A novel approach indirectly comparing benefit–risk balance across

European Heart Journal – Cardiovascular Pharmacotherapy (2015) 1, 15–28
doi:10.1093/ehjcvp/pvu007
ORIGINAL ARTICLE
Thrombosis and anti-thrombotic therapy
A novel approach indirectly comparing
benefit–risk balance across anti-thrombotic
therapies in patients with atrial fibrillation
Ariel Dogliotti 1,2 and Robert P. Giugliano 3*
1
Unidad de Epidemiologı´a Clı´nica y Estadı´stica, Grupo Oron˜o, Rosario, Argentina; 2Instituto Cardiovascular de Rosario, Rosario, Argentina; and 3Brigham and Women’s Hospital,
Harvard Medical School, TIMI Study Group, 350 Longwood Avenue, 1st Floor Offices, Boston 02115, MA, USA
Received 7 August 2014; revised 14 September 2014; accepted 7 October 2014
Background
Anti-thrombotic reduces thromboembolic events but increases bleeding in patients with atrial fibrillation (AF). We
evaluated the benefit –risk of anti-platelet and anti-coagulant therapies, weighing these conflicting effects of treatment.
.....................................................................................................................................................................................
Methods
Randomized controlled trials in patients with AF were identified from MEDLINE, Embase, and Cochrane Central Register
of Controlled Trials through April 2014. We performed a stochastic multi-criteria acceptability analysis, which allowed us
to compute a comprehensive benefit –risk profile. In the primary analysis, we used prior established rankings of mortality,
intracranial haemorrhage, ischaemic stroke, myocardial infarction, major extracranial haemorrhage, and systemic embolism based on utility functions. In sensitivity analyses, we explored: (i) rankings based on costs, (ii) bleeding ranked higher
than thromboembolism, and (iii) thromboembolism ranked higher than bleeding events.
.....................................................................................................................................................................................
Results
100 913 patients (21 studies) were allocated to placebo/control, aspirin and/or clopidogrel, vitamin K antagonists (VKAs),
or new oral anti-coagulants (NOACs). Based on utility, NOACs were better than VKA or anti-platelet therapy; dabigatran
150 mg was ranked highest (21% chance of being best). Ranked by cost, the 3 factor Xa inhibitors were very similar
(16 –18% chance of being best). When haemorrhagic events were weighted more than ischaemic events, edoxaban
30 mg was ranked higher (22%), while rivaroxaban (23%) was most preferred when ischaemic events were rated
worse than haemorrhagic events.
.....................................................................................................................................................................................
Conclusion
New oral anti-coagulants had a more favourable benefit–risk profile across a wide range of assumptions regarding
the relative importance of clinical events. Differences between NOACs were modest and depended upon the order
of ranking of clinical events.
----------------------------------------------------------------------------------------------------------------------------------------------------------Keywords
Anti-coagulant † Anti-platelet † Atrial fibrillation † Bleeding, † Haemorrhage † Novel oral anti-coagulant †
Stroke † Vitamin K antagonist
Introduction
In patients with atrial fibrillation (AF), Vitamin K antagonists (VKA),
such as warfarin, reduce the risk of stroke by 64% compared with
placebo, but increase the risk of haemorrhage.1 Anti-platelet agents
also significantly reduce the incidence of stroke,1,2 but are less effective than oral anti-coagulation. A meta-analysis of trials comparing
warfarin vs. aspirin (ASA) showed better stroke reduction with warfarin, but a higher rate of intracranial bleeding, with no difference in
total mortality.1 In a more recent randomized trial,3 anti-coagulant
therapy was superior to the combination of clopidogrel plus ASA
in reducing stroke, with no difference in bleeding.
Several novel oral anti-coagulants (NOACs) have been developed
in an attempt to overcome the many challenges faced by clinicians
and patients with the use of warfarin.4 – 6 Dabigatran,7 rivaroxaban,8
apixaban,9 and edoxaban10 were each ‘at least as effective as’ warfarin
in reducing the risk of stroke or embolism. Not only did these new
agents meet criteria for non-inferiority but in addition dabigatran
150 mg twice daily and apixaban 5 mg twice daily were superior to
warfarin in reducing the composite of stroke or systemic embolism.
Each of the NOACs markedly reduced intracranial and fatal bleeding
when compared with warfarin titrated to a target INR of 2.0 –3.0.
Finally, in a trial comparing apixaban with ASA in patients not eligible
or willing to take a VKA, apixaban reduced the risk of stroke or
* Corresponding author. Tel: +1 617 278 0145, Fax: +1 617 734 7320, Email: rgiugliano@partners.org
Published on behalf of the European Society of Cardiology. All rights reserved. & The Author 2015. For permissions please email: journals.permissions@oup.com
16
systemic embolism compared with ASA monotherapy without
significantly increasing the risk of major bleeding or intracranial
haemorrhage.11
In a recent meta-analysis of large phase III trials of NOAC compared with warfarin in patients with AF,12 statistically significant
reductions with NOAC were observed in the primary outcome of
stroke or systemic embolism (19% relative reduction), as well as allcause mortality (10%) and haemorrhagic stroke (51%). However,
standard meta-analyses are unable to integrate all available randomized evidence in one analysis that also includes indirect comparisons between therapies. In contrast, network meta-analysis allows
a unified, coherent analysis of all randomized controlled trials that
compare anti-thrombotic drugs head to head or with placebo/
control while fully respecting randomization.13 Recently, a network
meta-analysis14 that included three of the NOACs showed that
dabigatran 150 mg was ranked highest in prevention of stroke
(70% chance of being the best), the composite of ischaemic stroke
or systemic embolism (60%), and mortality (31%). Rivaroxaban
(30%) and ASA plus clopidogrel had the highest likelihood of major
bleeding (29%).
Several large clinical trials tested the efficacy and safety of different
anti-platelet and anti-coagulant strategies. In four of these studies:
ACTIVE (W),3 RE-LY,7 ARISTOTLE,9 and ENGAGE AF-TIMI 48,10
the benefit –risk was assessed by comparing the NOACs vs. warfarin
on net clinical outcomes that combined a variety of thromboembolic
events and bleeding with mortality.
However, use of net clinical outcomes can be an imprecise approach,
as it does not consider the weight or clinical importance of each
Figure 1 Flow diagram of literature search.
A. Dogliotti and R.P. Giugliano
outcome. Although such a measure is easy to interpret and implement
in clinical practice, drug benefit–risk analysis typically includes multiple
benefit and risk criteria and consequently must include value judgements. Moreover, when multiple alternatives (treatments) and different
criteria (efficacy and safety outcomes) are involved, a more complex
model is required. Choosing anti-thrombotic therapy is difficult
because each therapy may have different strengths and weaknesses.
We performed a network meta-analysis and a stochastic multicriteria acceptability analysis (SMAA), which allows to compute the
typical value judgement that support a decision, to quantify decision
uncertainly, and to compute a comprehensive benefit –risk profile to
answer the following question: What anti-thrombotic treatment in
patients with AF has the best balance of benefit and risk?
Material and methods
Search strategy
We searched Medline, Embase, and the Cochrane database of systematic
reviews through April 2014 with no language restriction using the following Medical Subject Heading and keywords: anti-coagulant, anti-platelet,
ASA, clopidogrel, warfarin, VKAs, dabigatran, apixaban, rivaroxaban,
edoxaban, AF, atrial arrhythmias, coupled with outcome searched
using the terms stroke, cerebrovascular accident, and transient ischaemic
attack. We also reviewed the reference lists of published meta-analyses of
anti-coagulant and anti-platelet therapies to prevent stroke and embolic
events in patients with AF. Overall, 21 manuscripts were included in our
analyses (Figure 1). We registered our review online with METCARDIO
(registration #1– 2014).
Benefit– risk balance of anti-thrombotics in atrial fibrillation
Eligibility of studies
Inclusion criteria for eligible studies were: (i) randomized controlled trials
of adjusted-dose VKA, ASA, clopidogrel, or NOACs in patients with nonvalvular AF; (ii) intention-to-treat analysis; (iii) minimum 1 year follow-up.
Data abstraction and quality assessment
Two independent reviewers performed data extraction. We used consensus to resolve discrepancies. The endpoints of interest were: mortality from any cause, ischaemic stroke, myocardial infarction, and systemic
embolism (efficacy outcomes); intracranial haemorrhage and major
extracranial haemorrhage (safety outcomes). We used data from the
intention-to-treat analyses.
Statistical analysis
Network meta-analysis was carried out in the Bayesian framework using
Markov Chain Monte Carlo simulation15 in the GeMTC R package.16 The
analysis pooled relative effects on the odds ratio scale using the binomial
likelihood and logit link function. GeMTC automatically specified vague
prior distributions for the trial baseline effects, the relative effects, and
the random-effects standard deviation.
