with those of the phase
III clinical trials comparing non-vitamin K antagonist oral
anticoagulants with warfarin, in which, with the exception
of ROCKET AF trial, composite outcomes of ischemic and
bleeding events often have been used as a measure of net
clinical benet.2-4

Weighed Net Clinical Benet


In the model of weighed net clinical benet we have presented, all non-vitamin K antagonist oral anticoagulants seem
to be associated with a signicantly lower rate of the composite event of ischemic stroke equivalentsand a lower
number needed to treatcompared with warfarin. This
means that all non-vitamin K antagonist oral anticoagulants
seem to be better than warfarin when we consider the overall
balance between ischemic and hemorrhagic events and
correct the impact of these events for subsequent mortality.
The weighed analysis allows all non-vitamin K antagonist
oral anticoagulants to perform better than warfarin in terms of
efcacy/safety balance probably because the estimated
weight of a hemorrhagic stroke is approximately 3 times
higher than the weight of ischemic stroke, and a substantial
part of the advantage of non-vitamin K antagonist oral

Renda et al

Net Clinical Benet of Non-Vitamin K Antagonist Oral Anticoagulants Versus Warfarin

Figure 2 Net clinical benet and 95% CI of all treatment


arms of non-vitamin K antagonist oral anticoagulants versus
warfarin tested in phase III clinical trials for the weighed
composite outcome of ischemic stroke systemic embolism
myocardial infarction hemorrhagic stroke adjusted major
bleeding (major bleeding minus hemorrhagic stroke). Net
clinical benet is expressed as ischemic stroke equivalents
prevented per 100 person-years using ischemic stroke as the
reference event (weight 1). CI condence interval; IS
ischemic stroke; NNT number needed to treat (to prevent all
grouped events included in the net clinical benet evaluation,
per year of treatment); NOAC non-vitamin K antagonist
oral anticoagulant.

anticoagulants versus warfarin relates to the prevention of


intracranial hemorrhage.
Because we could not calculate the weights of each event
considered in the specic trial populations, we chose the
weights from a recent analysis of net clinical benet using the
combined data from the ACTIVE A and RE-LY trial databases,21 in which all patients were treated with anticoagulants
or antiplatelet drugs. We specically chose the weights estimated for the patients taking an anticoagulant, because this
is the estimate more pertinent to the populations examined.
The weighing factor of 1.5 used for intracranial hemorrhage
in the ATRIA cohort,23 reecting the higher impact on
mortality of an intracranial hemorrhage while receiving
warfarin versus experiencing an ischemic stroke while not
receiving warfarin, is not appropriate for the populations in all
the trials we analyzed, because all received anticoagulation.
However, because the selection of weights could affect
analysis results, as recently discussed,25 we also performed a
sensitivity analysis using weights increased and decreased
by 25% and 50% compared with the main analysis (Presented
in Table E2, available online). These analyses yielded similar
results, conrming the robustness of our ndings.
Our analysis does not provide a statistical comparison of
the various treatment arms in the different studies, which
would be problematic in any case because of the different
populations studied. However, some differences among
non-vitamin K antagonist oral anticoagulants cannot go
unnoticed. Across the 4 trials we analyzed, apixaban
apparently confers the highest risk reduction (the lowest
number needed to treat) to prevent the composite weighed
outcome of ischemic stroke systemic embolism

1013

myocardial infarction hemorrhagic stroke adjusted


major bleeding. Closest to apixaban is edoxaban 30 mg,
followed by edoxaban 60 mg, dabigatran 150 mg, dabigatran 110 mg, and rivaroxaban. This goes qualitatively in the
same direction as the results of the previous unweighed
analyses but achieves formal statistical signicance for all
non-vitamin K antagonist oral anticoagulants at variance
from the unweighed analysis. Of note, the weighed net
clinical benet analysis yields edoxaban 30 mg (and not
edoxaban 60 mg) as the next most favorable composite
evaluation after apixaban, which would revert the impression of edoxaban 30 mg as an unfavorable choice compared
with warfarin if restricting the analysis to efcacy end points
only,3,18 highlighting the higher level of integrated information obtained with such analyses.
Finally, the practical message from this analysis is that
the choice of the antithrombotic drug in atrial brillation has
to integrate the efcacy of each drug in terms of prevention
of thromboembolic events and safety in terms of hemorrhagic complications, considering that bleeding may have an
important impact on survival.

