Clinical and economic impact of rivaroxaban on the burden of atrial fibrillation: The case study of Japan

Abstract

Objectives: Atrial fibrillation (AF) affects an estimated 1.5 million individuals in Japan, increasing their stroke risk and imposing considerable costs on the Japanese healthcare system. To reduce stroke incidence, guidelines recommend using anticoagulants in moderate-to-high risk non-valvular AF (NVAF) patients; however, many patients receive no treatment, aspirin only, or remain poorly-controlled on vitamin K antagonists (VKAs) due to high VKA discontinuation rates and non-adherence to guidelines. A prevalence-based Markov model was developed to estimate the clinical and budgetary impact of treating these patients with Xarelto^TM (rivaroxaban, Bayer AG) in Japan.

Methods: Population, baseline risk of events, and associated management costs were estimated using data from Japanese publications where available. Treatment efficacy and safety were derived from published data and the J-ROCKET AF trial. Drug and physician visit costs were based on data from the Ministry of Health, Labor, and Welfare, the J-ROCKET AF trial, and Japanese clinical guidelines.

Results: This model demonstrates that increased use of rivaroxaban in inadequately-managed NVAF patients could avoid 456 081 non-fatal ischemic strokes (IS) and 76 975 cardiovascular deaths over 10 years in Japan. This clinical benefit offsets the increased incidence of myocardial infarctions and anticoagulant-related bleeding. Decreased event costs could lead to a ¥188.4 billion decrease in net spending over the analysis time horizon.

Conclusions: Introducing rivaroxaban may decrease the burden of NVAF in Japanese society. From a clinical perspective, the reduction in IS and embolic events outweighs the increased risk of anticoagulant-related bleeding; from an economic perspective, reduced event costs offset drug and physician visit costs, resulting in cost savings.

Keywords:

• Atrial fibrillation
• stroke
• rivaroxaban
• anti-coagulation
• budget impact

Introduction

Atrial fibrillation (AF) is the most common sustained abnormality of heart rhythm¹. Approximately 1.5 million individuals were estimated to have AF in Japan in 2010, based on an extrapolation of AF prevalence reported in a community-based registry^2,3. This number is expected to increase further due to the aging of the population^1,4,5. AF significantly increases the risk of stroke 5-fold⁶. Stroke represents a major burden for Japanese society; one in five persons 65 years of age or older who require long-term care have suffered a stroke⁷. In 2015, it is estimated that there are 2.8 million prevalent cases of stroke and that 300,000 new patients will suffer from stroke in Japan⁸.

AF-related stroke is a type of cardioembolic stroke, which represents 27.0% of all ischemic strokes (IS)⁹ and imposes both greater mortality and disability than non-cardioembolic strokes^10–12. A prospective study in Japan showed that 13% of patients with cardioembolic stroke died during hospitalization¹³. Furthermore, a cohort study reported that 12.5% of patients with cardioembolic stroke died within 1 year following hospital discharge¹⁴. Cardioembolic stroke is also reported to impose the longest hospitalization, lowest levels of activities of daily living* (ADLs) both at admission and discharge, and highest medical cost compared to other types of stroke¹⁵. Given that in 2014 the Japanese healthcare system spent ¥1777 billion ($15 billion)¹⁶ managing cerebrovascular disease, there is clearly a need for improved control of conditions that contribute to, and exacerbate the humanistic and economic burden of cerebrovascular disease.

Non-valvular AF (NVAF) guidelines recommend anticoagulants in NVAF patients with moderate-to-high risk of stroke, as defined by risk scores such as the CHADS₂ or CHADS₂-VASc^17,18. Until recently, vitamin K agonists (VKAs) were the only anticoagulants used to manage AF, but they have a narrow therapeutic range¹⁹ and interact with a variety of foods²⁰ and commonly-used drugs²¹, making it difficult to maintain the international normalized ratio (INR) within the required range^22,23. Burdensome INR monitoring²⁴ and lifestyle changes required for VKA therapy often lead to poor persistence²⁵. Only half of NVAF patients are receiving VKAs, as demonstrated in several studies in Japan^2,26,27. Even patients receiving VKAs may not be adequately managed due to the difficulty of remaining in the target INR range. A prior study has shown that almost half of those receiving VKAs remained poorly controlled²⁸, highlighting the need for novel approaches of anti-coagulation.