First, a random-effects model network meta-analysis was performed.
The statistical analysis is based on binomial likelihoods with a logit link
function. GeMTC automatically specified vague prior distributions for
the trial baseline effects, the relative effects (normal with mean 0 and
standard deviation 37.5), and the random-effects standard deviation
(uniform in the interval 0–2.5). We used a technique known as ‘nodesplitting’17 to evaluate for inconsistency in the findings of the network
17
meta-analysis coming from direct vs. indirect evidence. Node-splitting
assesses whether direct and indirect evidence on a specific node
(the split node) are in agreement.
Second, we used a SMAA-218 – 23 as a new and more elaborate approach to drug benefit – risk analysis to overcome the limitations of the
net clinical benefit approach. Stochastic multi-criteria acceptability analysis was carried out based on the estimated incidence distributions of
each of the events of interest (criteria). The incidence distributions
were derived from relative effect estimates obtained from the network
meta-analyses, combined with baseline incidence estimates for Warfarin
treatment.
We performed two main analyses in a pre-specified order based on
the authors’ judgement. The first used preference information based
on utility functions.24 – 34 A utility function is a generic multi-attribute
preference-based measure of health status and health-related quality of
life that is widely used as an outcome measure in clinical studies, in population health surveys, in the estimation of quality-adjusted life years, and in
economic evaluations. Using utility functions, the events in order from
lowest (worst) to highest (best) were death from any cause, intracranial
haemorrhage, ischaemic stroke, myocardial infarction, major extracranial
haemorrhage, and systemic embolism.
The second main analysis ordered events based on medical costs.35 – 37
Here death had the highest cost and was followed by major extracranial
haemorrhage, ischaemic stroke, myocardial infarction, intracranial haemorrhage, and systemic embolism.
We also performed two sensitivity analyses that maintained death as
the lowest state and then ranked the non-fatal events as follows: (i) assigning the bleeding events lower health states than thromboembolic events
Figure 2 Network of eligible comparisons for the multiple-treatment meta-analysis for efficacy. The numbers represent the number of trials or
arms comparing each pair of treatment.
18
and (ii) assigning the thromboembolic events lower health states than
bleeding events. These were chosen as sensitivity analyses since the
two scenarios represent the extremes in ordering of the risks (bleeding)
and benefits (thromboembolic events) of anti-thrombotic therapy.
Within each group of bleeding and thromboembolic events, individual
endpoints were ordered according to their utility function.
We also ranked the various anti-thrombotic treatments in terms of
their likelihood of leading to the best results for each outcome (also
known as the ‘rank acceptability index’). In this approach, the share of
all possible values of the weight vector W and the joint random vector
X for which alternative i is ranked at place r is calculated. The result can
be interpreted as the probability that alternative i is ranked at place r,
where 0 indicates that the alternative will never obtain rank r and 1 indicates that alternative i will always obtain rank r. The rank acceptability
index is computed numerically as a multi-dimensional integral over the
criteria distributions and the favourable rank weights.
The preferred (best) alternatives are those with the highest probabilities for the best rank. The confidence factor (CF) is the probability for an
alternative to obtain the first rank when the central weight vector is
chosen. The CF is computed as a multi-dimensional integral over the criteria distributions.
Results
There were 100 913 patients with non-valvular AF, treated with one
of 10 different therapies: ASA, ASA plus clopidogrel, dabigatran
150 mg twice daily, dabigatran 110 mg twice daily, adjusted dose of
VKA, rivaroxaban, apixaban, edoxaban 30 mg once daily, edoxaban
60 mg once daily, and control/placebo (Figure 2). The main study
and baseline characteristics are listed in Table 1.
The 10 therapies resulted in 45 theoretical two-way comparisons
for each of six outcomes (yielding a total of 6 × 45 ¼ 270 evaluations). The following 14 comparisons were direct (Figure 2): VKA
vs. control/placebo, ASA vs. control/placebo, VKA vs. ASA, ASA vs.
ASA plus clopidogrel, ASA plus clopidogrel vs. VKA, VKA vs. apixaban, ASA vs. apixaban, VKA vs. dabigatran 150 mg, VKA vs. dabigatran
110 mg, dabigatran 150 mg vs. dabigatran 110 mg, rivaroxaban vs.
VKA, edoxaban 30 mg vs. VKA, edoxaban 60 mg vs. VKA, and edoxaban 30 mg vs. edoxaban 60 mg.
Findings
Network meta-analyses
We built a network meta-analysis in order to synthesize the outcomes for each treatment and endpoint. The summary results are
shown in Table 2, and details of the network meta-analysis are provided in Table A1.
Analyses with preference information
When utility values were used to order outcomes (Figure 3, Table A2),
dabigatran 150 mg obtained 20% of chance to rank first, followed
closely by the other five NOAC regimens. Meanwhile VKA, antiplatelet regimens, and placebo/control had a 0–1% probability of the
first rank. Placebo/control had a 74% chance to be ranked worst
(rank 10), while ASA alone and ASA plus clopidogrel had the highest
probabilities to be ranked 9th (49 and 23%, respectively), while VKA
occupied the central ranks (82% probability of being ranked 4th to 7th).
When the order of events was based on medical costs (Figure 4,
Table A2), the chance of been ranked first changed little: rivaroxaban,
A. Dogliotti and R.P. Giugliano
dabigatran 150 mg, and edoxaban 30 mg each achieved 18% of being
ranked first. Again, placebo/control and anti-platelet therapy ranked
worst. Vitamin K antagonist had an identical performance whether
using medical costs or patient utilities.
Figure 4A and B and Table A3 show the CFs. Again in both scenarios,
the NOACs were the best options. Based on utility, all NOACs were
similar, although dabigatran 150 mg obtained modestly the first probability to be the best alternative (21% chance of been the best).
Aspirin was the worst alternative (0%). Based on medical costs,
NOACs were closely similar: edoxaban 30 mg 19%, rivaroxaban
and dabigatran 150 mg 18%, VKA (1%), ASA (0%) or ASA plus clopidogrel (1%) were similar to placebo/control (1%). An analysis based
on the central weights and CFs showed that the first-rank acceptabilities are unlikely to change if more precise preference information
(e.g. an exact weighting) were given.
In a sensitivity analysis, ranking all the bleeding events higher than
thromboembolic events resulted in edoxaban 30 mg scoring the
highest CF (22%), while dabigatran 150 mg, rivaroxaban, apixaban
and edoxaban 60 mg each scored 15%. In the second sensitivity analysis that ranked ischaemic events highest, rivaroxaban (23%) and
dabigatran 150 mg (22%) were ranked highest, while dabigatran
110 mg and edoxaban 30 mg (both 10%) were ranked lowest
among the six NOAC regimens. In both sensitivity analyses, antiplatelet regimens and VKA rated equal or slightly worse than
placebo/control.
Consistency of network model
No major inconsistencies or qualitative differences (e.g. change in directionality of the estimate) were observed when we compared the
effect estimates based on direct vs. indirect evidence from the comparisons, supporting the robustness of the model.
Discussion
In this analysis of data from randomized clinical trials of antithrombotic agents in patients with non-valvular AF, we observed
that the benefit– risk including both efficacy and safety endpoints
favoured NOACs over either VKA or anti-platelet regimens. The
benefit of NOACs over other therapies was consistent whether clinical events were ordered based on utility, costs, with greatest weight
for thrombotic events, or with greatest weight for haemorrhagic
events.
Since individual trials used a composite of all stroke and systemic
embolism as the primary efficacy outcome, and major bleeding as
the principal safety outcome, performing a benefit–risk analysis
can be tricky. For example, the endpoint of ‘all stroke’ consists of ischaemic stroke (an event prevented by anti-thrombotic therapy) and
haemorrhagic stroke (an event that may be a consequence of
anti-thrombotic therapy). Moreover, major bleeding consists of
serious bleeding events that may be intra- or extracranial in location,
with all haemorrhagic stroke qualifying as a major bleed by definition.
This leads to the possibility of ‘double counting’ the importance of
haemorrhagic stroke if the traditional primary efficacy and safety endpoints are not carefully analysed. In our analysis, we have differentiated each of the major clinical events and also explored different
orders of clinical impact in order to provide a more accurate and
Main study and baseline characteristics
Clinical trial, year
Size (n)
Age (mean)
Male (%)
HTN (%)
DM (%)
Prior MI (%)
HF (%)
Prior TIA or
stroke (%)
Target INR
Median follow-up
(months)
Lost to
follow-up (%)
24
24
NR
0
.............................................................................................................................................................................................................................................
