Cost-effectiveness of apixaban and rivaroxaban in thromboprophylaxis of cancer patients treated with chemotherapy in Spain

Abstract Background Apixaban and rivaroxaban are two direct-acting oral anticoagulants (DOACs) recommended for thromboprophylaxis in cancer patients treated with chemotherapy in an ambulatory setting. We aimed to assess the cost-utility of thromboprophylaxis with apixaban and rivaroxaban vs no thromboprophylaxis in ambulatory cancer patients starting chemotherapy with an intermediate-to-high risk of venous thromboembolism (VTE), Khorana score ≥ 2 points. Methods A cost-effectiveness analysis was performed from the perspective of Spain’s National Health System (NHS) using an analytical decision model in the short-term (180 days) and a Markov model in the long-term (5 years). Transition probabilities were obtained from randomized, double-blind, placebo-controlled clinical trials of apixaban and rivaroxaban in adult ambulatory patients with cancer at risk for VTE, treated with chemotherapy (AVERT and CASSINI trials). The costs (€2,021) were taken from Spanish sources. The utilities of the model were obtained through the EQ-5D questionnaire. Deterministic (base case) and probabilistic (second-order Monte Carlo simulation) analyses were conducted. Results In the probabilistic sensitivity analysis, apixaban generated a cost per patient of €1,082 ± 187, with a 95% confidence interval (CI) of €713–1,442, while no prophylaxis produced a cost per patient of €1,146 ± 218, with a 95% CI of €700–1,491, with a saving of €64 per patient and a gain of 0.008 QALYs. Likewise, rivaroxaban provided a cost per patient of €993 ± 133, with a 95% CI of €748–1,310, while no prophylaxis produced a cost per patient of €872 ± 152, with a 95% CI of €602–1,250, with an additional expense of €121 per patient and a gain of 0.008 QALYs. Conclusions In thromboprophylaxis of cancer patients, the use of apixaban and rivaroxaban generated similar costs compared to non-prophylaxis, without the difference found being statistically significant, with a clinically insignificant QALY gain.


Introduction
Venous thromboembolism (VTE) is a frequent complication in cancer patients and has been progressively increasing in the last two decades 1 .Cancer-associated thrombosis (CAT) has been associated with a higher mortality, morbidity, healthcare costs, and, more recently, with a significantly negative impact on the quality-of-life of cancer patients 2 .
CAT rates vary widely among cancer patients 3 .The estimated CAT rate for average-risk patients is 13 per 1,000 person-years, while patients with metastatic disease or receiving thrombogenic therapies have a VTE rate of 68 per 1,000 person-years, with the highest risk reported in lung, pancreatic, and brain cancers (200 per 1,000 person-years).Currently, most venous thromboembolic events occur in the outpatient setting 4 .Different randomized clinical trials 5,6 and meta-analyses 7 have shown that ambulatory thromboprophylaxis is effective with an acceptable bleeding risk profile in this population.
Apixaban (APIX) and rivaroxaban (RIV) are the only direct oral anticoagulants (DOAC) recommended by the American Society of Clinical Oncology (ASCO) 8 and the International Initiative on Thrombosis and Cancer (ITAC) 9 in ambulatory cancer patients receiving systemic chemotherapy.These recommendations are based on two phase III randomized clinical trials: CASSINI 10 and AVERT 11 trials, that included a wide spectrum of prothrombotic cancers (gastrointestinal, genitourinary, lung, brain tumors and hematological neoplasms such as lymphoma and myeloma) in adult patients.Both studies have a similar design, they compared up to 6 months thromboprophylaxis with RIV (10 mg daily) or APIX (2.5 mg twice daily) to placebo in ambulatory cancer patients initiating systemic anti-cancer therapy with a Khorana score of 2 points or higher.Patient populations share similar clinical features (Supplementary Table S1 and S2).In the AVERT trial, apixaban was associated with a significantly lower incidence of VTE compared to placebo in the primary modified intention-to-treat analysis (mITT), 4.2% versus 10.2%, respectively (hazard ratio ¼ 0.41; 95% confidence interval [CI] ¼ 0.26-0.65;p < 0.001).The rate of major bleeding was significantly higher with apixaban than with placebo in the mITT (3.5% and 1.8%, respectively; HR ¼ 2.00; 95% CI ¼ 1.01-3.95),but not during the treatment period.In the CASSINI trial, VTE incidence was lower with rivaroxaban compared to placebo in the per-protocol analysis, 2.6% versus 6.4% (HR ¼ 0.40; 95% CI ¼ 0.20-0.80),but not in the primary intention-totreat analysis.Major bleeding occurred in 2% of patients receiving rivaroxaban and 1% receiving placebo (HR ¼ 1.96; 95% CI ¼ 0.59-6.49).This indication is not yet authorized in Spain, but, on the contrary, it is recommended by the Spanish Society of Medical Oncology (SEOM) 12 .Despite the availability of several validated risk assessment models for VTE in cancer patients and guideline recommendations, rates of VTE prophylaxis in high-risk patients are low in routine clinical practice.A recent survey by the Association of Community Cancer Centers reported that only 9% of oncologists use a risk assessment tool to identify high-risk patients in the outpatient setting and more than 90% of patients with very high-risk do not receive pharmacological thromboprophylaxis 13 .
Two recently published economic analyses 14,15 assessed cost-effectiveness of thromboprophylaxis with DOAC in cancer patients treated with chemotherapy in ambulatory patients with an intermediate to high risk of suffering VTE (Khorana score !2).In these studies thromboprophylaxis refers to the pharmacological thromboprophylaxis for patients with cancer with high risk of VTE based on Khorana risk score.However, neither of the two studies performed a single drug-based analysis and no formal drug-drug comparison was provided (Supplementary Table S3).
The main hypothesis of this study is that APIX and RIV may have a different cost-effectiveness profile.The aim of this analysis is to know if the use of APIX or RIV is a costeffective alternative in thromboprophylaxis for ambulatory cancer patients with intermediate-high risk of VTE (Khorana score ! 2) in Spain, based on the different results obtained in efficacy and safety with these drugs in placebo-controlled randomized clinical trials 9,10 and on different pharmacokinetic and pharmacodynamic profiles 16 .