The Japanese Circulation Society (JCS) updated the guidelines for pharmacotherapy of NVAF in January 2014¹⁸, promoting the use of anti-coagulants in moderate-to-high risk patients identified using the CHADS₂ score (i.e. CHADS₂ ≥1). Xarelto (rivaroxaban)* is a Factor Xa inhibitor that prevents thrombin formation²⁹. It has been studied in more than 14 000 NVAF patients in the ROCKET AF trial³⁰ and 1280 Japanese NVAF patients in the dose-adjusted J-ROCKET AF³¹ trial, and was included in the JCS guideline update¹⁸.

Assuming new oral anti-coagulants would improve adherence to treatment guidelines, we assessed the clinical and economic impact to the Japanese healthcare system of treating moderate-to-high risk patients currently inadequately-managed (i.e. those receiving no treatment, aspirin only, or poorly-controlled on VKA) with rivaroxaban, over 10 years.

Methods

Model design

A prevalence-based simulation was conducted from the perspective of the Japanese healthcare payers over a 10-year time horizon, taking into account drug acquisition, physician visit, and event costs associated with the treatment patterns before the introduction of novel oral anticoagulants (NOAC), and a revised treatment pattern in which rivaroxaban was used to simulate perfect adherence to treatment guidelines This perspective encompasses patient co-payments—in Japan, patients pay a proportion of direct medical costs, which varies based on their age and income level. The structure of our model is based on a cost-effectiveness model evaluating rivaroxaban for the prevention of stroke and systemic embolism (SE) in patients with AF. The initial model had been reviewed by the UK’s National Institute for Health and Care Excellence (NICE), and used to generate outcomes and cost estimates for Spain, Greece, Slovakia, and the UK^32–36. Our budget impact simulation was developed on the basis of this model by not factoring in treatment discontinuation, limiting the time horizon of the model, evaluating fewer comparators, tallying direct costs only, and considering outcomes, rather than quality-adjusted life years (QALYs), as a measure of treatment efficacy.

A Markov chain was used to model cardiovascular (CV) events and associated costs in the eligible patient population. IS, intracranial hemorrhage (ICH), or myocardial infarction (MI) resulted in long-term care, whereas SE and major extracranial hemorrhage (ECH) did not lead to long-term care (Figure 1). Moderate-risk patients—CHADS₂ score of 1—experiencing a non-fatal IS moved to the high-risk cohort, defined as a CHADS₂ score of 2 or higher. High-risk patients had the same baseline and recurrent risk of IS, which is based on the risk for patients with CHADS₂ score of 2 or higher. The size of the eligible patient population due to disease incidence and all-cause and CV mortality was reflected in the number of patients who incurred treatment, physician visit, and event costs. The risk of all-cause mortality was uniformly applied to patients in all states. Patients transitioned in the model in 1-year cycles. A half-cycle correction was applied to account for the timing of events as per the International Society for Pharmacoeconomics and Outcomes Research guidelines recommendations³⁷.

Figure 1. Markov chain. CV, cardiovascular; ECH, extracranial hemorrhage; ICH, intracerebral hemorrhage; IS, ischemic stroke; MI, myocardial infarction; SE, systemic embolism.

Model inputs

Population inputs

The number of patients initially eligible to receive rivaroxaban was based on the number of individuals aged 40 years and above estimated to reside in Japan in 2014 and taken from population projections from the Japanese National Institute of Population and Social Security Research³⁸ (Table 1). This source was selected as a representative estimate of the population size during the model’s start year. Within each age category, individual annual population growth estimates were applied year-on-year.

Table 1. Population, population growth, prevalence of NVAF and non-CV mortality, by age^2,38–41.

Age (years)	Population		AF		Non-CV mortality
	Number	Annual growth	Prevalence	NVAF
40−49	18,396,000	−1.2%	0.18%	76%	0.1%
50−59	15,429,000	+1.5%	0.49%	74%	0.3%
60−69	18,128,000	−2.4%	1.68%	84%	0.6%
70−79	14,183,000	+1.5%	4.29%	84%	1.5%
≥80	9,737,000	+3.1%	6.68%	81%	7.0%

AF, atrial fibrillation; CV, cardiovascular; NVAF, non-valvular atrial fibrillation.

The age-specific prevalence of AF was based on data from the Fushimi registry^2,39 (Table 1), a community-based survey of AF patients, which reported 3183 enrollees as of June 2012. The age-specific proportion of NVAF was based on another study evaluating the prevalence of AF⁴⁰.