AFASAK-I38, 1989
BAATAF39, 1990
1007
420
CAFA40, 1991
378
SPAF-I41, 1991
SPINAF42, 1992
421
525
EAFT43, 1993
439
SPAF-II44, 1994
AFASAK-II45, 1998
PATAF46, 1999
LASAF47, 1999
ACTIVE (W)3, 2006
JAST48, 2006
ACTIVE (A) 2, 2006
ATAFS49, 2006
BAFTA50, 2007
WASPO51, 2007
RE-LY7, 2009a
73.7
68
8.5
14.9
12.9
13.4
38.6
11.9
52
58.5
15.5
18.4
71.5
56
42.03
12.9
8.4
715
339
64
73.6
76
62
53
42
17
12
9
272
70.4
46
36.3
17
8.6
285
6706
66
70.2
52
66
51
82
8
21
38
14
65,5
67
74.79
31.5
51
72
100
871
67.7
53.5
72.52
65
70
7554
704
71
63.3
58
59.7
973
81.5
54
53.1
13.9
83
71
50
63
48
78.9
4
23.3
75
18 113
50.5
25.9
5
16.51
8
3.71
18.9
70
17
69.5
18
5
31
14
33
9
9.6
16.1
19.7
31.8
2.8–4.2
1.5–2.7
2–3
14.4
1.3
2–4.5
1.4–2.8
15.6
21
0
6
27.6
1.9
2–4.5
2–3
37.2
42
0.4
0
2.5–3.5
32.4
0
2–3
24
15.4
NR
0.33
25.2
NR
2–3
44.4
18.12
0.8
NR
13.1
2–3
32.4
0.82
20
2–3
2–3
13.2
24
NR
0.16
6
15
2.5–4
2.5
13
ROCKET-AF8, 2011
14 171
73
60.3
90.5
40
18
62
55
2–3
23.3
0.22
ARISTOTLE9, 2011
18 201
70
60.5
87.4
25
13.9
35
19.4
2–3
21.6
0.4
AVERROES11, 2011b
ENGAGE AF-TIMI 4810, 2013
5599
21 105
70
72
59
62
86.5
93.6
19
36.1
12
39
57.5
14
28.3
2–3
21
33.6
NR
0.05
Benefit– risk balance of anti-thrombotics in atrial fibrillation
Table 1
DM, diabetes mellitus; HF, heart failure; HTN, hypertension; NR, not reported; TIA, transient ischaemic attack.
a
RE-LY, corrected data.61
Patients were eligible for this study if their physicians considered VKA therapy to be unsuitable for them. The difficulty or anticipated difficulty of maintaining the INR in the therapeutic range was a major reason for the unsuitability of vitamin K
antagonist therapy in this study.
b
19
20
Table 2
Results of network meta-analyses
Drug, better than
Mortality
OR (95% CI)
Ischaemic stroke
OR (95% CI)
Systemic embolism
OR (95% CI)
Intracranial haemorrhage
OR (95% CI)
Apixaban vs.
Control
0.58 (0.34, 0.87)
ASA
ASA+ Clop
Control
0.39 (0.25, 0.64)
0.53 (0.26, 0.92)
0.28 (0.16, 0.47)
ASA
ASA+ Clop
Control
0.27 (0.10, 0.74)
0.25 (0.06, 0.73)
0.23 (0.07, 0.65)
VKA
0.42 (0.16, 0.92)
Control
Control
0.71 (0.47, 0.98)
0.53 (0.32, 0.95)
ASA+ Clop
0.41 (0.15, 0.88)
ASA+ Clop
VKA
0.12 (0.01, 0.57)
0.16 (0.01, 0.65)
VKA
0.31 (0.10, 0.88)
ASA+ Clop
VKA
0.23 (0.06, 0.97)
0.30 (0.10, 0.88)
.............................................................................................................................................................................................................................................
ASA vs.
ASA+ Clop vs.
Control
Dabigatran 110 vs.
Control
0.56 (0.29, 0.96)
Control
0.35 (0.18, 0.73)
Dabigatran 150 vs.
Control
0.55 (0.27, 0.93)
Edoxaban 30 vs.
Control
0.54 (0.28, 0.90)
ASA
ASA+ Clop
Control
Control
0.34 (0.19, 0.71)
0.45 (0.21, 0.94)
0.24 (0.12, 0.50)
0.46 (0.23, 0.91)
Edoxaban 60 vs.
Control
0.57 (0.28, 0.95)
ASA
Control
0.44 (0.25, 0.92)
Rivaroxaban vs.
VKA vs.
Control
Control
ASA
0.57 (0.29, 0.96)
0.61 (0.43, 0.81)
0.79 (0.59, 0.94)
ASA
ASA
ASA+ Clop
Control
0.31 (0.16, 0.64)
ASA
ASA+ Clop
Control
0.25 (0.07, 0.97)
0.24 (0.05, 0.89)
0.22 (0.06, 0.79)
0.41 (0.22, 0.85)
0.45 (0.35, 0.63)
0.59 (0.37, 0.93)
0.32 (0.22, 0.46)
ASA
ASA+ Clop
Control
0.41 (0.20, 0.82)
0.37 (0.13, 0.83)
0.34 (0.16, 0.71)
Drugs of the first column are better than the therapies listed in the other columns to the right. There were no statistically significant differences between therapies for the endpoints of myocardial infarction and major extracranial haemorrhage (as
detailed results in Appendix). ASA, aspirin; Clop, clopidogrel; OR (95% CI), odds ratio (credibility interval); VKA, vitamin K antagonist.
A. Dogliotti and R.P. Giugliano
Benefit– risk balance of anti-thrombotics in atrial fibrillation
21
Figure 3 Preference information. (A) Based on utility functions:24 – 34 (means ‘worse than’) mortality intracranial haemorrhage ischaemic stroke
myocardial infarction major extracranial haemorrhage systemic embolism. (B) Based on medical costs:35 – 37 mortality major extracranial haemorrhage ischaemic stroke myocardial infarction intracranial haemorrhage systemic embolism. The rank acceptabilities represent, given the preference
information, the probability (in per cent) that each treatment is best (rank 1), the second highest (rank 2), etc. Rank probabilities sum to 100%, both
within a rank over treatments and within a treatment over ranks. Figure shows the cumulative probability (plotted on the y-axis) of having a rank that is
equal to or less than the ranking on the x-axis.
fair assessment of the various anti-thrombotic regimens in patients
with AF.
We used a SMAA that ranks the individual therapeutic interventions from most likely to least likely of being the best. We found
that the NOACs were most likely to be the best active therapy
with respect to each of the six major efficacy-safety outcomes of
mortality from any cause, intracranial haemorrhage, ischaemic
stroke, myocardial infarction, major extracranial haemorrhage, and
systemic embolism.
Several other network meta-analyses of NOACs in patients with
AF have been published,14,52 – 60 but none shared our approach of
ordering clinical endpoints to permit comparison of net outcomes
with the various anti-thrombotic strategies that have been tested in
clinical trials. Furthermore, we included data from the largest and
most recent trial in AF (ENGAGE AF-TIMI 48).10 In contrast with
standard pairwise meta-analyses,12 our approach evaluated weighted
ordering of events (with three sensitivity analyses exploring differing
approaches to order the events), integrated indirect comparisons,
and included trials with various comparator arms.
There have been no large randomized clinical trials directly
comparing new oral anti-coagulants in patients with AF to date.
Furthermore, it is impractical to perform head-to-head comparisons of the 10 anti-thrombotic strategies that have been tested
thus far in clinical trials. Thus, indirect analyses represent an important tool to evaluate the 45 different pairwise comparisons
of these 10 options.
While NOACs appeared the most favourable strategy in our analyses, we recognize these drugs have limitations. Currently, there is
no widely available assay to measure their anti-coagulant effect. In
cases of life-threatening bleeding or need for urgent surgery, currently there are no antidotes or specific reversible agents available for
clinical use. Finally, the greater drug costs of NOACs limits their availability in countries where the majority of the drug costs is borne by
the patient.
When interpreting the results of these analyses and considering
how an individual patient should be treated, it is important for the
clinician to consider and weigh the relative balance of thromboembolic events and risk of bleeding for the individual patient.
22
A. Dogliotti and R.P. Giugliano
Figure 4 Confidence factor. The confidence is the probability that an alternative ranks first if the decision-maker’s preferences agree with the
alternative’s central weight vector. (A) Based on utility functions. (B) Based on medical costs.
