Background

Atrial Fibrillation (AF) is one of the most frequent arrhythmias in adult population. It’s estimated prevalence is 1 – 2 % in the general population and increases to 10 % in subjects > 65 years old [1, 2]. It is associated with a 5-fold increase in the risk of stroke and systemic embolic events (i.e. pulmonary embolism and myocardial infarction) [3]. The chance of having a stroke depends on several risk factors which are considered in scores like the CHADS2 (Cardiac heart failure, Hypertension, Age, Diabetes, and Stroke) [4] or, more recently, the CHA[2]DS[2]-VASC (cardiac failure or ejection fraction <40 %, high blood pressure, age 64 to 74 or ≥75 years, diabetes, previous stroke or transient ischemic attack or thromboembolic events, vascular disease, and female sex) [5–7]. These scores constitute the basis for the decision to use medication in order to reduce embolic risk. Until recently therapeutic options to reduce the risk of stroke in AF included oral vitamin K antagonists (VKAs), warfarin and acenocoumarol, and, in patients who were unsuitable for these drugs, aspirin, alone or in association with clopidogrel. Despite the proven efficacy of VKAs, they have several limitations. The limitations include failure to maintain the treatment range (an International Normalized Ratio between 2.00 to 3.00), which results in needing to perform regular coagulation tests and many drug-drug interactions which are cause of the underutilization of VKAs [8, 9]. A measure of the quality of anticoagulation is the time in treatment range (TTR) that indicates the time spent between an INR 2.00 to 3.00. The limitations resulted in the goal to develop alternative treatment options. New Oral Anticoagulants (NOACs), which have unique pharmacodynamic and pharmacokinetic features that result in more stable and predictable anticoagulant effect [10] are recent treatment options to reduce the risk of stroke in AF. Currently, there are four NOACs (dabigatran, apixaban, rivaroxaban, and edoxaban) that completed phase III research programs and proved their safety and efficacy [11]. All of these assets except edoxaban have received medical approval for the use in AF in both the United States and Europe. Apixaban, an oral factor Xa inhibitor is the most recent compound to receive medical approval for the prevention of thrombotic events in AF in US and Europe. In one clinical trial for apixaban, ARISTOTLE, Apixaban demonstrated that it is superior to dose-adjusted warfarin in patients suitable for oral anticoagulants [12]. ARISTOTLE revealed a 21 % relative risk reduction in the primary efficacy endpoint (stroke or systemic embolism) and a 31 % relative risk reduction in the safety endpoint (major bleeding). Apixaban was also compared with aspirin in patients who are unsuitable for oral anticoagulation in the AVERROES trial [13]. In this study, apixaban demonstrated a 55 % relative risk reduction in the primary efficacy endpoint (stroke or systemic embolism).

Beyond their efficacy and safety profile, the decision for adopting apixaban by health care decision-makers has been supported by several health economic evaluations. Apixaban received a positive assessment by National Institute of Health Care and Excellence (NICE) in 2013 [14] and several cost-effectiveness analysis have been published revealing that apixaban, compared to either warfarin or aspirin, is a cost-effective alternative given a variety of health care settings [15, 16]. Given that the estimates of about 290,000 patients with AF are suitable for oral anticoagulants in Argentina [17–20], it is important to know the economic impact of new treatment options. The aim of the study is to assess the cost-effectiveness of apixaban versus VKAs in Argentina in order to provide local data for decision-makers.

Methods

We performed an adaptation of a Markov-based cost-effectiveness model previously submitted by the manufacturers of apixaban (Bristol-Myers Squibb and Pfizer) to NICE in the UK including local epidemiological and clinical data.

Model

The model was developed according to Good Modeling Practices [21], a detailed description is available elsewhere [14, 15]. The model allows a comparison of apixaban against currently available treatment options, including: warfarin, aspirin, dabigatran, rivaroxaban, and aspirin + clopidogrel. For the present analysis we report data for apixaban versus warfarin for patients suitable for oral anticoagulant therapy. The model includes 18 mutually exclusive health states for a hypothetical cohort of patients with non-valvular atrial fibrillation (NVAF) considering the occurrence of stroke (both ischemic or heamorrhagic), systemic embolism (myocardial infarction, pulmonary embolism), bleeding (intracranial, major bleeding, clinically relevant non-major bleeding) and death (Fig. 1). Transition probabilities between health states were derived from the ARISTOTLE trial [12] and from the life expectancy table for Argentina obtained from the World Health Organization. [22] Patients were followed for a lifetime horizon with 6 week cycles, with only allowing for one event per cycle.

Fig. 1
figure 1

Non-valvular Atrial Fibrillation decision-tree used in the model

Full size image

Population

The model considered a hypothetical cohort of 1000 patients with AF suitable for the use of oral anticoagulants. Demographics and baseline stroke risk for the cohort (based on the CHADS2 score) were obtained from published reports (Table 1) [18–20, 23] and expert’s opinions obtained during the Delphi Panel. Anticoagulation quality for VKA users was considered using the average time in therapeutic range (TTR) [24] for centers in Argentina [Table 1] [25, 26].