Among patients with NVAF, only those with moderate- or high-risk, defined as CHADS₂ = 1 and CHADS₂ ≥2, respectively, were considered as per guideline recommendations^17,18. The proportion of patients with moderate (26%) and high risk (63%) is based on the Fushimi registry². Only patients aged 40 years and above were considered in the model; the proportion of the population younger than 40 years were assumed to be of limited interest due to their inherently lower risk of AF and lower CHADS₂ score.

The algorithm defining the eligible patient population is outlined in Figure 2.

Figure 2. Eligible patient population. AF, Atrial fibrillation; CHADS₂, Congestive heart failure, Hypertension, Age (≥75 years), Diabetes Mellitus, Prior Stroke or TIA; NVAF, Non-valvular atrial fibrillation.

All-cause mortality was based on the Handbook of Health and Welfare Statistics published in 2013 by the Japanese Ministry of Health, Labour and Welfare⁴¹. To avoid double-counting, CV mortality (categories Hi05 and Hi06) was subtracted from all-cause mortality.

Market share

In the base-case scenario, 54%, 3%, and 43% of NVAF patients with moderate risk and 46%, 4%, and 50% of patients with high-risk were assumed to receive no treatment, aspirin only, and VKA, respectively, based on treatment patterns in 273 Japanese acute care hospitals²⁶ (Table 2). Patients currently receiving VKA were assumed to be either well- or poorly-controlled, based on a multi-center study, in which 49% of moderate- and 32% of high-risk patients receiving VKA were poorly controlled²⁸. To model adherence to treatment guidelines in the revised treatment pattern, all patients with moderate-to-high risk of stroke, except those who were well controlled on VKA, were switched to rivaroxaban.

Table 2. Treatment patterns^26–28.

	CHADS2 score	Rivaroxaban	No treatment	Aspirin	VKA
					Well-controlled	Poorly-controlled
Treatment pattern before NOAC introduction	CHADS2 = 1	0%	54%	3%	26%	17%
	CHADS2 ≥ 2	0%	46%	4%	34%	16%
Perfect adherence to guidelines using rivaroxaban as an example	CHADS2 = 1	74%	0%	0%	26%	0%
	CHADS2 ≥ 2	66%	0%	0%	34%	0%

CHADS₂, Congestive heart failure, Hypertension, Age (≥75 years), Diabetes Mellitus, Prior Stroke or TIA; NOAC, Novel oral anti-coagulant; VKA, vitamin K antagonist.

Baseline risk of events

The model estimates the number of major and minor IS, SE, MI, ICH, and major ECH (Table 3). Major and minor strokes were defined as having a National Institutes of Health Stroke Scale (NIHSS) score above 16 and up to 15, respectively¹⁴. Major ECH was defined as clinically overt bleeding that was associated with a fall in hemoglobin ≥20 g/L, a transfusion of ≥2 units of packed red blood cells or whole blood, involved a critical site, or had a fatal outcome³¹. Minor bleeding events were not considered in this model.

Table 3. Baseline risk of events, by age^{14,31,42–44}.

Age (years)	IS		SE	MI	ICH		Major ECH
	Moderate	High			Moderate	High	Moderate	High
40−49	0.2%	1.1%	0.27%	0.1%	0.1%	5.5%	0.74%	1.68%
50−59	0.4%	2.6%		0.6%	0.1%	5.5%
60−69	1.4%	9.2%		1.4%	0.4%	6.9%
70−79	3.2%	21.7%		4.1%	0.6%	8.1%
≥80	3.2%	21.7%		4.1%	0.6%	8.8%

ECH, extracranial hemorrhage; ICH, intracranial hemorrhage; IS, ischemic stroke; MI, myocardial infarction; SE, systemic embolism.

Table 4. CV mortality associated with events^{14,31,45–47}.

	IS		SE	MI	ICH	Major ECH
	Minor	Major
Mortality rate following event	3.7%	23.1%	0.0%	22.9%	10.8%	6.7%

CV, cardiovascular; ECH, extracranial hemorrhage; ICH, intracranial hemorrhage; IS, ischemic stroke; MI, myocardial infarction; SE, systemic embolism.

