Economic Evaluation of First-Line Treatments for Metastatic Renal Cell Carcinoma: A Cost-Effectiveness Analysis in A Health Resource–Limited Setting

Background To estimate, from the perspective of the Chinese healthcare system, the economic outcomes of five different first-line strategies among patients with metastatic renal cell carcinoma (mRCC). Methods and Findings A decision-analytic model was developed to simulate the lifetime disease course associated with renal cell carcinoma. The health and economic outcomes of five first-line strategies (interferon-alfa, interleukin-2, interleukin-2 plus interferon-alfa, sunitinib and bevacizumab plus interferon-alfa) were estimated and assessed by indirect comparison. The clinical and utility data were taken from published studies. The cost data were estimated from local charge data and current Chinese practices. Sensitivity analyses were used to explore the impact of uncertainty regarding the results. The impact of the sunitinib patient assistant program (SPAP) was evaluated via scenario analysis. The base-case analysis showed that the sunitinib strategy yielded the maximum health benefits: 2.71 life years and 1.40 quality-adjusted life-years (QALY). The marginal cost-effectiveness (cost per additional QALY) gained via the sunitinib strategy compared with the conventional strategy was $220,384 (without SPAP, interleukin-2 plus interferon-alfa and bevacizumab plus interferon-alfa were dominated) and $16,993 (with SPAP, interferon-alfa, interleukin-2 plus interferon-alfa and bevacizumab plus interferon-alfa were dominated). In general, the results were sensitive to the hazard ratio of progression-free survival. The probabilistic sensitivity analysis demonstrated that the sunitinib strategy with SPAP was the most cost-effective approach when the willingness-to-pay threshold was over $16,000. Conclusions Our analysis suggests that traditional cytokine therapy is the cost-effective option in the Chinese healthcare setting. In some relatively developed regions, sunitinib with SPAP may be a favorable cost-effective alternative for mRCC.


Introduction
Renal cell carcinoma (RCC), the most common type of kidney cancer, accounts for about 3% of all human malignancies. It is estimated that nearly 30% of the patients with RCC have distant metastatic disease at presentation and that half of those with localised disease subsequently develop metastases during the course of their disease [1,2]. The median overall survival for metastatic RCC patients is 10 months, and the 5-year survival rate is 5-15%, even when all visible disease is cleared by metastasectomy and nephrectomy [3,4]. RCC is insensitive to traditional cytotoxic agents and radiation therapies. At present, the most widely used regimens for metastatic RCC (mRCC) are cytokine therapies, including interferon-alfa and interleukin-2, which in previous studies showed response rates of only 10-20% and resulted in debilitating adverse effects [5,6,7]. Studies of cytokine therapies with or without chemotherapy have shown short-term partial response rates of up to 20-35% [8,9]. New treatments are needed to develop strategies for controlling metastatic disease and improving quality of life.
Recently, targeted pharmacological therapies (e.g., sunitinib, sorafenib, bevacizumab, pazopanib, axitinib and temsirolimus) have been developed for RCC treatment [10]. Among them, sunitinib and bevacizumab (combined with interferon-alfa) have been accepted as first-line therapy options for patients with mRCC. Sunitinib is an orally administered multi-target inhibitor of vascular endothelial growth factor (VEGF) and platelet-derived growth factor (PDGF); bevacizumab is a humanised monoclonal antibody that selectively inhibits VEGF-A, which is involved in cancer angiogenesis. These targeted therapies can improve progression-free survival (PFS) and overall survival (OS) [11,12,13,14]. However, their substantial cost restricts their widespread use, especially in health resource-limited regions. An evaluation of the cost-effectiveness of these new therapies is important for improving resource allocation efficiency. A few economic studies of these new therapies have been reported [15,16,17,18]. However, there have been few cost-effectiveness analyses conducted in resource-limited settings, where the frontline therapy for mRCC still involves traditional chemotherapy and cytokine therapy.
In the current study, we evaluated the long-term economic outcomes of five first-line mRCC regimens based on clinical practice and recommendations: [19] sunitinib, bevacizumab plus interferonalfa, interferon-alfa, interleukin-2, and interleukin-2 plus interferonalfa. Because the follow-up times of most clinical trials do not focus on the lifetime course of the disease, head-to-head comparisons among several different therapies are rarely reported. Thus, mathematical modelling techniques must be used to supply decision making information. A perspective of Chinese healthcare system was adopted to assist in determining the direct economic value of the five different first-line therapies, given the willingness-to-pay threshold of Chinese and Shanghai residents that is associated with the per capita GDP per quality-adjusted life-year (QALY) gained. The analysis excluded indirect societal costs (i.e., productivity or caregiver costs).

