Clinical Therapeutics/Volume 36, Number 10, 2014

Cost-Utility Analysis of Oral Anticoagulants for Nonvalvular

Atrial Fibrillation Patients at the Police General Hospital,
Bangkok, Thailand
Siriporn Jarungsuccess, BScPharm, MPPM; and Satadon Taerakun, PharmD
Pharmacy Department, Police General Hospital, Bangkok, Thailand

ABSTRACT Key words: Asian, atrial fibrillation, cost utility,

Purpose: The genetic polymorphism was one of the new oral anticoagulant, stroke.
major considerations for adjusting doses of warfarin
in Thai individuals. As a result, new oral anticoagu-
lants (NOACs) were introduced to achieve thera-
INTRODUCTION
peutic goals in stroke prevention in atrial fibrillation
Ischemic stroke is the most common thromboembolic
(SPAF) patients. However, a cost-utility analysis in a
complication found in NVAF patients.1 Although the
population-specific model was lacking in Thailand.
prevalence of nonvalvular atrial fibrillation (NVAF) in
This study was performed to determine which NOACs
Asia is less than half that in the West,2 the risk of
yielded population-specific, cost-effective results for
thromboembolic events is 2 times higher.3 In particular,
SPAF compared with warfarin from both governmen-
neurologic deficits resulting from stroke events leading
tal and societal perspectives in Thailand.
to family dependence and financial burden represent
Methods: A simplified Markov health state model
the most concerning issue. Consequently, the Heart
was constructed to calculate the lifetime cost, life-
Association of Thailand has issued recommendations
years saved, and quality-adjusted life-years (QALYs)
on the use of anticoagulants to prevent stroke in
gained. Asia-specific clinical event parameters were
moderate- to high-risk patients.2
defined from systematic searches of PubMed. Cost
Even though warfarin, the standard anticoagulant
and utility input was obtained from hospital based
preferred in Thailand, has many advantages (eg,
data collection.
familiar to most physicians, comparatively inexpen-
Findings: Although NOACs produced more life-
sive, ability to predict the therapeutic efficacy and
years saved and QALYs gained resulting from the
safety from the international normalized ratio (INR),
base-case versus warfarin, the lifetime costs of new
availability of the exact antidote),2 the core limitations
alternatives increased to 41.4 times the comparative
of its use are interindividual variation including gen-
cost of warfarin. This caused an incremental cost-
etic polymorphism, slow onset and offset of action,
effective ratio that exceeded Thailand’s cost-
numerous drug and food interactions, and incon-
effectiveness threshold. The probabilistic sensitivity
venience of frequent INR monitoring.4 As a result,
analysis denoted the robustness of our model and
new oral anticoagulants (NOACs) were developed to
revealed that dose-adjusted warfarin was the most
overcome these limitations.5
cost-effective option in 499% of iterations. NOACs
In Thailand, there are currently 3 NOACs available
produced cost-effective results when the medication
for stroke prevention in atrial fibrillation (SPAF) that
unit cost was decreased by at least 85%.
have been approved by the US Food and Drug
Implications: According to the results of this first Administration: 1 is a direct thrombin inhibitor
cost-utility analysis in Thailand, warfarin is still the (dabigatran) and 2 are direct factor Xa inhibitors
most cost-effective medication for SPAF from any
perspective in Thailand at the threshold recommended Accepted for publication August 26, 2014.
by our health technology assessment guidelines. (Clin http://dx.doi.org/10.1016/j.clinthera.2014.08.016
Ther. 2014;36:1389–1394) & 2014 Elsevier HS Jour- 0149-2918/$ - see front matter
nals, Inc. All rights reserved. & 2014 Elsevier HS Journals, Inc. All rights reserved.