The baseline risk of IS, SE, MI, ICH, and ECH in untreated patients with high-risk of stroke was calculated by adjusting data from the J-ROCKET AF trial³¹, with results from a network meta-analysis (NMA) including 32 clinical trials (Data on file, Oxford Outcomes). The event rates in untreated patients were obtained on the basis of event rates in patients receiving warfarin or rivaroxaban. The J-ROCKET AF trial was conducted in patients whose CHADS₂ score was ≥2; therefore, the risk of IS in patients with moderate risk of stroke was calculated by dividing the rate of IS in high-risk patients by 6.77, based on the relative risk of stroke in moderate- and high-risk patients identified in a prior Japanese study⁴². The baseline rates of IS, MI, and ICH were adjusted by age using data from a Japanese multi-center population-based cohort study⁴³. Due to limitations in data acquisition, it was assumed that patients who were 80 years of age or older had similar baseline risks of events to those aged between 70–79 years of age—this assumption is conservative in that the true rate of events will likely be higher in the older patient sub-group.

The recurrent rate of IS in moderate risk patients is the same as the baseline risk of IS in high-risk patients as those patients move to the high-risk cohort. It is assumed that the recurrent risk of IS among high-risk patients is the same as the baseline risk. IS and ICH rates were adjusted by age group using age-specific cumulative event rates for all strokes from the same study⁴³; IS rates ranged from 0.2–3.2% and 1.1–21.7% in the moderate- and high-risk groups, respectively. ICH rates ranged from 0.1–0.6% and 5.5–8.8% in the moderate- and high-risk groups, respectively. MI rates were also adjusted by age group based on data from the same source⁴³, ranging from 0.1–4.1% (MI rates did not vary by moderate and high-risk patient groups as the model structure is based on CHADS₂ score). In the absence of relevant data, major ECH and SE rates were not adjusted by age group.

High-risk patients are assumed to have increased risk of ICH and major ECH considering the overlap between the factors affecting HAS-BLED risk score and the CHADS₂ score. However, as the model is based on CHADS₂ score, only the incidence of ischemic strokes was assumed to result in transition of patients from the moderate-to-high risk patient sub-groups in the model (Figure 1). In absence of relevant Japanese data, the baseline risk of major ECH among high-risk patients was calculated by extrapolating data from an American study⁴⁴.

CV mortality was estimated based on the proportion of the above-mentioned fatal events^{14,31,45–47} (Table 4). The breakdown of major and minor IS—41% and 59%, respectively—and the mortality associated with those events was based on data from a Japanese registry¹⁴. SE were assumed to be non-fatal, in line with an economic analysis of the best treatment pathways for patients with heart disease⁴⁵. The mortality rate of MI was based on the results from a cross-sectional study in Okinawa⁴⁶. The mortality of ICH and ECH were based on the Hisayama study and the J-ROCKET AF trial, respectively^31,47.

Treatment efficacy and safety

Treatment efficacy and safety data were based on the results from a NMA (including the J-ROCKET trial) comparing event rates with aspirin, VKA, and rivaroxaban to no treatment (Data on file, Oxford Outcomes; Table 5). It was assumed that increases and reductions in event rates associated with different treatments were the same across the risk groups, in line with findings from the ROCKET AF trial³⁰.

Table 5. Relative risk (RR) of events (including 95% CI) compared to no treatment.

Event	No treatment	Rivaroxaban	Aspirin	VKA
IS	1.00	0.14 (0.04–0.44)	0.63 (0.36–1.08)	0.36 (0.23–0.54)
SE		0.47 (0.01–36.24)	0.47 (0.06–2.23)	0.46 (0.10–1.58)
MI		1.61 (0.10–58.87)	0.49 (0.11–1.84)	0.42 (0.11–1.36)
ICH		2.24 (0.29–22.33)	2.08 (0.46–11.26)	4.21 (0.81–31.56)
ECH		4.46 (0.23–122.59)	3.13 (0.59–24.83)	4.40 (1.21–20.33)

CI, confidence interval; ECH, extracranial hemorrhage; ICH, intracranial hemorrhage; IS, ischemic stroke; MI, myocardial infarction; RR, relative risk; SE, systemic embolism; VKA, vitamin K antagonist.

Costs

All costs are evaluated and presented in 2014 Japanese Yen (¥)⁴⁸; no discounting was applied in the base case, as per ISPOR guidelines on budget impact analyses⁴⁹; however, discounting of costs was included within sensitivity analyses. The daily cost of treating patients with aspirin, VKA, or rivaroxaban was based on the daily recommended dose as per the J-ROCKET AF trial protocol (Data on File, Bayer), the J-RHYTHM registry⁵⁰, and post-marketing surveillance of rivaroxaban⁵¹, respectively, as well as on the 2014 revision of medical fees by the Ministry of Health, Labour, and Welfare⁵² (Table 6). It was assumed that 23.3% and 76.7% of patients receiving rivaroxaban received 10 mg and 15 mg doses, respectively, based on a post-marketing surveillance study⁵¹. Patients receiving no treatment were not associated with drug costs.