Table 1 Characteristics of the population considered in the model
Full size table

Clinical event risks and management

The risk and types of clinical events included in the model are presented in Table 2. Risks considered in the analysis were taken from the ARISTOLE trial [12]. Ischemic stroke risk was adjusted per each decade of life by a factor of 1.40 [27]. Due to inter-countries variation in medical management and treatment patterns, a Panel of Experts was convened using a Delphi method [28]. Two set of experts composed of 6 neurologists and 7 hematologists, representing the three health subsectors from Argentina (public, worker´s unions health care, and private), were consulted about clinical characteristics of patients with AF, treatment patterns, preferences for treatment change in case of bleeding events and resource utilization. All answers were revised in an open discussion and a final set of data was obtained.

Table 2 Type and risks of clinical events included in the model (reported per 100 patient/years)
Full size table

Utilities

Currently there is no local data regarding quality of life associated with AF, stroke or any other clinical outcome of interest. We therefore used the values from a UK catalogue of EQ-5D score [29] for each health state (Table 3).

Table 3 Utility and utility decrements associated with health states and treatments included in the model (measured by EQ-5D) from reference 30
Full size table

Costs and resource utilization

Direct costs and resource utilization for each health state included in the model were obtained from local data sources (Table 4) [30–33]. For data that was not available through a published report, we obtained the data from a Health Resource Cost Database. For the economic evaluation, we considered the perspective of the payer. Because the Argentinean health system comprises three health subsectors (public, worker’s unions health care, and private), each one with its own resource utilization pattern and prices, we reported cost as a weighed mean. For each item we considered the price for each health subsector multiplied by the proportion of subjects covered by the sector: 52.5 % by the public sector, 38.8 % by the worker’s union sector and 9 % by the private sector [34]. For example, the reported cost of stroke is: cost for the public sector x 0.522 + cost for the worker’s union sector x 0.388 + price for the private sector x 0.09.

Table 4 Drug and Event costs (in 2012 US Dollars)
Full size table

Drug costs were obtained from local formularies and adapted for each health subsector [30]. All prices were updated to the last quarter of 2012 and expressed in 2012 US Dollars (USD).

Economic analyses

We estimate the clinical effectiveness of apixaban versus warfarin in terms of events per 1000 treated patients using a lifetime scenario. Results were expressed as the incremental cost-effectiveness ratio (ICER) considering a cost effectiveness threshold between 1 to 3 per capita GDP (Gross Domestic Product) accordingly with the WHO-CHOICE recommendation [35]. The 2011 per capita GDP for Argentina was estimated as 11,558 USD [36], therefore the willingness to pay for every incremental QALY ranges from 11,558 USD to 34,674 USD. A 5 % discount rate was applied to both costs and events, as recommended by regional regulations in 2009 [37].

Deterministic sensitivity analysis was performed and presented as tornado graphic in order to assess the influence of key variables over the incremental cost-effectiveness ratio.

Probabilistic sensitivity analysis was conducted changing model’s parameters. Two thousand simulations were ran and plotted in a cost-effectiveness plane considering cost and QALY.

Sensitivity analyses based on the use of other AVK agent (acenocoumarol), cost’s discount rate and time horizon were also performed.

Results

Clinical effectiveness

In a VKA suitable cohort of 1000 patients with NVAF, compared with warfarin, the use of apixaban resulted in 24 fewer strokes (including first and recurrent ischemic and hemorrhagic) or systemic embolism, 41 fewer major bleeding events (including first and recurrent hemorrhagic stroke, other intracranial hemorrhage and other major bleeds), and 26 fewer cardiovascular-related deaths (Table 5). It is therefore estimated that 56 patients (number needed to treat) should be treated with apixaban over a lifetime in order to avoid one stroke (ischemic and hemorrhagic) compared with warfarin. In addition, apixaban is a better option than warfarin avoiding bleeding events (number needed to harm = 24). The use of apixaban in this target population resulted in an increase of 0.164 life years and 0.172 QALYs (Table 5).

Table 5 Clinical events in the cohort of NVAF patients treated with Apixaban and warfarin
Full size table

Costs

Table 6 presents the incremental costs of the use of apixaban compared with warfarin using a lifetime horizon. In every situation, the result falls below the willingness to pay threshold of 11558 USD (equivalent to one per capita GDP).

Table 6 Summary results of the cost-effectiveness analysis
Full size table

Sensitivity analysis

A one way sensitivity analysis (Fig. 2) revealed that the risk of stroke with apixaban is the main contributor to the cost-effectiveness. In Fig 3a and b are depicted the results of the probabilistic sensitivity analysis. The analysis presents that apixaban is more effective and more costly than warfarin. Considering a willingness to pay threshold of 11,558 USD per QALY gained, apixaban is a cost-effective alternative in 90 % of the cases. Using the upper threshold of 34,664 USD apixaban has a 95 % probability of being a cost-effective alternative compared to warfarin.