Limitations
We recognize that our approach to analysing benefit– risk has several
limitations. The findings based on clinical trial populations may not be
generalizable to clinical practice where patients tend to be older and
have more comorbidities (e.g. cancer, severe renal failure) that
exclude them from clinical trials. The studies in our analysis were heterogeneous with respect to the characteristics of the patients, the
definitions and assessment of some of the outcomes, the concomitant interventions allowed by the protocols, the length of, and differential loss in follow-up (Table 1). In addition, our analyses do not
account for other changes in clinical practice that have evolved
over the past 25 years. There were a limited numbers of some
events (e.g. systemic embolism) and certain therapies across the 21
studies were under-represented, which reduces the power to find
differences and may affect the precision of our results. Specifically,
given the small number of trials5 performed to date with the four
NOACs, our analyses are underpowered to determine whether important differences exist between individual NOACs. We did not
adjust for multiple comparisons, and thus difference of marginal statistical significance should be interpreted cautiously. We recognize
that events are not completely independent of one another, and
we did allow multiple events per patient. Nevertheless, we believe
that our approach offers clinicians a more comprehensive, quantitative,
and objective framework, as opposed to an intuitive approach, to evaluate the benefit–risk of 10 anti-thrombotic strategies in patients with AF.
Conclusions
In this benefit –risk analysis of anti-thrombotic therapies in patients
with AF using a multiple-treatment meta-analysis and stochastic
Benefit– risk balance of anti-thrombotics in atrial fibrillation
multi-objective acceptability procedures, we found that novel anticoagulants were the most preferred. None of the individual novel
agents should be considered superior to another based on these
data. Vitamin K antagonist and anti-platelet regimens appeared to
offer little advantage in efficacy or safety compared with NOACs.
Conflict of interest: R.G. is a member of the TIMI Study Group, who
has received research grant support from Johnson & Johnson and from
Daiichi-Sankyo related to clinical trials of anti-coagulants. He has received
honoraria for consultation and/or lectures from Bristol-Myers Squibb,
Daiichi-Sankyo, Johnson & Johnson, Pfizer and Portola.
References
1. Hart RG, Pearce LA, Aguilar MI. Meta-analysis: antithrombotic therapy to prevent
stroke in patients who have nonvalvular atrial fibrillation. Ann Intern Med 2007;
146:857 –867.
2. The ACTIVE Investigators. Effect of clopidogrel added to aspirin in patients with
atrial fibrillation. N Engl J Med 2009;360:2066 – 2078.
3. ACTIVE Writing Group of the ACTIVE Investigators. Clopidogrel plus aspirin versus
oral anticoagulation 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 –1912.
4. Zikria JC, Ansell J. Oral anticoagulation with factor Xa and thrombin inhibitors:
on the threshold of change. Curr Opin Hematol 2009;16:347 – 356.
5. Weitz JI. New oral anticoagulants in development. Thromb Haemost 2010;103:
62 –70.
6. Grip LT, Ruff CT, Giugliano RP. New oral antithrombotic strategies: 2013 update on
atrial fibrillation. Hot Topics 2013;31:7 –18.
7. Connolly SJ, Ezekowitz MD, Yusuf S et al. for the RE-LY Steering Committee
and Investigators. Dabigatran versus warfarin in patients with atrial fibrillation.
N Engl J Med 2009;361:1139 –1151.
8. Patel MR, Mahaffey KW, Garg J et al.. Rivaroxaban versus warfarin in nonvalvular atrial
fibrillation. N Engl J Med 2011;365:883–891.
9. Granger CB, Alexander JH, McMurray JJ et al.. Apixaban versus warfarin in patients
with atrial fibrillation. N Engl J Med 2011;365:981 – 992.
10. Giugliano RP, Ruff CT, Braunwald E, Murphy SA, Wiviott SD, Halperin JL, Waldo AL,
Ezekowitz MD, Weitz JI, Sˇpinar J, Ruzyllo W, Ruda M, Koretsune Y, Betcher J, Shi M,
Grip LT, Patel SP, Patel I, Hanyok JJ, Mercuri M, Antman EM, ENGAGE AF-TIMI
48 Investigators. Edoxaban versus warfarin in patients with atrial fibrillation.
N Engl J Med 2013;369:2093 –2104.
11. Connolly SJ, Eikelboom J, Joyner C, Diener HC, Hart R et al. Apixaban in patients with
atrial fibrillation. N Engl J Med 2011;364:806 –817.
12. Ruff CT, Giugliano RP, Braunwald E, Hoffman EB, Deenadayalu N, Ezekowitz MD,
Camm AJ, Weitz JI, Lewis BS, Parkhomenko A, Yamashita T, Antman EM. Comparison of the efficacy and safety of new oral anticoagulants with warfarin in patients with
atrial fibrillation: a meta-analysis of randomised trials. Lancet 2014;383:955 –962.
13. Thomas L. Network meta-analysis for indirect treatment comparisons. Stat Med
2002;21:2313 –2324.
14. Dogliotti A, Paolasso E, Giugliano R. Current and new antithrombotics in nonvalvular atrial fibrillation: a network meta-analysis of 79,808 patients. Heart 2014;
100:396 –405.
15. Salanti G, Higgins JP, Ades AE, Ioannidis JP. Evaluation of networks of randomized
trials. Stat Methods Med Res 2008;17:279–301.
16. van Valkenhoef G, Lu G, de Brock B, Hillege H, Ades A, Welton N. Automating
network meta-analysis. Res Syn Meth 2012;3:285 –299.
17. Dias S, Welton N, Caldwell D, Ades A. Checking consistency in mixed treatment
comparison meta-analysis. Stat Med 2010;29:932 –944.
18. Tervonen T, Lahdelma R. Implementing stochastic multicriteria acceptability analysis. Eur J Oper Res 2007;178:500 –513.
19. van Valkenhoef G, Tervonen T, de Brock J, Hillege H, Postmus D. ‘Multi-criteria
benefit-risk assessment using network meta-analysis’. J Clin Epidemiol 2012;65:
394 –403.
20. Tervonen T, Figueira JR. A survey on stochastic multicriteria acceptability analysis
methods. J Multi-Criteria Decis Anal 2008;15:1 –14.
21. Felli JC, Noel RA, Cavazzoni PA. A multiattribute model for evaluating the
benefit-risk profiles of treatment alternatives. Med Decis Making 2009;29:104–115.
22. Lahdelma R, Hokkanen J, Salminen P. SMAA – stochastic multiobjective acceptability
analysis. Eur J Oper Res 1998;106:137 – 143.
23. Lahdelma R, Salminen P. SMAA-2: Stochastic multicriteria acceptability analysis
for group decision making. Oper Res 2001;49:444 –454.
24. Post PN, Stiggelbout AM, Wakker PP. The utility of health states after stroke: a
systematic review of the literature. Stroke 2001;32:1425 –1429.
23
25. Teng TO, Wallace A. One thousand health-related quality-of-life estimates. Med
Care 2000;38:583 –637.
26. Lee HY, Hwang JS, Jeng JS, Wang JD. Quality-adjusted life expectancy (QALE) and
loss of QALE for patients with ischemic strokeand intracerebral hemorrhage: a
13-year follow-up. Stroke 2010;41:739 –744.
27. Lee S, Anglade MW, Pham D, Pisacane R, Kluger J, Coleman CI. Cost-effectiveness
of rivaroxaban compared to warfarin for stroke prevention in atrial fibrillation. Am
J Cardiol 2012;110:845–851.
28. Shah SV, Gage BF. Cost-effectiveness of dabigatran for stroke prophylaxis in atrial
fibrillation. Circulation 2011;123:2562 –2570.
29. Gage BF, Cardinalli AB, Owens D. The effect of stroke and stroke prophylaxis with aspirin or warfarin on quality of life. Arch Intern Med 1996;156:
1829 – 1836.
30. O’Brien CL, Gage BF. Costs and effectiveness of ximelagatran for stroke prophylaxis
in chronic atrial fibrillation. JAMA 2005;293:699 – 706.
31. Thomson R, Parkin D, Eccles M, Sudlow M, Robinson A. Decision analysis and
guidelines for anticoagulant therapy to prevent stroke in patients with atrial fibrillation. Lancet 2000;355:956 –962.
32. Fryback DG, Dasbach EJ, Klein R, Klein BE, Dorn N, Peterson K, Martin P. The Beaver
Dam Health Outcomes Study: initial catalog of health-state quality factors. Med Decis
Making 1993;13:89 –102.
33. Dyer MT, Goldsmith KA, Sharples LS, Buxton MJ. A review of health utilities using
the EQ-5D in studies of cardiovascular disease. Health Qual Life Outcomes 2010;
28:8 –13.
34. Schweikert B, Hunger M, Meisinger C, Ko¨nig HH, Gapp O, Holle R. Quality of life
several years after myocardial infarction: comparing the MONICA/KORA registry
to the general population. Eur Heart J 2009;30:436 –443.