October 2014 1389

 Clinical Therapeutics

(rivaroxaban and apixaban). Their primary advan- complications (eg, hydrocephalus, seizures, venous
tages over warfarin are that their thromboembolic thrombotic events, hyperglycemia, increased blood
preventive effects are at least equal to those of pressure, fever, infections), extracranial hemorrhage
warfarin but with no INR monitoring required.5 (eg, major gastrointestinal bleeding with/without com-
Conversely, the budget gain from medical expenses plications such as hypovolemia and shock); nondis-
was often realized as a financial burden in Thailand. abling major bleeding (defined by mRS scores of 0–2);
Finally, there were no population-specific cost-effec- disabling major bleeding (defined by mRS scores of 3–
tiveness studies available, either in Thailand specifi- 5); myocardial infarction with/without complications
cally or in Asia in general. (eg, acute heart failure, arrhythmia); full recovery
Therefore, this study was conducted to evaluate the from myocardial infarction (defined as a successful
cost-utility analysis of NOACs compared with war- percutaneous coronary intervention); and death (from
farin for SPAF in the Thailand using the Asia-Pacific any health state including natural cause of death). The
subgroup analysis parameters from 3 main studies assumptions were set into the model as (1) the treat-
which had been submitted to the US Food and Drug ment effect exhibited immediately after starting and
Administration for approval: the RE-LY (Randomized remaining constant throughout life, (2) the adherence
Evaluation of Long-Term Anticoagulation Therapy),6 to each alternative was similar, and (3) drug was
ROCKET-AF (Rivaroxaban Once Daily Oral Direct discontinued until the patient died.
Factor Xa Inhibition Compared with Vitamin K
Antagonism for Prevention of Stroke and Embolism Clinical Treatment Effect and Transitional
Trial in Atrial Fibrillation),7 and ARISTOTLE Probability Parameters
(Apixaban for Reduction in Stroke and Other A systematic review was conducted in MEDLINE
Thromboembolic Events in Atrial Fibrillation).8 via PubMed to gather clinical model input. Three
notable published studies (subgroup analysis of RE-
METHODS LY, ROCKET-AF, and ARISTOTLE) were recruited
Model Structure and their relevant data extracted to obtain the relative
A Markov health state model was adapted from risk of such alternatives (Supplemental Table I). The
related health technology assessment literature9,10 and baseline annual rate of important clinical events while
reviewed by a cardiologist (Supplemental Figure 1). taking warfarin were calculated by the pooled mean
The model cohorts were patients older than 65 years and SD from Asia-Pacific regional data according to the
of age with newly diagnosed NVAF, a moderate to formula reported elsewhere.11 The pooled estimation of
high risk of stroke (CHADS2 score [Congestive heart clinical relevance events of NOACs was derived from a
failure, Hypertension, Age Z75, Diabetes mellitus, meta-analysis using Review Manager software version
and prior Stroke or transient ischemic attack 5.2. The mortality rate was multiplied by factors of 3.7,
(doubled)] Z2), and no history of stroke. Each 3.7, and 1.05 after ischemic stroke, intracranial hemor-
patient included in the model was assigned to one of rhage, and myocardial infarction, respectively.9 The
the following strategies: dose-adjusted warfarin (target transitional probabilities of health outcomes were
INR of 2–3), dabigatran 150 mg BID, dabigatran 110 obtained from local literature (Thailand journal) to
mg BID, rivaroxaban 20 mg/day OD, or apixaban 5 reflect our context (Supplemental Table I).
mg BID. Starting from a well state, each selected Because Asian individuals are more prone to
patient could then be in any one of 10 states of health experiencing dyspepsia than white individuals,3 that
in 1-year cycles for 30 years or until death. The 10 was the only adverse drug reaction (ADR) in this
health states included in the model were as follows: model. The ADR event rate of NOACs other than
well/full recovery from any health state; ischemic dabigatran was assumed to be equal to that of
stroke with/without complications (eg, pneumonia, warfarin. Minor bleeding rates and consequences
seizure, urinary tract infection, pressure sore); nondis- were excluded from our analysis because most of
abling ischemic stroke (defined by modified Rankin our patients were advised in self-management by a
Scale [mRS] scores of 0–1); disabling ischemic stroke team of clinical pharmacists so fewer patients pre-
(defined by mRS scores of 2–5); major bleeding sented to a hospital for further investigation or treat-
(defined by intracranial hemorrhage) with/without ment in such cases.

