The Markov model (Figure 1) used in this evaluation is based on those used in economic studies on the introduction of dabigatran in Canada16 and the United Kingdom,17 adapted to the Portuguese context. The model simulates patients’ clinical course in 3-month cycles to the end of their lives, applying a specified probability of the mutually exclusive occurrence of the major events in this population: death, ischemic and hemorrhagic stroke, transient ischemic attack (TIA), systemic embolism, myocardial infarction (MI), and intra and extracranial bleeding. Minor bleeding, which can occur concomitantly with the other events, is also simulated.

TIA: transient ischemic attack.'> TIA: transient ischemic attack.' title='Markov model. a History of stroke is modeled but not represented in the figure; b discontinuation due to minor bleeding. HS: hemorrhagic stroke; IB: intracranial bleeding; IS: ischemic stroke; MI: myocardial infarction; TIA: transient ischemic attack.'/>

Figure 1.

Markov model. a History of stroke is modeled but not represented in the figure; b discontinuation due to minor bleeding. HS: hemorrhagic stroke; IB: intracranial bleeding; IS: ischemic stroke; MI: myocardial infarction; TIA: transient ischemic attack.

(0.38MB).

An individual who has suffered a cerebral event (ischemic or hemorrhagic stroke or intracranial bleeding) may lose their independence. Such patients are classified according to the modified Rankin scale (for stroke) or the Glasgow Outcomes scale (for intracranial bleeding) as independent, moderately dependent, or completely dependent. Treatment can be discontinued, permanently in the case of hemorrhagic stroke or intracranial bleeding, or temporarily due to extracranial bleeding or other reasons.

Such a simulation of a patient cohort can be used to estimate the costs and clinical consequences associated with different therapeutic options. In this case the model was used to predict costs, life years, and quality-adjusted life years (QALY) for the alternatives under analysis and to determine costs per life year and per QALY for the treatment with the best clinical results.

The choice of comparator (the therapeutic option to be replaced by the drug under assessment) is a crucial element in the design of an economic evaluation. Dabigatran is an alternative not only to warfarin but also to other therapeutic options for AF patients, particularly aspirin and no antithrombotic medication.

In the FAMA study,3 as stated above, only 38% of eligible patients were medicated with warfarin and 40% were receiving no treatment. For the construction of our model it was assumed that the rest were taking aspirin, and so the comparator used in our study is a weighted combination of warfarin, aspirin and no treatment.

The model uses the results of the RE-LY trial,14,15 a direct comparison between warfarin and dabigatran in over 18 000 patients with non-valvular AF and moderate to high stroke risk randomized to 110 mg or 150 mg dabigatran twice daily or warfarin at doses adjusted for a target INR of 2.0–3.0 for a median of two years. The study took place between December 2005 and March 2009 in 44 countries around the world, including Portugal, where nine centers with 117 patients participated.

The study analyzed the occurrence of events in individuals taking warfarin and the relative risks of those taking dabigatran. The results showed that 110 mg dabigatran reduced the primary outcome (stroke or systemic embolism) by 10%, although it was not significantly superior to warfarin, with a better safety profile (20% less major bleeding than warfarin, p=0.003). The 150 mg dose was superior to warfarin, reducing the primary outcome by 35% (p<0.001) and major bleeding by 7% (without statistical significance). Dabigatran significantly reduced the incidence of intracranial bleeding, by 70% for the 110 mg dose and by 59% for the 150 mg dose (p<0.001).15

The only adverse effect that was significantly more common with dabigatran than with warfarin was dyspepsia (including abdominal pain), which occurred in 5.8% of patients in the warfarin group and in 11.8% and 11.3% in the 110 mg and 150 mg dabigatran groups, respectively (p<0.001 for both comparisons). Overall adverse event rates were similar between the treatment groups.

There has also been a meta-analysis18 that estimated the relative risk of various events in patients treated with aspirin and no treatment, patients’ ischemic stroke risk being stratified according to their CHADS2 score.18Table 1 shows the results, which differ from those in the RE-LY trial publications in accordance with the drug's approval by the European Medicines Agency, which recommends 150 mg dabigatran twice daily for patients aged under 80 and 110 mg twice daily for those aged 80 or over.19

The table shows that compared to warfarin, dabigatran significantly reduces the events under consideration, particularly ischemic and hemorrhagic stroke, systemic embolism, TIA, intra- and extracranial bleeding (the latter two only in patients aged under 80), and minor bleeding. These clinical advantages are naturally more marked in relation to aspirin and no treatment. A slightly lower rate of MI was also observed in the warfarin group, although without statistical significance.15

The epidemiological data from the FAMA study were used to estimate the proportion of men to women with AF in Portugal and their mean age,3 giving a figure of 55% males in the population aged under 80 and 26% in the older age-group.

The estimation of QALYs associated with each option is based on the assumption that quality of life diminishes in the presence of disease and with the occurrence of events. Analysis of the initial quality of life of the study population and of the impact of events is thus essential.

