Cost-effectiveness of Apixaban for Stroke Prevention in Patients with Atrial Fibrillation in Algeria

Background: Atrial fibrillation (AF) is a chronic sustained heart rhythm disorder associated with an increased risk of stroke. Apixaban, a new oral anticoagulant, was approved by the European Medicines Agency for prevention of stroke in patients with AF. The efficacy of apixaban has been investigated in randomised controlled trials. Objectives: The objective of this study was to estimate the economic implications of using apixaban compared to other anti-coagulations to reduce the risk of stroke in patients with AF from the perspective of the Algerian payer. Methods: A previously published Markov model was adapted to the Algerian setting. The model included patients for whom vitamin K antagonist (VKA) treatment is suitable and could initiate on acenocoumarol, rivaroxaban or apixaban, and those unsuitable for VKA treatment who could initiate on aspirin or apixaban. Over a lifetime time horizon, costs were estimated in Algerian dinars (DZD) and outcomes included life-years (LYs), quality-adjusted life-years (QALYs) and incremental cost-effectiveness ratios (ICERs). Results: In the VKA suitable population, apixaban was estimated to be a dominant treatment option over rivaroxaban, providing a higher number of QALYs at lower costs, while when compared with acenocoumarol, an ICER of 3 672 059 DZD per QALY gained was estimated. Amongst those unsuitable for VKA therapy, the ICER was 2 061 863 DZD per QALY gained. Conclusion: Apixaban was found to be a cost-effective choice for stroke prevention in patients with AF in Algeria compared to acenocoumarol and rivaroxaban in the VKA suitable population and compared to aspirin in the VKA unsuitable population.


BACKGROUND
Atrial fibrillation (AF) is a chronic sustained heart rhythm disorder most common among the elderly. 1 The burden of AF at age 65 years is nearly seven times that at age 45 years, and only one-quarter that at age 80 and older. 2 Patients with AF have a five-fold increased risk of stroke, and it is estimated that up to 25% of all strokes in the elderly are a consequence of AF. 1 Aging populations suffer from an increasing morbidity, mortality and economic burden of non-communicable diseases such as cardiovascular disease, AF and stroke.A rapid 11% increase in life expectancy seen since 1990 in the North Africa and Middle East (NAME) region, including Algeria, contributed to the modest increase of the prevalence of AF (2.6% [uncertainty intervals: -24.5 to 42.2] median change between 1990 and 2010) and stroke (5.6% between 2000 and 2012). 2,3Based on the projection of the World Health Organization (WHO), stroke-related mortality in the NAME region is expected to increase by 23% between 2015 and 2030. 4Overall, health spending per capita in the NAME region is estimated to be 956 and 932 United States dollar (USD) (PPP, 2005), respectively. 3cording to the 2016 European Society of Cardiology (ESC) Guidelines for the management of AF, when oral anticoagulation is initiated in a patient with AF who is eligible for a non-vitamin K antagonist oral anticoagulant (NOAC) drugs (apixaban, dabigatran, edoxaban, or rivaroxaban), a NOAC is recommended in preference to a vitamin K antagonist.Antiplatelet monotherapy is not recommended for stroke prevention in patients with AF, regardless of stroke risk. 5In a recently published international cross-sectional survey on the use of antithrombotic therapy in patients with AF, the results showed that patients with the greatest risk of stroke (i.e., CHADS 2 ≥2) in the Middle East and Africa region had the highest oral anticoagulant use (66.7%). 6ixaban is a direct and highly selective active site inhibitor of factor Xa that has been approved by the European Medicines Agency for stroke prevention in patients with AF.The efficacy and safety of apixaban (5 mg twice daily [b.i.d]) versus dose-adjusted warfarin and aspirin has been studied in two large randomised, multicentre, double-blind, Phase III trials (ARISTOTLE 7 and AVERROES, 8 respectively) in patients with nonvalvular atrial fibrillation (NVAF) and one or more additional risk factors (i.e., prior stroke, age, symptomatic heart failure, hypertension and/or diabetes mellitus).Apixaban was found to significantly reduce the risk of stroke and systemic embolism, major bleeding and all-cause mortality compared to warfarin.In patients who were unsuitable for vitamin K antagonist therapy, apixaban significantly reduced the risk of stroke and systemic embolism without significantly increasing the risk of major bleeding compared with aspirin.
An assessment of the cost-effectiveness of available therapies for stroke prevention in patients with AF has not been previously performed in Algeria.To inform decision making around optimal treatment, it is important to consider the costs and benefits for alternative therapies.To this end, the objective of this study was to estimate the economic implications of using apixaban compared to other anti-coagulations to reduce the risk of stroke in patients with AF from the perspective of the Algerian payer.

