FormalPara Key Points

The COMPASS trial demonstrated that rivaroxaban 2.5 mg twice daily in combination with aspirin 100 mg once daily was more effective than aspirin 100 mg once daily in preventing CV death, stroke or MI in patients with CAD or PAD.

In our analysis, rivaroxaban plus aspirin was cost-effective (high value) in the USA.

 

1 Introduction

Cardiovascular disease continues to be a growing problem, impacting hundreds of millions of people globally, and even with secondary prevention strategies, some patients with cardiovascular disease can experience recurrent cardiovascular events each year [1]. In the USA the economic burden of this has been estimated to be over half a trillion dollars in direct and indirect costs [2]. In the Cardiovascular Outcomes for People Using Anticoagulation Strategies (COMPASS) trial, rivaroxaban 2.5 mg orally (BID) plus aspirin 100 mg daily was shown to be clinically superior to aspirin alone for reducing the risk of cardiovascular (CV) death, stroke or myocardial infarction (MI) in participants with either stable coronary artery disease (CAD) or peripheral artery disease (PAD); rivaroxaban 5 mg twice daily alone was not beneficial [3]. In light of the strong clinical benefits of rivaroxaban 2.5 mg BID plus acetylsalicylic acid (ASA), an understanding of the cost implications and cost-effectiveness of the strategy would be useful for healthcare providers and decision-makers. In this analysis, we report our findings on the cost implications and cost-effectiveness of the COMPASS trial from a US Medicare perspective using a methodology we have used prior for Canada, France and Germany [4].

2 Methods

2.1 Clinical Trial

The COMPASS trial set out to compare rivaroxaban 2.5 mg BID plus aspirin 100 mg once daily (OD), rivaroxaban 5 mg (alone) BID and aspirin 100 mg alone OD. Participants in the trial were also randomized to receive either pantoprazole 40 mg OD or a matching placebo [3, 5]. To be eligible for the trial, participants had to be younger than 65 years with coronary artery disease as well as documentation of atherosclerosis involving at least two vascular beds or at least two additional risk factors. The COMPASS trial ended early following strong positive results after review of interim data by an independent data and safety monitoring board.

In the COMPASS trial the primary outcome (a composite of CV death, stroke or MI) was reduced by 24% for rivaroxaban plus aspirin participants compared with participants receiving aspirin alone; the secondary composite outcome of ischemic stroke, MI, acute limb ischemia or CV death was also reduced by approximately 27%. However, these benefits were offset by more major bleeding events (Supplementary Table S1). Both the rivaroxaban 5 mg alone BID arm as well as the addition of pantoprazole were not included in our analysis, as neither showed any clinical benefits compared with their respective placebos.

2.2 Economic Analysis Design

We looked at the costs and outcomes associated with rivaroxaban plus aspirin compared with aspirin alone in a two-part analysis. An in-trial cost–impact analysis was performed for the first part of our analysis. For the second, we performed a lifetime cost-effectiveness analysis by looking at clinical outcomes during the trial and then projecting these outcomes up to a further 33 years (mean age of 100 years). All costs are reported in 2017 USD. This is a similar approach to one we used in our analysis of the COMPASS trial in Canada, France and Germany [4]. To restate our approach, this analysis was performed assuming intention to treat and only considered direct healthcare system costs. Costs related to outpatient visits, investigations (e.g. outpatient imaging or blood work) or indirect costs such as productivity costs were not included. While a societal perspective may be preferred for economic analyses, other perspectives are acceptable and, in some cases, could even be considered more appropriate [6].

We hypothesized that rivaroxaban 2.5 mg BID plus aspirin 100 mg OD would be cost-saving or cost-neutral compared with aspirin 100 mg alone. We considered the intervention to be cost-neutral if the incremental cost was between a range that was ±5% of the total cost of aspirin alone as per our previously published definition [7,8,9]. If the incremental cost was below this range, it was considered cost-saving and more expensive if it was above. If more expensive, we will test for high value, which we defined as using a willingness to pay of $50,000 per QALY as suggested in American College of Cardiology (ACC)/American Heart Association (AHA) guidelines [10].

2.3 Statistical Analysis: In-Trial Analysis

Unit costs for events, procedures and study medication were collected for the USA and applied to all randomized participants in the COMPASS trial. Data from the trial case report forms (CRFs) were used to determine the frequency of events and procedures that occurred during the trial. We included CV events and procedures and bleeding events captured in the COMPASS CRFs. While captured in the CRFs, other events of a non-CV nature were not considered, as they had a similar frequency of occurrence in both groups and were unrelated to the use of rivaroxaban and/or aspirin. Every instance of an event or procedure captured by the study CRFs was counted, not just those used for the primary outcome.