35. Deitelzweig S, Amin A, Jing Y, Makenbaeva D, Wiederkehr D, Lin J, Graham J. Medical
cost reductions associated with the usage of novel oral anticoagulants vs. warfarin
among atrial fibrillation patients, based on the RE-LY, ROCKET-AF, and ARISTOTLE
trials. J Med Econ 2012;15:776 –785.
36. Deitelzweig SB, Pinsky B, Buysman E, Lacey M, Makenbaeva D, Wiederkehr D,
Graham J. Bleeding as an outcome among patients with nonvalvular atrial fibrillation
in a large managed care population. J Clin Ther 2013;35:1536 –1545.
37. Deitelzweig S, Amin A, Jing Y, Makenbaeva D, Wiederkehr D, Lin J, Graham J. Medical
costs in the US of clinical events associated with oral anticoagulant (OAC) use compared to warfarin among non-valvular atrial fibrillation patients ≥75 years of age,
based on the ARISTOTLE, RE-LY, and ROCKET-AF trials. J Med Econ 2013;16:
1163 –1168.
38. Petersen P, Boysen G, Godtfredsen J, Andersen ED, Andersen B. Placebocontrolled, randomised trial of warfarin and aspirin for prevention of thromboembolic complications in chronic atrial fibrillation. The Copenhagen AFASAK
study. Lancet 1989;1:175 –179.
39. The effect of low-dose warfarin on the risk of stroke in patients with nonrheumatic
atrial fibrillation. The Boston Area Anticoagulation Trial for Atrial Fibrillation Investigators. N Engl J Med 1990;323:1505 –1511.
40. Connolly SJ, Laupacis A, Gent M, Roberts RS, Cairns JA, Joyner C. Canadian Atrial
Fibrillation Anticoagulation (CAFA) study. J Am Coll Cardiol 1991;18:349 –355.
41. Stroke Prevention in Atrial Fibrillation study. Final results. Circulation 1991;84:
527 –539.
42. Ezekowitz MD, Bridgers SL, James KE, Carliner NH, Colling CL, Gornick CC,
Krause-Steinrauf H, Kurtzke JF, Nazarian SM, Radford MJ. Warfarin in the prevention
of stroke associated with nonrheumatic atrial fibrillation. Veterans Affairs Stroke
Prevention in Nonrheumatic Atrial Fibrillation Investigators. N Engl J Med 1992;
327:1406 –1412.
43. Secondary prevention in non-rheumatic atrial fibrillation after transient ischemic
attack or minor stroke. EAFT (European Atrial Fibrillation Trial) study group.
Lancet 1993;342:1255 –1262.
44. Warfarin versus aspirin for prevention of thromboembolism in atrial fibrillation:
Stroke Prevention in Atrial Fibrillation II study. Lancet 1994;343:687 –691.
45. Gullov AL, Koefoed BG, Petersen P, Pedersen TS, Andersen ED, Godtfredsen J,
Boysen G. Fixed minidose warfarin and aspirin alone and in combination vs.
adjusted-dose warfarin for stroke prevention in atrial fibrillation: second Copenhagen Atrial Fibrillation, Aspirin, and Anticoagulation study. Arch Intern Med 1998;
158:1513 – 1521.
46. Hellemons BS, Langenberg M, Lodder J, Vermeer F, Schouten HJ, Lemmens T, van
Ree JW, Knottnerus JA. Primary prevention of arterial thromboembolism in nonrheumatic atrial fibrillation in primary care: randomised controlled trial comparing
two intensities of coumarin with aspirin. BMJ 1999;319:958 –964.
47. Posada IS, Barriales V. Alternate-day dosing of aspirin in atrial fibrillation. LASAF Pilot
study group. Am Heart J 1999;138:137 –143.
48. Sato H, Ishikawa K, Kitabatake A, Ogawa S, Maruyama Y, Yokota Y, Fukuyama T,
Doi Y, Mochizuki S, Izumi T, Takekoshi N, Yoshida K, Hiramori K, Origasa H,
Uchiyama S, Matsumoto M, Yamaguchi T, Hori M. Low-dose aspirin for prevention
24
49.
50.
51.
52.
53.
54.
A. Dogliotti and R.P. Giugliano
of stroke in low-risk patients with atrial fibrillation: Japan Atrial Fibrillation Stroke
Trial. Stroke 2006;37:447 – 451.
Antithrombotic Therapy in Atrial Fibrillation Study Group. [The randomized study
of efficiency and safety of antithrombotic therapy in nonvalvular atrial fibrillation:
warfarin compared with aspirin]. Zhonghua Xin Xue Guan Bing Za Zhi 2006;34:
295 –298.
Mant J, Hobbs FD, Fletcher K, Roalfe A, Fitzmaurice D, Lip GY, Murray E. BAFTA
Investigators; Midland Research Practices Network (MidReC). Warfarin versus
aspirin for stroke prevention in an elderly community population with atrial fibrillation (the Birmingham Atrial Fibrillation Treatment of the Aged study, BAFTA): a randomised controlled trial. Lancet 2007;370:493 – 503.
Rash A, Downes T, Portner R, Yeo WW, Morgan N, Channer KS. A randomised controlled trial of warfarin versus aspirin for stroke prevention in octogenarians with
atrial fibrillation (WASPO). Age Ageing 2007;36:151 – 156.
Roskell NS, Lip GY, Noack H, Clemens A, Plumb JM. Treatments for stroke prevention in atrial fibrillation: a network meta-analysis and indirect comparisons versus
dabigatran etexilate. Thromb Haemost 2010;104:1106 –1115.
Harenberg J, Marx S, Diener HC, Lip GY, Marder VJ, Wehling M, Weiss C. Comparison
of efficacy and safety of dabigatran, rivaroxaban and apixaban in patients withatrial fibrillation using network meta-analysis. Int Angiol 2012;31:330–339.
Baker WL, Phung OJ. Systematic review and adjusted indirect comparison
meta-analysis of oral anticoagulants in atrial fibrillation. Circ Cardiovasc Qual Outcomes
2012;5:711 –719.
55. Biondi-Zoccai G, Malavasi V, D’Ascenzo F, Abbate A, Agostoni P, Lotrionte M,
Castagno D, Van Tassell B, Casali E, Marietta M, Modena MG, Ellenbogen KA,
Frati G. Comparative effectiveness of novel oral anticoagulants for atrial fibrillation:
evidence from pair-wise and warfarin-controlled network meta-analyses. HSR Proc
Intensive Care Cardiovasc Anesth 2013;5:40– 54.
56. Schneeweiss S, Gagne JJ, Patrick AR, Choudhry NK, Avorn J. Comparative efficacy
and safety of new oral anticoagulants in patients with atrial fibrillation. Circ Cardiovasc
Qual Outcomes 2012;5:480 – 486.
57. Lip GY, Larsen TB, Skjøth F, Rasmussen LH. Indirect comparisons of new oral anticoagulant drugs for efficacy and safety when used for stroke prevention in atrial fibrillation. J Am Coll Cardiol 2012;60:738 –746.
58. Rasmussen LH, Larsen TB, Graungaard T, Skjøth F, Lip GY. Primary and secondary
prevention with new oral anticoagulant drugs for stroke prevention in atrial fibrillation: indirect comparison analysis. BMJ 2012;345:e7097.
59. Mantha S, Ansell J. An indirect comparison of dabigatran, rivaroxaban and apixaban
for atrial fibrillation. Thromb Haemost 2012;108:476 – 484.
60. Skjøth F, Larsen TB, Rasmussen LH, Lip GY. Efficacy and safety of edoxaban in
comparison with dabigatran, rivaroxaban and apixaban for stroke prevention in
atrial fibrillation. An indirect comparison analysis. Thromb Haemost 2014;111:
981 –988.
61. Connolly SJ, Ezekowitz MD, Yusuf S, Reilly PA, Wallentin L; Randomized Evaluation
of Long-Term Anticoagulation Therapy Investigators. Newly identified events in
the RE-LY trial. N Engl J Med 2010; 363; 1876– 1876.
Tables A1 –A3.