Methods
This cost-effectiveness analysis was performed from the perspective of the Spanish National Health Service, so only healthcare costs (direct medical costs) were included.A discount rate of 3% per year was applied to costs and outcomes.The primary outputs of the model included costs, quality-adjusted life years (QALYs), and incremental cost-effectiveness ratio (ICER).The model was constructed via TreeAge Pro 2018 (TreeAge Software, Inc, Willianston, MA).

Model structure and analysis type
The structure of the model was based on previously published models [17][18][19][20] .Figure 1 shows the structure of the economic model.It consists of a short-term (with a 180-day time horizon) decision analysis (DA) and subsequent long-term Markov model (with a 5-year time horizon) and a cycle length of 1 year.The short-term DA includes the following health states: (i) actively treated cancer; (ii) minor bleeding events (MIBE); (iii) major bleeding events (MABE); (iv) deep venous thrombosis (DVT); (v) pulmonary embolism (PE); or (vi) death from any cause.Patients with DVT were expected to evolve in the long-term (Markov model) to the following states: (i) actively treated cancer; (ii) post-thrombotic syndrome (PTS); or (iii) death from any cause.Similarly, the patient with PE was assumed to evolve (Markov) to the following states: (i) actively treated cancer; (ii) chronic thromboembolic pulmonary hypertension (CTPH); or (iii) death from any cause (Figure 1).Deterministic and probabilistic analysis (through 1.000 second-order Monte Carlo simulations) were performed.Probabilistic analyses allow the assessment of the uncertainty of the variables 21 .One-way (tornado) deterministic sensitivity analyses were also performed, modifying probabilities, costs, and utilities.A willingness-to-pay in Spain of EUR 25,000 per QALY gained was considered.

Probabilities of the model
Table 1 shows the short-and long-term probabilities considered in the model.The short-term probabilities were mainly obtained from the economic studies of Du et al. 14 and Li et al. 15 , who, in turn, had taken them from phase III, randomized, placebo-controlled, double-blind clinical trials of APIX (AVERT study) 10 , and RIV (CASSINI study) 11 in cancer, as well as from another previously published analysis 22 .The design and results of the AVERT and CASSINI trials, as well as the baseline demographic and clinical characteristics of the patients, are detailed in Supplementary Tables S1 and S2.The probabilities considered in the long-term Markov model (with 1 year cycles) were obtained from clinical and economic studies published earlier 14,17,[21][22][23][24][25] .Transition probabilities of the model were calculated according to the formula Pt ¼ 1 -e -rt , r being the event rate and t the time in which such rate takes place 26 .

Costs of the model
The costs of the model are presented in Table 2.For the calculation of the mean cost of prophylaxis with DOAC, the mean duration of prophylaxis (in days) was calculated using the medians from the AVERT study 10 according to the formula from Pudar et al. 27 .Thus, a mean duration of prophylaxis of 140 days was considered, between a minimum of 70 and a maximum of 150 days.The cost of the events (DVP, PE, MABE, MIBE, PTS, and CTPH) in Spain were obtained from the public prices of health services of the Basque Health Service for the year 2020 28 .The costs of the following diagnostic-related groups (DRG): were considered: DVP (197); PE (134); MABE (44); MIBE (253); PTS (197), and CTPH (207).