In our analysis we used a price of $6.06/day for rivaroxaban obtained from a published paper which factored in discounts, rebates, concessions to wholesalers and distributors and patient assistance programs [11]. US costs were from study leaders who provided costs based on 2015 Medicare Provider Analysis and Review (MEDPAR) data. The cost for each event was derived from the weighted average of the relevant diagnosis-related groups and inflated to 2017 USD using consumer price index data from the US Bureau of Labour Statistics [12].

While most of our costs for events and procedures reflect resource utilization during the acute hospitalization, the costs used for strokes and limb amputations were an exception. To better represent the greater amount of care and, by extension, resource utilization that these conditions require, a 1-year cost was used for these events [7, 8, 13].

2.4 Lifetime Cost-effectiveness Analysis

The results of the in-trial analysis were projected up to an additional 33 years using a two-state, patient-level, microsimulation Markov Model (Supplementary Figure S1), the details of which have been previously published [14, 15]. Rather than begin our projections following the complete duration captured in the COMPASS database, we chose to begin and base our projections on data 30 months (2.5 years) following randomization. This decision allowed us to use more stable data for our projections. COMPASS participants began in the “alive” state if they were alive at the end of the COMPASS trial. For each Markov cycle (1 year), a patient’s probability of dying or remaining alive was based on annual mortality probabilities from the 2013 US life tables [16]. These probabilities were age- and gender-specific and were considered an individual patient’s base probability of death. A patient’s life expectancy was projected each cycle in our model using a half cycle correction [17].

The individualized effects of MIs, ischemic strokes and International Society on Thrombosis and Haemostasis major bleeds that occurred during the trial [3] as well as continued use of study medication were modelled with time-dependent Cox regression models, and when necessary they were multiplied by a patient’s base probability of death to yield an individualized probability of death that accounts for a participant’s experience during the trial.

Compliance during our projected period was based on the compliance at the end of the COMPASS trial (83.5% for the rivaroxaban plus aspirin group and 84.3% for aspirin alone). Simulations were run 200 times for each patient and the mean of these replicates was used. The number of life years was determined using restricted meant survival times (RMST) [18, 19]. Costs were estimated for the projected period by first determining the annual mean cost for each participant during the trial and then multiplying these costs by the number of years of post-trial survival to yield the total cost of each participant. The average cost per participant for each treatment group was then determined by taking the sum of the total cost of all participants and dividing by the sample size of the treatment group. A discount rate of 3% per annum was used for both costs and effects [20].

2.5 Utilities and Quality of Life

European Quality of Life 5 Dimensions 3 Level Version (EQ-5D-3L) questionnaires were administered at baseline as well as the final visit during the COMPASS trial. Results were then scored using a US population value set to obtain utility values and the mean of a participant’s baseline and final visit was used to calculate quality adjusted life years (QALYs) [21]. If a participant was missing both baseline and final visit responses, a linear regression model, informed using data from the COMPASS trial, was used to impute. This model considered factors such as age, gender, race, medical history (diabetes, hypertension, heart failure and PAD) and MIs and ischemic strokes that occurred during the trial as well as their timing. The continued use of rivaroxaban plus aspirin was not included in our imputation regression model, as it was not significantly related to utility. The derived utilities were then applied to survival data from our model to determine QALYs which were obtained through RMST. Bootstrapping (1000 samples) was used for determination of life years and QALYs gained as well as their corresponding 95% confidence intervals (CIs).

2.6 Subgroup Analyses

Subgroup analyses were performed on several of the subgroups reported by the COMPASS trial. For these analyses all outcomes were determined in a similar manner as the base case; however, the adjusted survival and utility data for the overall study group were utilized because of the subgroups having a much smaller sample size.

2.7 Scenario and Sensitivity Analyses

The cost and compliance of rivaroxaban are key variables in this economic analysis. With that principle in mind, we create a sensitivity analysis that looked at a range of costs for rivaroxaban to better represent the variety of rivaroxaban pricing in the USA. We also incorporated a sensitivity analysis of the annual discontinuation rates using 1% or 2.7% in addition to our base case, as well as 100% compliance throughout our projected period.

Lastly, we expect a generic version of rivaroxaban to be made available in the USA in the next few years. To better understand the impact of this, we reduced the cost of rivaroxaban in our model by 75% after 7.5 years from the beginning of our analysis.