Fig. 2
figure 2

Tornado sensitivity analysis of the incremental cost-effectiveness ratio of Apixaban compared with warfarin. The solid vertical line represents the base-case incremental cost-effectiveness ratio. In the vertical lines are depicted the range obtained for a variable while the others are constant. MI: myocardial infarction; CV: cardiovascular hospitalization; ICH: intracranial hemorrhage; Monitoring visit: applied only to patients treated with warfarin; AFtrialRate: treatment´s discontinuation rate; OMB: other major bleed

Full size image
Fig. 3
figure 3

a Probabilistic sensitivity analysis of the incremental cost-effectiveness ratio of Apixaban compared with warfarin. b Acceptability curve for Apixaban compared with warfarin. a Upper threshold per QALY gained: 34664 USD; Lower Threshold per QALY gained: 11558 USD. b Probability of being accepted as a cost-effectiveness alternative considering the upper and lower thresholds showed in figure 3a

Full size image

Because warfarin is not the only AVK available in Argentina, we perform a separate analysis considering the cost for acenocoumarol. In this case the cost per life year gained was 745.20 USD and the cost per QALY gained was 711.52 USD.

We also assessed if changes in costs’ discounting rate would alter our results. Because there is no specific long-term investment rate in Argentina we run the analysis raising the discount rate to 10 % and 15 %. For the 10 % cost discount rate the cost per life year gained was 798.26 USD and the cost per QALY gained was 762.18 USD. For the 15 % cost discount rate the cost per life year gained was 811.28 USD and the cost per QALY gained was 774.61 USD.

Finally, we assessed different scenarios based on time horizon. For a 2 year scenario the cost per life year gained was 17043.00 USD and the cost per QALY gained was 6153.35 USD. For a 5 year scenario cost per life year gained was 3624.11 USD and the cost per QALY gained was 2234.97 USD.

Discussion

Our study estimated that apixaban is a cost-effective option, compared to warfarin, for the management of NVAF with an ICER of 786.08 USD/QALY. Local epidemiological data regarding AF in Argentina revealed that subjects suitable for oral anticoagulants are older and have a higher stroke risk (assessed by the CHADS2 score) than patients included in clinical trials [12]. We also found that the TTR is less than optimal in more than two third of the centers in Argentina. These findings are particularly relevant when the clinical effectiveness of apixaban is assessed versus vitamin K antagonist warfarin. In fact, considering the results of the model, apixaban use resulted in fewer thrombotic and bleeding events than warfarin, leading to less health resource utilization. Therefore, in spite of the much higher drug cost for apixaban, the strategy of adopting this novel anticoagulant results in an incremental cost of only 135 USD in a lifetime scenario. Even considering the cost for acenocoumarol instead of warfarin in a sensitivity analysis, the ICERs obtained were consistent with the cost-effectiveness of apixaban over all other options.

The definition of a willingness to pay threshold for incremental effectiveness is a matter of debate in countries that lack a defined value, as happens in the United Kingdom [38]. The adoption of the WHO Choice rule, using the GDP per capita as a parameter to set thresholds, is a valuable strategy in developing countries [35]. In the case of Argentina, which has three different payers’ sub-sectors, this method for establishing a threshold provides a wide reference for decision-makers. Therefore, the use of sensitivity analysis as a way to manage uncertainty provides a range of costs-effectiveness ratios that improves availability of data for decision-makers [39].

Our results are concordant with published cost-effectiveness analysis from other countries reporting that apixaban is a cost-effectiveness alternative. Canestaro et al. [16] reported for the United States, that apixaban is the optimal anticoagulant resulting in a net effectiveness increment of 0.41 QALY compared with warfarin. Accordingly, using the same model, Dorian et al. revealed that in the United Kingdom, apixaban improved both life expectancy and quality adjusted-life years compared to warfarin. Both studies, which used different cost data, revealed that apixaban is the most cost-effective therapeutic alternative.

The development of local data is of capital importance in the process of health technology assessment. Moreover, considering that most physicians in Argentina require that a medical technology is fully tested before adopting it [40], the availability of data which includes both clinical and economic aspects will certainly contribute to the decision making process.

Our study has many limitations. Local data are scarce and it is difficult to obtain needed data in published reports. This reflects problems regarding scientific publications in Argentina [41]. As a consequence, the relative weight of local experts’ opinion in validating information is much higher than in other countries. This issue has many potential consequences over results. The introduction of biases in expert’s responses (such as anchoring effects, absolute and relative judgements) are one of the most important factors to be taken into account when results are considered [42]. Finally, in recent years Argentina developed economic instability which could represent an objection with the discount rate adopted in this report despite that there have not been any discounting modifications in regional recommendations for conducting health economic evaluations.

Conclusions

In our study, using local epidemiological estimates and based on randomised clinical trials data, apixaban resulted a cost-effectiveness alternative to warfarin according to local willingness to pay thresholds.