Utilities of the model
The utilities considered in the model were the ones used in the recent economic analysis of Du et al. 13 (Table 3).These utilities derive from a study conducted in the United Kingdom using the EQ-5D instrument 29 , excluding the CTPH utility, that was obtained from a study performed in the USA 30 .

Clinical validation of the model
The economic model was clinically validated by two Spanish clinical experts with experience in thromboprophylaxis in cancer patients (Dr.Enrique Gallardo, Oncology Department, Hospital Universitario Parc Taul ı, Sabadell, Spain; Dr. Andr es J. Muñoz, Medical Oncology Department, Hospital General Universitario Gregorio Marañ on, Madrid, Spain).

Sensitivity analysis
A univariable and probabilistic sensitivity analyses were performed to assess the robustness of the model and the uncertainty in parameter estimation.In this study, a probabilistic deterministic sensitivity analysis (PDSA) expressed in incremental net benefit was performed instead of a normal univariant sensitivity analysis, because PDSA offers advantages versus a classic deterministic sensitivity analysis for providing insight on the effect of uncertainty in individual parameters on the estimate cost-effectiveness 31 .

APIX vs no prophylaxis
APIX would be dominant (greater effectiveness with lower costs) compared to no prophylaxis, with a gain of 0.005 quality adjusted life years (QALY) and a cost difference of -e59.49(Table 4).It is estimated that for every 1,000 patients treated with APIX instead of RIV, seven episodes of DVT and 11 episodes of PE would be prevented (Supplementary Table S4).

RIV vs no prophylaxis
RIV would not be dominant compared with no prophylaxis.However, it would be cost-effective.The cost of gaining a QALY with RIV would be e18,746.77(below a willingness to pay of e25,000-60,000 per QALY gained) 32 .If one considered that the QALY gain was not clinically relevant, with RIV, an additional cost of e116.23 per patient would be generated versus no thromboprophylaxis (Table 4).(0.442-0.702)Ã Of prophylaxis versus no prophylaxis, for a willingness to pay of e25,000 per QALY gained.If all iteration with less QALY as negatives are taken into account, the probability API versus No prophylaxis will be 49.1%, and the probability RIV versus No prophylaxis will be 41.2%.Abbreviations: QALYs, quality-adjusted life years.

APIX vs no prophylaxis
As shown in Table 5 and Figure 2a, the probabilistic analysis revealed savings of around e64 per patient treated with APIX and a gain of 0.008 QALY per patient treated with APIX compared to no prophylaxis.The probability of APIX being costeffective, for a willingness-to-pay of e25,000 per QALY gained, was 62.2%. Figure 3 shows the cost-effectiveness acceptability curve (CEAC) of APIX vs no prophylaxis.

RIV vs no prophylaxis
As noted in Table 5 and Figure 2b, the probabilistic analysis showed an additional cost per patient treated with RIV of around e121 and a gain of 0.008 QALY per patient treated with RIV in comparison with no prophylaxis.The probability of RIV being cost-effective, for a willingness to pay of e25,000 per QALY gained, was 51.9%. Figure 4 shows the CEAC of RIV vs no prophylaxis.

Probabilistic deterministic sensitivity analyses
Incremental net monetary benefit (INMB) was positive (costeffective) for a willingness to pay of e25,000 per QALY gained for all variables in the case of RIV, with two exceptions: the probability of cancer mortality with or without prophylaxis, in which case the risk of a negative IBNM (not cost-effective) would be 60.9% and 36.2%,respectively.In the case of APIX, the risk of a negative IBNM for the probability of cancer mortality with or without prophylaxis would be 64.0%and 30.8%, respectively.One-way deterministic sensitivity analyses are described in detailed for APIX and RIV in Table 6.