1390 Volume 36 Number 10

 S. Jarungsuccess and S. Taerakun

Utility Values analysis of NOACs unit costs was conducted to obtain

Utility values were derived from population-specific the optimal cost that yielded cost-effectiveness at 1
articles through a systematic search of MEDLINE via gross domestic product (GDP) per capita.
PubMed for published articles and via both the Thai
Library Integrated System and Thai Theses Online for RESULTS
unpublished articles. Unless the utility parameters of Base-Case Analysis
any state were present, collection of hospital based From the base-case analysis, dabigatran 150 mg
data using the Thai version of the EuroQol 5- BID and apixaban 5 mg BID were associated with the
dimension questionnaire was performed.12 Then highest QALYs gained compared with other anti-
quality-adjusted life-years (QALYs) were projected coagulants (Table I), whereas the warfarin treatment
from life-years saved multiplied by utility values. group generated the lowest QALYs gained. Apixaban
5 mg was the most expensive alternative in any
Costs and Related Input perspective. In a subgroup analysis stratified by
All direct medical and direct nonmedical costs were CHADS2 score, the results of both expected cost and
obtained from 5-year hospital based data extraction. QALYs gained were consistent with the base-case
Disease-specific costs, number of hospital visits, cost outcomes. According to base-case and subgroup re-
per unit of anticoagulants, and laboratory and inves- sults, there are no NOACs that yield an ICER below
tigation costs selected by the International Statistical the Thailand threshold at 1 GDP per capita in the
Classification of Diseases and Related Health Prob- analysis year.
lems 10th Revision were calculated. Because apixaban
5 mg is not now available in Thailand, two 2.5-mg Probabilistic Sensitivity Analysis
tablets were used to obtain a 5-mg dose. The average The probabilistic sensitivity analysis (PSA) results,
warfarin daily dose required to maintain the INR in which compared NOACs with warfarin, are displayed
the target range (2–3) and for the informal care of in a cost-effectiveness plane (Figure 1). Following the
post-stroke patients was obtained from published threshold of 1 and 3 times the GDP per capita, all
articles elsewhere.13,14 Out-of-pocket expenses for iterations were located above the threshold. There was
transportation, food, and accommodation was based only 1 in 5,000 iterations that was located below the
on the Standard Cost List for Health Technology threshold when 5 times of GDP per capita was used as
Assessment.15 the willingness-to-pay (WTP).
From the government perspective, the direct non-
medical cost was excluded from analysis but was Threshold Analysis of NOAC Unit Cost
included when the societal perspective was evaluated. The WTP was set at 1 GDP per capita in 2013 Thai
If any input was expressed in 2013 Thai baht, Thai- baht per QALYs gained as the cost-effective threshold.
land’s consumer price index was used to inflate them. NOAC unit costs were varied until the base-case ICER
at threshold was obtained. To achieve cost-effectiveness
Analysis in the market, NOACs need to reduce their current
The incremental cost-effective ratio (ICER) in market prices by at least 85% (Supplemental Table II).
both base-case and subgroup analysis stratified by
CHADS2 score was derived by dividing the cost diffe- Budget Impact Analysis
rence by the QALY difference among each NOAC and The prevalence of NVAF in our setting was
warfarin. Following our Thailand Health Technology estimated at 0.04% yearly; the total number of
Assessment Guideline, cost and health outcomes were identically matched cases in Thailand was 2,659
discounted at a rate of 3%. A probability sensitivity patients. The percentage of anticoagulants prescribed
analysis was performed using Microsoft Office Excel for NVAF was estimated at 15% to 60%, and the
2013 (Microsoft Corp, Redmond, Washington) to proportion of patients considered unsuitable to receive
review the impact of input variables on the study warfarin was assumed at between 0.14 and 0.44.
results. Log-normal, β, and γ distributions were used Following these parameters, the maximum budget of
for relative risk, transitional probability, and utility NOACs by National List of Essential Medicines policy
and resource use parameters, respectively. A threshold was
168,364,400.53 Thai baht in the first year and

October 2014 1391

 Clinical Therapeutics

Table I. Expected cost and QALYs in base-case and subgroup analyses.