3 Results

3.1 In-Trial and Lifetime Costs

In the USA rivaroxaban plus aspirin was associated with lower mean costs per participant for events and procedures compared with aspirin alone (−$683; Table 1) due to fewer CV events and procedures during the trial (Supplementary Table S2). However, the addition of the acquisition cost of rivaroxaban ($3962 per participant) absorbed all these savings, resulting in a total mean difference in cost of $3253 per participant after the entire in-trial period.

Table 1: Mean in-trial costs per patient in USD
Full size table

Similar to the in-trial period, following a lifetime period the cost per participant for hospitalized events and procedures were lower for participants in the rivaroxaban plus aspirin group (−$8581 per participant). As with the in-trial period, adding the cost of rivaroxaban resulted in higher overall costs: $27,255 per participant (95% CI: $27,347 to $28, 566; Table 2).

Table 2 Lifetime cost-effectiveness for patients continuing rivaroxaban and aspirin (in USD)
Full size table

3.2 Life Years Gained and QALYs

Utility scores in the COMPASS trial were the same between both groups (0.87) at baseline as well as at the end of the trial. During the in-trial period rivaroxaban plus aspirin participant gained 15.5 days (0.042 QALYs) more compared with ASA alone. In the projected period rivaroxaban plus aspirin resulted in 411.7 days (1.128 QALYs) gained after adjusting for in-trial MIs, strokes, major bleeds and continuation of the study regimen. Combining the in-trial and projected periods yielded a total gain of 1.17 QALYs per participant (95% CI: 0.98–1.36) in favour of rivaroxaban plus aspirin.

3.3 Lifetime Cost-Effectiveness Analysis

Using an incremental QALY of 1.17 (95% CI: 0.98 to 1.36) gave us an incremental cost-effectiveness ratio (ICER) of $23,295 per QALY gained. From our bootstrap replicates, 90% of these samples were below a threshold of $25,000 per QALY (Fig. 1). PAD and polyvascular disease subgroups were the most attractive of all subgroups analysed ($15,438 and $9419 per QALY respectively)

Fig. 1
figure 1

Joint distribution of incremental costs and projected incremental QALYs. Joint distribution of incremental costs and projected incremental QALYs for rivaroxaban plus ASA versus ASA is based on 1000 bootstrap samples. The red dots represent the point estimate. The purple line represents a cost-effectiveness threshold of $25,000/QALY, and the green line represents a cost-effectiveness threshold of $50,000/QALY

Full size image

3.4 Sensitivity Analyses

The cost and compliance of the study regimen are key elements in this economic analysis, as the savings from events and procedures are outweighed by the cost of rivaroxaban. We have conducted a sensitivity analysis that is based not only on a range of prices for rivaroxaban but a variety of patient’s compliance as well. Our base case used $6.06/day as the price for rivaroxaban which included discounts, rebates, concessions to wholesalers and distributors and patient assistance programs, which was obtained from a published report [11]. We performed a sensitivity analysis using $8.00/day as advertised in some drugstores that gave an ICER of $37,709 per QALY gained but also $13.67 per day as distributed by Bayer with an ICER of $68,284 per QALY gained (Fig. 2).

Fig. 2
figure 2

Daily drug cost versus incremental cost OR incremental cost-effectiveness. Daily cost of rivaroxaban versus total lifetime cost (primary y-axis) and lifetime incremental cost/QALY (secondary y-axis). A is the daily cost of rivaroxaban in scenario analysis, B is the daily cost of rivaroxaban in base case, C is the daily cost of rivaroxaban with coupon at certain pharmacies and D is the daily cost of rivaroxaban based on list price of $13.67. Linear equations: Base Case − y = 4608.8x − 4641.1; 1% Annual Discontinuation − y = 5005.9x − 5267.9; 2.7% Annual Discontinuation − y = 5771.1x − 6460.9; 100% Compliance − y = 4099.8x − 3812.1

Full size image

For our base case analysis, we adopted the same compliance (permanent discontinuation) rate as the clinical trial (16.5% and 15.7% for rivaroxaban plus aspirin and aspirin alone, respectively). Increasing this by an additional 1% or 2.7% per annum increased the ICER to $25,075 and $28,518 per QALY gained respectively. Assuming 100% compliance after the trial, the QALYs gained increased to 1.33, and the ICER decreased to $23,125 per QALY gained (Table 3 and Fig. 2). Discounting had minimal impact on our findings, and as expected using a lower discounting rate led to better ICERs.

