Cardiovasc Drugs Ther (2014) 28:575–585
DOI 10.1007/s10557-014-6558-1
ORIGINAL ARTICLE
Cost-effectiveness of Dabigatran and Rivaroxaban Compared
with Warfarin for Stroke Prevention in Patients with Atrial
Fibrillation
Ye Wang & Feng Xie & Ming Chai Kong & Lai Heng Lee &
Heng Joo Ng & Yu Ko
Published online: 16 October 2014
# Springer Science+Business Media New York 2014
Abstract
Purpose This study aimed to evaluate the cost-effectiveness
of dabigatran and rivaroxaban compared with warfarin for the
prevention of stroke in patients with atrial fibrillation (AF) in
Singapore.
Methods A Markov model was constructed to compare the
lifetime costs, quality-adjusted life-years (QALYs) and incre-
mental cost-effectiveness ratios (ICERs) of dabigatran 110
and 150 mg, rivaroxaban 20 mg and adjusted-dose warfarin
from the perspective of the Singapore healthcare system,
using clinical data from published studies, utilities from a
patient-reported survey and costs from hospital databases.
The target population was a hypothetical cohort of 65-year-
old AF patients with no contraindications to anticoagulation.
Results In the base-case analysis, the QALYs were 8.75 with
warfarin, 8.73 with dabigatran 110 mg, 8.82 with dabigatran
Y. Wang (*)
Center for Surgery and Public Health, Brigham and Women’s
Hospital, Harvard Medical School, One Brigham Circle, 1620
Tremont Street, 4-020, Boston, MA 02120, USA
e-mail: YEWANG@research.bwh.harvard.edu
F. Xie
Department of Clinical Epidemiology & Biostatistics, McMaster
University, Hamilton, Canada
M. C. Kong
Department of Pharmacy, Singapore General Hospital, Outram Park,
Singapore
L. H. Lee : H. J. Ng
Department of Hematology, Singapore General Hospital, Outram
Park, Singapore
Y. Ko (*)
Department of Pharmacy, Faculty of Science, National University of
Singapore, Level 3, Block S4A, 18 Science Drive 4,
Singapore 117543, Singapore
e-mail: nancykotw@gmail.com
150 mg, and 9.33 with rivaroxaban. The costs were Singapore
dollar (SG$) 34,648 for warfarin, SG$54,919 for dabigatran
110 mg, SG$50,484 for dabigatran 150 mg and SG$51,975 for
rivaroxaban. The ICER of rivaroxaban versus warfarin was
SG$29,697 (US$26,727) per QALY. Rivaroxaban and warfarin
had extended dominance over the high-dose dabigatran. The
low-dose dabigatran was dominated by warfarin. Deterministic
sensitivity analyses showed that the ICER of rivaroxaban ver-
sus warfarin was sensitive to cost of rivaroxaban and utilities
for rivaroxaban and warfarin. Probability sensitivity analysis
demonstrated that the probability of rivaroxaban being the
optimal choice was 97.8 % and 99.5 % at a willingness-to-
pay threshold of SG$65,000 (US$58,500) and SG$130,000
(US$117,000) per QALY, respectively.
Conclusion Rivaroxaban may be a cost-effective alternative
to warfarin for the prevention of stroke in patients with AF in
Singapore.
Keywords Anticoagulation . Atrial fibrillation .
Cost-effectiveness . Dabigatran . Rivaroxaban . Warfarin
Introduction
Atrial fibrillation (AF) is the most common type of cardiac
rhythm disorders [1]. The prevalence of AF increases as age
increases, affecting approximately 10 % of people age 80 or
older [1]. Patients with AF are at a four- to five-fold increased
risk of stroke, the second leading cause of global mortality,
across all age groups [1, 2]; indeed, these patients account for
almost one-fifth of stroke events [3]. The disease burden of
stroke in patients with AF is likely to increase in Singapore
because Singapore’s population has been aging and its life
expectancy has been prolonged over the past a few years [4].
To prevent patients with AF from experiencing stroke,
warfarin has been used as a prophylactic measure for more
576
than half a century [5]. However, a major concern of warfarin
lies in its narrow therapeutic range and considerable variabil-
ity in inter-individual responses, which may lead to the occur-
rences of out-of-range international normalized ratio (INR)
values in clinical practice [5], subjecting patients to an in-
creased risk of thromboembolism and bleeding that can lead
to chronic disabilities and even death [6–8]. Given warfarin’s
adverse effects, new oral anticoagulants have been used for
the prevention of stroke in patients with AF in several coun-
tries; these new drugs not only can prevent thromboembolism,
but also have lower bleeding risks than warfarin [9, 10].
A novel oral anticoagulant is dabigatran, which is a direct
thrombin inhibitor. The efficacy and safety of dabigatran were
assessed in the Randomized Evaluation of Long-Term
Anticoagulation Therapy (RE-LY) trial [9]. It was found that,
compared with warfarin, dabigtran 150 and 110 mg twice
daily had lower and similar risks of stroke or systemic embo-
lism, respectively. Moreover, the two dose regimens of
dabigatran had significantly lower risks of intracranial hem-
orrhage (ICH) than warfarin. Another novel oral anticoagulant
is rivaroxaban, which is a direct factor Xa inhibitor. The
Rivaroxaban Once Daily Oral Direct Factor Xa Inhibition
Compared with Vitamin K Antagonism for Prevention of
Stroke and Embolism Trial in Atrial Fibrillation (ROCKET-
AF) study found that rivaroxaban was non-inferior to warfarin
for the prevention of stroke or systemic embolism, and had a
lower risk of ICH in the as-treated safety population [10].