Government Perspective Societal perspective

Strategy Cost ( ) QALYs ICER Cost ( ) QALYs ICER

Base-case
Dabigatran 150 mgBID 186,641.60 2.34 2,268,738.48 335,804.35 2.34 2,252,938.19
Dabigatran 110 mg BID 187,653.46 2.29 46,426,823.22 337,267.66 2.29 46,286,254.56
Rivaroxaban 20 mg OD 173,149.61 2.31 5,050,231.84 322,713.63 2.31 5,030,280.45
Apixaban 5 mg BID 299,536.42 2.33 5,583,860.99 448,747.82 2.33 5,565,388.48
Dose-adjusted Warfarin 71,184.43 2.29 — 221,151.27 2.29 —
CHADS2 score r 2
Dabigatran 150 mg BID 186,748.20 2.33 2,535,524.53 335,959.69 2.33 2,518,951.89
Dabigatran 110 mg BID 187,668.28 2.29 65,944,679.23 337,289.23 2.29 65,748,864.69
Rivaroxaban 20 mg OD 173,193.96 2.31 5,686,276.58 322,778.63 2.31 5,664,973.82
Apixaban 5 mg BID 300,099.13 2.30 14,015,073.58 449,534.87 2.30 13,982,557.34
Dose-adjusted Warfarin 71,184.43 2.29 — 221,151.27 2.29 —
CHADS2 score Z 3
Dabigatran 150 mgBID 186,960.42 2.32 3,308,592.48 336,268.88 2.32 3,289,777.82
Dabigatran 110 mgBID 187,193.14 2.31 4,543,093.27 336,597.37 2.31 4,521,060.53
Rivaroxaban 20 mg OD 173,284.06 2.30 7,639,459.04 322,910.70 2.30 7,614,004.55
Apixaban 5 mg BID 299,677.37 2.32 6,575,136.64 448,944.98 2.32 6,555,015.63
Dose-adjusted Warfarin 71,184.43 2.29 — 221,151.27 2.29 —

decreased in the subsequent year. When the unsuit- by differences in the model input. First, both ischemic
ability of warfarin was implied, the maximum budget and hemorrhagic stroke rates were higher than in
for apixaban 5 mg was 74,080,336.23 Thai baht in other reports. Second, comorbidity treatment cost was
the first year (Supplemental Table III). included in our analysis. Finally, the apixaban unit
cost was assumed by multiplying the unit cost of 2.5
DISCUSSION mg by 2. These factors caused the lifetime cost in the
This is the first health technology assessment in Thai- present study to be higher than in previously pub-
land that evaluated the cost utility of NOACs com- lished literatures.
pared with warfarin in SPAF by using population- NOACs dominated warfarin in 490% of itera-
matched parameters. The base-case analysis revealed tions when decision makers had a WTP of
that warfarin produced higher cost-effectiveness prob- Z4,000,000 Baht per QALY gained. Consequently,
ability than NOACs in any scenario. This is consistent policymakers should not include NOACs for SPAF in
with the findings of the PSA and indicates the robust- all NVAF patients. When WTP was set at 1 GDP per
ness of the model. capita following Thailand’s guideline, all NOAC costs
As far as we know, the findings of this analysis are should be negotiated.
inconsistent with those of any previously published The methods of patient selection or conditioning
studies. Coyle et al10 showed that dabigatran 150 mg used in this study are of interest to policymakers. If the
and apixaban dominated warfarin at a WTP threshold budget allocation for health technology is dependent
of $50,000 per QALY gained. Harrington et al9 on financial implications, then the most valuable way
reported that NOACs (dabigatran 150 mg, riva- to use NOACs is to prescribe them only in cases in
roxaban, and apixaban) were cost-effective alterna- which warfarin is unsuitable. In clinical practice,
tives when using a WTP of $50,000 per QALY warfarin-unsuitable patients account for 14% to
gained. This contradiction could be explained in part 44% of total SPAF patients.4 One of the most

1392 Volume 36 Number 10

 S. Jarungsuccess and S. Taerakun

115,000.00
110,000.00
105,000.00
100,000.00
95,000.00
90,000.00
85,000.00 5 GDP per capita
80,000.00
Incremental cost (Thai Baht)

75,000.00
70,000.00
65,000.00
60,000.00
55,000.00
3 GDP per capita
50,000.00
45,000.00
40,000.00
35,000.00
30,000.00
25,000.00
20,000.00
15,000.00 1 GDP per capita
10,000.00
5,000.00
0.00
–0.02 0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14
Incremental QALYs gained