Discussion
In the last three decades the overall incidence of VTE has increased by 3-fold.It's estimated the risk of VTE in cancer patients is 12-fold higher than in the overall population, and in patients receiving systemic anti-cancer therapy reaches 23-fold 1 .VTE in cancer patients provokes a significant increase in VTE recurrence and major bleeding during anticoagulant therapy compared to non-cancer patients 33 .Furthermore, VTE is associated with temporary or definitive interruption of anticancer therapy with a potential impact on cancer prognosis.These data suggest that prevention of VTE in patients with cancer who receive systemic therapy in an ambulatory setting has a potentially high clinical value.Despite all these facts and available evidence from placebocontrolled randomized clinical trials showing outpatient thromboprophylaxis in cancer patients is effective and safe, it is still seldom used in clinical practice, and the cost-effectiveness relationship has not been well established yet.
In the present economic analysis thromboprophylaxis in cancer patients treated with systemic therapy would be more efficient with APIX than with RIV.APIX is dominant (greater effectiveness with lower costs) compared to no prophylaxis.Nevertheless, the gain of 0.005 QALY per patient might not be clinically relevant, since clinical relevance in utility gain (the smallest difference in quality-of-life that a patient can perceive) is generally considered to be between 0.03 and 0.04 QALY gained 34,35 .If this premise is accepted, one would save up to e59.49 per patient with APIX compared to no thromboprophylaxis.This fact shows that in addition to the clinical value derived from the use of APIX and RIV as thromboprophylaxis, the use of these alternatives will generate resource savings for the National Health System in the case of APIX and will be a cost-effective option in the case of RIV.It should be taken into account that there are no  studies with this objective from the perspective of the Spanish National Health System.
Overall, data coming from cost-effectiveness analysis of ambulatory thromboprophylaxis of cancer patients (including DOAC, low-molecular-weight heparins, or vitamin K antagonists) from western European countries are scarce.These results are quite reassuring in order to expand outpatient thromboprophylaxis with low dose DOAC.In addition, it is expected to achieve a better cost-effectiveness result if other risk assessment models of VTE recently published, that have shown to significantly improve the capability of VTE risk prediction compared to Khorana score, are implemented in clinical practice 36,37 .Another significant finding of this analysis, and to our knowledge not described before in this setting, is the different cost-effectiveness profile of APIX and RIV.These results are in line with other clinical findings in cancer patients, mainly in safety, with a better bleeding risk profile for APIX compared to RIV (gastrointestinal tract bleeding and gastrointestinal cancers) 38,39 , that suggest these drugs are not directly interchangeable.Lastly, these results have a direct local applicability from Spanish National Health System perspective and are in line with the recommendation of the Spanish National Guideline of Thrombosis and Cancer (Spanish Society of Medical Oncology, SEOM) 12 .
This result should be assessed considering the strengths and weaknesses of the study.One of the main strengths would be the clinical validation of the model by two expert physicians with experience in thromboprophylaxis in cancer patients.Secondly, a second-order Monte Carlo simulation allowed us to analyze the uncertainty of the data (probabilities, costs, utilities) included in the analysis.This mathematical method allows the replication of simultaneous and random changes in the model parameters (probabilities, costs, utilities) in an attempt to simulate clinical evolution of patients in real life 40,41 .The confidence of the results was analyzed by the confidence intervals of 95%, with probabilities of cost-effectiveness of 62.2% and 51.9% in patients treated with APIX or RIV compared to no prophylaxis, respectively.In both cases, the probability of cost-effectiveness versus no thromboprophylaxis would be low (under 70%) for the indicated willingness to pay.
As far as the limitations of the study are concerned, first of all, it has to be taken into account that the theoretical model developed is, by definition, a simplified simulation of reality.The fact that there was no direct comparison between thromboprophylaxis with APIX and RIV can be considered a study limitation.Nonetheless, this was not possible for several reasons.Firstly, there are no clinical trials that compare these two drugs directly.The available clinical trials concerning DVT prophylaxis in cancer patients compared APIX (AVERT study) 10 and RIV (CASSINI study) 11 with a placebo control (no prophylaxis).Secondly, the results obtained in the placebo group of patients differed considerably between both studies, in particular, the probability to suffer symptomatic VTE, minor bleedings and VTE-related mortality, although this point did not invalidate the results of this study.However, the patients from both studies (AVERT and CASSINI) had similar baseline characteristics regarding age, gender, Khorana score, and previous thromboembolism rate 8,11 .The fact that only two clinical trials are available should be considered a major limitation of the study.Despite there not being any indications or signals that there would be subgroups of patients in which RIV was more costeffective, we can't exclude that in certain subgroups RIV may be a better option.This type of subgroup analysis of cancer patients would require dedicated studies.Finally, quality-oflife studies in this setting are needed and should be performed to complement cost-effectiveness analysis.

Figure 4 .
Figure 4. Cost-effectiveness acceptability curve of RIV versus no prophylaxis.

Table 1 .
Probabilities of the economic model in the short-and long-term.

Table 4 .
Base-case analysis.With apixaban, QALYs are gained and savings are generated compared to no prophylaxis.Abbreviation.QALYs, quality-adjusted life years.