OUP user menu

Cost-effectiveness of dabigatran compared with warfarin for patients with atrial fibrillation in Sweden

Thomas Davidson, Magnus Husberg, Magnus Janzon, Jonas Oldgren, Lars-Åke Levin
DOI: http://dx.doi.org/10.1093/eurheartj/ehs157 177-183 First published online: 24 June 2012


Aims Patients with atrial fibrillation have a significantly increased risk of thromboembolic events such as ischaemic stroke, and patients are therefore recommended to be treated with anticoagulation treatment. The most commonly used anticoagulant consists of vitamin K antagonist such as warfarin. A new oral anticoagulation treatment, dabigatran, has recently been approved for stroke prevention among patients with atrial fibrillation. The purpose of this study was to estimate the cost-effectiveness of dabigatran as preventive treatment of stroke and thromboembolic events compared with warfarin in 65-year-old patients with atrial fibrillation in Sweden.

Methods and results A decision analytic simulation model was used to estimate the long-term (20-year) costs and effects of the different treatments. The outcome measures are the number of strokes prevented, life years gained, and quality-adjusted life years (QALYs) gained. Costs and effect data are adjusted to a Swedish setting. Patients below 80 years of age are assumed to start with dabigatran 150 mg twice a day and switch to 110 mg twice a day at the age of 80 years due to higher bleeding risk. The price of dabigatran in Sweden is €2.82 (Swedish kronor 25.39) per day for both doses. The cost per QALY gained for dabigatran compared with warfarin is estimated at €7742, increasing to €12 449 if dabigatran is compared with only well-controlled warfarin treatment.

Conclusion Dabigatran is a cost-effective treatment in Sweden, as its incremental cost-effectiveness ratio is below the normally accepted willingness to pay limit.

  • Cost-effectiveness
  • Atrial fibrillation
  • Anticoagulants
  • Stroke

See page 166 for the editorial comment on this article (doi:10.1093/eurheartj/ehs204)


Atrial fibrillation (AF) is the most common type of cardiac arrhythmia, occurring in >1% of the population1 and causing large costs for the health care system.24 Patients with AF often suffer a decreased quality of life,5,6 and having AF increases the risk of thromboembolic events such as ischaemic stroke.7 The treatment of AF varies for different patients and different types of AF, but is usually given in combination with anticoagulation treatment to prevent thromboembolic events. The most commonly used anticoagulant consists of vitamin K antagonists such as warfarin. This treatment requires accurate dosing and careful monitoring, to reduce the risk of stroke without greatly increasing the risk of bleedings. Too high a dose causes a risk of serious complications in the form of bleedings including intracranial hemorrhagic stroke, and too low a dose does not protect against thromboembolic events. An alternative to warfarin is treatment with acetylsalicylic acid (ASA), though this is less effective than warfarin in preventing thromboembolic events and also increases the risk of bleeding.7 Warfarin is recommended to the vast majority of AF patients, whereas ASA might be considered only in the small group of AF patients at very low risk for stroke.8

A new oral anticoagulation treatment, dabigatran, has recently been approved as preventive treatment of stroke and systemic embolism among patients with AF. Dabigatran has been used previously for the prophylaxis of deep vein thrombosis in hip and knee surgery. In AF indication, dabigatran has been studied in the Randomized Evaluation of Long-Term Anticoagulation Therapy (RE-LY) trial. Two dabigatran doses were studied, 150 and 110 mg twice a day, and compared with warfarin. The results showed that dabigatran 150 mg twice a day reduced stroke and intracranial bleeding without any significant change in overall major bleeding compared with warfarin, whereas dabigatran 110 mg twice a day was non-inferior at reducing risk of stroke, but reduced intracranial and major bleeding.911 However, the drug cost of dabigatran is higher than the drug cost of warfarin, so it is important to study its cost-effectiveness. In order for drugs to be included in the reimbursement system in Sweden, they have to be proven cost-effective in comparison with the most relevant comparator.

The purpose of this study was to estimate the cost-effectiveness of dabigatran as preventive treatment of stroke and thromboembolic events compared with warfarin in 65-year-old patients with AF in Sweden.


