A Cost Effectiveness Analysis of Rivaroxaban Compared to Warfarin for Deep Vein Thrombosis (DVT) Treatment in Ethiopia

Introduction

Background

Deep vein thrombosis (DVT) and pulmonary embolism (PE) are known by the collective name venous thromboembolism.¹ DVT is the third most common cardiovascular disorder in the world, which affects 2 million people every year in the United States. Its overall prevalence rate is generally 1–2 cases per 1000 population in the western world, and from these 0.5 cases per 1000 population develop PE.^1,2 The disease is also prevalent in Africa including Ethiopia, besides limited studies that showed its epidemiology.³

If left untreated, DVT is a life-threatening disorder, and will became a major problem when the clot formed in the invisible deep veins reaches the lung artery. In one study, previous history of DVT/PE (25%), surgery (20%), trauma (12%), cancer (4–6%) and immobility (8%) were the main risk factors while 25–50% of cases were idiopathic.^3,4 Bleeding and PE due to medications are the major complications of DVT.⁴

Alleviating the clot, promoting cure and preventing complications are the main objectives of DVT treatment. Unfractionated heparin (UFH) with warfarin is still the standard first line treatment in Ethiopia. Rivaroxaban provides a simple, fixed-dose regimen for treating DVT and for continued treatment, without the need for laboratory monitoring.⁵ Warfarin based therapy is less costly, less effective, with slower effect, had many food and drug interactions, require regular monitoring, and had severe bleeding, but the reverse is true for rivaroxaban^6,7 and there are no previous cost effectiveness (CE) studies in the Ethiopian setting comparing the two strategies. These differences gave us initiation to do a CE analysis in order to compare the economic and clinical outcomes of rivaroxaban with the current standard (UFH with warfarin) by constructing a Markov model in a cohort of individuals following an initial DVT diagnosis. This study aimed to assess the CE of rivaroxaban compared to UFH/Enoxaparin with Warfarin for treatment of DVT in patients aged 40 from the restricted societal perspective in Ethiopia.

Methods

Study population

The target population was hypothetical adult DVT patients with no contraindication, comorbid disease or concomitant therapy at age of 40 Years (the age of high prevalence and sex distribution in Ethiopia).³

Perspective and audience

The analysis was conducted from the restricted societal perspective. Direct medical cost, patient cost, and costs covered by government were included. The reason behind conducting such analysis was to assist decision makers (MOH and Health institutions) to choose the best anticoagulant strategy for DVT patients.

Comparators

Two possible alternatives, either rivaroxaban (newly introduced) 15 mg PO bid for three weeks and then 20 mg PO per day for 1 year or a standard strategy (first line), UFH/Enoxaparin for 5–7 days plus Warfarin 5 mg PO per day for one year was considered. The international normalized ratio monitoring for warfarin is assumed to be done every other day for one week, every week for three weeks, and every two weeks for one month, then every month for 1 year was considered.

Time horizon

The clinical outcomes and economic costs of DVT are difficult to determine early. So, the time horizon of analysis was used with 1-year cycle length for 24 years (until life expectancy of a healthy Ethiopian).

Discount rate

Discounting was necessary since the study had time horizon of lifetime, both costs and benefits were discounted with global discount rate, 3%.

Choice of outcomes measures

Since DVT affects QOL and mortality of patients, health outcomes in terms of effectiveness, cost, QALY and incremental cost effectiveness ratio (ICER), was done.

Measurement of effectiveness

The primary measure of treatment effectiveness was QALYs gained. Clinical trials, pooled analysis and other articles from PubMed were used as a source of data to measure QALY. QALY is common measure of effectiveness that takes into account the quantity of years lived adjusted to its quality.⁸

Resources and costs

Currency, price date and conversion

The cost data were collected in Ethiopian Birr and converted to United States Dollar (US$) with the exchange rate of 1US$/39.2652v Ethiopian birr on January 5, 2021 from National Bank of Ethiopia.

Cost

Since our perspective is restricted societal, all direct medical costs associated with each treatment like cost of medications, hospitalization, laboratory (monitoring, and diagnostic if complication), professional service, one-time treatment of cost of complication and side effects were included.

