Cost-Effectiveness of Pembrolizumab plus Axitinib Versus Sunitinib as First-Line Therapy in Advanced Renal Cell Carcinoma in the U.S.

BACKGROUND
The data from phase 3 clinical trial KEYNOTE-426 has indicated that pembrolizumab plus axitinib compared to sunitinib could generate clinical benefits in patients with previously untreated advanced renal cell carcinoma (RCC). Given the incremental clinical benefits, we aim to examine the potential cost-effectiveness of pembrolizumab plus axitinib versus sunitinib in the first-line setting for patients with advanced RCC from the United States (US) payers' perspective.


MATERIALS AND METHODS
Cost and health outcomes were estimated at a willingness-to-pay (WTP) threshold of $100,000 to $150,000 per quality adjusted life year (QALY). One-way and probabilistic sensitivity analyses were performed by varying potentially modifiable parameters, and additional subgroup analyses were performed as well.


RESULTS
Upon our analyses, the total treatment costs in the pembrolizumab plus axitinib and sunitinib groups were $522,796 and $348,424, and the QALYs gained 2.90 and 1.72, respectively. In the base-case analysis, compared with receiving sunitinib, advanced RCC patients receiving pembrolizumab plus axitinib gained 1.18 more QALYs at a cost-effectiveness ratio (ICER) of $148,676/QALY. The results of subgroup analyses demonstrated that pembrolizumab plus axitinib was most cost-effective for patients who had one organ with metastasis.


CONCLUSION
First-line treatment with pembrolizumab plus axitinib, compared to sunitinib, is a cost-effective strategy when the value of WTP is from $100,000 to $150,000 per QALY in advanced RCC patients. For patients with one-organ metastasis and those in poor IMDC risk group, first-line treatment with pembrolizumab plus axitinib is more cost-effective than others.


IMPLICATIONS FOR PRACTICE
We performed the first study to examine the cost-effectiveness of pembrolizumab plus axitinib versus sunitinib in advanced RCC. Our study find first-line treatment with pembrolizumab plus axitinib is a cost-effective strategy when the value of WTP is from $100,000 to $150,000 per QALY in advanced RCC patients from the United States payers' perspective.


INTRODUCTION
Renal cancer is the third most common genitourinary malignancy in the world, and its incidence is on the rise [1]. Roughly 85% of renal cancer is renal cell carcinoma (RCC), which leads to more than 140,000 deaths worldwide per year [2][3][4]. In the 20% of patients with RCC, and these patients have a 5-year relative survival rate of only 12.3% [3,4,7].
Over the past decade, drugs targeting the vascular endothelial growth factor receptor (VEGFR) or the mammalian target of rapamycin pathway, such as sunitinib and everolimus, have become the standard treatment for advanced RCC. Sunitinib, as the current standard treatment in the first-line setting, has demonstrated prolonged progression-free survival (PFS), but the overall survival (OS) showed no obvious gain over that of interferon alfa [8][9][10][11].
The VEGFR tyrosine kinase inhibitor axitinib showed antitumor activity in patients with previously untreated clear-cell advanced RCC [12]. Pembrolizumab is a human immunoglobulin G4 monoclonal antibody that binds with the programmed death 1 (PD-1) receptor and restores T-cell immune activity. After applying the combination of pembrolizumab and axitinib, the results of the KEYNOTE-426 trial showed that among patients with previously untreated advanced RCC, treatment with pembrolizumab plus axitinib resulted in significantly longer OS (hazard ratio [HR], 0.53; 95% confidence interval [CI], 0.38-0.74; p < .0001) and PFS (HR, 0.69; 95% CI, 0.57-0.84; p < .001), across the International Metastatic Renal Cell Carcinoma Database Consortium (IMDC) risk groups and regardless of programmed death ligand 1 (PD-L1) expression, than treatment with sunitinib [13]. The patients in the trial did well in adherence, and the incidence of grade 3 or higher adverse events of pembrolizumab plus axitinib was slightly higher than that of sunitinib (75.8% vs. 70.6%) [13]. Subsequently, the pembrolizumab plus axitinib strategy was granted approval for advanced RCC by the U.S. Food and Drug Administration, and National Comprehensive Cancer Network [14] and European Society for Medical Oncology [15] guidelines both recommended pembrolizumab plus axitinib as a firstline treatment for advanced RCC.
Because of the high costs of pembrolizumab, it is unclear whether the promising pembrolizumab-axitinib combination treatment would be cost-effective in the first-line setting for patients with advanced RCC. The current study investigated the economic outcomes of pembrolizumab plus axitinib versus sunitinib for patients with previously untreated advanced RCC in the context of the U.S. health care system by using the published results of the phase III KEYNOTE-426 trial.