The structure and guidelines of a cost-effectiveness analysis are used in this study,12 and the recommendations made by the Swedish Dental and Pharmaceutical Benefits Board (TLV) are followed.13 A societal perspective was chosen, but productivity losses are not estimated due to the age of the patients (>65 years). The long-term (20-year) costs and effects of dabigatran and warfarin are estimated with a decision analytic model.

The RE-LY trial

Dabigatran is an oral direct thrombin inhibitor. It is administered as the pro-drug dabigatran etexilate, which is converted to dabigatran in the body. The effect is predictable, and the dabigatran dose need not be regularly monitored or adjusted. Dabigatran was compared with warfarin in the RE-LY trial, which was a randomized study comparing two different doses of dabigatran (150 and 110 mg twice a day) with warfarin. In total, 18 113 patients from 44 countries were included. The study was blinded for the two doses of dabigatran but open in comparison with warfarin.9 The main results are presented later in this article.


The most commonly used risk classification system for ischaemic stroke is CHADS2, which gives the patient a rating on a scale from 0 to 6, relating to the risk based on predefined risk factors.14 Patients across all CHADS2 scores were eligible in the RE-LY trial, although the group with CHADS2 score 0 was small (n= 452). CHADS2 is used to classify the patients in the calculations made in this article, and is used for subgroup analyses, with patients divided into three groups with CHADS2 scores of 0–1, 2, and 3–6, respectively.

Decision analytic model

Having AF means having a lifelong increased risk of stroke and other thromboembolic events, and so the costs and effects of AF and anticoagulation treatment occur throughout the patient’s lifetime. Estimating the costs only during the short follow-up periods found in clinical trials would greatly misjudge the true outcomes. Therefore, a decision analytic model is used in this article to study the economic impact over a time horizon of 20 years.

Depending on the patients’ CHADS2 score and type of treatment, an annual risk of complications is applied in the model. The complications that can occur in the model are ischaemic and hemorrhagic strokes, gastrointestinal (GI) bleedings, other major bleedings, minor bleedings, and myocardial infarction. Ischaemic and hemorrhagic strokes cause lifelong costs and deterioration of quality of life. Other complications in the model are assumed to only affect costs and quality of life in the year in which they occur.

The decision model follows a Markov design. In such a model, the patient is always in one of several specified states. The patients can move between the specified states annually with certain probabilities. The states are defined in accordance with their anticoagulation treatment (dabigatran, warfarin, ASA, or no anticoagulation). For each state there is also a breakdown in risk of stroke based on CHADS2. This risk classification can in the model change when the patient either reaches the age of 75 years or has a stroke. In addition, because patients can switch between treatments, this is enabled in the model with certain annual probabilities. Finally, patients may end up in the ‘dead’ state. The model structure is illustrated in Figure 1. The reason for choosing states based on type of treatment and risk score is that these aspects are the most important ones for the risk of complications. Furthermore, all complications can occur annually. This means that the model does not follow the progression or severity of the complications but assumes average severity levels of complications in terms of costs, outcomes, and mortality. The model therefore represents the average of a large population.

Figure 1

Structure of the decision analytic model.

The data underlying the calculations are mostly drawn from the RE-LY trial, supplemented by local data to be relevant for a Swedish population. The most important parameters used in the modelling are presented in Table 3. Patients below 80 years of age are assumed to start with dabigatran 150 mg twice a day and switch to 110 mg twice a day at the age of 80 years due to higher bleeding risk.


All costs are calculated in Swedish kronor (SEK) 2010. The costs are presented in Euros (€), where €1 = SEK 9.