Table A1
Odds ratio and credibility interval of drug named in column vs. row
Mortality from any cause
Apixaban
1.36 (0.96, 2.13)
1.23 (0.72, 2.08)
1.73 (1.14, 2.90)
0.96 (0.49, 1.82)
0.97 (0.47, 1.79)
0.94 (0.47, 1.74)
0.97 (0.51, 1.84)
0.97 (0.51, 1.82)
1.07 (0.71, 1.56)
0.73 (0.47, 1.04)
0.82 (0.48, 1.38)
ASA
1.10 (0.78, 1.73)
0.91 (0.58, 1.28)
ASA+ clopidogrel
1.26 (0.96, 1.72)
1.41 (0.94, 2.35)
0.72 (0.38, 1.18)
0.78 (0.40, 1.52)
0.70 (0.36, 1.16)
0.78 (0.39, 1.49)
0.69 (0.36, 1.14)
0.76 (0.39, 1.42)
0.72 (0.38, 1.19)
0.80 (0.40, 1.52)
0.73 (0.39, 1.17)
0.80 (0.42, 1.52)
0.79 (0.59, 0.94)
0.88 (0.58, 1.27)
0.58 (0.34, 0.87)
0.80 (0.58, 1.04)
0.71 (0.43, 1.07)
Control
0.56 (0.29, 0.96)
0.55 (0.27, 0.93)
0.54 (0.28, 0.90)
0.57 (0.28, 0.95)
0.57 (0.29, 0.96)
0.61 (0.43, 0.81)
1.04 (0.55, 2.04)
1.04 (0.56, 2.12)
1.38 (0.85, 2.67)
1.42 (0.86, 2.78)
1.28 (0.66, 2.48)
1.28 (0.67, 2.59)
1.77 (1.04, 3.49)
1.81 (1.07, 3.68)
Dabigatran 110
1.01 (0.61, 1.78)
0.99 (0.56, 1.64)
Dabigatran 150
0.97 (0.46, 2.01)
0.97 (0.47, 2.09)
1.02 (0.48, 2.07)
1.02 (0.49, 2.19)
1.02 (0.49, 2.14)
1.02 (0.48, 2.21)
1.11 (0.65, 1.88)
1.12 (0.66, 1.95)
1.07 (0.58, 2.11)
1.46 (0.88, 2.75)
1.31 (0.70, 2.58)
1.86 (1.11, 3.63)
1.03 (0.50, 2.16)
1.04 (0.48, 2.14)
Edoxaban 30
1.04 (0.62, 1.75)
1.04 (0.52, 2.21)
1.14 (0.69, 1.93)
1.03 (0.54, 1.98)
1.03 (0.55, 1.94)
1.38 (0.84, 2.65)
1.37 (0.85, 2.57)
1.25 (0.66, 2.48)
1.25 (0.66, 2.41)
1.75 (1.06, 3.52)
1.75 (1.04, 3.47)
0.98 (0.48, 2.09)
0.98 (0.47, 2.05)
0.98 (0.46, 2.04)
0.98 (0.45, 2.07)
0.96 (0.57, 1.60)
0.96 (0.45, 1.94)
Edoxaban 60
1.00 (0.48, 2.07)
1.00 (0.48, 2.08)
Rivaroxaban
1.10 (0.65, 1.86)
1.09 (0.65, 1.84)
0.94 (0.64, 1.41)
1.27 (1.06, 1.70)
1.14 (0.78, 1.73)
1.64 (1.24, 2.33)
0.90 (0.53, 1.54)
0.89 (0.51, 1.51)
0.87 (0.52, 1.46)
0.91 (0.54, 1.54)
0.91 (0.54, 1.53)
VKA
1.14 (0.75, 1.88)
Ischaemic stroke
Apixaban
2.56 (1.56, 3.97)
1.90 (1.09, 3.79)
3.60 (2.11, 6.43)
1.27 (0.62, 2.80)
0.85 (0.43, 1.96)
1.64 (0.80, 3.56)
1.14 (0.57, 2.47)
1.05 (0.51, 2.33)
0.39 (0.25, 0.64)
ASA
0.75 (0.52, 1.27)
1.42 (1.02, 2.12)
0.50 (0.27, 1.04)
0.34 (0.19, 0.71)
0.64 (0.36, 1.30)
0.44 (0.25, 0.92)
0.41 (0.22, 0.85)
0.45 (0.35, 0.63)
0.53 (0.26, 0.92)
0.28 (0.16, 0.47)
1.34 (0.79, 1.94)
0.71 (0.47, 0.98)
ASA+clopidogrel
0.53 (0.32, 0.95)
1.90 (1.05, 3.13)
Control
0.67 (0.31, 1.43)
0.35 (0.18, 0.73)
0.45 (0.21, 0.94)
0.24 (0.12, 0.50)
0.87 (0.40, 1.75)
0.46 (0.23, 0.91)
0.60 (0.28, 1.24)
0.31 (0.16, 0.64)
0.55 (0.25, 1.12)
0.29 (0.15, 0.59)
0.59 (0.37, 0.93)
0.32 (0.22, 0.46)
0.78 (0.36, 1.62)
2.00 (0.96, 3.69)
1.49 (0.70, 3.28)
2.83 (1.36, 5.63)
Dabigatran 110
0.68 (0.37, 1.28)
1.28 (0.56, 2.97)
0.89 (0.38, 2.08)
0.83 (0.36, 1.91)
0.89 (0.49, 1.64)
1.18 (0.51, 2.35)
0.61 (0.28, 1.25)
2.96 (1.41, 5.39)
1.55 (0.77, 2.81)
2.20 (1.06, 4.84)
1.15 (0.57, 2.49)
4.25 (1.99, 8.26)
2.19 (1.10, 4.30)
1.48 (0.78, 2.73)
0.78 (0.34, 1.78)
Dabigatran 150
0.52 (0.23, 1.22)
1.93 (0.82, 4.37)
Edoxaban 30
1.33 (0.58, 3.03)
0.69 (0.39, 1.23)
1.22 (0.52, 2.83)
0.64 (0.28, 1.43)
1.32 (0.71, 2.38)
0.69 (0.39, 1.25)
0.88 (0.41, 1.77)
2.26 (1.08, 4.04)
1.67 (0.80, 3.57)
3.18 (1.56, 6.20)
1.12 (0.48, 2.60)
0.75 (0.33, 1.73)
1.45 (0.81, 2.56)
Edoxaban 60
0.93 (0.40, 2.04)
1.00 (0.55, 1.78)
0.95 (0.43, 1.96)
0.88 (0.53, 1.34)
2.42 (1.18, 4.46)
2.23 (1.59, 2.85)
1.81 (0.89, 3.95)
1.68 (1.08, 2.72)
3.42 (1.71, 6.85)
3.16 (2.17, 4.58)
1.21 (0.52, 2.81)
1.12 (0.61, 2.05)
0.82 (0.35, 1.93)
0.76 (0.42, 1.40)
1.57 (0.70, 3.51)
1.45 (0.80, 2.57)
1.08 (0.49, 2.52)
1.00 (0.56, 1.83)
Rivaroxaban
0.93 (0.51, 1.66)
1.08 (0.60, 1.98)
VKA
1.42 (0.82, 2.36)
ASA
1.26 (0.66, 2.50)
0.90 (0.56, 1.54)
1.95 (0.88, 4.72)
1.40 (0.71, 2.93)
1.37 (0.61, 2.89)
0.95 (0.47, 1.99)
1.33 (0.58, 2.87)
0.93 (0.46, 1.94)
1.29 (0.57, 2.65)
0.92 (0.44, 1.81)
1.02 (0.44, 2.06)
0.72 (0.35, 1.42)
0.84 (0.38, 1.74)
0.60 (0.29, 1.18)
1.05 (0.62, 1.67)
0.74 (0.51, 1.07)
Benefit– risk balance of anti-thrombotics in atrial fibrillation
Appendix
Myocardial infarction
Apixaban
0.70 (0.42, 1.22)
0.79 (0.40, 1.52)
1.11 (0.65, 1.78)
ASA+ clopidogrel
1.57 (0.69, 3.67)
1.08 (0.47, 2.37)
1.05 (0.45, 2.29)
1.03 (0.43, 2.13)
0.81 (0.34, 1.68)
0.67 (0.29, 1.45)
0.83 (0.48, 1.33)
0.51 (0.21, 1.14)
0.73 (0.35, 1.63)
0.72 (0.34, 1.40)
1.05 (0.50, 2.13)
0.64 (0.27, 1.46)
0.93 (0.42, 2.14)
Control
1.46 (0.57, 3.99)
0.68 (0.25, 1.77)
Dabigatran 110
0.67 (0.25, 1.74)
0.98 (0.52, 1.82)
0.66 (0.24, 1.59)
0.95 (0.39, 2.09)
0.53 (0.19, 1.25)
0.74 (0.31, 1.68)
0.43 (0.16, 1.04)
0.62 (0.26, 1.43)
0.53 (0.24, 1.08)
0.77 (0.42, 1.40)
0.75 (0.35, 1.71)
1.07 (0.52, 2.19)
0.95 (0.44, 2.22)
1.49 (0.58, 3.99)
1.02 (0.55, 1.92)
Dabigatran 150
0.97 (0.41, 2.17)
0.76 (0.31, 1.73)
0.63 (0.27, 1.49)
0.79 (0.43, 1.45)
0.77 (0.38, 1.75)
0.98 (0.49, 2.25)
1.08 (0.55, 2.30)
1.39 (0.71, 2.83)
0.97 (0.47, 2.32)
1.24 (0.59, 2.90)
1.52 (0.63, 4.25)
1.90 (0.80, 5.38)
1.05 (0.48, 2.54)
1.36 (0.60, 3.18)
1.03 (0.46, 2.45)