Model Design
A previously developed and validated Markov model9,10 was used to evaluate the long-term clinical and economic outcomes in Algerian patients with AF receiving anticoagulant treatment over a lifetime.This evaluation was conducted from the perspective of the Algerian payer.The model included two cohorts of patients -those suitable for VKA treatment and those unsuitable for VKA treatment.VKA-suitable patients could initiate on either of the following treatments: dose-adjusted acenocoumarol, rivaroxaban 20 mg once daily or apixaban 5 mg b.i.d.Patients who were VKA unsuitable could initiate on aspirin or apixaban.Patients transitioned through health states including: NVAF, ischaemic stroke and haemorrhagic strokes (mild, moderate, severe and fatal), intracranial haemorrhage (ICH) other than haemorrhagic strokes (referred to as other ICH), systemic embolism, myocardial infarction, other major bleeds (OMB; non-ICH major bleeds), clinically relevant non-major (CRNM) bleeds and NVAF with subsequent aspirin treatment or death (Figure 1).

Model Inputs
Patient population characteristics including age, gender and CHADS 2 distribution, as well as clinical event rates, were taken from the ARISTOTLE and AVERROES trials. 7,8Event rates for acenocoumarol were assumed the same as warfarin, obtained from the ARISTOTLE trial.For rivaroxaban, clinical event rates were taken from indirect treatment comparison in the absence of head-to-head clinical trial data. 9,10Tables 1 and 2 show a summary of inputs included in the model.Direct costs in Algerian dinars (DZD), at 2015 prices, specific to the Algerian market were applied to the model (Table 3).The cost estimates of clinical events were obtained from a local Algerian private clinic and the university hospital CHU Mustapha.Health and cost outcomes were discounted at an annual rate of 3.5% as in prior published studies in the United Kingdom (UK). 9,10Patients were assigned utilities according to their health state (where a utility of 1 denotes full health and 0 denotes death) as presented in Table 4.In the absence of Algerian-specific utility estimates, the model utilised estimates from a UK-based EuroQol-5 dimensions (EQ-5D) catalogue as those used in earlier models. 9,10,12tility decrements associated with the use of treatments were applied, with the highest disutility applied to acenocoumarol due to monitoring and drug interactions based on prior warfarin studies.Apixaban or aspirin 0.0020 (0.00-0.04) Acenocoumarol* 0.0120 (0.00-0.08)

13
CI: onfidence interval * Utility decrements in this analysis are assumed to be the same as warfarin.

Analyses
Apixaban was compared with acenocoumarol and rivaroxaban among the VKA-suitable population.Among the VKA-unsuitable population, apixaban was compared with aspirin.For a cohort of 1,000 patients, the model estimated the total clinical benefit in terms of number of clinical events, estimated life-years (LYs), qualityadjusted life-years (QALYs), as well as costs, over a lifetime time horizon for each treatment.The relative clinical benefit of apixaban compared to other treatments was assessed using incremental cost-effectiveness ratios (ICERs), which denoted the cost per QALY gained per average patient for the adoption of apixaban over the comparator treatments.
In addition to the analyses using the primary inputs -known as base-case -univariate sensitivity analyses were conducted to explore the impact of various input parameters on the ICERs including: varying discount rates, event risks for each treatment, the HRs for acenocoumarol, rivaroxaban, and aspirin versus apixaban, as well cost and utility inputs.These input parameters were varied by their 95% confidence intervals where available.Scenario analyses were also conducted to examine the impact of key assumptions in the model.The evaluated scenarios included: 1) variation in the CHADS 2 distribution, 2) variation in the quality of international normalised ratio (INR) control (i.e., the centre time in therapeutic range [cTTR] distribution), 3) variations in population characteristics and INR control estimated for the Algerian population, 4) assumptions on second-line treatment (i.e., no treatment), and 5) assumptions on treatment discontinuation (i.e., 0 risk of discontinuation, or same as apixaban).
Furthermore, probabilistic sensitivity analyses were conducted by running 2000 iterations of a cohort of 1000 patients, with the model parameters drawn randomly from probability distributions in each iteration.The results of the probabilistic sensitivity analyses were presented as a scatterplot of the incremental QALYs versus incremental costs for apixaban versus comparators in both the VKA-suitable and VKA-unsuitable populations.
A cost-effectiveness acceptability curve (CEAC) was generated representing the proportion of the iterations for which each treatment was considered the most cost-effective alternative at a given threshold of willingness to pay for a QALY gained.Given no established thresholds exist in Algeria, the analysis assessed the costeffectiveness of apixaban versus each of the comparators using threshold values commonly used in the United States ($50 000/QALY equal to DZD 5 430 721, based on exchange rates in 2015 where $1=DZD 108.6), and the UK (£30 000/QALY equal to DZD 4 779 190, based on exchange rates in 2015 where £1=DZD 159.3).