The cost of dabigatran is €2.82 (SEK 25.39) per day for both doses, giving an annual cost of €1030. By also including one yearly visit to a primary care doctor at a cost of €199, the annual cost of dabigatran treatment is €1229. The cost of warfarin and its monitoring, also including one annual visit to a doctor, was estimated to €776 per patient per year.2 The cost of treatment with ASA was estimated to €431, also including one annual visit to a doctor.2

Ghatnekar et al.15 studied the cost of stroke in Sweden in 2004 using an incidence cost approach; the direct costs amounted to €65 652 and indirect costs to €15 986 in 2010 values. This estimate includes admission costs, re-stroke admission costs, outpatient costs, and costs for social services. These values, divided into annual costs, are used for the calculations of both ischaemic and hemorrhagic stroke in this study. Studies have found that the costs of stroke are higher for patients with a history of AF compared with patients without an AF history.16,17 The estimate of the stroke cost used in our analyses is for that reason conservative.

Existing data on the cost of bleedings are poor, partly due to the difficulty of defining major and minor bleedings. On the basis of registry data, the average cost of major bleeding has been estimated to be €2773.18 A Canadian study estimated the cost of GI bleeding to be €3303,19 based on the cost of hospitalization and outpatient care. These costs of bleedings are also used in the analysis in this study. The cost of minor bleedings is estimated to be €33, and the cost of myocardial infarction is set at €20 457.20

Risk of complications

Absolute risks for complications with warfarin used in these calculations are taken from the RE-LY trial.911 A sub-study from the RE-LY trial has presented risks of total stroke or systemic embolism, intracranial bleedings, and major bleedings divided by the background risk of stroke according to CHADS2.21 Table 1 presents the absolute annual risks of complications arising from warfarin that are used in the model.

View this table:
Table 1

Absolute annual risks for complications with warfarin for patients with atrial fibrillation divided by CHADS2 score

CHADS2= 0–1 (%)CHADS2= 2 (%)CHADS2= 3–6 (%)Reference
Stroke or systemic embolism1.081.382.7321
Intracranial bleeding0.540.691.0721
Myocardial infarction0.640.640.6411
Major bleeding2.843.304.6021
GI bleeding1.071.071.0711
Minor bleeding16.3716.3716.3711

Another subgroup analysis from the RE-LY trial studied the risks for complications with warfarin divided by the different study centres’ average international normalized ratio level.22 Centres were divided into quartile groups according to their mean time in therapeutic range (TTR). This is important, as the Swedish TTR in the RE-LY trial (77%)22 is higher than the average TTR in the RE-LY trial (64%).9 The quartile group with the best-controlled warfarin treatment in the RE-LY trial had a mean TTR of >72.6%.22 The calculations in this article assume that this corresponds to well-controlled warfarin patients. For poorly controlled warfarin patients, the second worst quartile is used, with TTR ranging between 57.1 and 65.5%. Risks for stroke and major bleedings divided by well-controlled warfarin treatment and poorly controlled warfarin are reported in Table 2. However, this adjustment of warfarin treatment absolute risks was performed after the risk for complications with dabigatran had been estimated (which uses data from the whole RE-LY study sample).

View this table:
Table 2

Absolute annual risks for complications with well-controlled and poorly controlled warfarin

Well-controlled warfarin (%)Poorly controlled warfarin (%)Reference
Ischaemic stroke and systemic embolism0.971.6322
Intracranial bleeding0.770.9322
Major bleeding3.114.1322
GI bleeding0.871.6022

Quality-adjusted life year weights and mortality rates

Weights for quality-adjusted life years (QALY weights) are used in order to take into account the quality of life of the different health states. On the basis of the instrument EQ-5D health profiles can be created, which has been equipped with QALY weights based on a general public’s preferences.23 A study from Sweden used the EQ-5D to estimate QALY weights for various ages in Sweden.24 These QALY weights were used as a basis for the patients in the model, but including a decrement of 0.05 due to having AF. Reductions in the QALY weights were also made for stroke25 and for myocardial infarction.26

The model uses age-based standard mortality for Sweden in 2010 according to data from Statistics Sweden.27 Increased mortality due to stroke is modelled using the results of a study by Henriksson et al.,28 which presented the stroke-associated mortality rates in Sweden based on data from the national register ‘Riks-Stroke’; stroke patients with CHADS2 scores of 1, 2, and 3–6 had mortality rates of 20.7, 26.9, and 39.2%, respectively (after combining the rates of CHADS2 3–6). These data are used for the first year after the stroke, and standard mortality rates are used thereafter.27 This is a conservative assumption as long-term survival for patients who has had a stroke likely is lower than for patients with no stroke history (Table 3).