The sites of data collection were Tikur anbessa specialized hospital (TASH). The prices of the medicines (UFH/warfarin and Rivaroxaban) were obtained from TASH database and Ethiopian Pharmaceutical supply agency (EPSA) price list. The cost of hospitalization, laboratory, complications and side effects were derived from TASH which is the main cardiology center in Ethiopia. Professional service cost was estimated based on daily salary of each professional per the number of patients served per day. The average hospital stay was obtained based on expert’s opinion and the laboratory cost for regular monitoring and diagnostic (complications) were collected from TASH laboratory department. Cost of blood was estimated from Ethiopian Red Cross Association.

To estimate the total costs for treating DVT per cycle length, the costs of the drug after hospitalization and regular laboratory monitoring costs were added to cost of professional service, and complications. The total cost of treating DVT in warfarin and rivaroxaban arm per patient per year were estimated before discounting. The detailed calculation of each component of cost was explained according to the practice at TASH.

The cost of follow up is the cost incurred during the follow up for 1 year. It includes drug cost, monitoring laboratory cost, and service cost. The cost of bleeding is also high which includes the cost of blood and other supplementary therapies during bleeding management.

Model overview

A Markov model was designed to follow the two identical cohorts of hypothetical DVT patients. Each patient was presented with DVT diagnosis without any complication or comorbidity. The patients were followed for 24 years starting from treatment initiation. In each cycle of the treatment/follow-up, different health states of the disease including the treatments will occur according to the known history of disease and drugs. From these, recurrent DVT with or without PE and bleeding is one.

Each treatment option might result in one of the health states, No DVT, DVT, death or intermediate states (PE and Bleeding). Patients may transit to death from any health state. Bleeding and other complications increase the risk of death in the model. These health states were chosen in this model, as they were the most common complications reported in the clinical trials and are highly expensive to treat. Model creation and analysis were performed by using Tree-Age – 2018 software and the Markov model structure is depicted in Figure 1.

Figure 1 State diagram for the economic model.

Key assumptions

Several assumptions were made in the design of the model:

1. The two hypothetical cohorts had uniform case distribution and condition with ENSTIEN-DVT/PE clinical trials. Each group received either rivaroxaban or Standard therapy only. No patients in the sample population were contraindicated to the treatments, have comorbid diseases or not take concomitant medication that will affect the efficacy of the two strategies because these conditions create dilemma in the outcome of the model.
2. Costs of minor side effects and complications was similar across both treatments and for major complications and adverse events (bleeding and PE), average costs of one-time total treatment/hospitalization cost for that event were considered.
3. All patients at first and during complications were considered as admitted because the disease is severe, acute and asymptomatic at first.
4. We assumed there will be no lost from treatment. Adherence rates for both treatment alternatives were assumed similar. If this is not true, the difference in the outcome may be due to adherence or left from treatment.
5. Generic drug was used for the treatment and costing, since the cost of brand drugs is very expensive.
6. Patients develop the adverse event or complication once through one cycle period.
7. Willingness to pay (WTP) was set to be US$783 in Ethiopia.

Base case values

The inputs of this analysis were mainly obtained from the ENSTIEN-DVT/PE clinical trial and other supplementary reviews and literatures.^5,9

Probabilities

The probabilities of adverse events and complications were based on data from the ENSTIEN-DVT/PE clinical trial.^5,9 Mortality rates of DVT, PE and bleeding were estimated from previous literature (Table 1). Probability of No DVT is similar between patients that develop bleeding and non-bleeding and PE does not affect the probability of bleeding.¹⁰

Table 1 Base-Case Probability Values and Ranges, Base-Case Utility Values and Ranges, and Cost of Treatment Ranges Used in Sensitivity Analysis (95% CI or ± 20%)

Utility

The type of anticoagulation therapy may affect patient’s quality of life or utility. Rivaroxaban has advantages over warfarin as the patients do not require continuous monitoring or food restriction and had fewer drug-drug interactions.