Model Structure
A Markov model was developed to estimate the costs and effectiveness of treatment for advanced RCC with pembrolizumab plus axitinib or sunitinib. The model structure included three states to represent the progression of advanced RCC: PFS, progressive disease (PD), and death (supplemental online Fig. 1). All patients started with PFS and were treated with pembrolizumab plus axitinib or sunitinib until disease progression. Patients could experience subsequent treatment until death or when the disease progression or unacceptable toxic effects occurred. The models for each simulation used the mean age of 62 years (range 26-90) from the KEYNOTE 426 study.
The model cycle length was 6 weeks, and the time duration was lifetime. An annual 3% discount rate for both costs and outcomes was adopted [16]. The primary outputs of the models included the total costs, life-years (LYs), qualityadjusted life-years (QALYs), and incremental cost-effectiveness ratios (ICERs). Based primarily on the results of the KEYNOTE-426, the model structure and data were also supplemented by data retrieved from publicly available databases and published literature. The model was constructed via TreeAge Pro 2018 (TreeAge Software, Inc., Williamstown, MA).

Model Survival and Progression Risk Estimates
The estimates of OS for the pembrolizumab plus axitinib and sunitinib groups were based on the results of KEYNOTE-426. First, the GetData Graph Digitizer (version 2.25; http://www. getdata-graph-digitizer.com/index.php) was applied to collecting data points from the OS Kaplan-Meier curve for sunitinib announced by Rini et al. [13], and these data points were then fit into the parametric survival model. The PFS probability for sunitinib was obtained by the same approach. Given its flexible and wide application in cancer survival analyses, the Weibull distribution was used to match the number of patients in the three states over time [17,18]. Next, the two parameters, shape (γ) and scale (λ), were calculated through this fitting and applied to the Kaplan-Meier curve by means of the R software package (http://www.r-project.org) and the method proposed by Hoyle and Henley [19].
The time-dependency transition probabilities in each Markov cycle were calculated based on the following formula: tp(t u ) = 1 − exp{λ(t − u) γ − λt γ } (λ > 0, γ > 0), where u is the Markov cycle and t u represents the arrival at state t after u Markov cycles [20].
After applications of the HRs for OS and PFS to the simulated sunitinib curve, the OS and PFS risks of pembrolizumab plus axitinib were calculated [21].