Study Limitations
For our analysis, we have used only published data from the
4 phase III trials comparing non-vitamin K antagonist oral
anticoagulants with warfarin for the prevention of thromboembolism in nonvalvular atrial brillation. We had no
access to undisclosed information; thus, this is not a patientlevel analysis, but an analysis based on published data.
However, different approaches are not likely to change the
main conclusions of our study.
We arbitrarily chose some composites of ischemic and
hemorrhagic events, which are the most relevant ones in such
trials. However, not all events were reported similarly across
the 4 trials. In such cases, as described in the Materials and
Methods section, we indirectly determined the number of
events not overtly reported in the trials and then calculated the
event rates according to the median follow-up of each trial.
Because the denition of life-threatening bleeding was
available only in the RE-LY and ENGAGE AF-TIMI 48
trials, we considered the most similar denitions for the 2
other trials, which were Global Strategies for Opening
Occluded Coronary Arteries severe bleeding (GUSTO) for
ARISTOTLE and critical bleeding for ROCKET-AF. Minor
differences among the events included under each heading
would not substantially affect the outcomes, robustness, and
nal messages of our analysis.
For the determination of net clinical benet, we could not
directly calculate the weights of events in the specic trial
populations because we did not know the rates of subsequent death associated with the event considered. Therefore,
we used the weights from a recent analysis of net clinical
benet of the 2 doses of dabigatran in the RE-LY trial,21
assuming that the impact of each event on mortality was
comparable in the 4 trial populations. This may affect the
estimates, although most likely not to a substantial extent.

1014

The American Journal of Medicine, Vol 128, No 9, September 2015

CONCLUSIONS
The assessment of a weighed net clinical benet according
to the impact of ischemic and hemorrhagic events on subsequent mortality indicates that all non-vitamin K antagonist oral anticoagulants have a better efcacy/safety prole
than warfarin in patients with atrial brillation. Among
non-vitamin K antagonist oral anticoagulants, apixaban and
edoxaban 30 mg seem to confer the lowest risk of a composite outcome, including ischemic stroke systemic
embolism myocardial infarction hemorrhagic stroke
adjusted major bleeding.

References
1. Husted S, de Caterina R, Andreotti F, et al. Non-vitamin K antagonist
oral anticoagulants (NOACs): no longer new or novel. Thromb Haemost. 2014;111:781-782.
2. Connolly SJ, Ezekowitz MD, Yusuf S, et al. Dabigatran versus
warfarin in patients with atrial brillation. N Engl J Med. 2009;361:
1139-1151.
3. Giugliano RP, Ruff CT, Braunwald E, et al. Edoxaban versus warfarin
in patients with atrial brillation. N Engl J Med. 2013;369:2093-2104.
4. Granger CB, Alexander JH, McMurray JJ, et al. Apixaban versus
warfarin in patients with atrial brillation. N Engl J Med. 2011;365:
981-992.
5. Patel MR, Mahaffey KW, Garg J, et al. Rivaroxaban versus warfarin in
nonvalvular atrial brillation. N Engl J Med. 2011;365:883-891.
6. De Caterina R, Husted S, Wallentin L, et al. New oral anticoagulants in
atrial brillation and acute coronary syndromes: ESC Working Group
on Thrombosis-Task Force on Anticoagulants in Heart Disease position
paper. J Am Coll Cardiol. 2012;59:1413-1425.
7. Ruff CT, Giugliano RP, Braunwald E, et al. Comparison of the efcacy
and safety of new oral anticoagulants with warfarin in patients with
atrial brillation: a meta-analysis of randomised trials. Lancet.
2014;383:955-962.
8. Camm AJ, Lip GY, De Caterina R, et al. 2012 focused update of the
ESC Guidelines for the management of atrial brillation: an update of
the 2010 ESC Guidelines for the management of atrial brillation.
Developed with the special contribution of the European Heart Rhythm
Association. Eur Heart J. 2012;33:2719-2747.
9. January CT, Wann LS, Alpert JS, et al. 2014 AHA/ACC/HRS guideline for the management of patients with atrial brillation: a report of
the American College of Cardiology/American Heart Association Task
Force on Practice Guidelines and the Heart Rhythm Society. Circulation. 2014;130:e199-e267.
10. De Caterina R, Connolly SJ, Pogue J, et al. Mortality predictors and
effects of antithrombotic therapies in atrial brillation: insights from
ACTIVE-W. Eur Heart J. 2010;31:2133-2140.
11. Connolly S, Pogue J, Hart R, et al. Clopidogrel plus aspirin versus oral
anticoagulation for atrial brillation in the Atrial brillation Clopidogrel Trial with Irbesartan for prevention of Vascular Events
(ACTIVE W): a randomised controlled trial. Lancet. 2006;367:
1903-1912.