In our study, the mean utility of patients on warfarin therapy was used in the model. The utility of patients on rivaroxaban was greater than warfarin, because rivaroxaban does not require routine monitoring. By definition, utility of dead patients is zero. Utilities of patients experiencing each health state were obtained from literature (Table 1).

Efficacy/Effectiveness

The effectiveness of the two treatment strategies/drugs were obtained from a medical literature.¹¹ (Table 2).

Table 2 Base-Case LOS and Effectiveness (95% CI) Values and Ranges for Analysis

Hospitalization stay

Data on hospitalization and length of stay was reported from EINSTEIN-DVT/PE^5,9 but due to variations between settings, local experts at TASH were consulted (Table 2).

Sensitivity analysis

Several sensitivity analyses were performed to test the uncertainty of the estimate values, model assumptions, and how it affects the result of the outcome.

One-way sensitivity analysis was performed on highly influential variables which include utility of No DVT, effectiveness and cost of rivaroxaban over plausible ranges presented in the above tables. The values of ranges were obtained from previous literatures by using 95% CI if available, or by calculating a range of 20% in each direction.

In addition, two-way sensitivity analysis was performed between cost of warfarin and cost of rivaroxaban. Two-way sensitivity analysis allows us to demonstrate impact of the two variables when changing their values simultaneously within given ranges.

Scenario analysis

A scenario analyses was also conducted, focusing on cost of warfarin monitoring laboratory. This variable was selected because international normalized ratio monitoring is done for warfarin therapy only and we want to know what will happen if cost of monitoring was not included, ie the influence of this cost on the overall CE.

Calibration/Validation

Before analyzing the results of this model, it was ensured that the results were logically believable and could be explained accurately (face validity). This was done by consulting experts and TASH cardiology residents. The model was also assessed for logical inconsistencies by evaluating it under hypothetical conditions. The mathematical calculations were confirmed to be accurate and consistent with the specifications of the model (internal validity).

Operational definitions

Cost: direct medical cost of treatment, monitoring, hospitalization, and service.

Effectiveness: the capacity of producing the desired outcome/result at the end of treatment.

Length of stay: duration of time in which a certain health state remains.

No DVT: when patients become free from DVT.

Standard therapy: represents UFH/Warfarin treatment.

Warfarin: in the document, warfarin represents standard therapy.

Results

Base case analysis

In the base case analysis, rivaroxaban therapy resulted in higher QALYs with a value of 16.78, while warfarin-based treatment resulted in 16.34 QALYs (Table 3). Total lifetime costs were $988.58 for rivaroxaban and $932.92 for UFH/warfarin. Therefore, rivaroxaban resulted in a gain of 0.443 QALYS at an additional cost of $55.661. Rivaroxaban resulted in $58.91 per QALY while warfarin resulted $57.10 only, which is acceptable cost increase per QALY. The ICERs for rivaroxaban compared with warfarin was $125.68 per QALY saved which is less than willingness to pay threshold of $783 per QALY saved. Warfarin resulted in a NMB of $11,859.72 while that of rivaroxaban is $12,150.82 meaning rivaroxaban is cost-effective. Therefore, for DVT treatment in the Ethiopian setting, rivaroxaban is considered the more cost-effective choice (Figure 2).

Table 3 Base Case Cost Effectiveness Table

Figure 2 Base case cost effectiveness graph.

One-way sensitivity analyses

From the tornado diagram (Figure 3) with the results of a series of one-way sensitivity analyses, the variables with the most influential impact were, Utility of No DVT, Effectiveness of rivaroxaban and Cost of rivaroxaban, respectively. Based on the tornado diagram, one-way sensitivity analyses were done for these influential variables with the higher impact on the model over plausible ranges.

Figure 3 Tornado diagram that demonstrates influence of each variable on the base case results.

Varying the utilities of No DVT had an impact on NMB value of both strategies. If the value for utility of No DVT changed from 0.8 through one, NMB value for warfarin and rivaroxaban increases simultaneously but the increase in rivaroxaban is greater throughout as it is the cost-effective strategy (Table 4). NMB graph enables us to identify exactly the threshold in which below/above it the result will change. The analysis values showed that at all values of the utility of No DVT in the plausible range, rivaroxaban is a cost-effective strategy (Figure 4).