Table 6. Drug costs^50–52.

	Rivaroxaban	Aspirin	VKA
Average daily dose	1 (10/15mg)	1 (100mg)	3 (3mg)
Average daily cost	¥464.95	¥5.60	¥28.80

VKA, vitamin K antagonist.

The annual physician visit cost for well- and poorly-controlled VKA patients were based on the cost of a follow-up consultation including prothrombin-time check and diagnosis fee (¥2150), as outlined in the Medical Fee Ministerial Notification No.57 of the Ministry of Health, Labour, and Welfare⁵³ and the average number of visits per year. Following recommendations from guidelines¹⁸, well-controlled patients were estimated to have an average of 12 monitoring visits, while poorly-controlled patients had 15 such visits per year based on the assumption that poorly-controlled patients were managed similarly to VKA-naïve patients—four visits during the month of drug initiation, followed by one visit per subsequent month (Data on file, Bayer). As a result, well- and poorly-controlled patients commanded an annual monitoring cost of ¥25 800 and ¥32 250, respectively. The annual physician-visit cost of patients receiving aspirin or rivaroxaban was assumed to be the same (¥8640), assuming one physician visit per month based on the cost of follow-up visits (¥720) outlined by the Ministry of Health, Labour, and Welfare⁵³.

The first year cost of a major IS, MI, or ICH was based on 3 months of acute care, followed by 9 months of long-term care; the cost of event management in subsequent years was based on 12 months of long-term care (Table 7). The cost of minor IS, SE, major ECH, and CV deaths included 3 months of acute care. The acute costs of managing minor and major IS, SE, MI, or ICH were based on a prior study of Diagnosis Procedure Combination* (DPC) data²⁶. The cost of managing an acute, major ECH and the cost of 1 year of long-term care was based on a Japanese pharmacoeconomic study⁵⁴. Long-term care costs for IS and ICH were based on costs of managing patients based on their modified Rankin Scale (mRS) score, as reported by Hattori et al.⁵⁵, weighted by the proportion of patients with each mRS score as identified in the Stroke Data Bank 2009⁵⁶. Long-term care costs for MI were based on costs reported in a Japanese retrospective study⁵⁷. All costs were inflated to 2014 Japanese Yen using the Japanese consumer price index⁴⁸.

Table 7. First and subsequent year costs^{26,48,54–57}.

Event	First year	Subsequent year	CV death
Major IS	¥4,616,921	¥4,307,905	¥1,385,992
Minor IS	¥763,271	—	¥763,271
SE	¥1,169,439	—	¥1,169,439
MI	¥2,764,925	¥1,002,192	¥2,013,281
ICH	¥4,288,283	¥3,338,187	¥1,784,642
Major ECH	¥310,759	—	¥310,759

CV, cardiovascular; ECH, extracranial hemorrhage; ICH, intracranial hemorrhage; IS, ischemic stroke; MI, myocardial infarction; SE, systemic embolism.

Sensitivity analyses

One-way sensitivity analyses were performed by modifying the following inputs by ±10%: proportion of patients with moderate or high-risk of stroke; baseline risk of IS, MI, and ICH; cost per dose of aspirin and VKA; annual physician visit cost of VKAs (well- and poorly-managed), aspirin, and rivaroxaban; and the acute and long-term care costs of major IS, MI, and ICH. The relative risk of IS due to VKA and rivaroxaban was varied based on the 95% confidence interval (CI) identified in the NMA; relative risks for other events were not amended as none were statistically significant for rivaroxaban. Lower and upper values for the cost per dose of rivaroxaban were obtained by assuming that all patients received low- or high-dose of rivaroxaban—costing ¥383.00 and ¥545.60—respectively.

In addition, results for the following scenario analyses were calculated: using pooled data from the ROCKET AF³⁰ and J-ROCKET AF³¹ trials as part of the NMA; decreasing and increasing all acute and long-term care costs of events by 10%; treating only high-risk (rather than moderate-to-high risk) patients with rivaroxaban; discounting costs by 2% and 4% in line with Japanese guidelines for health economic evaluation⁵⁸.