12. Connolly SJ, Ezekowitz MD, Yusuf S, Reilly PA, Wallentin L. Newly
identied events in the RE-LY trial. N Engl J Med. 2010;363:1875-1876.
13. Connolly SJ, Pogue J, Hart RG, et al. Effect of clopidogrel added to
aspirin in patients with atrial brillation. N Engl J Med. 2009;360:
2066-2078.
14. Connolly SJ, Eikelboom JW, Ng J, et al. Net clinical benet of adding
clopidogrel to aspirin therapy in patients with atrial brillation for
whom vitamin K antagonists are unsuitable. Ann Intern Med.
2011;155:579-586.
15. ROCKET AF Study Investigators. Rivaroxaban-once daily, oral, direct
factor Xa inhibition compared with vitamin K antagonism for prevention
of stroke and Embolism Trial in Atrial Fibrillation: rationale and design
of the ROCKET AF study. Am Heart J. 2010;159:340-347, e1.
16. Ezekowitz MD, Connolly S, Parekh A, et al. Rationale and design of RELY: randomized evaluation of long-term anticoagulant therapy, warfarin,
compared with dabigatran. Am Heart J. 2009;157:805-810, 10 e1-2.
17. Lopes RD, Alexander JH, Al-Khatib SM, et al. Apixaban for reduction
in stroke and other ThromboemboLic events in atrial brillation
(ARISTOTLE) trial: design and rationale. Am Heart J. 2010;159:
331-339.
18. Ruff CT, Giugliano RP, Antman EM, et al. Evaluation of the novel
factor Xa inhibitor edoxaban compared with warfarin in patients with
atrial brillation: design and rationale for the Effective aNticoaGulation
with factor xA next GEneration in Atrial Fibrillation-Thrombolysis In
Myocardial Infarction study 48 (ENGAGE AF-TIMI 48). Am Heart J.
2010;160:635-641.
19. Boehringer Ingelheim Dabigatran Brieng Document. Silver Spring,
MD: Food and Drug Administration; August 27, 2010.
20. Rothman K. Epidemiology: An Introduction. 2nd ed. Oxford: Oxford
University Press; 2012.
21. Eikelboom JW, Connolly SJ, Hart RG, et al. Balancing the benets and
risks of 2 doses of dabigatran compared with warfarin in atrial brillation. J Am Coll Cardiol. 2013;62:900-908.
22. R Development Core Team. R: A Language and Environment for
Statistical Computing 2011; Vienna, Austria: the R Foundation for
Statistical Computing.
23. Singer DE, Chang Y, Fang MC, et al. The net clinical benet of
warfarin anticoagulation in atrial brillation. Ann Intern Med.
2009;151:297-305.
24. Banerjee A, Lane DA, Torp-Pedersen C, Lip GY. Net clinical benet
of new oral anticoagulants (dabigatran, rivaroxaban, apixaban) versus
no treatment in a real world atrial brillation population: a modelling
analysis based on a nationwide cohort study. Thromb Haemost.
2012;107:584-589.
25. Ciolino JD, Carter RE. Reanalysis or redenition of the hypothesis?
Eur Heart J. 2015;36:340-341.

APPENDIX
Supplemental gure and tables accompanying this article
can be found in the online version at 10.1016/j.amjmed.
2015.03.034.

Renda et al

Net Clinical Benet of Non-Vitamin K Antagonist Oral Anticoagulants Versus Warfarin

APPENDIX

Figure E1 RR and 95% CI of all treatment arms versus


warfarin for the composite outcome including unweighed
ischemic stroke systemic embolism myocardial
infarction hemorrhagic stroke adjusted major bleeding
(major bleeding minus hemorrhagic stroke) all-cause death
in the 4 phase III trials of a non-vitamin K antagonist oral
anticoagulant versus warfarin in nonvalvular atrial brillation.
Contrary to the analyses presented in the article, the current
analysis has to be considered exploratory, because it may
include nonmutually exclusive events. NOAC non-vitamin
K antagonist oral anticoagulant.