View this table:
Table 3

Main parameters in the model

ParameterEstimate (CI)DistributionReference
RR for events dabigatran 150 mg b.i.d. compared with warfarin
 Ischaemic stroke0.76 (0.59–0.97)Log-normal11
 Intracranial bleeding0.41 (0.28–0.60)Log-normal11
 Major bleeding0.93 (0.81–1.07)Log-normal11
 GI bleeding1.48 (1.18–1.85)Log-normal11
 Minor bleeding0.91 (0.86–0.97)Log-normal11
 Myocardial infarction1.27 (0.94–1.71)Log-normal11
RR for events dabigatran 110 mg b.i.d. compared with warfarin
 Ischaemic stroke1.11 (0.88–1.39)Log-normal11
 Intracranial bleeding0.30 (0.19–0.45)Log-normal11
 Major bleeding0.80 (0.70–0.93)Log-normal11
 GI bleeding1.08 (0.85–1.38)Log-normal11
 Minor bleeding0.79 (0.74–0.84)Log-normal11
 Myocardial infarction1.29 (0.96–1.75)Log-normal11
Costs (€)
 Cost of warfarin treatment per year776Deterministic2
 Number of monitoring visits (warfarin)16.94Gamma2
 Cost of ASA treatment per year431Deterministic2
 Cost of dabigatran treatment per year1229Deterministic
 Stroke, admission costs8969Normal15
 Stroke, outpatient costs, 1st year3933Normal15
 Stroke, outpatient costs, second year onward2363Normal15
 Stroke, social services, first year5354Normal15
 Stroke, social service costs, second year onward4311Normal15
 Major bleeding2773Deterministic18
 GI bleeding3303Deterministic19
 Minor bleeding33Deterministic
 Myocardial infarction20 457Deterministic20
 Costs for an added year of life, 65–74 years old14 813Deterministic31
 Costs for an added year of life, 75–84 years old18 005Deterministic31
QALY weights
 65–69 years with AF0.81Deterministic24
 70–74 years with AF0.78Deterministic24
 75–79 years with AF0.76Deterministic24
 80–84 years with AF0.71Deterministic24
 QALY weight decrement; atrial fibrillation0.05Deterministic
 QALY weight decrement; ischaemic stroke0.15Beta25
 QALY weight decrement; hemorrhagic stroke0.30Beta25
 QALY weight decrement; myocardial infarction0.19Deterministic26
  • RR, relative risks; b.i.d., twice a day.

Transition probabilities between different treatments

In the RE-LY trial, 16% of participants treated with dabigatran 150 mg twice a day discontinued their treatment within the first year, and another 5% stopped within the second year. For patients who received warfarin, the discontinuation figures were 10 and 7%, respectively.9

It is likely that the discontinuation rates decrease annually. The model assumes that the proportion of patients who discontinue treatment decreases at the same rate as during the first 2 years. We do not have any information on how those patients who discontinue their treatment were treated thereafter, and so the model assumes that patients who discontinue warfarin treatment switch to ASA or no treatment in equal proportions, and that patients who discontinue dabigatran treatment switch to warfarin, ASA, or no treatment in equal proportions.


The analyses are undertaken for 65-year-old patients, followed for 20 years, starting their treatment with either dabigatran 150 mg twice a day or warfarin. Patients using dabigatran switch to 110 mg twice a day at the age of 80 years.

The effects are measured in number of prevented strokes, number of life years gained, and number of QALYs gained. Both costs and effects (life years and QALYs) are discounted by 3% annually.13

To analyse how different assumptions and simplifications in the base case analysis affect the results, several one-way sensitivity analyses were performed where key assumptions were varied.