Table 4 Net Monetary Benefit Table for Sensitivity Analysis of Utility of No DVT, Cost of Rivaroxaban and Effectiveness of Rivaroxaban

Figure 4 Net benefit graph for one-way sensitivity analysis of utility of No DVT.

A one-way sensitivity analysis was also conducted on cost of rivaroxaban. Varying values of cost of rivaroxaban on plausible ranges influence the ICER values compared to warfarin therapy from the base case analysis. Below the base case value, warfarin was dominated but becomes undominated when cost of rivaroxaban increases from the base case. NMB graph for one-way sensitivity analysis of cost of rivaroxaban showed that if the cost of rivaroxaban increased, NMB of rivaroxaban decreased and that of warfarin was unaffected (Figure 5). At all points in the plausible range under WTP, rivaroxaban resulted in greater NMB, which confirms it is the cost-effective choice (Table 4).

Figure 5 Net benefit graph for one-way sensitivity analysis on cost of rivaroxaban.

Another one-way sensitivity analysis was performed for the effectiveness of rivaroxaban and the result indicates that it had an impact on ICER value. As effectiveness increases, the ICER value decreases and at maximum effectiveness, warfarin becomes dominated. When the effectiveness of rivaroxaban reaches 0.9784, the NMB of the two strategies is equal and for higher effectiveness values from the threshold (Table 5), warfarin is more cost effective (Figure 6). The NMB table also shows that the NMB of rivaroxaban increases through the range and after threshold effectiveness, its value becomes greater than warfarin (Table 6).

Table 5 Threshold Analysis for Sensitivity Analysis of Effectiveness of Rivaroxaban

Table 6 Cost Effectiveness Table for Scenario Analysis

Figure 6 Net benefit graph for one-way sensitivity analysis on effectiveness of rivaroxaban.

Two-way sensitivity analysis

The two-way sensitivity analysis for the cost of warfarin and cost of rivaroxaban indicated that warfarin treatment would have a chance of being cost effective when cost of warfarin is less than $1342.647 and cost of rivaroxaban is greater than $1943.854 otherwise, rivaroxaban is the cost-effective choice. These two points are the approximate thresholds taken from the graph of the two cost values that determine cost effective choice (Figure 7).

Figure 7 Net benefit graph for two-way sensitivity analysis on cost of warfarin and rivaroxaban.

Scenario analysis

Scenario 1: If cost of monitoring for warfarin was not considered

In this scenario, we go through to check what will happen if cost of monitoring for warfarin was missed and the results are presented in Table 6. The yearly monitoring cost of warfarin is $32.39. So, this scenario assesses whether this cost affects the CE plane. According to the results, rivaroxaban is still the cost-effective alternative with $988.58 per 16.78 QALYs while warfarin has $912.67 per 16.34 QALYs. Rivaroxaban has $12,150.82 NMB which is greater than warfarin (only 11,879.97) under WTP threshold. Since the ICER is less than WTP threshold, rivaroxaban with high NMB and effectiveness becomes cost effective.

Acceptability curve

Rivaroxaban was the preferred (cost effective) strategy at all willingness-to-pay thresholds according to the acceptability curve result (Figure 8). Therefore, rivaroxaban is an acceptable strategy in WTP threshold.

Figure 8 Monte Carlo Simulation cost effectiveness Acceptability curve at WTP.

Probabilistic sensitivity analysis

Probabilistic sensitivity analysis is used to determine uncertainties in the model. The result shows that warfarin is cost effective until a WTP threshold of 156.6 while rivaroxaban is the acceptable strategy in the long term above a threshold of 156.6 (Figure 9).

Figure 9 Probabilistic sensitivity analysis acceptability curve.

Discussions

In this study, we assessed CE of rivaroxaban compared to UFH/warfarin for DVT treatment. For DVT patients, it is more cost effective to treat them with rivaroxaban than warfarin in Ethiopian setting under WTP threshold of $783. Rivaroxaban therapy resulted in higher QALYs with a value of 16.78, while warfarin-based treatment resulted in 16.34 QALYs. The analysis also showed that rivaroxaban was cost effective which resulted in $12,150.82 NMB while that of warfarin is $11,859.72 only. The ICERs for rivaroxaban compared with warfarin was $125.683 per QALY saved which is acceptable based on WTP threshold of $783 per QALY saved.