Base Case
Among a cohort of 1000 patients with AF who were VKA-unsuitable, patients treated with apixaban were predicted to experience 58 fewer strokes than patients treated with aspirin.The number of bleeds increased in patients treated with apixaban, with 32 additional major bleeds (haemorrhagic strokes, OMB and ICHs) and 56 additional CRNM bleeds compared to patients treated with aspirin (Table 5).The reduction in strokes associated with apixaban treatment translated into 0.205 additional discounted QALYs and 0.237 additional discounted LYs.
In the cohort of VKA-suitable patients with AF, those treated with apixaban were predicted to experience 43 fewer major bleeds (haemorrhagic strokes, OMB and ICHs) than patients treated with acenocoumarol or rivaroxaban.CRNM bleeds were also reduced by 41 and 52 events compared to acenocoumarol and rivaroxaban, respectively (Table 5).The reduction in clinical events among patients treated with apixaban was associated with 0.14 and 0.05 additional discounted LYs, and an additional 0.14 and 0.04 discounted QALYs compared to acenocoumarol and rivaroxaban, respectively.
Treatment costs are presented by event costs, anticoagulant treatment and management costs, routine care and monitoring in Table 5.In the VKA-unsuitable population, the model predicted an average lifetime discounted total cost per patient of 920 612 DZD for patients treated with apixaban, and a lifetime discounted cost of 496 716 DZD for patients treated with aspirin.Among the VKA-suitable population, the average lifetime discounted total cost per patient was 978 530 DZD for patients treated with apixaban; lifetime discounted costs for patients treated with acenocoumarol and rivaroxaban were 441 714 DZD and

Scenario Analyses
The results from the scenario analyses are detailed in Table 6.These showed that the ICER for the comparison of apixaban to aspirin in the VKA unsuitable population was mostly sensitive in low-risk patients (i.e., those with CHADS 2 score less than 1) with the least favourable ICER (97.53% increase from base case) whereas in the VKA suitable population, results were mostly sensitive to the quality of INR control cTTR distribution, with the most favourable ICER (33.02% decrease from base case) in patients with poorly control INR (i.e., cTTR < 52.38%).In the VKA-suitable population, in all scenarios tested, apixaban dominated rivaroxaban, providing higher number of QALYs at lower costs.

Probabilistic Sensitivity Analyses
The incremental outcomes in terms of QALYs gained were plotted against incremental costs of apixaban versus other comparators on the cost-effectiveness planes for all 2000 iterations, shown in Figures 3a and 3b for the VKA-unsuitable and VKA-suitable populations, respectively.In the VKA-unsuitable population, the costeffectiveness plane suggested that in the majority of iterations apixaban was more effective and more costly than aspirin and acenocoumarol.When compared to rivaroxaban, the cost-effectiveness plane suggested that apixaban was more effective and less costly in most of the iterations.When compared to aspirin, the treatment of VKA-unsuitable patients with apixaban was an optimal treatment choice, representing the maximum net benefit at a willingness to pay threshold above 2 198 335 DZD ($20 242, £13 800; based on exchange rates in 2015, where $1=DZD 108.6 and £1=DZD 159.3) per QALY (Figure 4a).At a willingness to pay threshold of 5 430 721 DZD (i.e., $50 000), apixaban was considered a cost-effective option in 96% of iterations when compared to aspirin.At a willingness to pay threshold of 4 779 190 DZD (i.e.£30 000) per QALY, apixaban was considered cost-effective in 94% of iterations.According to the CEAC for VKA-suitable patients (Figure 4b), apixaban was an optimal treatment choice over acenocoumarol and rivaroxaban at a willingness to pay threshold above 3 869 070 DZD ($35 627, £24 288; based on exchange rates in 2015, where $1=DZD 108.6 and £1=DZD 159.3) per QALY.Below this threshold, acenocoumarol was an optimal treatment option.