Figure 1. Cost-effective plan calculated from 5,000 simulations that compared new oral anticoagulants with
warfarin. GDP ¼ gross domestic product

important limitations in prescribing warfarin is herbal healing, which is common in Thailand, such real-
hemorrhagic complications. Hori et al3 documented world practices should be taken into consideration by
that the bleeding rate in Asians was much higher policymakers during their decision making. In addition,
than in non-Asians. In the study conducted by Shah all direct medical costs were collected solely in the Police
and Gage,16 NOACs were cost-effective when the General Hospital. This could cause different outcomes
CHADS2 score and risk of hemorrhage predicted by when generalized to other settings.
the HEMORR2HAGES score were Z3. Therefore, the
risk of bleeding should be evaluated and documented CONCLUSIONS
in all cases before NOACs are prescribed. According to this analysis, the government-supported
One other point that has been of concern among medication for SPAF in patients 65 years or age and
some clinicians recently is genetic polymorphism, older with a moderate to high risk of stroke (CHADS2
particularly in Thais. From Asian base-case input,3 score Z2) in Thailand should continue to be treated
the mean time in the therapeutic range was 54.5%, with warfarin. NOACs were determined as likely to
which was higher than in our setting by
20%. As a be cost-effective only when their prescription is limited
result, the budget for NOACs could be much lower to patients for whom warfarin is unsuitable or if the
and the role of genotype-guided therapy increased. unit cost of NOACs is reduced by 485%.
There were a number of limitations to our analysis.
There was no patient-level data available for obtaining ACKNOWLEDGMENTS
the true mortality rate of Asian NVAF patients as a Research question, study design, data collection and
survival function. Consequently, Thai life tables and analysis was done by first author. The second
multiple factors were used to predict the death risk of author responsible in report writing and corre-
each cycle. The heterogeneity of the parameters was spondence. We express our great appreciation to
another limiting factor due to data being extracted from Dr. Yot Teerawattananon and the Health Interven-
a variety of different sources. Because of poor adherence tion and Technology Assessment Program team for
to drug regimens and the popular reliance on traditional their guidance and encouragement. We are grateful