Analytical overview
With the R software package (version 2.13.0; R Development Core Team, Vienna, Austria), we used a previously developed Markov model to simulate the transition of a cohort of individuals with mRCC based on the clinical course [20]. We used this model to estimate and compare the lifetime direct medical costs and health outcomes associated with different first-line strategies for mRCC from the perspective of the Chinese healthcare system. The future costs and benefits were discounted using a 3% annual discount rate.
Although multiple agents for mRCC have been evaluated in clinical trials, the most commonly used first-line strategies are interferon-alfa, interleukin-2, sunitinib and bevacizumab [19]. Therefore, the cost and effectiveness of the five first-line strategies (sunitinib, bevacizumab plus interferon-alfa, interferon-alfa, interleukin-2 and interleukin-2 plus interferon-alfa) were evaluated and compared. Because of the absence of head-to-head clinical trials comparing these five first-line strategies, an indirect comparison was performed following a well-established approach [15,17,21]. Because the affordability of sunitinib in China can be a challenge, the Sunitinib Patient Assistance Program (SPAP) was introduced to make sunitinib available to eligible patients. Currently, the SPAP requires RCC patients to pay for three cycles of sunitinib, after which they will receive donations of sunitinib until the end of their treatment [22]. Therefore, the scenario analyses included the importance of SPAP for sunitinib.
Four types of parameters were inputs for the model: transition probabilities, which reflect the probabilities of moving between health states at each cycle; event proportions, which govern the ratios of events; direct medical costs, which were estimated based on health resource consumption; and health state utilities, which project the health-related quality of life for different health states. These data were derived from the published literature or from local health systems.
Cost-effectiveness ratios were calculated to evaluate the outcomes of the different strategies. The main health outcomes were presented by quality-adjusted life-year (QALY). The results were presented as an incremental cost-effectiveness ratio (ICER).

Decision Model Structure
A decision-analytic Markov model was used to compare the lifetime clinical progression, costs and utilities of treating mRCC with sunitinib, bevacizumab plus interferon-alfa, interferon-alfa, interleukin-2 and interleukin-2 plus interferon-alfa. Three discrete health states reflecting different characteristics of the disease were identified: progression-free survival (PFS), progressed survival (PS) and death. The structure of the model is shown in Figure 1. To be consistent with other economic evaluations in the literature, a 10year time horizon was used to determine the lifetime outcomes [16,17]. In the Markov model, the cycle length was 6 weeks and the entry state was progression-free survival. During each 6-week cycle, the patients either remained in their assigned health state or progressed to a new health state. The hypothetical patient demographics, when entering the model, matched those of the patients in the pivotal clinical trials: histologically proven renal cell carcinoma, advanced metastatic disease, finding of at least one measurable lesion, a World Health Organization (WHO) or Eastern Cooperative Oncology Group (ECOG) performance status (PS) of 0 or 1, and adequate organ function [5,11,12].