Results

Base case scenario

The model followed a population of 1.37 million Japanese patients with NVAF, with 353,345 at moderate (CHADS₂ = 1) and 867,550 at high (CHADS₂ ≥ 2) risk of stroke, respectively, over a period of 10 years. Of those, 589,879 received no treatment, 45,302 were treated with aspirin alone, and 585,713 were treated with VKAs, of whom 198,877 were poorly managed.

Treating inadequately-managed Japanese NVAF patients with rivaroxaban led to a 60% decrease in the number of non-fatal IS, with 456 081 fewer patients experiencing a non-fatal IS (Figure 3). Treatment with rivaroxaban also resulted in 6622 (34%) fewer non-fatal SE. However, the revised treatment pattern was associated with additional events, notably 15,638 non-fatal ICH, 132,535 non-fatal MI, and 207,523 non-fatal major ECH. Applying event-specific mortality rates to overall events showed that 76,975 CV deaths were avoided by adequately treating NVAF patients, representing a 7% reduction in the number of CV deaths.

Figure 3. Breakdown of events experienced over the analysis time horizon. CV, cardiovascular; ECH, extracranial hemorrhage; ICH, intracranial hemorrhage; IS, ischemic stroke; MI, myocardial infarction; SE, systemic embolism.

Treating inadequately-managed patients with rivaroxaban led to a net spending of ¥7242.8 billion (US $60.1 billion) on drug acquisition, physician visits, and event management over the analysis time horizon. This represented a decrease in costs of ¥188.4 billion (US $1.6 billion) compared to net spending under the treatment pattern before the introduction of NOAC (¥7431.2 billion; US $61.7 billion). Although treating patients with rivaroxaban in early years led to an increase in costs (mainly attributable to higher drug acquisition costs), decreased event costs in the long-term led to cost savings at 3 years that were maintained over the model time horizon (Figure 4). The reduction in event management costs was driven primarily by the avoidance of major IS, and resulted in a considerable reduction in costs of care in the first and subsequent years. Annual cost savings began to present in the third year of the analysis, and a net cumulative saving occurred from year 6 onward (at 5 years, the cumulative budget impact was 1.22%). Annual savings reduced over time due to discounting and reduction in the size of the surviving eligible patient population (Figure 4).

Figure 4. Net budget impact over the analysis time horizon.

Increases in treatment and physician visit costs were completely offset by a 25% reduction in event management costs (¥1822.0 billion; US $15.1 billion), as shown in Figure 5.

Figure 5. Breakdown of net spending over the analysis time horizon.

Sensitivity analyses

The reduction in the risk of IS associated with rivaroxaban was the most influential factor affecting the budget impact of all model inputs (Figure 6). The relative risk of IS for rivaroxaban was the only input value that changed the direction of overall results to an overall cost increase (using the upper bound value). All other inputs maintained the cost savings registered in the base case scenario. The cost per dose of rivaroxaban and cost of IS per subsequent year were the next most impactful inputs, whereas the cost per dose of aspirin and VKA were among the least impactful input values in the model.

Figure 6. One-way sensitivity analysis. ASA, Acetylsalicylic acid; ICH, intracranial hemorrhage; IS, ischemic stroke; MI, myocardial infarction; RR, relative risk; VKA, vitamin K antagonist.

Scenario analyses

All scenario analyses consistently yielded cost savings; the lowest cost saving arose from decreasing all long-term event management costs by 10%, leading to a saving of ¥130.5 billion (US $1.1billion). Indeed, cost savings increased compared to the baseline scenario when only high-risk patients (CHADS₂ ≥ 2) received rivaroxaban (¥693.6 billion; US $5.8 billion). This is consistent with these patients’ higher risk of stroke and the benefit of rivaroxaban in preventing IS and the associated costs. When relative risk data were based on pooled data from the global ROCKET-AF and J-ROCKET AF trials, cost savings were increased compared to the base case (¥316.6 billion; US $2.6 billion, Figure 7).

Figure 7. Budget impact over analysis time horizon, scenario analysis. AF, atrial fibrillation.

Discussion

Using a prevalence-based, deterministic budget impact model, we found that treating inadequately-managed Japanese NVAF patients with rivaroxaban could have a positive clinical and economic impact over the analysis time horizon. Indeed, the revised treatment pattern resulted in a 60% decrease in non-fatal IS, corresponding to a reduction of 186 993 major and 269 088 minor non-fatal IS events. It was also associated with a 34% decrease in non-fatal SE. A 76% increase in non-fatal MI was observed; this was due to an increased relative risk reported in the J-ROCKET-AF study³¹, and thus the NMA upon which risk reductions in CV events in our analysis were based. It should be noted, though, that the increased MI rates from this study contrast with results from the ROCKET-AF study³⁰, in which a non-statistically significant trend towards reduced MI compared to VKA was observed³⁰. An increase in non-fatal ICH and non-fatal major ECH was also observed. Overall, a 7% decrease in the number of CV deaths was observed. These results apply to a modelled population of 1.37 million patients with NVAF over 10 years.