Utility and Cost Estimates
To estimate the total QALYs, the survival time was adjusted by health-related quality of life (QoL). Utility was applied to measuring patients' preferences for living at a particular health state. It reflected the impacts of the disease-related health state. The values 0.76 and 0.66 were used as the mean health utility value for the PFS and the PD state (supposing the utility values of the two groups were the same after progression), respectively; these were derived from the previously published literature [22]. Similarly, according to the published literature, the utility value of the first-line treatment using sunitinib was 0.73 [21]. The disutility values due to grade 3 and 4 adverse effects (AEs) were included in the model [23].
Only direct medical care costs, including drug costs, administration costs, AE costs (assuming that AEs appeared only once in the PFS and the PD state), terminal care [24], and costs for laboratory and radiographic tests, were studied during the analyses. According to the KEYNOTE-426 trial [13], there were two treatments in the first-line setting, pembrolizumab plus axitinib and sunitinib. Pembrolizumab was administered intravenously at a dose of 200 mg once every 3 weeks, and the maximum cycle was 35. The dosage of axitinib was 5 mg twice daily. Sunitinib was administered orally at a dose of 50 mg daily for the first 4 weeks in each 6-week cycle. The median daily doses of axitinib (9.8 mg) and sunitinib (50.0 mg) were calculated. Fifty percent of the patients in the pembrolizumab-axitinib group and 60.7% of the patients in the sunitinib group received subsequent anticancer therapy [13]; details on subsequent therapy are shown in Table 1, which also shows the prices of the drugs based on the Centers for Medicare & Medicaid Services and previously published literature [21,25]. Grade 3 and 4 AEs were included in the model, which had notably different probabilities between the arms of the -KEYNOTE-426 trial. The incidence rates of AEs in the model came from the KEYNOTE-426 trial data [13]. The costs of AEs were based on the published literature [21,[26][27][28][29]. The follow-up costs covered fees for computed tomography [30] (baseline estimate; week 12; every 6 weeks in first 45 months; every 12 weeks thereafter) and laboratory evaluations [31] (regularly throughout the treatment period and for 30 days thereafter). Costs were adjusted to present 2019 U.S. dollars, and the U.S. Consumer Price Index was adopted as a reference for inflation. Information on these costs were obtained, shown in Table 1. The willingness-to-pay (WTP) was set at a value of $150,000 [24].

Sensitivity Analysis
A series of sensitivity analyses were performed to evaluate the robustness of the model and to address uncertainty in variable estimates. The ranges of the parameters were obtained from published literature; when there were no available published data, a range of AE20% of the base-case value was considered, in accordance with established approaches [32]. This study assigned a log normal distribution to HRs, a gamma distribution to costs, and a beta distribution to transition probabilities and health state utilities [21,33]. The values, variation ranges, and distributions of the parameters are shown in Table 1. During the one-way sensitivity analyses, this study used varied values of a certain parameter within its defined range and examined the individual effects of this parameter on the ICERs. In addition, probabilistic sensitivity analyses were completed to assess the variations in multiple parameters at once. We resorted to 10,000 Monte Carlo repetitions, randomly setting all parameters simultaneously before each repetition. A cost-effectiveness acceptability curve of each treatment strategy was generated to calculate the probability of this strategy being the most cost-effective one at different WTP thresholds.
We also took into account all patient subgroups presented in the forest plot of the KEYNOTE-426 trial. In the absence of sufficient data for each patient subgroup, the study adopted the same baseline sunitinib survival curve for all patients in the sunitinib group, and their pembrolizumab plus axitinib survival curves were produced based on the subgroup-specific HRs according to the approach taken by Hoyle et al. [34]. The subgroup-specific HRs for OS and PFS are listed in supplemental online Table 1. We calculated QALY in first 24 months and the rest of life as well because the KEYNOTE-426 trial was less than 24 months.

RESULTS
The Weibull models validated well, and model-derived survival curves were close to the Kaplan-Meier curves for OS and PFS in the KEYNOTE-426 trial (supplemental online Fig. 2).

Base-Case Results
The model projected that the life expectancy of patients receiving pembrolizumab plus axitinib was 4.34 LYs, which was 1.74 LYs more than those of patients receiving sunitinib. Considering the QoL, patients receiving pembrolizumab plus axitinib gained 2.90 QALYs, 1.18 QALYs more than those of patients receiving sunitinib. Compared with the treatment using sunitinib, the use of pembrolizumab plus axitinib cost an additional $174,372, resulting in an ICER of $100,504 ($68,403-$220,604) per LY, or $148.676 ($90,628-$278,411) per QALY (Table 2).