Both dabigatran and rivaroxaban have been approved for
use in Singapore; however, concerns have arisen due to their
higher costs than warfarin. Given the increasing healthcare
expenditures and budget constraints, the clinical use of novel
drugs can be influenced by their economic impact. As such, it
is important to assess the cost-effectiveness of new drugs to
testify if the benefits gained are worth the high costs. This
study aimed to evaluate the cost-utility of dabigatran and
rivaroxaban compared with warfarin for the prevention of
stroke in patients with AF in Singapore.
Methods
Decision Model
Four treatment strategies, i.e., dabigatran 150 and 110 mg
twice daily, rivaroxaban 20 mg once daily and adjusted-dose
warfarin, were compared for the prevention of stroke in pa-
tients with AF, using a Markov model (Fig. 1). The base case
was a hypothetical cohort of 65-year-old patients with newly
diagnosed AF and no contraindications to anticoagulation.
The starting age was 65 years because patients above this
age and with AF are at a high risk of stroke, including those
without any other risk factors for stroke [1]. The health states
in the model included: well with AF, transient ischemic attack
Cardiovasc Drugs Ther (2014) 28:575–585
(TIA), ischemic stroke, ICH, extracranial hemorrhage (ECH),
myocardial infarction (MI), combined events of ischemic
stroke and ICH, and death. All patients entered the model in
the “well with AF” health state and transitioned to other health
states or remained in the “well with AF” state in the next cycle.
A cycle length of 1 month and a lifetime horizon with a
maximum of 20 years were used. Evaluated outcomes includ-
ed direct medical costs and quality-adjusted life years
(QALYs). The willingness-to-pay (WTP) threshold was equal
to Singapore’s 2012 per-capita gross domestic product, i.e.,
Singapore dollar (SG$) 65,000 per QALY [11], which was
US$58,500 per QALY according to purchasing power parities
[12]. As recommended by the World Health Organization, this
threshold can be used to indicate if the intervention is highly
cost-effective compared with its comparator [13]. The model
development and analyses were performed using TreeAge Pro
Suite 2013 (TreeAge Software, Inc., Williamstown, MA).
Clinical Inputs
The base-case estimates for rates of clinical events while on
warfarin (i.e., ischemic stroke, ICH, ECH and MI) were
obtained from the RE-LY trial (Table 1) [9]. The range for
the sensitivity analysis of ischemic stroke rate on warfarin was
derived from published studies to cover the varied rates of
ischemic stroke with CHADS2 scores ranging from 0 to 6 [9,
10, 14–16]. The ranges for sensitivity analyses of other pa-
rameters of warfarin used a wider range obtained from the RE-
LY and ROCKET-AF trials [9, 10]. The base-case estimates
and ranges for the relative risks or hazard ratios of clinical
events on dabigatran and rivaroxaban were derived from the
RE-LY and ROCKET-AF trials, respectively [9, 10]. The rate
of a clinical event on dabigatran or rivaroxaban was derived
by multiplying the rate of event on warfarin by the relative risk
or hazard ratio of the event on dabigatran or rivaroxaban from
the RE-LY or ROCKET-AF trials, respectively.
The percentages of clinical events with different severity
levels were derived from published cost-effectiveness studies
[14, 17]. Ischemic stroke was classified into four categories:
fatal, major, minor and reversible. It was assumed that 28 % of
ischemic neurologic events were TIAs [17]. The rates of
ischemic stroke and TIA were assumed to increase by 1.4-fold
per decade of life [16]. We classified ICH into fatal and non-
fatal, and assumed that the rate of ICH increased by 1.97-fold
per decade of life [18]. In addition, ECH was classified as non-
fatal minor, non-fatal major, and fatal major. Patients with an
ICH or major ECH were assumed to stop anticoagulation and
resume the same anticoagulant after 1 month. Similarly, MI
may be fatal and non-fatal. The rate of MI was assumed to
increase 1.3-fold per decade of life [14, 17].
Mortality rates were adjusted for age (beginning at age
65 years) [19]. Compared with the general population, the
mortality rates increased 1.3-fold in patients with AF and 2.3-
Cardiovasc Drugs Ther (2014) 28:575–585
Fig. 1 Representation of the Markov model. The four treatment options
are shown on the left. “M” represents a Markov process with 7 health
states. These health states are identical for each treatment option. All
patients remain in the “well with AF” state until one of the six events
occurs, which are TIA, ischemic stroke, ICH, ECH, MI and non-event
death. The branch from “well with AF” illustrates these events. The
fold in patients with AF and prior stroke [20–23]. It was
assumed that the event rates of other conditions not included
in the model were similar across all treatments.
Utility Inputs
To calculate QALYs, health utilities for different health states
(Table 1) were multiplied by the time spent in each state. To
obtain the utility scores of different health states in the Markov
model, a patient survey was conducted in a group of 100
Singaporean patients who were at risk of stroke and taking
warfarin using the standard gamble technique, which can
reflect the risks of therapeutic failure and adverse effects
involved in the anticoagulation therapies (details of the health
state descriptions are provided in theAppendix). The mean
utility scores derived from the survey were used as the base-
case estimates while the 10th and 90th percentiles were used
as the ranges of parameters in sensitivity analysis. The QALYs
were discounted at an annual rate of 3 %.
Cost Inputs
Costs reflected the perspective of the Singapore healthcare
system. Therefore, only direct medical costs (i.e., costs of
577
probabilities of these events depend on the treatment. Branches from
the other health states (not shown) have a similar structure. AF atrial
fibrillation; ECH extracranial hemorrhage; ICH intracranial hemorrhage;
MI myocardial infarction; RIND reversible ischemic neurological deficit;
TIA transient ischemic attack
anticoagulation therapies and complications) were included
(Table 1). We projected the cost for each treatment for pa-
tients’ life time. Costs were inflated to 2012 Singapore dollars
and discounted at an annual rate of 3 %.