The observed increase in the incidence of non-fatal ICH and non-fatal major ECH is mostly due to the increased use of rivaroxaban in patients currently receiving no treatment, and should be considered in the context of decreased ischemic events and CV mortality. Additionally, clinical trials evaluating novel oral anti-coagulants compared to VKAs have tended to show a reduction in ICH⁵⁹, as confirmed by results from both J-ROCKET AF and ROCKET AF trials^30,31.

Following Japanese NVAF treatment guidelines¹⁸ also had a positive economic impact, with a decrease in net spending on drug acquisition, physician visit, and event management costs of ¥188.4 billion (US $1.6 billion) over the analysis time horizon. These savings were driven by reductions in event management costs (¥1822.0 billion; US $15.1 billion), due in part to the avoidance of major ischemic events and the associated long-term care costs. This suggests that adequately managing NVAF patients may have a positive impact not only on the clinical outcomes of moderate-to-high risk NVAF patients, but also on the costs of managing this condition over patients’ lifetime.

A range of one-way sensitivity and scenario analyses showed that the model was particularly sensitive to inputs affecting the number and costs of IS, which are the most prevalent and costly events in the model, and, therefore, a main driver of costs. The costs of long-term care for IS were particularly impactful, confirming the fact that decreased events associated with rivaroxaban drive the cost savings identified in the model, while physician visit costs and the cost per dose of aspirin and VKA have a limited impact on results. The cost of rivaroxaban was the third most impactful input in the model. Although using the upper limits from the 95% CI of the relative risks reported in the NMA suggested substantial additional costs instead of the savings shown in the base case scenario, it should be noted that these relative risks ranged from 0.44–122.59 across all events, and that only two were significant compared to no treatment (namely, the relative risk of IS associated with VKA and rivaroxaban). Another scenario analysis showed that savings were maintained when only high-risk patients (rather than those with moderate-to-high risk of stroke) received rivaroxaban, illustrating the need for adequate management of patients who otherwise remain at significant risk of life-altering yet preventable events.

We could find no budget impact analyses directly comparable to our own (especially considering the Japanese setting) within the literature. There are, however, numerous economic analyses assessing the cost-effectiveness of rivaroxaban specific to stroke-prevention in AF across a range of countries and settings. Wang et al.⁶⁰ conducted an analysis from the perspective of the Singaporean healthcare system, concluding that rivaroxaban offered a cost-effective treatment option compared to warfarin, and dominated both high- and low-dose treatment with dabigatran. The cost-effectiveness of rivaroxaban compared to warfarin was reflected in analyses from other country settings. A US-based analysis from a payer/Medicare perspective⁶¹ found rivaroxaban to be highly cost-effective compared to warfarin, as did analyses from the German Statutory Health Insurance perspective⁶² and the Belgian healthcare payer (National Institute for Health and Disability Insurance, and patient) perspective⁶³. Morais et al.⁶⁴ conducted an analysis from a Portuguese societal perspective, yielding similar conclusions compared to both warfarin and a mix of comparators based on real-world Portuguese prescribing patterns. Kourlaba et al.⁶⁵ concluded rivaroxaban to be a dominant alternative (both cost saving and improving quality-of-life), from a Greek third-party payer perspective. These studies report rivaroxaban to be cost-effective across a broad range of settings, which, while difficult to place in context of our own analysis (which does not consider quality-of-life), does support the concept of economic offsets achieved through reductions in the rate of CV events through the use of rivaroxaban. However, it is also important to note that the results in our analysis are specific to Japan, and cannot be directly generalized to other countries.

Limitations

Our analysis has several limitations. First, our model assumes that well- and poorly-controlled VKA patients are equally at risk of stroke. This assumption may be conservative, as suggested by the GARFIELD-AF registry data⁶⁶, and thereby under-estimate benefits arising from the increased use of rivaroxaban.