October 2014 1393

 Clinical Therapeutics

Dr. Kasem Ratanasumawong, Head of the Depart- 7. Patel MR, Mahaffey KW, Garg J, et al. Rivaroxaban versus
ment of Medicine, Medical College and Vajira Warfarin in Nonvalvular Atrial Fibrillation. N Engl J Med.
Hospital, Bangkok, Thailand, for reviewing and 2011;365:883–891.
checking the model. 8. Granger CB, Alexander JH, McMurray JJV, et al. Apixaban
versus Warfarin in Patients with Atrial Fibrillation. N Engl J
CONFLICTS OF INTEREST Med. 2011;365:981–992.
9. Harrington AR, Armstrong EP Jr, PEN, et al. Cost-
The authors have indicated that they have no conflicts
Effectiveness of Apixaban, Dabigatran, Rivaroxaban, and
of interest regarding the content of this article.
Warfarin for Stroke Prevention in Atrial Fibrillation. Stroke.
2013;44:1676–1681.
SUPPLEMENTAL MATERIALS 10. Coyle D, Coyle K, Cameron C, et al. Cost-Effectiveness of
Supplemental figure and tables accompanying this New Oral Anticoagulants Compared with Warfarin in
article can be found in the online version at http:// Preventing Stroke and Other Cardiovascular Events in Pa-
dx.doi.org/10.1016/j.clinthera.2014.06.021. tients with Atrial Fibrillation. Value Health. 2013;16:498–506.
11. Bisson J, Dix P, Ehlers A, et al. Appendix 10 Methods for
calculating means and standard deviations for pooled
REFERENCES treatment groups. In post-traumatic stress disorder
1. Camm AJ, Kirchhof P, Lip GYH, et al. Guidelines for the (PTSD): The management of PTSD in adults and children
management of atrial fibrillation. The Task Force for the in primary and secondary care: Gaskell and the British
Management of Atrial Fibrillation of the European Psychological Society. 2005.
Society of Cardiology (ESC). Eur Heart J. 2010;31: 12. Tongsiri S. The Thai Population Based Preference Scores
2369–2429. for EQ5D Health States. Thailand: Health Intervention
2. Sitthisook S. Thai Guideline for Diagnosis and Treatment of and Technology Assessment Program; 2009. Available
Atrial Fibrillation. Bangkok: The Heart Association of Thai- from http://www.hitap.net/en/research/10525. [Accessed
land under the Royal Patronage of H.M. the King; 2012. January 18, 2013].
3. Hori M, Connolly SJ, Zhu J, et al. Dabigatran Versus 13. Riewpaiboon A, Riewpaiboon W, Ponsoongnern K, et al.
Warfarin: Effects on Ischemic and Hemorrhagic Strokes Economic valuation of informal care in Asia: A case study
and Bleeding in Asians and Non-Asians With Atrial of care for disabled stroke survivors in Thailand. Social
Fibrillation. Stroke. 2013;44:1891–1896. Science & Medicine. 2009;69:648–653.
4. Lam Y-Y, Ma TKW, Yan BP. Alternatives to chronic 14. Sangviroon APD, Tassaneeyakul W, Namchaisiri J. Pharma-
warfarin therapy for the prevention of stroke in patients cokinetic and Pharmacodynamic Variation Associated with
with atrial fibrillation. Int J Cardiol. 2011;150:4–11. VKORC1 and CYP2C9 Polymorphisms in Thai Patients Taking
5. Camm AJ, Lip GYH, Caterina RD, et al. 2012 focused Warfarin. Drug Metab Pharmacokinet. 2010;25:531–538.
update of the ESC Guidelines for the management of 15. Standard Cost List for Health Technology Assessment.
atrial fibrillation. An update of the 2010 ESC Guidelines Health Intervention and Technology Assessment: HITAP.
for the management of atrial fibrillation. Eur Heart J. Ministry of Public Health. 2010. Available from http://www.
2012;33:2719–2747. hitap.net/costingmenu/. [Accessed January 20, 2013].
6. Connolly SJ, Ezekowitz MD, Yusuf S, et al. Dabigatran 16. Shah SV, Gage BF. Cost-Effectiveness of Dabigatran for
versus warfarin in patients with atrial fibrillation. N Engl J Stroke Prophylaxis in Atrial Fibrillation. Circulation. 2011;
Med. 2009;361:1139–1151. 123:2562–2570.

Address correspondence to: Satadon Taerakun, PharmD, Pharmacy

Department, Police General Hospital, No. 492/1 Rama I Road, Pathum
Wan sub-area, Pathum Wan Area, Bangkok, Thailand 10330.. E-mail:
[email protected]

1394 Volume 36 Number 10

 S. Jarungsuccess and S. Taerakun

Warfarin (INR 2–3)

M
Dabigatran 150 mg BID
M M
NVAF Dabigatran 110 mg BID
M
Rivaroxaban 20 mg OD WELL
M
Apixaban 5 mg BID
M
Ischemic stroke Major bleeding

Complication Death Complication

Partial Partial
recovery recovery

Myocardial infarction

Complication

Supplemental Figure. Model structure and relevance health state.

SUPPLEMENTAL MATERIALS
Supplemental Figure 1Supplemental Table I–III

October 2014 1394.e1

Clinical Therapeutics

Supplemental Table I. Model input variables: base-case values and ranges used in probabilistic sensitivity
analysis.
Variables Base-case 95% CI Reference