Costs of Anticoagulation Therapies
The costs of anticoagulants, lab tests and professional consul-
tations were retrieved from hospital electronic databases of the
anticoagulation clinic (ACC) of Singapore General Hospital
(SGH), the largest tertiary hospital in Singapore. Frequencies
of professional consultations were retrieved from patients’
charts by reviewing the number of clinical visits in patients
who had AF and were taking warfarin, dabigatran or
rivaroxaban in 2012. Patients taking warfarin had an average
of two physician visits and one pharmacist visit in the initia-
tion phase (i.e., 1st month), and a physician visit every
6 months and a pharmacist visit every 2 months in the main-
tenance phase. On average, patients had an INR test during
every visit. Patients taking dabigatran or rivaroxaban had an
average of one physician visit in the initiation phase (i.e., 1st
month), and one physician visit every 3 months in the main-
tenance phase.
578 Cardiovasc Drugs Ther (2014) 28:575–585

Table 1 Base-case values and ranges used in sensitivity analyses

Variable Base-case value Range Reference

Ischemic stroke
Rate of ischemic stroke with warfarin (per 100 patient-years) 1.21 0.61–5.82 [9, 10, 14–16]
Relative risk of ischemic stroke [9]
Dabigatran 110 mg vs warfarin 1.11 0.88–1.39
Dabigatran 150 mg vs warfarin 0.76 0.59–0.97
Hazard ratio of ischemic stroke [10]
Rivaroxaban vs warfarin 0.94 0.75–1.17
Percentage of ischemic stroke with warfarin, dabigatran or rivaroxaban (%) [17]
Fatal (within 30 days) 8.20 5.50–10.90
Major 40.20 35.30–45.10
Minor 42.50 37.60–47.40
Reversible 9.10 NA
ICH
Rate of ICH with warfarin (per 100 patient-years) 0.76 0.59–0.90 [9, 10, 14]
Relative risk of ICH [9]
Dabigatran 110 mg vs warfarin 0.30 0.19–0.45
Dabigatran 150 mg vs warfarin 0.41 0.28–0.60
Hazard ratio of ICH [10]
Rivaroxaban vs warfarin 0.67 0.47–0.93
Percentage of ICH with warfarin, dabigatran or rivaroxaban (%) [14, 32]
Fatal (within 30 days) 36.40 28.30–45.20
Non-fatal 63.60 NA
ECH
Rate of ECH with warfarin (per 100 patient-years) 2.84 2.35–2.93 [9, 14]
Relative risk of ECH [9]
Dabigatran 110 mg vs warfarin 0.94 0.81–1.10
Dabigatran 150 mg vs warfarin 1.07 0.92–1.24
Hazard ratio of ECH [14]
Rivaroxaban vs warfarin 1.04 0.90–1.20
Rate of minor ECH with warfarin (per 100 patient-years) 16.37 10.0–17.0 [9, 10, 14, 17]
Relative risk of minor ECH [9]
Dabigatran 110 mg vs warfarin 0.79 0.74–0.84
Dabigatran 150 mg vs warfarin 0.91 0.86–0.97
Hazard ratio of minor ECH [10]
Rivaroxaban vs warfarin 1.04 0.96–1.13
Percentage of major ECH with warfarin, dabigatran or rivaroxaban (%) [14]
Fatal (within 30 days) 1.47 1.00–4.00
Non-fatal 98.53 NA
MI
Rate of MI with warfarin (per 100 patient-years) 0.64 0.51–1.31 [10, 14, 33, 34]
Relative risk of MI [9]
Dabigatran 110 mg vs warfarin 1.29 0.96–1.75
Dabigatran 150 mg vs warfarin 1.27 0.94–1.71
Hazard ratio of MI [10]
Rivaroxaban vs warfarin 0.81 0.63–1.06
Percentage of MI with warfarin, dabigatran or rivaroxaban (%) [14, 35]
Fatal (within 30 days) 16.60 15.80–17.40
Non-fatal 83.40 NA
Cardiovasc Drugs Ther (2014) 28:575–585 579

Table 1 (continued)

Variable Base-case value Range Reference

Death
Relative risk of non-event death
NVAF 1.30 1.00–1.50 [20, 21, 23]
NVAF and prior stroke 2.30 1.30–3.00 [22]
Health utilities Patient survey
Healthy (i.e., AF without a stroke or bleed)
Warfarin 0.86 0.63–0.98
Dabigatran 0.83 0.58–0.98
Rivaroxaban 0.90 0.73–0.98
Ischemic stroke
Major deficit 0.22 0.00–0.68
Minor deficit 0.69 0.29–0.98
Bleeding
ICH 0.17 0.00–0.62
Major ECH 0.49 0.00–0.90
Minor ECH 0.67 0.06–0.98
TIA 0.64 0.00–0.97
MI 0.46 0.00–0.93
Costs
Monthly cost of anticoagulant (SG$) Hospital databases
Warfarin 9.06 4.72–13.25
Dabigatran 204.54 102.00–307.00
Rivaroxaban 176.54 88.00–265.00
Cost of INR test (SG$) 13.30 NA Hospital databases
Cost of physician consultation (SG$) 71.00 NA Hospital databases
Cost of pharmacist consultation (SG$) 21.60 NA Hospital databases
One-time cost of neurological event (SG$) Hospital databases
Major ischemic stroke 11,147 7,720–17,268
Minor ischemic stroke 3,421 1,710–5,562
TIA 2,153 1,588–2,718
ICH 22,917 15,882–36,786
One–time cost of other event (SG$) Hospital databases
Major ECH 4,481 3,175–15,064
Minor ECH 3,804 1,696–5,959
MI 4,018 565–8,317
Monthly cost of ischemic stroke and ICH (SG$) 1,635 938–2,332 Hospital database and expert opinions

AF atrial fibrillation; ECH extracranial hemorrhage; ICH intracranial hemorrhage; INR international normalized ratio; MI myocardial infarction; NA not
applicable; NVAF non-valvular atrial fibrillation; SG Singapore; TIA transient ischemic attack

Costs of Complications
One-time costs of ischemic stroke, TIA, ICH, ECH and MI
were estimated based on the costs of hospitalization of patients
in the corresponding diagnosis-related groups in the ACC of
SGH in the past 10 years, including facility costs for hospital-
ization, professional costs for a surgeon and anesthetist, and
costs for drugs and lab tests (e.g., magnetic resonance imaging
and x-ray) due to the complications. The monthly cost of
ischemic stroke and ICH was estimated based on the costs
of rehabilitation services at the SGH. The fixed cost of a visit
to the rehabilitation services center was obtained from the
hospital database. The average monthly frequency of visits
was estimated based on expert opinions.