Previous CE studies have concluded that rivaroxaban is more cost effective compared to warfarin.⁶ concluded rivaroxaban strategy costs less ($3195 vs $6188) and was more effective than warfarin (9.29 QALYs vs 9.14 QALYs). Our result is also coherent with this study.¹² In this study, Rivaroxaban was dominant (cheaper and more effective) and, therefore, cost-effective, in both patients with DVT and PE in three-treatment durations (3, 6 or 12 months), and was cost-effective in patients requiring lifelong anticoagulation (ICERs: $8677 per QALY and $7072 per QALY in patients with DVT and PE, respectively). Our study is supported by different studies in other countries and settings. In addition, a cross sectional study done in Brazil, considering thromboprophylaxis showed both rivaroxaban and enoxaparin showed equivalence in effectiveness and safety but enoxaparin was highly expensive resulting in rivaroxaban being a cost-effective strategy.¹⁴

Similar study done in China also showed a supportive result to this study. Rivaroxaban therapy resulted in an increase of 0.008 QALYs and was related with lower total costs compared with warfarin (US$4744.4 vs US$5572.4, respectively). There rivaroxaban was the most cost saving strategy.¹⁵ Similar pattern was observed in Greece for DVT and PE patients, there rivaroxaban was a dominant (less costly, more effective) and cost-effective (ICER: €177 per QALY gained).¹⁶ Another study done in Japan among non-valvular atrial fibrillation patients for the treatment of stroke prevention identified that Rivaroxaban was cost-effective compared to warfarin (ICER= €24,446.42/QALY).¹⁷ Even in different disease in US citizens, Warfarin is less cost effective compared with Apixaban for treatment of atrial fibrillation. Apixaban compared with Warfarin resulted in ICER of $53,925/QALY.¹⁸

In contrast to this, studies showed that rivaroxaban is associated with greater costs but fewer QALYs than LMWH/VKA,¹⁹ and Rivaroxaban was a dominated less cost-effective option compared with oral anticoagulants in Colombia for atrial fibrillation therapy.²⁰ This difference may be due to duration of therapy (<6 month), patient age groups used (65) and from assumptions about the patient population being treated and the setting of treatment in different countries.

In addition to the findings of this study, there are other reasons to limit its use in replacing warfarin in real world practice. One is that most evidences for its use comes from specific populations of DVT patients aged 40 years old with no contraindications, complication and concomitant medication, which is not true in the real world that elderly patients have complications and comorbidities. The second limitation is that the data was collected from studies done in other country settings which results in unreliability. Third, we assumed that all patients in both treatment options had the same medication adherence rates. This might differ from the real world, because each option differed in dosage regimen, safety profile, and monitoring. Although this study was done from restricted societal perspective, we only captured direct medical costs associated with each treatment option and some governmental costs only, and the long-term effect of both treatments was not addressed due to the lack of data regarding long-term effect of rivaroxaban and government covered cost.

Despite the limitations, this study also has a number of strengths. The study modeled the decision problem by communicating clinical experts and most of the data were obtained from reliable stage three clinical trials and systematic reviews. So, it can be used as a starting topic for further research. The study can also be used for policy decision making.

Conclusion

In conclusion, treatment of DVT with rivaroxaban is an economical and effective strategy for patients aged 40 years old with higher risk of PE in Ethiopia. This result is highly affected by utility of No DVT, effectiveness and cost of rivaroxaban. Because this model was built upon clinical trial data, future studies may be needed to assess effectiveness, safety and long-term effects of rivaroxaban compared to warfarin in real world settings.

Our finding showed that rivaroxaban is a more cost-effective alternative than warfarin-based therapy. Therefore, we recommend that MOH should facilitate scholars to research the safety profile of rivaroxaban for long-term use, and in collaboration with EPSA, the drug should be availed at each health facility and should replace warfarin therapy for DVT patients.