Clinical outcome parameters

Incidence of ischemic stroke (IS):
Annual probability: INR-in range warfarin 0.02 0.02–0.03 [3, 16–18]
Relative risk: dabigatran 150 mg BID 0.55 0.32–0.95 [3]
Relative risk: dabigatran 110 mg BID 1.01 0.63–1.61 [3]
Relative risk: rivaroxaban 20 rag OD 0.82 0.55–1.22 [7]
Relative risk: apixaban 5 mg BID 0.65 0.32–0.98 [8]
Relative risk of CHADS2 r 2: dabigatran 150 mg BID 0.60 0.16–1.05 L6J
Relative risk of CHADS2 Z 3: dabigatran 150 mg BID 0.70 0.29–1.12 [6]
Relative risk of CHADS2 r 2: dabigatran 110 mg BID 1.02 0.63–1.41 [6]
Relative risk of CHADS2 Z 3: dabigatran 110 mg BID 0.79 0.39–1.19 [6]
Relative risk of CHADS2 r 2: rivaroxaban 20 mg OD 0.84 0.37–1.32 [7]
Relative risk of CHADS2 Z 3: rivaroxaban 20 mg OD 0.89 0.72–1.06 [7]
Relative risk of CHADS2 r 2: apixaban 5 mg BID 0.88 0.64–1.13 [8]
Relative risk of CHADS2 Z 3: apixaban 5 mg BID 0.71 0.45–0.97 [8]
IS outcomes:
Transition probability: fatal stroke 0.26 0.26–0.27 [19]
Transition probability: non-disabling 0.71 0.68–0.74 [19]
Transition probability: disabling 0.03 0.00–0.07 [19]
Incidence of major bleeding (MB):
Annual probability: INR-in range warfarin 0.04 0.03–0.04 [3, 17]
Relative risk: Dabigatran 150 mg BID 0.57 0.38–0.84 [3]
Relative risk: Dabigatran 110 mg BID 0.57 0.39–0.85 [3]
Relative risk: Rivaroxaban 20 mg OD 0.66 0.41–1.04 [7]
Relative risk: Apixaban 5 mg BID 0.53 0.20–0.87 [8]
MB outcomes:
Transition probability: fatal bleeding 0.30 0.11–0.49 [20–23]
Transition probability: non-disabling 0.48 0.40–0.57 [20–23]
Transition probability: disabling 0.22 0.04–0.40 [20–23]
Incidence of myocardial infarction (MI):
Annual probability: INR-in range warfarin 0.007 0.004–0.009 |3J
Relative risk: Dabigatran 150 mg BID 0.89 0.80–0.98 [3]
Relative risk: Dabigatran 110 mg BID 0.90 0.01–1.80 [3]
Relative risk: Rivaroxaban 20 mg OD 0.80 0.54–1.06 [7]
Relative risk: Apixaban 5 mg BID 0.88 0.60–1.16 [8]
MI outcomes:
Transition probability: fatal MI 0.17 0.13–0.20 L24–30J
Transition probability: fully recovery from MI 0.83 0.78–0.88 [31, 32]
Incidence of dyspepsia:
Annual rate: INR-in range warfarin 0.06 0.02–0.10 [3, 33]
Annual rate: dabigatran 150 mg BID 0.14 0.04–0.23 [3, 33]
Annual rate: dabigatran 110 mg BID 0.14 0.04–0.23 [3, 33]
Annual rate: rivaroxaban 20 mg OD 0.06 0.02–0.10 Assumption
Annual rate: apixaban 5 mg BID 0.06 0.02–0.10 Assumption
Death risk:
Natural death risk over 65-70 years 0.02 [34]
Natural death risk over 71-75 years 0.03 L34J
Natural death risk over 76-80 years 0.05 [34]
Natural death risk over 81-85 years 0.08 [34]
Natural death risk over 86-90 years 0.13 [34]
Natural death risk over 91 years 0.20 [34]
Cost parameters (in Thai baht)
Direct medical cost:
Average daily dose of warfarin 3.7 3.42–3.98 [14]
Average number of hospital visit per year 6 5.45–6.55 Hospital based data
Annual cost of warfarin 1,830.38 1,258.16–2,402.60 Hospital based data

(continued)

1394.e2 Volume 36 Number 10

 S. Jarungsuccess and S. Taerakun

Supplemental Table I. (continued).