For this purpose, we used the results of a meta-analysis20 of 20 studies estimating quality of life for stroke, and of one study that estimated utility values for various chronic diseases using a representative sample of the US population based on data from the Medical Expenditure Panel Survey.21 The former was used to establish weights for levels of dependence after stroke, while the latter was used to evaluate the impact of events on quality of life (Table 2).

The resources consumed for treatment and follow-up of events were estimated on the basis of information provided by a panel of experts composed of six specialists with proven clinical experience (two cardiologists, two general clinicians and two neurologists).

Based on the resources identified by the panel, together with their respective unit costs22–24 (net of Value Added Tax [VAT]1), the costs incurred during acute treatment and follow-up were estimated. The cost of both doses of dabigatran (110 mg and 150 mg) are based on the recommended price of the patent-holding company. These are presented in Table 3.

The figures below were calculated from a societal perspective, considering overall costs, whether paid by the State or by patients. Costs accruing to the State were also estimated, taking a figure of 70% of the above figures for incremental costs and thus cost-effectiveness ratios, reflecting the fact that the State's contribution to the cost of dabigatran is 69%.

The results naturally reflect the available clinical evidence. Dabigatran reduces the incidence of almost all the events considered, particularly ischemic stroke, systemic embolism, TIA, hemorrhagic stroke and intracranial bleeding, while it is associated with a higher incidence of MI and extracranial bleeding due to the greater risk in patients aged 80 or over taking 110 mg dabigatran.

The lower event rate is reflected in gains for patients medicated with dabigatran in terms of both life years and QALYs: 0.415 life years and 0.439 QALYs per patient when therapy begins before the age of 80, and 0.143 life years and 0.166 QALYs for therapy begun after the age of 80. Assuming that 69% of the population with AF is aged under 80,4 dabigatran confers mean gains of 0.331 life years and 0.354 QALYs.

The daily cost of dabigatran is higher than that of warfarin or aspirin. However, the cost of treatment and follow-up of patients after suffering an event reduces the incremental cost by 20–30%. The cost of such preventive therapy thus rises by 5572 euros for patients aged under 80 and 2300 euros for those aged 80 or over, a mean of 4558 euros; the net cost increases by 3763 euros, 1231 euros and 2978 euros, respectively.

Integrating the clinical and economic data enables the incremental cost-effectiveness ratio (ICER) to be calculated per QALY and per life year (Table 4). Although there is no fixed willingness-to-pay threshold in Portugal per life year or QALY gained, the figure obtained – around 10000 euros per life year or QALY – is acceptable in the Portuguese context.

All economic evaluations are based on hypotheses that are used to construct the main scenario. However, since these hypotheses can affect the results, it is important to analyze the effect of alternative values for the relevant parameters, usually through analysis of the sensitivity of the final result to the hypotheses adopted.

It should be borne in mind that there are two types of parameter: those that can be associated with a statistical distribution to reflect the level of uncertainty, such as the probability of the events under consideration; and those that cannot, such as discount rate or time horizon.

With regard to the first type, a probabilistic sensitivity analysis can be performed by simultaneously simulating different occurrence rates for the different parameters. By way of example, Figure 2 presents the incremental costs and life years from 1000 simulations of the population beginning therapy before the age of 80. One advantage of this type of sensitivity analysis is that the proportion of simulations in which the cost-effectiveness ratio is less than a particular willingness-to-pay threshold can be determined (Table 5). Based on the values usually taken as the willingness-to-pay threshold for health gains, the adoption of dabigatran is likely to be cost-effective: in almost all the 1000 simulations performed, dabigatran has an ICER of less than 15 000 euros, around half the threshold generally used in Portugal.

Figure 2.

Probabilistic sensitivity analysis.

(0.14MB).

For the other parameters, a univariate sensitivity analysis is more useful, assessing the effect of different values in the main scenario. Table 6 presents the results for different discount rates, time horizons and costs. Both the discount rate (5%) and the time horizon (lifetime) used in the main scenario are those stipulated in the methodological guidelines in force in Portugal.25 However, while there is genuine uncertainty concerning the optimum value of the discount rate in this analysis, varying the time horizon merely reflects the time-frame during which the drug has an effect, since in methodological terms it is agreed that the differences between therapeutic options should be evaluated up to the time when these differences cease to exist, which in the case of chronic medication is the moment of death.

The costs of treatment and follow-up could theoretically be included in the probabilistic sensitivity analysis. However, since they were estimated on the basis of figures agreed by an expert panel, univariate sensitivity analysis would seem to provide a better assessment of their effect on the final result.

The univariate analysis showed that none of the parameters considered, except the time horizon, had a significant effect on cost-effectiveness ratios, since altering these parameters resulted in variations of only around 10%. The larger impact of the time horizon is to be expected, as shorter periods would mean a significant proportion of dabigatran's health gains would be excluded; this is often the case in economic evaluations of preventive therapies. Naturally, given the shorter life expectancy of patients aged over 80, the impact of reducing the time horizon on results for this population is considerably less.

The authors declare that no experiments were performed on humans or animals for this study.

The authors declare that no patient data appear in this article.

The authors declare that no patient data appear in this article.