In order to take into account the uncertainty in the parameter estimates derived from clinical studies, the model was run probabilistically. This means that the analysis was repeated 1000 times, each time with the parameters being selected randomly from the distribution of the uncertainty surrounding the parameters. In this way, the total statistical uncertainty is illustrated. A log-normal distribution is used for relative risk reduction parameters. A normal distribution is used for the cost parameters, because there are a very large number of observations underlying the cost estimates. The number of monitoring visits for warfarin treatment uses a γ distribution and the QALY weights use a β distribution.


The base case compares dabigatran with warfarin for all patients who currently receive warfarin in Sweden. The results (presented in Table 4) show that dabigatran leads to an incremental cost-effectiveness ratio (ICER) of €7699 per life year gained and €7742 per QALY gained compared with warfarin.

View this table:
Table 4

Cost-effectiveness of dabigatran compared with warfarin, costs in €

CostsNumber of strokesLife yearsQALYCost per life year gainedCost per QALY gained
Warfarin24 7970.5211.838.31
Dabigatran27 0090.3712.118.6076997742

Figure 2 illustrates the cost-effectiveness plane with the results from 1000 iterations. In all cases, the ICER is less than €50 000 per QALY gained.

Figure 2

Cost-effectiveness plane illustrating probabilistic analysis with 1000 iterations of dabigatran compared with warfarin, costs in €.

Sensitivity analyses

Table 5 presents one-way sensitivity analyses of alternative scenarios. If dabigatran is compared with only those patients who are well-controlled with warfarin, the cost per QALY gained increases to €12 449. This analysis uses data on warfarin from the quartile of the study centres in the RE-LY trial that had the highest TTR. Analyses for patients with different CHADS2 scores revealed that patients with CHADS2 = 3–6 generated a better cost-effectiveness of dabigatran compared with patients with lower CHADS2 scores. The assumptions made for the transitions between the treatments were also relevant to the analysis. If only the transitions that occurred within the time frame of the RE-LY trial (2 years) were included, the ICER for dabigatran increased to €10 287 per QALY gained.

View this table:
Table 5

One-way sensitivity analyses of cost-effectiveness for dabigatran compared with warfarin, costs in €

Alternative scenariosCost per QALY gained
Base case7742
Comparison with well-controlled warfarin patients12 449
Comparison with poorly controlled warfarin patientsDominant
Only CHADS2= 0–120 929
Only CHADS2= 28216
Only CHADS2= 3–62652
No transitions between treatments after second year10 287
QALY loss from warfarin and ASA included5540
Willingness to pay to replace warfarin includedDominant
Costs of added life years included27 514
Discount rate (costs and QALYs): 0%6085
Discount rate (costs and QALYs): 5%9077

Warfarin treatment has been shown to decrease the patients’ quality of life, and decrements in the QALY weight of 0.013 for warfarin and 0.002 for ASA have been found.29 Including these patient preferences improves the ICER for dabigatran. Another study showed that patients in Sweden are willing to pay €85 per month to use a medicine that is slightly better than warfarin, and to avoid monitoring,30 and when this willingness to pay value is included on the cost side of the ICER, dabigatran becomes dominant in comparison with warfarin.

There are recommendations in Sweden to also include the costs of added life years if a treatment prolongs life.13 When these costs are included according to Swedish values (see Table 5),31 the ICER increases to €27 514 per QALY gained. Discount rates do not have a major influence on the cost-effectiveness of dabigatran.


We have estimated the cost-effectiveness of dabigatran as preventive treatment of stroke and thromboembolic events compared with warfarin in Sweden. We found that dabigatran is expected to increase the total costs but also to lead to fewer strokes, increased life expectancy, and an increase in QALYs. In the base case analysis, the cost per QALY gained with dabigatran is €7742. In comparison with only well-controlled warfarin patients, the cost per QALY gained for dabigatran was €12 449, and in comparison with poorly controlled warfarin patients the dabigatran treatment was dominating with both lower total costs and gained QALYs. The probabilistic analysis undertaken showed that the ICER for dabigatran is always below €50 000. Even though no threshold value per QALY exists in Sweden, a limit of €50 000 is often used, meaning that this study has found dabigatran to be cost-effective.