Sensitivity Analyses
One-way sensitivity analyses were performed on all parame-
ters included in the model over a set of plausible ranges of
values that were determined a priori (Table 1). Two-way
580
sensitivity analyses were performed over the combinations of
parameters that influenced the decision of the optimal strategy,
including rate of ischemic stroke on warfarin, cost of
rivaroxaban, and utilities for dabigatran and rivaroxaban. To
evaluate the impact of uncertainty on all the parameter values
simultaneously, a probabilistic sensitivity analysis was per-
formed using a Monte Carlo simulation with 10,000 iterations.
We assumed a log-normal distribution for event rates, a beta or
dirichlet distribution for percentage parameters with two or
four categories, respectively, a beta distribution for utilities,
and a gamma distribution for costs. The parameterization of
distributions is shown in Table 1A in the Appendix.
Results
Model Validation
The 5-year cumulative probabilities of ischemic stroke and
ICH on warfarin projected by the model (i.e., 4.2 % and 2.0 %,
respectively) were comparable to those reported from the
Atrial Fibrillation Follow-up Investigation of Rhythm
Management (AFFIRM) study with more than 4,000 patients
with AF [24], which reported the 5-year probabilities of
ischemic stroke and ICH to be 5.5 and 1.9 %, respectively.
Base-case Analysis
The base-case analysis showed that the QALYs were 8.75
with warfarin, 8.73 with dabigatran 110 mg, 8.82 with
dabigatran 150 mg and 9.33 with rivaroxaban (Table 2). The
total costs were SG$34,648 for warfarin, SG$54,919 for
dabigatran 110 mg, SG$50,484 for dabigatran 150 mg and
SG$51,975 for rivaroxaban. Dabigatran 100 mg gained fewer
QALYs at a higher cost than warfarin, and therefore it was
dominated. The ICER of dabigatran 150 mg versus warfarin
was SG$218,762 (US$196,886) per QALY whereas the ICER
of rivaroxaban versus warfarin was SG$29,697 (US$26,727)
per QALY. Therefore, rivaroxaban and warfarin had extended
dominance over dabigatran 150 mg. Extended dominance is a
method to eliminate from consideration a strategy (i.e.,
dabigatran 150 mg in this study), when a mixed strategy of
Table 2 Projected costs, QALYs and ICERs in base-case analysis
Treatment Cost (SG$) Effectiveness (QALY)
Warfarin 34,648 8.75
Dabigatran (150 mg) 50,484 8.82
Rivaroxaban 51,975 9.33
Dabigatran (110 mg) 54,919 8.73
ICER incremental cost-effectiveness ratio; NA not applicable; QALY quality-adjusted life year; SG Singapore
Cardiovasc Drugs Ther (2014) 28:575–585
two other alternatives (i.e., a proportion of patients were given
rivaroxaban and other patients were given warfarin) can gain
the same or more QALYs with a lower or same cost [25].
Rivaroxaban was cost-effective in the base-case scenario.
One-way Sensitivity Analyses
One-way sensitivity analyses showed that several parameters
influenced the cost-effectiveness of rivaroxaban versus war-
farin. The model was most sensitive to the utility for
rivaroxaban. At a utility value lower than 0.87, warfarin was
the optimal choice. Rivaroxaban was dominated by warfarin
(i.e., having higher costs and lower QALYs than warfarin)
when the utility value was lower than 0.84, leading to a
negative ICER. In addition, the utility for warfarin also influ-
enced the ICER to a great extent. At a utility value higher than
0.90, the ICER of rivaroxaban versus warfarin exceeded
SG$65,000 (US$58,500) per QALY. Rivaroxaban was dom-
inated by warfarin (i.e., having higher costs and lower QALYs
than warfarin) when the utility for warfarin was higher than
0.92. Moreover, when the monthly cost of rivaroxaban was
higher than SG$260, the ICER of rivaroxaban versus warfarin
exceeded SG$65,000 (US$58,500) per QALY. When the
values of other model parameters were varied across plausible
ranges, no substantial influence on the ICER of rivaroxaban
versus warfarin was found. The varied ICER values remained
less than SG$45,000 (US$40,500) per QALY and varied by
less than SG$2,400 (US$2,160) per QALY.
The model was more robust to the variation in parameter
values for the ICERs of dabigatran 150 and 110 mg versus
warfarin. Dabigatran 150 mg was the optimal therapy if the
utility for dabigatran was higher than 0.87. When values of
other parameters varied, dabigatran 150 mg was dominated.
Dabigatran 110 mg was dominated over the entire ranges of
parameter values.