Sensitivity Analysis
The one-way sensitivity analyses, shown in Figure 1, revealed that the variable with the strongest influence on the ICER was HR for OS (ranging from 0.38 to 0.74, with the ICER increasing from $129,280 per QALY to $206,136 per QALY), followed by the costs of pembrolizumab, the costs of axitinib, the utility of PD state, the utility of PFS state in pembrolizumab plus axitinib, and HR for PFS.
The results of the probabilistic sensitivity analyses are shown in supplemental online Figure 3. The ICER scatterplot displays the results of the 10,000 Monte Carlo repetitions with all parameters randomly varied. These results demonstrated that the probability of pembrolizumab plus axitinib being cost-effective, compared with that of sunitinib, was 38.88% and 53.65% at a willingness-to-pay threshold of $100,000 and $150,000 per QALY.
The acceptability curves comparing the two strategies indicated that pembrolizumab plus axitinib, compared with sunitinib, was the more effective and cost-effective choice at a WTP threshold of from $100,000 to $150,000 per QALY (Fig. 2).
The results of subgroup analyses implied that pembrolizumab plus axitinib was most cost-effective with a probability of 66.3% for patients who had one organ with metastasis, followed by patients under the age of 65. In the IMDC risk groups, treatments for patients in the poor risk group were likelier to be cost-effective than those in the favorable or intermediate groups (supplemental online Table 1). In the pembrolizumab-axitinib group, patients gained 1.13 QALYs in the first 24 months and 1.77 QALYs in rest of life, and 0.98 QALYs and 0.74 QALYs in sunitinib group, respectively.