Variables Base-case 95% CI Reference

Annual cost of dabigatran 150 mg and 110 mg BID 41,011.40 Hospital based data
Annual cost of rivaroxaban 20 mg OD 36,222.60 Hospital based data
Annual cost of apixaban 5 mg BID 78,110.00 Hospital based data
Annual cost of 1NR monitoring 420.00 381.58–458.42 Hospital based data
Annual cost of service fee for OFD visit 300.00 272.56–327.44 Hospital based data
Annual cost of EKG monitoring 200.00 Hospital based data
Admission cost per 1 event of ischemic stroke 24,604.16 21,686.02–27,522.30 Hospital based data
Admission cost per 1 event of major bleeding 33,683.87 28,020.74–39,347.00 Hospital based data
Admission cost per 1 event of myocardial infarction 93,882.44 75,380.85–112,384.03 Hospital based data
Treatment cost per 1 event of dyspepsia 197.71 180.71–214.71 Hospital based data
Co-morbidity treatment per year: Diabetes mellitus 3,597.71 3,538.71–3,656.71 Hospital based data
Co-morbidity treatment per year: Heart failure 2,239.16 2,077.24–2,401.08 Hospital based data
Co-morbidity treatment per year: Hypertension 10,080.78 9,973.59–10,187.98 Hospital based data
Co-morbidity treatment per year: Myocardial infarction 2,239.23 2,161.22–2,317.24 Hospital based data
Direct non-medical cost:
Transportation cost for hospital visit per year 955.99 819.55–1,092.43 [15]
Food cost for hospital visit per year 352.15 289.23–415.07 [15]
Informal care cost per month for severe disabling stroke 12,260.59 4,208.91–20,312.27 [13]
Informal care cost per month for mild to moderate disabling stroke 5,023.93 3,359.71–6,688.15 [13]
Food and room cost per 1 admission due to ischemic stroke 8,568.32 7,146.78–9,989.86 [15]
Food and room cost per 1 admission due to major bleeding 11,726.36 9,362.09–14,090.63 [15]
Food and room cost per 1 admission due to myocardial infarction 6,285.11 5,023.88–7,546.34 [15]
Utility parameters
Well/Full recovery from any health state 0.79 0.70–0.88 Hospital based data
Non-disabling stroke, unspecified 0.61 0.58–0.64 [35]
Disabling stroke, unspecified 0.14 0.13–0.15 [35]
Myocardial infarction 0.77 0.75–0.79 [36]

Supplemental Table II. Threshold analysis of medication unit cost depending on societal perspective.

ICER Unit Cost Optimal Cost Percentage of

Strategy (Thai Baht/QALYs) (Thai Baht) (Thai Baht) Reduction

Dabigatran 150 mg 2,252,938.19 56.18 8.23 85.35

Dabigatran 110 mg 46,286,254.56 56.18 4.15 92.61
Rivaroxaban 20 mg 5,030,280.45 99.24 10.66 89.26
Apixaban 2.5 mg (double dose) 5,565,388.48 53.5 3.76 92.97

October 2014 1394.e3

1394.e4

Clinical Therapeutics
Supplemental Table III. Budget impact analysis according to societal perspective.
Maximum Budget impact
(Warfarin Unsuitable
Budget Impact* (National List of Essential Medicines Policy) Policy†)

Year Warfarin Dabigatran 150 mg Dabigatran 110 rag Rivaroxaban 20 mg Apixaban 5 mg New Oral Anticoagulants

1 20,755,255.64–83,021,022.55 31,511,656.081–26,046,624.31 31,609,643.781–26,438,575.12 30,297,898.681–21,191,594.73 42,091,100.131–68,364,400.53 74,080,336.23

2 14,411,158.96–57,644,635.83 21,873,476.96–87,493,907.82 21,952,140.03–87,808,560.12 21,029,276.94–84,117,107.76 29,220,741.241–16,882,964.95 51,428,504.58
3 9,603,175.583–8,412,702.33 14,599,389.26–58,397,557.04 14,661,649.77–58,646,599.08 14,018,290.32–56,073,161.29 19,511,909.77–78,047,639.07 34,340,961.19
4 6,139,181.47–24,556,725.90 9,345,046.11–37,380,184.44 9,396,957.45–37,587,829.81 8,962,230.38–35,848,921.51 12,497,470.57–49,989,882.30 21,995,548.21
5 3,751,477.30–15,005,909.20 5,711,753.94–22,847,015.78 5,755,365.96–23,021,463.85 5,473,822.36–21,895,289.45 7,644,487.22–30,577,948.88 13,454,297.51

*
Percentage of warfarin prescribed in NVAF was estimated at 15 to 60 [4].
†
Warfarin unsuitability was defined by the published criteria elsewhere and assumed in a proportion of 14-44% of indicated cases [4].
Volume 36 Number 10