Two-way Sensitivity Analyses
As the cost of rivaroxaban increased, warfarin and dabigatran
150 mg were the optimal therapies in patients with low and
high rates of ischemic stroke on warfarin, respectively
(Fig. 2a). When utility for rivaroxaban was low, warfarin
Incremental cost per QALY gained (SG$ (US$) / QALY)
ICER of a novel anticoagulant versus warfarin Sequential ICER
NA NA
218,762 (196,886) Extended dominance
29,697 (26,727) 29,697 (26,727)
Dominated Dominated
Cardiovasc Drugs Ther (2014) 28:575–585
Fig. 2 Two-way sensitivity analyses on the incremental cost-effectiveness
of rivaroxaban versus warfarin. a Monthly cost of rivaroxaban (SG$) and
rate of ischemic stroke on warfarin (per 100 patient-years) are varied
simultaneously. Dabigatran 110 mg was dominated. SG Singapore. b Utility
for rivaroxaban and rate of ischemic stroke on warfarin (per 100 patient-
years) are varied simultaneously. Dabigatran 110 mg was dominated. c
Utility for dabigatran and rate of ischemic stroke on warfarin (per 100
patient-years) are varied simultaneously. Dabigatran 110 mg was dominated
and dabigatran 150 mg were the optimal therapies in patients
with low and high rates of ischemic stroke on warfarin,
581
respectively (Fig. 2b). In addition, it was also found that
dabigatran 150 mg was the optimal choice if utility for
dabigatran and rate of ischemic stroke on warfarin were high
(Fig. 2c).
Probabilistic Sensitivity Analysis
The probabilities of rivaroxaban, warfarin, dabigatran 150 and
110 mg being the optimal treatment strategy were 97.8, 1.7,
0.5 and 0.0 % , respectively, using a WTP threshold of
SG$65,000 (US$58,500) per QALY, and changed to 99.5,
0.1, 0.4 and 0.0 %, respectively, if the threshold increased to
SG$130,000 (US$117,000) per QALY (Fig. 3).
Discussion
To the best of our knowledge, this is the first study that
conducted a cost-utility analysis of oral anticoagulants from
the perspective of the Singapore healthcare system. Despite its
importance, such study has rarely been conducted in the
Singapore healthcare setting. The findings of this study may
assist clinical decision-making and also provide useful infor-
mation for future cost-effectiveness studies in Singapore as
well as other countries with similar healthcare settings.
In the base-case analysis, both high- and low-dose
dabigatran were dominated and rivaroxaban was a cost-
effective therapy compared with warfarin. The robustness of
the results was further confirmed by the probabilistic sensitiv-
ity analysis over a wide range of values for the model inputs.
The cost-effectiveness of rivaroxaban versus warfarin was
most sensitive to the cost of rivaroxaban and utilities for
rivaroxaban and warfarin.
The findings of our study were similar to those reported in a
previous study [14], which found that the ICER of rivaroxaban
versus warfarin was US$27,498 per QALY and that rivaroxaban
was cost-effective in 80.1 and 91.4 % of the Monte Carlo
simulations, using a WTP threshold of US$50,000 and
US$100,000 per QALY, respectively. The findings of determin-
istic sensitivity analyses in our study were also similar to those
reported in the previous study [14]. The risk of ischemic stroke
on warfarin with CHADS2 scores ranging from 0 to 6 did not
affect the cost-effectiveness of rivaroxaban versus warfarin. The
ICER was primarily driven by the cost of rivaroxaban, which
increased as the cost increased.
Another study evaluated the cost-effectiveness of
rivaroxaban, dabigatran 150 mg and apixaiban versus warfa-
rin and found that the ICER of rivaroxaban versus warfarin
was US$3,190 per QALY [26], which was lower than that of
the present study. One explanation could be that the previous
study adapted a societal perspective and included costs of
patient time for clinical visits in the estimation of the total
582
Fig. 3 Cost-effectiveness
acceptability curves. The dotted
line represents the cost-
effectiveness threshold of
SG$65,000 (US$58,500) per
QALY. QALYquality-adjusted life
year; SG Singapore; WTP
willingness-to-pay
costs of oral anticoagulation therapies. As the annual cost of
patient time spent on INR tests for warfarin (i.e., US$
1,750.92) was much higher than that for other oral anticoag-
ulants (i.e., US$ 229.36), the difference in costs between
warfarin and other anticoagulant therapies was reduced, which
may have decreased the ICER of rivaroxaban versus warfarin.
Although the ICER of rivaroxaban versus warfarin was low in
the base-case analysis of the previous study, the probabilistic
sensitivity analysis found that rivaroxaban was the optimal
treatment strategy in only 14.9 and 4.4 % of the Monte Carlo
simulations, using a WTP threshold of US$50,000 and
US$100,000 per QALY, respectively. One explanation of the
low probabilities could be the inclusion of apixaban, which
had lower rates of all adverse events than rivaroxaban in the
model. Therefore, the probability of rivaroxaban being the
optimal strategy was reduced. Another explanation could be
the difference in utilities for anticoagulants. In our study, the
mean utility value of dabigatran was lower than that of
rivaroxaban. However, in the previous study, dabigatran was
assumed to have the same utility as rivaroxaban (i.e., decre-
ment in utility for anticoagulation in the base case=− 0.0105;
sensitivity range=−0.0110 to −0.0090), which further reduced
the probability of rivaroxaban being the optimal strategy.