1014.e1

Ischemic Stroke Hemorrhagic Stroke

Disabling Stroke Life-threatening Bleeding

Ischemic Stroke Hemorrhagic


Stroke Myocardial Infarction
Systemic Embolism Adjusted
Major Bleeding

Treatment

Cardiovascular Mortality

All-cause Mortality

Cardiovascular Mortality

All-cause Mortality

Cardiovascular Mortality

All-cause Mortality

Dabigatran 150 mg

0.78 (0.68-0.89)
<.001
0.91 (0.80-1.02)
.128
0.85 (0.74-0.98)
.025
0.85 (0.75-0.96)
.007
0.87 (0.79-0.95)
.004
0.95 (0.87-1.04)
.284

0.82 (0.74-0.92)
<.001
0.91 (0.81-1.01)
.086
0.83 (0.72-0.95)
.006
0.87 (0.79-0.95)
.004
0.91 (0.83-0.98)
.025
0.94 (0.87-1.02)
.181

0.83 (0.74-0.93)
.002
0.85 (0.76-0.95)
.006
0.79 (0.68-0.92)
.002
0.75 (0.66-0.85)
<.001
0.82 (0.74-0.91)
<.001
0.83 (0.74-0.92)
<.001

0.86 (0.77-0.95)
.003
0.87 (0.78-0.96)
.006
0.77 (0.67-0.89)
<.001
0.81 (0.73-0.89)
<.001
0.88 (0.80-0.96)
.005
0.85 (0.77-0.93)
<.001

0.90 (0.83-0.98)
.034
0.92 (0.84-1.01)
.092
0.91 (0.83-1.00)
.064
0.81 (0.75-0.89)
<.001
0.87 (0.81-0.93)
<.001
0.85 (0.79-0.92)
<.001

0.91 (0.84-0.98)
.035
0.92 (0.85-1.00)
.068
0.90 (0.82-0.98)
.024
0.83 (0.77-0.90)
<.001
0.89 (0.83-0.96)
.001
0.86 (0.81-0.92)
<.001

Dabigatran 110 mg
Rivaroxaban
Apixaban
Edoxaban 60 mg
Edoxaban 30 mg

Table E2

Sensitivity Analysis of Net Clinical Benet of Non-vitamin K Antagonist Oral Anticoagulants Versus Warfarin, Using Increased and Decreased Weights by 25% and 50%
Weights of the Main Analysis

25% Decreased Weights

25% Increased Weight

50% Decreased Weights

50% Increased Weights

Treatment

IS equivalents (95% CI)

IS equivalents (95% CI)

IS equivalents (95% CI)

IS equivalents (95% CI)

IS equivalents (95% CI)

Dabigatran 150 mg
Dabigatran 110 mg
Rivaroxaban
Apixaban
Edoxaban 60 mg
Edoxaban 30 mg

1.02
0.82
0.74
1.36
1.04
1.29

0.84
0.58
0.59
1.04
0.78
0.84

1.21
1.06
0.90
1.68
1.30
1.74

0.66
0.35
0.43
0.72
0.52
0.38

1.39
1.30
1.05
2.00
1.56
2.19

(1.56
(1.37
(1.29
(1.80
(1.48
(1.72

to
to
to
to
to
to

0.48)
0.27)
0.17)
0.92)
0.61)
0.86)

(1.32
(1.07
(1.08
(1.43
(1.17
(1.23

to
to
to
to
to
to

0.37)
0.10)
0.09)
0.65)
0.40)
0.46)

(1.78
(1.63
(1.48
(2.15
(1.76
(2.20

to
to
to
to
to
to

0.63)
0.48)
0.30)
1.21)
0.84)
1.29)

(1.07
(0.70
(0.85
(1.06
(0.86
(0.73

Net clinical benet is expressed as ischemic stroke equivalents prevented per 100 person-years using ischemic stroke as the reference event (weight 1).
CI condence interval; IS ischemic stroke.

to
to
to
to
to
to

0.24)
0.00)
0.01)
0.38)
0.18)
0.04)

(2.0 to 0.77)
(1.91 to 0.68)
(1.69 to 0.43)
(2.51 to 1.50)
(2.06 to 1.08)
(2.68 to 1.72)

The American Journal of Medicine, Vol 128, No 9, September 2015

RR and 95% CI of each treatment arm versus warfarin are reported in each case, with P value immediately below, for the various composite outcomes. Statistically signicant P values are in bold.
Adjusted major bleeding major bleeding  hemorrhagic stroke.

1014.e2

Table E1 Rate Ratio and Corresponding 95% Condence Interval for Each Treatment Arm of the Various Trials Versus Warfarin for the Various Composite Outcomes Combined with
Cardiovascular and All-cause Mortality