Our results can be compared with other cost-effectiveness analyses which also base their calculations on the findings from the RE-LY trial.3236 The main differences between our study and the other ones are that our study uses the most relevant data for a Swedish setting. Data on costs of complications, number of visits to monitor warfarin, and costs of both warfarin treatment and dabigatran are adjusted to reflect the real-world situation in Sweden. Furthermore, the model used in our study is unique and created within the project. Sorensen et al.36 estimated the cost per QALY gained for dabigatran in Canada, using a model run over the patient's lifetime and a price of Can$ 3.20 per day for dabigatran. The model also allowed for various severity levels of stroke. The ICER of dabigatran in comparison with warfarin was Can$ 10 440 per QALY gained, very similar to the findings in this analysis. Freeman et al.32 estimated the cost-effectiveness of dabigatran in the US using an estimated price of US$ 13 per day for dabigatran 150 mg and US$ 9.50 per day for dabigatran 110 mg. The cost per QALY gained was US$ 45 372 for dabigatran 150 mg twice a day compared with warfarin and US$ 51 229 for dabigatran 110 mg twice a day compared with warfarin. In an updated analysis with the price of US$ 8 per day for dabigatran 150 mg, the ICER decreased to US$12 386 per QALY gained.37 Shah and Gage35 estimated the cost-effectiveness of dabigatran using a daily price of US$9. They found that dabigatran 150 mg BID led to an average ICER at $86 000 per QALY gained in comparison with warfarin and below $50 000 per QALY gained among patients who had a high risk of hemorrhagic or ischaemic stroke, unless the warfarin treatment could be excellently monitored. Pink et al.34 estimated the cost per QALY gained with dabigatran 150 mg twice daily compared with warfarin to £22 082, and to £46 386 if compared only to well-controlled warfarin patients. The main reasons for the higher ICER in this study compared with our findings is probably the older starting age (71 years), lower costs applied for stroke, and lower monitoring costs for warfarin. Finally, Kamel et al.33 estimated the cost per QALY gained for dabigatran in comparison with warfarin to $25 000, using a daily price of US$6.75 for dabigatran.

The calculations in our analysis have some limitations. We chose not to consider treatment of symptomatic AF, such as antiarrhythmic drugs or ablation technique, as these treatments are given in combination with anticoagulation in patients at risk for stroke. Furthermore, a decision analytic model is intended to provide information for decision makers, and not to describe the clinical situation in detail. The model therefore included some simplifications, for example only allowing patients to change CHADS2 score based on age and stroke. Furthermore, the model did not allow patients who started with warfarin to switch to dabigatran as this was not possible in RE-LY.

One challenge of using the RE-LY trial results in a Swedish setting is that the average TTR of warfarin in Swedish RE-LY sites (77%) is higher than the average TTR in RE-LY (64%). We have adjusted for this in our sensitivity analyses by using risk data on warfarin treatment from the quartile in RE-LY with the highest TTR. Nevertheless, <50% of patients with AF at risk for stroke are treated with warfarin in Sweden38 and the others are either untreated or treated with ASA, which leads to an increased risk for stroke and thromboembolic events. For these patients, dabigatran would lead to reduced costs and increased QALYs, making dabigatran a dominant treatment. It is worth remembering that RE-LY was a well-controlled clinical study and that other risks for side-effects and complications may occur in a real-life setting, for both dabigatran and warfarin.

The main conclusion from this study is that dabigatran has been found to be cost-effective as a preventive treatment of stroke and thromboembolic events compared with warfarin in Sweden. The use of dabigatran should, however, be continuously studied so that more relevant data for the Swedish setting can be found.


This work was supported by Boehringer Ingelheim and the County council of Östergötland , Sweden.

Conflict of interest: T.D., lecture fees from Boehringer Ingelheim; J.O., institutional grant support from Boehringer Ingelheim; consulting fees and lecture fees from Bayer, Boehringer-Ingelheim, Bristol-Myers Squibb and Pfizer; L.-Å.L., lecture fees from Boehringer Ingelheim.


View Abstract