Table 3 Sensitivity analysis and lifetime scenario analysis results
Full size table

3.5 Scenario Analysis

It is expected that rivaroxaban will be generic within a few years. Our scenario analysis considered a generic version of rivaroxaban entering the market during the fifth year of our projections would lower lifetime costs from $71,397 to $50,583 in the USA (Table 3). As a result of this reduction, rivaroxaban plus aspirin is considered very attractive at an additional cost per participant of only $5506 per QALY gained.

4 Discussion

Rivaroxaban 2.5 mg BID plus aspirin was shown to decrease the incidence of major adverse cardiovascular events in COMPASS participants with stable atherosclerotic vascular disease. In this analysis we show that, while rivaroxaban plus aspirin was not cost-saving or cost-neutral in the USA, it was of high value due to its ICER of $23,295/QALY gained. This is substantially higher than our previous findings in Canada, France and Germany, where we reported ICERs of $4231, $10,005 and $10,174, respectively [4], and can be directly attributed to the higher cost of rivaroxaban in the USA, which is roughly double the cost compared with Canada, France or Germany. The daily rivaroxaban cost in this analysis was $6.06 USD compared with $2.48 USD in Canada, $3.24 USD in France and $3.22 USD in Germany.

The incremental cost-effectiveness ratio or cost per QALY represents a matrix of comparisons of a decision-maker’s willingness to pay for an improvement in quality-adjusted survival. In most countries these thresholds are not well established; however, in the UK the National Institute for Health and Care Excellence (NICE) has often supported a threshold of £20,000/QALY as cost-effective, and in the USA the ACC/AHA have suggested an ICER less than $50,000 is of “high value” and ICERs less than $100,000 are often considered cost-effective [10, 20, 22]. In the USA there are some health economists and politicians who do not approve of the use of QALYs, as it could be considered prejudicial against certain vulnerable populations [23]. One alternative that has been proposed is the equal value life year gained (evLYG) [24]. The evLYG assumes that all life years gained are at a utility value of 0.851. As this outcome is relatively new, its precise methodology is a bit unclear [25]; however, in the COMPASS trial a mean utility of 0.87 was found for both groups at both baseline and the end of the trial. As this is only slightly more than what is recommended for evLYG calculations, we feel the ICERs we have presented here could also be considered reasonable approximations for evLYG for those who prefer this approach.

The acquisition cost of the new intervention is usually the main driver in these ratios. The healthcare system in the USA is decentralized, and the price of new medicines varies widely. The wholesale acquisition price used in the USA included discounts from the manufacturer that are normally negotiated with payers and healthcare organizations. This is an important consideration, as without these rebates the ICER would be substantially less appealing. However, generic pricing of rivaroxaban will be available in a few years and will result in much lower ICERs in the USA, as confirmed in our scenario analysis.

In our base case analysis, we used a discount rate of 3% for costs and effects, as suggested by the Second Panel on Cost-Effectiveness in Health and Medicine [20]. As an additional sensitivity analysis, we also used a discounting rate of 1% for effects while maintaining the same rate for costs. This avoids penalizing health benefits that occur in the future too harshly and is in line with work published by Paulden and Claxton [26].

It is important to acknowledge the limitations of our analysis. As with many cost analyses there could be some variability with the unit costs we used. However, the cost inputs in our model were provided by local experts using publicly available government data. In addition, we focused our efforts on costs that could be directly attributed to hospitalizations; out of hospitals costs, caregiver costs or indirect costs such as productivity costs were not included in our analysis, as they were not recorded in the COMPASS trial. Since there were significantly fewer events and procedures in the rivaroxaban plus aspirin arm, we feel our approach is conservative. The model we have developed is also only comprised of two states and as such it cannot model the impact additional CV events would have on mortality and utilities. However, as our costs are projected separately based on data during the trial period, we are still able to estimate an aggregate cost for events and procedures post-trial. Furthermore, the probabilities we have used to represent the background mortality in our model are from US lifetables. As these represent the general population and not the disease population, the mortality probabilities we have used are likely lower than what is actually experienced by the target population. In a complementary study [14], after the addition of estimated post-trial CV events using 20-month risks from the REACH registry [27], it was demonstrated that our estimated gains in life years, relying on US life tables, were more conservative. In addition, as this applied to both treatment arms, we do not believe it was biased towards any one particular group. Projections of each patient’s utility values were based on data collected during the trial, and the projected utilities we use in our analysis may be higher than expected, as documenting changes in utility as a result of CV events was not one of the focuses of the COMPASS trial.