DISCUSSION
This study assessed the cost-effectiveness of apixaban compared to other available treatments, including aspirin, acenocoumarol and rivaroxaban, from an Algerian payer perspective.Among the patients suitable for VKA, fewer thromboembolic events and bleedings were estimated with apixaban compared to acenocoumarol and rivaroxaban, and these translated into LYs and QALYs gained.Apixaban was estimated to be a dominant treatment option over rivaroxaban, providing a higher number of QALYs at lower costs, while when compared with acenocoumarol, an ICER of 3 672 059 DZD ($37 054, £23 051) per QALY gained was estimated.
Among those unsuitable for VKA, although apixaban was estimated to slightly increase in the number of bleedings, it still led to LYs and QALYs gained due to a significant reduction in the number of thromboembolic events, with an estimated ICER of 2 061 863 DZD ($35 627, £24 288) per QALY gained.
Lower medical care costs were observed with apixaban due to the reduction in overall clinical events.Treatment costs increased with apixaban due to the higher acquisition cost compared to aspirin and acenocoumarol, and longer use with apixaban given the increased in life expectancy and lower discontinuation rates as observed in the clinical trials.
To date, our study is the first published economic evaluation assessing the cost-effectiveness of apixaban for stroke prevention in patients with NVAF from an Algerian payer's perspective.There are, however, some important limitations to these analyses.In Algeria, the standard anticoagulant used in patients eligible for VKA is acenocoumarol.Due to the unavailability of efficacy and safety evidence to link acenocoumarol to the apixaban clinical trial, the study used results extrapolated from clinical trials for warfarin.5][16] Another limitation of the current study relates to the unavailability of Algerian-specific utilities.The analysis employed utilities based on a UK EQ-5D catalogue. 12Since these were taken from a European population, it is plausible to assume that they would be similar for the Algerian population.8][19] In addition, results from univariate sensitivity analysis, varying utility for each clinical event by its' 95% confidence intervals, demonstrated that the ICERs were altered by only less than 4%.
Previous cost-effectiveness analyses for apixaban have been conducted in Europe, the US and Latin America using different willingness to pay thresholds established in the countries involved in order to determine costeffectiveness. 20In Algeria, there is no such established threshold, therefore, these analyses considered threshold values that have been established in other countries, namely the US and the UK.An alternative approach for determining cost-effectiveness in low and middle income countries is the threshold of one-to-three times per capita income for averting a disability-adjusted life-year recommended by WHO. 21Although this approach has been frequently adopted in prior studies, the lack of empirical evidence to support this rule has been widely criticised. 22,23In addition, the outcomes assessed in the model used in this study included QALYs as opposed to DALYs, therefore, the WHO threshold was not considered in the present analyses.It is widely acknowledged that a broad range of factors are shaping decision making criteria especially in the context of low and middle income countries. 23,24The authors believe that budgetary constraints and priorities set within the current political and institutional context in Algeria should ultimately be taken into account for appropriate decision making.

CONCLUSION
The study demonstrated that apixaban is a cost-effective choice for stroke prevention in patients with AF compared to acenocoumarol and rivaroxaban in the VKA-suitable population and compared to aspirin in the VKA-unsuitable population.

DISCLOSURE
This study was funded by Pfizer.Thitima Kongnakorn and Evie Merinopoulou are full-time employees of Evidera and served as paid consultants to Pfizer in association with conducting this study and with the development of this manuscript.Mohamed Said Bettayeb and Sid Ahmed Kherraf are full-time employees of Pfizer with ownership in stocks.

Figure 3 .
Figure 3. Results of Probabilistic Sensitivity Analyses for the VKA-unsuitable Population a) Cost-effectiveness plane for apixaban versus aspirin (incremental costs and QALYs)

Figure 4 .
Figure 4. Results of Probabilistic Sensitivity Analyses for the VKA-suitable Population a) Cost-effectiveness plane for apixaban versus acenocoumarol and apixaban versus rivaroxaban (incremental costs and QALYs)

Table 1 .
Population Demographic and Clinical Characteristics

Table 2 . Clinical Event Rates by Treatment VKA Unsuitable VKA Suitable Apixaban Aspirin Source Apixaban Acenocoumarol* Rivaroxaban Source Event Rate of events per 100 patient years HR vs. apixaban (95% CI)
11resource use and cost inputs were adapted to the Algerian setting.Background mortality was derived from age-and gender-specific life tables of the Algerian population taken from WHO.11

Table 3 .
Model Cost Inputs Adapted to the Algerian Market

& other CV hospitalisation Acute care (per episode) Source
CRNM: clinically relevant non-major; CV: cardiovascular; GI: gastrointestinal; ICH: intracranial haemorrhage; CHU Mustapha: centre hospitalier universitaire-university hospital Mustapha *Acute period for strokes and HS (mild, moderate and severe) were to be 2 weeks.

Table 4 .
Model Inputs on Utilities

Table 5 .
Base-case results over lifetime: clinical events predicted among 1 000 patients, mean LYs, QALYs and costs per patient Univariate Sensitivity AnalysesResults from univariate sensitivity analyses are presented as tornado diagrams (Figures2a-2b