Previous studies found that dabigatran was another cost-
effective anticoagulant compared with warfarin [17, 27]. A study
that used clinical event rates derived from the RE-LY trial found
that the ICER of dabigatran 150 mg versus warfarin was
US$45,372 per QALY and that dabigatran 150 mg was the
optimal therapy in 53 and 68 % of the Monte Carlo simulations,
using a WTP threshold of US$50,000 and US$100,000 per
QALY, respectively [17]. Another study found that the cost-
effectiveness of high-dose dabigatran was sensitive to patients’
INR control, and that high-dose dabigatran was only cost-
effective compared with warfarin in patients with low time in
the therapeutic range (TTR) [27]. However, despite the similar
model used in our study, it was found that dabigatran 150 mg
Cardiovasc Drugs Ther (2014) 28:575–585
was dominated. One explanation could be that the local utilities
used in our study were different from those used in the previous
studies [17, 27]. For example, when eliciting the utility for
dabigatran in our study, participants were asked to consider
dyspepsia as an adverse effect of dabigatran [9, 28, 29], which
may have led to the lower mean utility for dabigatran than
warfarin. However, in previous studies, the utility for dabigatran
was 0.994, which was derived from the utility for ximelagatran
that was elicited by expert opinions [30] and higher than that of
warfarin (i.e., 0.987) [17, 27].
The values of clinical parameters in this model were
extracted from the RE-LY and ROCKET-AF trials.
Differences in the clinical characteristics of participants
such as CHADS2 scores and TTR in the two trials
challenged the cross-trial comparison, as patients with
different clinical characteristics may have had different
rates of clinical events. In our study, the base-case
analysis used clinical data derived from the RE-LY trial
whereas those from the ROCKET-AF study were used
only when the data from the RE-LY trial were not
available. This decision was made because the average
TTR of participants in the RE-LY trial was very close
to the mean TTR of patients from which the cost and
utility data were derived in this study. Moreover, the
clinical parameter values derived from the RE-LY trial
were comparable to those used in previous cost-
effectiveness studies that compared multiple novel oral
anticoagulants with warfarin [26, 31].
This study has several limitations. First of all, the rates of
clinical events in the model were derived from clinical trials,
which may have overestimated the efficacy and safety of
therapies because participants enrolled in the trials may have
had higher medication adherence and more intensive moni-
toring than patients in general practice. Second, the rates of
clinical events were derived from clinical trials with a median
follow-up period of approximately 2 years and extrapolated
Cardiovasc Drugs Ther (2014) 28:575–585 583

over the lifetime horizon. Nevertheless, the real rates of clin- Well on rivaroxaban
ical events over time could be different from the extrapolated
rates. Third, the event rates used in this study were from the You need to take an oral medicine once daily with a fixed
trials conducted in western populations. Whether they can be dose. Generally you are in your good health state.
generalized to Asian populations remains unknown. Fourth,
due to the availability of cost data, the costs of complications Part 2
were derived from patients in an ACC and may not be gener-
alizable to patients in the inpatient setting. However, as the
study focused on the incremental cost-effectiveness of thera- Table 3 Parameterization of distributions for the probabilistic sensitivity
analysis
pies, its results may not have been significantly impacted by
the use of complication costs from a single healthcare setting. Variable Distribution (parameters)
In conclusion, this study found that rivaroxaban was a cost- Clinical inputs Log-normal (u, sigma)a
effective therapy compared with warfarin for the prevention of
Ischemic stroke
stroke in patients with AF. Both high- and low-dose
Rate of ischemic stroke with 0.191, 0.575
dabigatran therapies were dominated. The findings were ro- warfarin (per 100 patient-years)
bust in the deterministic and probabilistic sensitivity analyses. Relative risk of ischemic stroke
Dabigatran 110 mg vs warfarin 0.104, 0.117
Acknowledgments We would like to acknowledge the National Uni- Dabigatran 150 mg vs warfarin −0.274, 0.127
versity of Singapore for providing us with a research grant. Hazard ratio of ischemic stroke
Rivaroxaban vs warfarin −0.062, 0.113
Conflict of Interest The authors declare that they have no conflict of
Percentage of ischemic stroke with Dirichlet (alpha)
interest.
warfarin, dabigatran or
rivaroxaban (%)
Fatal (within 30 days) 80
Appendix
Major 350
Minor 340
Part 1
Reversible 70
ICH
Descriptions of health states
Rate of ICH with warfarin −0.274, 0.108
(per 100 patient-years)
Descriptions of health states (available upon request) Relative risk of ICH
were developed based on preference assessment guide- Dabigatran 110 mg vs warfarin −1.204, 0.220
lines, medical textbooks, published literature and expert Dabigatran 150 mg vs warfarin −0.892, 0.194
opinions (i.e., two researchers with Ph.D. degrees in Hazard ratio of ICH
pharmacy, two physicians with master’s degrees in med- Rivaroxaban vs warfarin −0.400, 0.174
icine, a clinical pharmacist with a master’s degree in Percentage of ICH with warfarin, Beta (alpha, beta)
pharmacy and a Ph.D. candidate in pharmacy). dabigatran or rivaroxaban (%)
Important attributes of the health states were described. Fatal (within 30 days) 44.974, 78.580
Examples are shown below. Non-fatal –
ECH
Rate of ECH with warfarin 1.044, 0.056
Well on warfarin (per 100 patient-years)
Relative risk of ECH
You need to take an oral medicine once daily with occasional Dabigatran 110 mg vs warfarin −0.062, 0.078
dose adjustments. You need to attend the outpatient clinic at a Dabigatran 150 mg vs warfarin 0.068, 0.076
hospital or polyclinic about once a month for a blood test. You Hazard ratio of ECH
are not able to drink too much alcohol. You may bruise more Rivaroxaban vs warfarin 0.039, 0.073
easily, but generally you are in your good health state. Rate of minor ECH with warfarin 2.795, 0.135
(per 100 patient-years)
Relative risk of minor ECH
Well on dabigatran
Dabigatran 110 mg vs warfarin −0.236, 0.032
Dabigatran 150 mg vs warfarin −0.094, 0.031
You need to take an oral medicine twice daily with a fixed
Hazard ratio of minor ECH
dose. You may have dyspepsia (i.e., stomach discomfort or
Rivaroxaban vs warfarin 0.039, 0.042
burning pain), but generally you are in your good health state.