1.32 (0.58, 3.20)
Edoxaban 30
1.27 (0.70, 2.27)
0.78 (0.44, 1.43)
Edoxaban 60
0.66 (0.29, 1.55)
0.84 (0.37, 2.00)
0.81 (0.45, 1.49)
1.04 (0.58, 1.89)
1.19 (0.57, 2.61)
1.68 (0.85, 3.45)
1.50 (0.69, 3.42)
2.32 (0.96, 6.32)
1.62 (0.70, 3.84)
1.58 (0.67, 3.77)
1.51 (0.65, 3.49)
1.19 (0.50, 2.74)
Rivaroxaban
1.24 (0.67, 2.27)
0.96 (0.60, 1.62)
1.35 (0.93, 1.98)
1.20 (0.75, 2.10)
1.88 (0.93, 4.12)
1.29 (0.71, 2.41)
1.26 (0.69, 2.35)
1.24 (0.67, 2.20)
0.96 (0.53, 1.72)
0.80 (0.44, 1.49)
VKA
25
Continued
26
Table A1
Continued
Systemic embolism
Apixaban
3.69 (1.35, 10.14)
4.03 (1.36, 16.20)
4.31 (1.53, 14.39)
1.07 (0.25, 5.49)
1.16 (0.29, 5.68)
1.77 (0.51, 7.87)
0.95 (0.25, 4.23)
1.22 (0.30, 6.05)
1.45 (0.64, 4.01)
0.27 (0.10, 0.74)
ASA
1.11 (0.55, 2.86)
1.19 (0.59, 2.57)
0.30 (0.08, 1.23)
0.32 (0.09, 1.34)
0.50 (0.15, 1.75)
0.25 (0.07, 0.97)
0.33 (0.09, 1.33)
0.41 (0.20, 0.82)
0.25 (0.06, 0.73)
0.23 (0.07, 0.65)
0.90 (0.35, 1.81)
0.84 (0.39, 1.68)
ASA+ clopidogrel
0.92 (0.36, 2.82)
1.08 (0.35, 2.76)
Control
0.26 (0.06, 1.15)
0.25 (0.06, 1.06)
0.28 (0.07, 1.20)
0.26 (0.07, 1.13)
0.43 (0.10, 1.64)
0.42 (0.12, 1.45)
0.24 (0.05, 0.89)
0.22 (0.06, 0.79)
0.30 (0.07, 1.27)
0.28 (0.07, 1.13)
0.37 (0.13, 0.83)
0.34 (0.16, 0.71)
0.93 (0.18, 4.05)
3.32 (0.82, 12.14)
3.91 (0.87, 17.59)
4.03 (0.94, 15.67)
Dabigatran 110
1.11 (0.33, 3.59)
1.65 (0.32, 7.88)
0.89 (0.16, 4.18)
1.16 (0.21, 5.56)
1.37 (0.39, 4.23)
0.86 (0.18, 3.45)
3.13 (0.74, 10.96)
3.61 (0.83, 15.01)
3.78 (0.88, 13.80)
0.90 (0.28, 3.05)
Dabigatran 150
1.54 (0.31, 7.33)
0.84 (0.15, 3.95)
1.08 (0.21, 5.35)
1.28 (0.37, 3.78)
0.56 (0.13, 1.97)
1.05 (0.24, 3.97)
2.01 (0.57, 6.58)
3.92 (1.03, 13.66)
2.30 (0.61, 9.63)
4.21 (1.12, 19.79)
2.38 (0.69, 8.37)
4.47 (1.26, 17.07)
0.61 (0.13, 3.12)
1.13 (0.24, 6.29)
0.65 (0.14, 3.20)
1.19 (0.25, 6.78)
Edoxaban 30
1.89 (0.66, 5.86)
0.53 (0.17, 1.52)
Edoxaban 60
0.69 (0.15, 3.32)
1.27 (0.28, 6.50)
0.81 (0.29, 2.23)
1.53 (0.53, 4.71)
0.82 (0.17, 3.36)
3.01 (0.75, 10.78)
3.32 (0.79, 14.89)
3.63 (0.89, 13.35)
0.87 (0.18, 4.72)
0.92 (0.19, 4.79)
1.44 (0.30, 6.88)
0.79 (0.15, 3.59)
Rivaroxaban
1.22 (0.36, 3.71)
2.72 (1.20, 7.53)
2.97 (1.42, 6.30)
0.73 (0.24, 2.58)
0.78 (0.26, 2.69)
1.24 (0.45, 3.48)
0.65 (0.21, 1.88)
0.82 (0.27, 2.76)
VKA
0.69 (0.25, 1.56)
2.44 (1.22, 4.96)
Intracranial haemorrhage
Apixaban
1.26 (0.46, 2.98)
3.05 (1.02, 10.34)
0.39 (0.03, 1.82)
0.72 (0.18, 3.13)
0.94 (0.25, 3.95)
0.71 (0.19, 3.08)
1.07 (0.29, 4.87)
1.50 (0.42, 6.70)
2.37 (1.08, 6.17)
0.79 (0.34, 2.16)
0.33 (0.10, 0.98)
ASA
0.41 (0.15, 0.88)
2.47 (1.13, 6.72)
ASA+ clopidogrel
0.31 (0.02, 1.31)
0.12 (0.01, 0.57)
0.57 (0.17, 2.56)
0.24 (0.06, 1.03)
0.74 (0.23, 3.22)
0.30 (0.08, 1.37)
0.56 (0.17, 2.48)
0.23 (0.06, 0.97)
0.85 (0.26, 3.86)
0.35 (0.09, 1.47)
1.19 (0.38, 5.30)
0.49 (0.12, 2.13)
1.88 (0.97, 4.67)
0.75 (0.32, 2.03)
2.58 (0.55, 37.87)
3.24 (0.76, 41.87)
8.12 (1.77, 119.74)
Control
1.91 (0.32, 32.94)
2.50 (0.44, 44.91)
1.89 (0.36, 31.39)
2.82 (0.53, 47.71)
3.97 (0.71, 68.05)
6.25 (1.53, 89.72)
1.39 (0.32, 5.50)
1.06 (0.25, 4.03)
1.76 (0.39, 5.84)
1.36 (0.31, 4.44)
4.19 (0.97, 18.11)
3.37 (0.73, 13.25)
0.52 (0.03, 3.10)
0.40 (0.02, 2.27)
Dabigatran 110
0.78 (0.24, 2.32)
1.28 (0.43, 4.15)
Dabigatran 150
0.97 (0.22, 4.76)
0.77 (0.16, 3.58)
1.45 (0.34, 7.57)
1.16 (0.23, 5.32)
2.10 (0.49, 10.20)
1.67 (0.34, 7.64)
3.23 (1.13, 10.11)
2.53 (0.83, 7.45)
1.40 (0.32, 5.29)
1.78 (0.40, 5.84)
4.35 (1.03, 17.11)
0.53 (0.03, 2.81)
1.04 (0.21, 4.63)
1.29 (0.28, 6.25)
Edoxaban 30
1.50 (0.51, 4.59)
2.11 (0.47, 9.66)
3.33 (1.14, 9.59)
0.93 (0.21, 3.46)
0.67 (0.15, 2.41)
1.17 (0.26, 3.78)
0.84 (0.19, 2.60)
2.88 (0.68, 11.03)
2.06 (0.47, 8.14)
0.35 (0.02, 1.88)
0.25 (0.01, 1.42)
0.69 (0.13, 2.98)
0.48 (0.10, 2.06)
0.86 (0.19, 4.27)
0.60 (0.13, 2.91)
0.67 (0.22, 1.95)
0.47 (0.10, 2.13)
Edoxaban 60
0.71 (0.16, 3.31)
1.41 (0.30, 6.27)
Rivaroxaban
2.22 (0.75, 6.24)
1.58 (0.52, 4.54)
0.42 (0.16, 0.92)
0.53 (0.21, 1.03)
1.34 (0.49, 3.14)
0.16 (0.01, 0.65)
0.31 (0.10, 0.88)
0.40 (0.13, 1.20)
0.30 (0.10, 0.88)
0.45 (0.16, 1.33)
0.63 (0.22, 1.92)
VKA
1.14 (0.32, 2.79)
1.54 (0.27, 9.35)
0.48 (0.07, 1.80)
1.10 (0.14, 5.05)
1.25 (0.16, 5.78)
0.59 (0.08, 2.94)
1.03 (0.14, 4.98)
1.29 (0.18, 6.32)
1.00 (0.32, 2.85)
ASA
1.53 (0.36, 6.50)
0.47 (0.09, 1.53)
1.10 (0.15, 4.36)
1.24 (0.18, 4.88)
0.59 (0.09, 2.39)
1.02 (0.16, 4.18)
1.26 (0.21, 5.37)
1.11 (0.40, 2.19)
0.65 (0.11, 3.72)
2.08 (0.56, 13.97)
0.65 (0.15, 2.79)
2.11 (0.65, 11.67)
ASA+ clopidogrel
3.18 (0.66, 32.41)
0.31 (0.03, 1.52)
Control
0.72 (0.06, 4.93)
2.13 (0.40, 16.01)
0.81 (0.07, 5.60)
2.44 (0.48, 19.06)
0.38 (0.04, 2.72)
1.20 (0.24, 8.99)
0.66 (0.06, 4.61)
2.05 (0.44, 15.44)
0.83 (0.08, 6.12)
2.54 (0.51, 20.35)
0.73 (0.11, 3.11)
2.31 (0.80, 8.93)
0.91 (0.20, 7.23)
0.91 (0.23, 6.67)
1.40 (0.20, 17.50)
0.47 (0.06, 2.48)
Dabigatran 110
1.14 (0.26, 5.12)
0.55 (0.07, 4.36)
0.95 (0.13, 7.24)
1.17 (0.16, 9.86)
1.05 (0.27, 4.79)
0.80 (0.17, 6.10)