Clinical data and adjusted indirect comparison
We performed a literature search in the following electronic databases for the pivotal clinical trials pertaining to mRCC treatments: PubMed, EMBASE, CINAHL, AltHealthWatch, the Cochrane Library, and the National Library of Science and Technology. The search covered the periods from the databases inceptions to the end of July 2011. However, no clinical trial directly comparing these five strategies was identified. Therefore, an indirect comparison of key clinical trials was performed. Clinical effectiveness data, including HR (hazard ratio), were separately extracted from the four multicenter, randomisedcontrolled clinical trials that had interferon-alfa as a common comparator; each of these trials constituted level 1 evidence [23].
The MRC RE04/EORTC GU 30012 trial enrolled 1,006 advanced mRCC patients who were randomly allocated (1 to 1) by minimisation to receive interferon-alfa alone or combination therapy with interferon-alfa, interleukin-2, and fluorouracil. After a median follow-up of nearly 37 months, the median OS was 18.8 months for the patients receiving interferon-alfa versus 18.6 months for those receiving the combination therapy (HR 1.05, 95% CI 0.90-1.21, p = 0.55). There was no evidence that the median PFS differed between the treatment groups (5.5 vs. 5.3 months, HR 1.02, 95% CI 0.89-1.16, p = 0.81) [5].
The clinical trial reported by Groupe Francais d'Immunotherapie enrolled 425 mRCC patients who were randomly assigned to receive interleukin-2, interferon-alfa alone or both. There was no significant difference in OS among the three groups. However, the PFS in the combined cytokine treatment group was significantly higher than in the other two groups (P = 0.01) [6]. The HR of the interleukin-2 group compared with the interferon-alfa group was derived from a previous economic study ( [17]. Indirect comparisons of the five strategies were conducted using a hypothetical average interferon-alfa (reference) survival rate. Weibull survival models were fitted to the Kaplan-Meier PFS and OS data for interferon-alfa. The estimated Weibull parameters, scale (l) and shape (c), and their SEs and correlation coefficients are shown in Table 1. The variations in the interferon-alfa survival rates may be attributable to variations in baseline demographics or disease severities. Based on the Weibull model, the PFS and OS rates for interferon-alfa in the four trials were calculated at each cycle. The reference survival rates at each cycle were calculated and weighted according to patient numbers [21]. Once a survival rate was calculated, the survival rates for the active strategies were adjusted by the following formula: S active strategies = (S interferon-alfa(reference) ) HR . All HR data is presented in Table 2. We did not use the Weibull model to fit the Kaplan-Meier data of the other alternative strategies, because there are two important advantages to using the HRs to derive the survival curves, as mentioned by Hoyle, M. et al [16]. First, the method allows for the number of patients at risk on the Kaplan-Meier curve. Second, it allows for the analysis of uncertainty in clinical effectiveness by changing the HRs. The PFS and OS HRs between the alternative strategies and interferon-alfa were derived from the previously mentioned  published studies. The final adjusted Weibull PFS and OS survival rates for the five strategies are shown in Figure 2.

Medical health resources and utilities
A Chinese healthcare system perspective was used to estimate the costs in the current study. Direct medical costs, such as first and second-line treatment-related medicines, radiological and laboratory examinations, management of serious adverse events (SAEs), physician visits, and BSC in the terminally ill, were included. Indirect costs (e.g., lost productivity or premature death) were not included [17]; the costs were converted into 2011 US dollars (Table 3). All the unit costs of the health resources were estimated using data from the local health system or the National Development and Reform Commission (NDRC) of China [24].
The drug costs associated with each strategy were estimated according to the following schedules. The interferon-alfa was assumed to be administered by subcutaneous injection three times per week in first cycle at 3 MU/dose in the first week, 6 MU/dose in the second week, and 9 MU/dose thereafter. The subsequent cycles involved three 9-MU/dose injections. The sunitinib was assumed to be administered orally at 50 mg once daily for 4 weeks, followed by 2 weeks off treatment. Bevacizumab 10 mg/kg or placebo was assumed to be administered intravenously every 2 weeks. Interleukin-2 was assumed to be administered intravenously as 18 MU6body-surface area (m 2 ) daily for 5 days, once every 3 weeks. Sorafenib (a second-line treatment) was assumed to be administered orally at 400 mg twice per day. To estimate the dosage of the therapeutic agents, we assumed that a typical patient weighed 65 kg and had a height of 1.64 m, resulting in a bodysurface area (BSA) of 1.72 m 2 . We assumed that the unused drugs in the open vials were discarded.
Based on a previous study [17], we assumed that 66% of patients would receive second-line treatment when disease progressed. Of those, nearly 33% received second-line targeted therapy (14.2%, sunitinib; 12.2% sorafenib; 1.7% bevacizumab; and 4.9% other targeted agents) [25]; 33.5% received either interferon-alfa or interleukin-2 [17], regardless of the first-line treatment. Because other targeted agents, such as everolimus and axitinib, are unavailable in the Chinese market, we assumed that the 4.9% receiving other targeted agents would switch to sorafenib, based on expert opinion. Because no detailed BSC treatment was reported in the clinical trial, we assumed that the major cost would be pain medications (morphine sulphate) and that the mean dosage would be 300 mg taken twice daily.
A previous study has shown that hematologic toxicities and fatigue/asthenia are the main drivers of management costs in sunitinib treatment and that proteinuria and fatigue/asthenia are the main drivers in bevacizumab plus interferon-alfa treatment [26]. Our model incorporated these treatment-related AEs. Other high incidence toxicities, such as nausea and hypertension, were also included. SAEs (grade$3) management strategies were based on clinical practice and expert opinions. The incidences of SAEs were sourced from clinical trials ( Table 2). The unit costs of treating SAEs were estimated based on patient records in local hospitals.
The utility values of the PFS and PS states were derived from previous published studies; 0.60 and 0.45 were assigned for PFS and PS, respectively. Their standard errors were estimated at 10% of the mean in our sensitivity analysis [16].