There have been other economic evaluations that have looked at the COMPASS trial or rivaroxaban 2.5 mg in patients with CAD or PAD more generally. A group from Australia has published two manuscripts regarding the cost-effectiveness of rivaroxaban plus aspirin versus aspirin alone. The first of these was focused on patients with stable CV disease and reported an ICER of approximately $25,931 USD per QALY gained after 20 years [28]. The second was in patients with peripheral or carotid artery disease and reported that rivaroxaban plus aspirin resulted in $22,081 USD per QALY gained over a lifetime [29]. In a Markov model published by Bayer, which was also used in its application to the NICE, it was reported that rivaroxaban plus aspirin in the UK resulted in an ICER of $24,082 per QALY over a lifetime horizon [30]. This model was also recently adapted for the Netherlands and Italy and reported discounted ICERs of $14,199 USD and $19,496 USD, respectively [31, 32]. An analysis from the Institute for Clinical and Economic Review in the USA found that, over a lifetime, rivaroxaban plus aspirin resulted in an incremental cost per QALY of $36,000 per QALY [11]. In eastern countries, results are varied. In China, rivaroxaban plus aspirin was associated with an ICER of $7932 USD at the end of 13 modelled years [33]. Conversely, an analysis in Taiwan found that after 30 years rivaroxaban plus aspirin had an ICER of $83,459 USD per QALY in patients with stable cardiovascular disease [34]. All these analyses utilized a more involved Markov model that used five events at most and depended largely on data from the published COMPASS trial paper as well as external sources to obtained other inputs not captured during the clinical trial. This method is less specific to the COMPASS trial itself and more about the application of a class of medicine on a general disease population for healthcare planners. The approach we have adopted is different in many ways. Our model is a simpler two-state Markov model similar to a previously published economic analysis [35]. We also have direct access to the COMPASS database and made full use of it. As stated earlier, the COMPASS trial was halted by the DMSB on 6 February 2017 following strong evidence of the overwhelming positive efficacy of rivaroxaban plus aspirin. The data reviewed by the DSMB was also used in the published analysis of the clinical trial [3]. However, our analysis used the COMPASS database after it was officially closed a few months later, as all participants were still required to have a final visit and be officially taken off the study medication. A large number of events were collected at this time that were not included in the clinical manuscript. As a result of these additional data, all-cause mortality increased from 3.4% to 4.1% for the rivaroxaban plus aspirin group compared with 4.1–4.9% for aspirin alone (Supplementary Table S3). As we had direct access to the COMPASS database, we were able to include data from participants who had more than one event, and by extension, the higher costs these events would generate (Supplementary Table S4). We were also able to obtain quality of life data from the trial directly instead of estimating it from other publications. Our model, as with the other published works on rivaroxaban, depends largely on all-cause mortality and the treatment effect. A lower all-cause mortality rate reduces the difference between both treatment groups and in turn reduces the life years gained by rivaroxaban.

We chose to project our results up to 33 years since the anticipated survival of a cohort of 68 years, such as in COMPASS, would be 16 years. As this is a mean value, half of the participants in our model will still be alive after 16 years. To not ignore this group of participants, we extended our projections up to 33 years when almost all participants would be deceased. There is no evidence to suggest that the efficacy of rivaroxaban plus aspirin would decrease over time. This is supported by the fact that in the COMPASS trial rivaroxban plus aspirin had a constant effect regardless of age at the beginning of the trial. The incremental life years gained each year in our projected period peaked around 15–20 years and began to decline in the final 10 years, as there are fewer participants left alive (Supplementary Figure S2).

Finally, we have displayed the total incremental costs and corresponding ICER outcomes of our base case and sensitivity analyses in a figure that conveniently displays our findings across a range of daily rivaroxaban costs to better assist decision-makers and administrators in identifying what their local costs would look like using our approach.

The current US administration has recently announced that Medicare would begin negotiations with pharmaceutical companies to lower drug prices [36]. Rivaroxaban was one of the drugs that was selected for the first wave of these negotiations, and it is anticipated that a lower price will go into effect in 2026. As the outcome of these negotiations is unknown at the time of writing, the linear regression model we have presented (Fig. 2) can provide some insight into how this new price might impact our findings. However, once negotiations have concluded, we look forward to reviewing our findings using the new costs available.

In conclusion, our analysis is a better reflection of mortality, resource utilization, compliance and quality of life in the COMPASS trial, as we obtained these data directly from the final study database; no unnecessary external assumptions were used in our analysis. We also conducted sensitivity analyses on two of the key variables in our analysis: the acquisition cost of rivaroxaban and participant compliance. For these reasons, we believe our analysis is more representative of the realities of the COMPASS trial.

5 Conclusion

Rivaroxaban 2.5 mg BID plus aspirin alone is cost-effective (high value) in the USA at $23,295 per QALY gained.