Second, this model does not take medication discontinuation and adherence into account. However, as bleeding rates are reasonably similar in VKA and rivaroxaban³¹, it is unlikely that this omission favors rivaroxaban. In fact, economic benefits of rivaroxaban compared to VKA may even be under-estimated due to discontinuation rates and adherence not being modelled, since rivaroxaban’s simple, one-tablet, once-daily dosing and no need for routine monitoring may lead to fewer discontinuations, as suggested by real-world evidence⁶⁷, and better adherence than with VKA^24,25,68.

Third, the treatment pattern of patients with different stroke risks was based on NVAF patients in 273 Japanese acute care hospitals²⁶. The use of VKA tends to be higher in hospitalized patients than in the general NVAF population. The proportion of well- and poorly-managed patients, however, was based on patients from six institutions and one clinic across five prefectures in Japan²⁸. As a result, our model may over-estimate the percentage of treated patients and under-estimate the proportion of VKA patients who are poorly-managed, as the data may reflect a higher-quality of management than could be expected for the general population. Overall, treatment patterns may vary amongst hospitals, although evidence suggests an over-arching trend towards inadequate treatment^2,26,69,70.

Fourth, the prevalence of AF was based on the community-based Fushimi AF registry, which may not represent the prevalence of AF in Japan. However, the age distribution in the Fushimi area was similar to that of the general population in Japan.

Finally, our analysis required some structural assumptions which must be noted, all of which were applied uniformly to all patients, regardless of the treatment they received, and were, therefore, unlikely to favor any specific treatment. First, no increase in the risk of recurrent event for MI, ICH, SE, major ECH, and IS (beside moderate risk patient transitioning to the high-risk cohort following an IS) was applied. Indeed, the model uses a Markov-based approach which does not track individual patients’ history of events. Second, in the absence of relevant data, it does not differentiate between moderate- and high-risk patients’ baseline risk of experiencing a SE and MI. Also, changes in patient age (i.e. movement of patients to the next age category, and, thus, to a higher baseline risk of CV events) was not simulated over the model’s 10-year time horizon; this assumption is in fact conservative in nature, as the risk of CV events in some patients will be under-estimated. Finally, due to the inherent lack of model ‘memory’ attributable to the Markov-based approach in general, patients who experience an event followed by long-term rehabilitation, followed by an event with no long-term rehabilitation, will see their rehabilitation cost-effectively cancelled (as the model is unable to return them to their previous state automatically); again, this assumption conservatively under-estimates the true costs associated with CV events. Last, this is a simulation to test the potential effect of introducing rivaroxaban in the Japanese healthcare system in light of the current unmet medical need. Although it is unlikely that all inadequately-managed NVAF patients will receive rivaroxaban, it does not critically influence the conclusion of the analysis, which emphasizes the importance of stroke prevention in AF.

Conclusion

Over one million Japanese individuals are estimated to have NVAF^5,27, the majority of which are currently receiving no treatment, aspirin only²⁶, or remain poorly-controlled despite receiving VKA²⁸. Optimal treatment for these patients, many of whom have a CHADS₂ score pertaining to an increased risk of preventable stroke²⁷, represents an issue that Japanese guidelines seek to address by recommending the adequate use of anti-coagulation therapy in NVAF patients¹⁸.

Using the example of rivaroxaban, our model demonstrates that adequate treatment of NVAF patients could avoid a large number of non-fatal IS and CV deaths, the clinical benefit of which may offset concomitant increases in the incidence of non-fatal ICH, MI, and major ECH. Increases in treatment and physician visit costs through the use of rivaroxaban may be offset by savings associated with decreased event management and long-term care costs, such that treating inadequately-managed patients with rivaroxaban could be associated with an overall decrease in costs over the analysis time horizon. In view of the positive clinical and economic impact demonstrated in this model, novel oral anti-coagulants should be considered as a viable option by payers and clinicians in the management of moderate-to-high risk NVAF patients, especially among inadequately-managed patients.

Transparency

Declaration of funding

This study was funded by Bayer Pharma AG.

Declaration of financial/other relationships

SI, KO, SM, and YK are paid consultants to Bayer Yakuhin, Ltd. CM was an employee of Covance at the time of the analysis. LR is an employee of Covance. EWF and BR are employees of Bayer Yakuhin, Ltd. TE and JBB are employees of Bayer Pharma AG. JME peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Notes

*ADLs consist of self-care tasks one normally performs in daily living, such as feeding oneself, bathing, and dressing, and are often used as part of functional assessments in the elderly population.

*Xarelto is a registered trademark of Bayer Pharmaceuticals, Inc., Berlin, Germany

*DPC is a fixed reimbursement system per hospitalization day based on diagnostic categories.