DISCUSSION
Immunotherapy checkpoint inhibitors have dramatically changed the therapeutic canvas for patients with advanced RCC, particularly in first-line treatment. The results recently published inferred that the combinations of immune agents (i.e., nivolumab plus ipilimumab) or immune-checkpoint inhibitors with antivascular agents (i.e., avelumab plus axitinib, pembrolizumab plus axitinib, and atezolizumab plus bevacizumab) serve well to yield overall improvements in treatment outcomes over the antivascular tyrosine kinase inhibitor monotherapy in first-line setting [13,[35][36][37].
Based on the data of CheckMate 214 trial [37], the combination of nivolumab (the anti-PD-1 agent, Bristol-Myers Squibb) and ipilimumab (the anticytotoxic T-lymphocyte antigen 4 agent, Bristol-Myers Squibb) was authorized as a firstline treatment of patients with metastatic RCC. According to this trial, a study reported that nivolumab plus ipilimumab was cost-effective for intermediate-and poor-risk patients with metastatic RCC in a willingness-to-pay threshold range from $100,000 to $150,000 per QALY in the U.S. [21]. These results suggested that we should further assess which patients with advanced RCC really would gain both curative effect and the cost-effectiveness benefits from the immunotherapy combination.
Our study is the first Markov model-based study to assess the health and economic outcomes of pembrolizumab plus axitinib versus sunitinib as a first-line treatment for patients with previously untreated advanced RCC in the U.S., according to the data of the phase III KEYNOTE-426 trial. The results suggest that compared with the standard sunitinib treatment, the pembrolizumab plus axitinib therapy produces an additional 1.18 QALYs and costs $148,676 per additional QALY. The probabilistic sensitivity analyses suggested that it is more likely that the pembrolizumab plus axitinib treatment is a cost-effective therapeutic option in the U.S. Nonetheless, it is worth noting that the ICER is only slightly below the WTP of $150,000, and the probability of being cost-effective is only 38.88% and 53.65% at a willingness-to-pay threshold of $100,000 and $150,000 per QALY; future studies would be needed to reduce the uncertainty.
The one-way sensitivity analyses showed that the HR for OS is the factor that most influences the ICER. Therefore, we further analyzed the data of the subgroups to explore in which subgroup of patients pembrolizumab plus axitinib would be more cost-effective. The results imply that the combination of pembrolizumab plus axitinib is the most cost-effective therapy for patients with one-organ metastasis. It is noteworthy that the patients older than 65 or in the favorable risk group of IMDC might not obtain the costeffective benefit. Moreover, treating patients in the poor risk group would be more cost-effective than treating those in the intermediate risk group. It has been reported that patients may continue to respond after stopping immunotherapy [38], but whether pembrolizumab has a long-term efficacy in RCC remains uncertain. We await more mature data to see if there is a plateau on the tail of the survival curve. In the model, patients were assigned to receive pembrolizumab for a maximum of 35 cycles, which can solve this uncertainty.
Of particular interest, our subgroup analysis finds that treatments of patients with PD-L1 expression greater than 1% are less cost-effective than those with PD-L1 expression less than 1%. On the one hand, many clinical trials demonstrated nivolumab's superiority for patients with advanced RCC regardless of the expression of PD-L1, although a previous study involving patients with metastatic RCC accepting nivolumab assumed that the positive expression level of PD-L1 may predict the efficacy of nivolumab [39]. It suggests that PD-L1 expression is not an ideal biomarker for anti-PD-1/PD-L1 therapies, especially in combination therapies [40]. On the other hand, treatment with pembrolizumab would obtain the cost-effective benefit in patients with non-small cell lung cancer with PD-L1 expression greater than 50%, but not for those with PD-1 expression greater than 1% in our previous research [41]. Because of the lack of sufficient data for each patient, especially because there are no subgroups with PD-L1 expression greater than 50% and 20%, it might be an important reason leading to the results of our subgroup analysis. In addition, more in-depth research is needed on other factors affecting the ICER of subgroups besides the HR for OS based on more detailed data.
In the analysis by Chen et al. [42], pembrolizumab plus axitinib is not cost-effective when compared with sunitinib for patients with previously untreated advanced RCC in China. Although different country-or region-specific economic conditions may result in different cost-effectiveness outcomes, the research of Chen et al. is based entirely on the KEYNOTE-426 trial, which does not truly reflect the clinical utility and health status in the Chinese population. Considering the disutility values due to AEs, we used the mean health utility value before progression for advanced RCC, rather than the utility value of sunitinib in PFS state used by Chen et al., to evaluate the utility value of the first-line treatment using pembrolizumab plus axitinib, and only grade 3 and 4 AEs were factored into our analysis, which have a strong correlation with QoL [22]. These differences revealed that our study could more accurately reflect the utilities of different treatments. Moreover, we also explored the effect of PD-L1 expression on the results, which may facilitate more in-depth studies of PD-L1 expression.
This study has several limitations. First, similar to many cost-effectiveness studies, this study used previously published data from the KEYNOTE-426 trial rather than prospectively collected data, and any biases within the trial will be mirrored in the study. Second, an inevitable limitation was the use of a Weibull distribution to project consequences beyond the lifetime horizon of the KEYNOTE-426 trial. Third, because of the lack of complete QoL data to calculate the utility values, we corrected the utility values by considering the disutility values of AEs, and only grade 3 and 4 AEs were included, which might lead to overestimates or underestimates of the utility values. Fourth, because of the absence of head-to-head trials, other competing strategies were not included, such as pazopanib, a more cost-effective treatment than that using sunitinib [7,43,44]. Fifth, this trial-based study could not wholly track the natural disease course in the real world. This approach could not adequately reflect effectiveness and resource consumption in routine clinical practice. Finally, the KEYNOTE-426 trial did not provide the Kaplan-Meier curve for each subgroup, making it impossible to run the model completely for each subgroup, and the original group balance produced by randomization may not exist in the subgroups; thus, the results of the subgroup analyses should be interpreted with caution.

CONCLUSION
This economic evaluation projected that, compared with sunitinib, pembrolizumab plus axitinib was estimated to be costeffective for patients with previously untreated advanced RCC at a willingness-to-pay threshold of from $100,000 to $150,000 per QALY in the U.S. For patients with one-organ metastasis and those in poor IMDC risk group, first-line treatment with pembrolizumab plus axitinib is more cost-effective than in other subgroups that were analyzed.