1.68 (0.34, 12.45)
0.80 (0.21, 5.66)
1.70 (0.42, 10.76)
1.23 (0.18, 13.98)
2.60 (0.37, 27.91)
0.41 (0.05, 2.08)
0.83 (0.11, 4.13)
0.88 (0.20, 3.91)
1.82 (0.23, 13.69)
Dabigatran 150
2.09 (0.26, 16.29)
0.48 (0.06, 3.81)
Edoxaban 30
0.83 (0.12, 6.45)
1.73 (0.42, 7.64)
1.03 (0.15, 7.92)
2.13 (0.29, 17.17)
0.91 (0.23, 4.04)
1.91 (0.45, 7.99)
0.97 (0.20, 7.17)
0.98 (0.24, 6.17)
1.51 (0.22, 15.74)
0.49 (0.06, 2.29)
1.06 (0.14, 7.93)
1.21 (0.16, 8.53)
0.58 (0.13, 2.36)
Edoxaban 60
1.24 (0.18, 9.45)
1.12 (0.28, 4.51)
0.78 (0.16, 5.49)
0.88 (0.36, 3.11)
0.79 (0.19, 4.71)
0.90 (0.46, 2.52)
1.21 (0.16, 12.09)
1.37 (0.32, 8.87)
0.39 (0.05, 1.95)
0.43 (0.11, 1.25)
0.85 (0.10, 6.12)
0.96 (0.21, 3.68)
0.97 (0.13, 6.83)
1.09 (0.25, 4.38)
0.47 (0.06, 3.43)
0.52 (0.13, 2.21)
0.80 (0.11, 5.46)
0.89 (0.22, 3.58)
Rivaroxaban
1.12 (0.27, 4.77)
0.90 (0.21, 3.68)
VKA
Interventions are reported in alphabetical order. Results are the ORs in the column-defining treatment compared with the ORs in the row-defining treatment. For efficacy, ORs lower than one favour the column-defining treatment (i.e. the first in
alphabetical order). To obtain ORs for comparisons in the opposite direction, reciprocals should be taken. Significant results are in bold. VKA, vitamin K antagonist.
A. Dogliotti and R.P. Giugliano
Major extracranial haemorrhage
Apixaban
1.00 (0.35, 3.16)
27
Benefit– risk balance of anti-thrombotics in atrial fibrillation
Table A2 Rank acceptabilities
Rank 1
(%)
Based on utility functions
Apixaban
17
Rank 2
(%)
Rank 3
(%)
Rank 4
(%)
Rank 5
(%)
Rank 6
(%)
Rank 7
(%)
Rank 8
(%)
Rank 9
(%)
Rank 10
(%)
18
17
15
12
10
7
2
1
0
ASA
0
0
0
0
1
4
9
33
49
3
ASA+
clopidogrel
0
1
1
2
3
6
13
35
23
17
Control
Dabigatran 110
1
13
0
16
1
14
1
14
1
12
2
12
2
11
5
5
13
3
74
1
Dabigatran 150
20
18
16
12
10
10
8
4
2
1
Edoxaban 30
Edoxaban 60
15
16
16
16
14
15
14
14
12
12
11
11
12
9
4
4
2
2
1
1
Rivaroxaban
18
13
13
12
11
11
11
5
3
2
Vitamin K
antagonist
0
2
8
17
24
24
17
4
3
1
Apixaban
ASA
16
0
17
0
17
0
14
1
12
2
10
5
9
11
4
31
1
47
1
3
ASA+
clopidogrel
0
1
2
3
4
8
13
32
21
16
Control
1
1
2
2
2
2
3
7
12
68
Dabigatran 110
Dabigatran 150
11
18
14
16
14
14
13
12
12
10
13
9
12
9
6
5
4
4
3
3
Edoxaban 30
18
17
15
12
11
10
11
4
2
1
Edoxaban 60
Rivaroxaban
16
18
17
14
14
12
12
11
11
10
11
10
9
11
5
6
3
4
2
4
Vitamin K
antagonist
1
4
11
22
26
23
11
2
1
0
Based on medical costs
Preference information: based on utility functions:24 – 34 mortality intracranial haemorrhage ischaemic stroke myocardial infarction major extracranial haemorrhage systemic
embolism. Based on medical costs:35 – 37 mortality major extracranial haemorrhage ischaemic stroke myocardial infarction intracranial haemorrhage systemic embolism. The rank
acceptabilities represent, given the preference information, the probability (in per cent) that each treatment is best (rank 1), the second highest (rank 2), etc. Rank probabilities sum to
100%, both within a rank over treatments and within a treatment over ranks. In grey are the most relevant results.
28
A. Dogliotti and R.P. Giugliano
Table A3 Central weight vectors and confidence factors
Based on utility
functions
Confidence
factor
Mortality
Intracranial
haemorrhage
Ischaemic
stroke
Myocardial
infarction
Extracranial
haemorrhage
Systemic
embolism
Apixaban
ASA
17%
0%
0.39
0.33
0.24
0.24
0.16
0.19
0.1
0.12
0.02
0.01
0.06
0.07
ASA+
clopidogrel
0.3%
0.54
0.19
0.12
0.07
0.01
0.04
Control
0.6%
0.37
0.25
0.15
0.1
0.03
0.07
Dabigatran 110
Dabigatran 150
13%
21%
0.41
0.4
0.24
0.24
0.15
0.16
0.09
0.1
0,00
0.02
0.06
0.05
Edoxaban 30
16%
0.42
0.24
0.15
0.09
0.02
0.05
Edoxaban 60
Rivaroxaban
16%
17%
0.4
0.4
0.24
0.24
0.16
0.16
0.1
0.1
0.02
0.02
0.06
0.06
Vitamin K
antagonist
Based on
medical costs
Apixaban
0.3%
0.4
0.23
0.16
0.1
0.03
0.06
Confidence
factor
17%
Mortality
0.39
Extracranial
haemorrhage
0.23
Ischaemic
stroke
0.16
Myocardial
infarction
0.1
Intracranial
haemorrhage
0.06
Systemic
embolism
0.02
ASA
0%
0.53
0.21
0.08
0.08
0.05
0.02
ASA+
clopidogrel
0.3%
0.41
0.24
0.15
0.1
0.05
0.02
Control
Dabigatran 110
0.8%
13%
0.38
0.41
0.26
0.24
0.15
0.15
0.1
0.09
0.06
0.05
0.02
0.02
Dabigatran 150
17%
0.41
0.23
0.15
0.1
0.06
0.02
Edoxaban 30
Edoxaban 60
17%
17%
0.42
0.4
0.24
0.23
0.15
0.16
0.09
0.1
0.05
0.06
0.02
0.02
Rivaroxaban
18%
0.4
0.23
0.15
0.1
0.06
0.02
Vitamin K
antagonist
0.3%
0.39
0.23
0.16
0.1
0.06
0.02
Preference information based on utility functions:24 – 34 mortality intracranial haemorrhage myocardial infarction ischaemic stroke major extracranial haemorrhage systemic
embolism. Based on medical costs:35 – 37 mortality major extracranial haemorrhage ischaemic stroke myocardial infarction intracranial haemorrhage systemic embolism. Central
weights represent a typical weighting that favours a specific alternative. The CF is the probability that an alternative rank first if the decision-maker’s preferences agree with the
alternative’s central weight vector.