584 Cardiovasc Drugs Ther (2014) 28:575–585

Table 3 (continued) Table 3 (continued)

Variable Distribution (parameters) Variable Distribution (parameters)

Clinical inputs Log-normal (u, sigma)a Clinical inputs Log-normal (u, sigma)a

Percentage of major ECH with Beta (alpha, beta) TIA 74.149, 0.034
warfarin, dabigatran or ICH 25.936, 0.001
rivaroxaban (%)
One-time cost of other event (SG$)
Fatal (within 30 days) 3.621, 242.675
Major ECH 5.019, 0.001
Non-fatal –
Minor ECH 2.316, 0.001
MI
MI 4.036, 0.001
Rate of MI with warfarin −0.446, 0.241
(per 100 patient-years) Monthly cost of ischemic 16.701, 0.010
Relative risk of MI stroke and ICH (SG$)
Dabigatran 110 mg vs warfarin 0.255, 0.153
AF atrial fibrillation; ECH extracranial hemorrhage; ICH intracranial
Dabigatran 150 mg vs warfarin 0.239, 0.153 hemorrhage; INR international normalized ratio; MI myocardial infarc-
Hazard ratio of MI tion; NA not applicable; NVAF non-valvular atrial fibrillation; SG Singa-
Rivaroxaban vs warfarin −0.211, 0.133 pore; TIA transient ischemic attack
a
Percentage of MI with warfarin, Beta (alpha, beta) Clinical inputs were assumed to follow a log-normal distribution unless
dabigatran or rivaroxaban (%) otherwise stated
Fatal (within 30 days) 1379.311, 6929.790
Non-fatal –
Death
Relative risk of non-event death
NVAF 0.262, 0.103
NVAF and prior stroke 0.833, 0.213 References
Health utilities Beta (alpha, beta)
Healthy (i.e., AF without a 1. Singer DE, Albers GW, Dalen JE, Go AS, Halperin JL, Manning WJ.
stroke or bleed) Antithrombotic therapy in atrial fibrillation: the seventh ACCP con-
Warfarin 3.702, 0.618 ference on antithrombotic and thrombolytic therapy. Chest.
Dabigatran 3.404, 0.678 2004;126:429S–56S.
2. World Health Organization. Fact sheet—the top 10 causes of death in
Rivaroxaban 2.559, 0.278
2011. 2012. http://www.who.int/mediacentre/factsheets/fs310/en/
Ischemic stroke index.html . Accessed 1 Aug 2013.
Major deficit 0.479, 1.740 3. Albers GW, Amarenco P, Easton JD, Sacco RL, Teal P.
Minor deficit 1.699, 0.771 Antithrombotic and thrombolytic therapy for ischemic stroke: the
seventh ACCP conference on antithrombotic and thrombolytic ther-
Bleeding apy. Chest. 2004;126:483S–512S.
ICH 0.328, 1.637 4. Singapore Ministry of Health. Singapore health facts: population and
Major ECH 1.406, 1.458 vital statistics in 2012. 2013. http://www.moh.gov.sg/content/moh_
web/home/statistics/Health_Facts_Singapore/Population_And_
Minor ECH 1.623, 0.782
Vital_Statistics.html . Accessed 26 Aug 2013.
TIA 1.672, 0.936 5. Ansell J, Hirsh J, Poller L, Bussey H, Jacobson A, Hylek E. The
MI 1.387, 1.621 pharmacology and management of the vitamin K antagonists: the
Costs Gamma (alpha, lambda) seventh ACCP conference on antithrombotic and thrombolytic ther-
apy. Chest. 2004;126:204S–33S.
Monthly cost of anticoagulant
6. Hylek EM, Go AS, Chang Y, Jensvold NG, Henault LE, Selby JV,
(SG$)
et al. Effect of intensity of oral anticoagulation on stroke severity and
Warfarin 5.130, 0.566 mortality in atrial fibrillation. N Engl J Med. 2003;349:1019–26.
Dabigatran 20.660, 0.101 7. Levine MN, Raskob G, Beyth RJ, Kearon C, Schulman S.
Rivaroxaban 15.391, 0.087 Hemorrhagic complications of anticoagulant treatment: the seventh
ACCP conference on antithrombotic and thrombolytic therapy.
Cost of INR test (SG$) – Chest. 2004;126:287S–310S.
Cost of physician consultation – 8. Merli GJ, Tzanis G. Warfarin: what are the clinical implications of an
(SG$) out-of-range-therapeutic international normalized ratio. J Thromb
Cost of pharmacist consultation – Thrombolysis. 2009;27:293–9.
(SG$) 9. Connolly SJ, Ezekowitz MD, Yusuf S, Reilly PA, Wallentin L.
One-time cost of neurological Newly identified events in the RE-LY trial. N Engl J Med.
event (SG$) 2010;363:1875–6.
Major ischemic stroke 32.680, 0.003 10. Patel MR, Mahaffey KW, Garg J, Pan G, Singer DE, Hacke W, et al.
Minor ischemic stroke 18.283, 0.005 Rivaroxaban versus warfarin in nonvalvular atrial fibrillation. N Engl
J Med. 2011;365:883–91.
Cardiovasc Drugs Ther (2014) 28:575–585
11. Singapore Department of Statistics. Singapore’s 2012 per capita
GDP. 2013. http://www.singstat.gov.sg/statistics/latest_data.html .
Accessed 3 July 2013.
12. The World Bank. 2012 PPP conversion factor (GDP) to market
exchange rate ratio. 2013. http://data.worldbank.org/indicator/PA.
NUS.PPPC.RF . Accessed 2 Feb 2014.
13. World Health Organization. CHOosing Interventions that are Cost
Effective (WHO-CHOICE): cost-effectiveness thresholds. 2013.
http://www.who.int/choice/costs/CER_thresholds/en/index.html .
Accessed 4 Dec 2013.
14. 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:
845–51.
15. Gage BF, Waterman AD, Shannon W, Boechler M, Rich MW,
Radford MJ. Validation of clinical classification schemes for
predicting stroke: results from the National Registry of Atrial
Fibrillation. JAMA. 2001;285:2864–70.
16. Laupacis. Risk factors for stroke and efficacy of antithrombotic
therapy in atrial fibrillation: analysis of pooled data from five ran-
domized controlled trials. Arch Intern Med. 1994;154:1449–57.
17. Freeman JV, Zhu RP, Owens DK, Garber AM, Hutton DW, Go AS,
et al. Cost-effectiveness of dabigatran compared with warfarin for
stroke prevention in atrial fibrillation. Ann Intern Med. 2011;154:1–11.
18. Ariesen MJ, Claus SP, Rinkel GJ, Algra A. Risk factors for intrace-
rebral hemorrhage in the general population: a systematic review.
Stroke. 2003;34:2060–5.
19. Singapore Department of Statistics. Population trends in 2013. 2014.
http://www.singstat.gov.sg/publications/publications_and_papers/
population_and_population_structure/population2013.pdf .
Accessed 25 Jan 2014.
20. Dries DL, Exner DV, Gersh BJ, Domanski MJ, Waclawiw MA,
Stevenson LW. Atrial fibrillation is associated with an increased risk
for mortality and heart failure progression in patients with asymp-
tomatic and symptomatic left ventricular systolic dysfunction: a
retrospective analysis of the SOLVD trials—studies of left ventricular
dysfunction. J Am Coll Cardiol. 1998;32:695–703.
21. Wyse DG, Love JC, Yao Q, Carlson MD, Cassidy P, Greene LH,
et al. Atrial fibrillation: a risk factor for increased mortality—an
AVID registry analysis. J Interv Card Electrophysiol. 2001;5:267–73.
22. Dennis MS, Burn JP, Sandercock PA, Bamford JM, Wade DT,
Warlow CP. Long-term survival after first-ever stroke: the
Oxfordshire community stroke project. Stroke. 1993;24:796–800.
23. Yuan Z, Bowlin S, Einstadter D, Cebul RD, Conners Jr AR, Rimm
AA. Atrial fibrillation as a risk factor for stroke: a retrospective cohort
585
study of hospitalized medicare beneficiaries. Am J Public Health.
1998;88:395–400.
24. Sherman DG, Kim SG, Boop BS, Corley SD, DiMarco JP, Hart RG,
et al. Occurrence and characteristics of stroke events in the Atrial
Fibrillation Follow-up Investigation of Sinus Rhythm Management
(AFFIRM) study. Arch Intern Med. 2005;165:1185–91.
25. Cantor SB. Cost-effectiveness analysis, extended dominance,
and ethics: a quantitative assessment. Med Dec Making.
1994;14:259–65.
26. Harrington AR, Armstrong EP, Nolan Jr PE, Malone DC. Cost-
effectiveness of apixaban, dabigatran, rivaroxaban, and warfarin for
stroke prevention in atrial fibrillation. Stroke. 2013;44:1676–81.
27. Shah SV, Gage BF. Cost-effectiveness of dabigatran for stroke pro-
phylaxis in atrial fibrillation. Circulation. 2011;123:2562–70.
28. NHS Choices. Anticoagulant medicines: pradaxa side-effects. 2013.
http://www.nhs.uk/conditions/anticoagulant-medicines/pages/
MedicineSideEffects.aspx?condition=Blood%
20clotting&medicine=Pradaxa&preparation=Pradaxa%20110mg%
20capsules . Accessed 20 Jan 2013.
29. MedlinePlus. Drugs and supplements: dabigatran. 2013. http://www.
nlm.nih.gov/medlineplus/druginfo/meds/a610024.html . Accessed
20 Jan 2013.
30. O’Brien CL, Gage BF. Costs and effectiveness of ximelagatran for
stroke prophylaxis in chronic atrial fibrillation. JAMA. 2005;293:
699–706.
31. Coyle D, Coyle K, Cameron C, Lee K, Kelly S, Steiner S, et al. Cost-
effectiveness of new oral anticoagulants compared with warfarin in
preventing stroke and other cardiovascular events in patients with
atrial fibrillation. Value Health. 2013;16:498–506.
32. Hylek EM, Singer DE. Risk factors for intracranial hemor-
rhage in outpatients taking warfarin. Ann Intern Med. 1994;
120:897–902.
33. Hohnloser SH, Oldgren J, Yang S, Wallentin L, Ezekowitz M, Reilly
P, et al. Myocardial ischemic events in patients with atrial fibrillation
treated with dabigatran or warfarin in the RE-LY (randomized eval-
uation of long-term anticoagulation therapy) trial. Circulation.
2012;125:669–76.
34. You JHS, Tsui KKN, Wong RSM, Cheng G. Cost-effectiveness of
dabigatran versus genotype-guided management of warfarin therapy
for stroke prevention in patients with atrial fibrillation. PLoS ONE.
2012;7:e39640.
35. Krumholz HM, Merrill AR, Schone EM, Schreiner GC, Chen J,
Bradley EH, et al. Patterns of hospital performance in acute myocar-
dial infarction and heart failure 30-day mortality and readmission.
Circ Cardiovasc Qual Outcome. 2009;2:407–13.