Sensitivity Analyses
One-way deterministic sensitivity analyses were used to identify key model input parameters over the low/high values, which are listed and illustrated in Table 2 and Table 3. The results are presented as a tornado diagram based on the impact of the variable on the incremental net health benefit, using 36the per capita GDP of China as the threshold according to World Health Organization(WHO) guidelines for cost-effectiveness analysis13,290 [27,28,29]. Probabilistic sensitivity analyses (PSA) were used to evaluate the impact of uncertainty across all the parameters simultaneously. The values of the input parameters were sampled from lognormal distributions for costs and from b distributions for utility values and probabilities or proportions with an assumed standard deviation of 10% from mean values. Using these distributions, 1,000 iterations of 1,000 simulated patients were run for our analysis. The results are shown on a cost-effectiveness plane. The outcomes projected from all 1,000 simulations were used to plot acceptability curves, which estimated the willingness to pay (WTP) threshold for an incremental unit of effectiveness.

Base Case Result
The base case cost-effectiveness results (Table 4) were estimated with a 10-year time horizon. Much longer periods of PFS achieved by the targeted therapies (sunitinib and bevacizumab plus interferon-alfa) resulted in longer survival times with more QALYs than could be achieved with cytokine therapies (interferon-alfa, interleukin-2 and interleukin-2 plus interferon-alfa). The sunitinib strategy produced an average of 1.71 years in the PFS health state, compared to 0.59 years, 1.04 years, 0.67 years and 0.58 years for those receiving interferon-alfa, bevacizumab plus interferon-alfa, interleukin-2 and interleukin-2 plus interferon-alfa, respectively. The sunitinib strategy gained the greatest number of QALYs over the course of the disease (1.40), compared to 1.11 QALYs for the interferon-alfa strategy and 1.23 for the bevacizumab plus interferon-alfa strategy. The acquisition PFS time of sunitinib largely explains its higher QALY output.  Figure 3A). Regarding the SPAP, the scenario analysis showed that the sunitinib strategy achieved dominance; the bevacizumab plus interferon-alfa and interleukin-2 plus interferon-alfa strategies were dominated, and the interferon-alfa strategy was extended dominated (Table 4 and Figure 3B).

Sensitivity Analyses
The tornado diagram (Figure 4) revealed that the net health benefit of sunitinib vs. that of interleukin-2 was sensitive to some of  the input parameters; the PFS HR of sunitibinb was the most influential factor. Changing the PFS HR for sunitinib vs. interferon-alfa in the range of the 95% CI had the effect of changing the net health benefit significantly. At the upper boundary of the HR, which resulted in a shorter PFS with sunitinib, the net health benefit decreased to 26.26 QALYs (WTP = $13,290). A longer PFS for sunitinib was observed at the lower boundary of the HR, with the net health benefit increasing to 23.81 QALYs. The other important drivers of the model were the OS and PFS HR for interleukin-2, discount rate and utilities. Other factors, such as the costs of managing SAEs, had little impact.
The plot data from the PSA of 1,000 simulations revealed the probabilities of meeting the ICER thresholds of $13,290 per additional QALY for sunitinib over bevacizumab plus interferonalfa, interferon-alfa, interleukin-2 and interleukin-2 plus interferonalfa. The results are shown in Figure 5. With the SPAP, the probabilities of achieving cost-effectiveness with sunitinib for interferon-alfa, bevacizumab plus interferon-alfa, interleukin-2 and interleukin-2 plus interferon-alfa were 75.2%, 100%, 45.1% and 95.3%, respectively, under the $13,290 threshold. Without the sunitinib patient assistant program, the probabilities of achieving cost-effectiveness with sunitinib were all zero when compared to all other options, except for bevacizumab plus interferon-alfa (,99%). The cost-effectiveness acceptability curves (CEACs) showed the preferred first-line strategies for mRCC when accounting for a range of cost-per-QALY thresholds. The CEAC plot shows that when no patient assistant program was offered, interleukin-2 achieved a nearly 89% likelihood, when the threshold level was $13,290 ( Figure 6A). When the program was offered, sunitinib achieved likelihoods of nearly 40% ( Figure 6B). Sunitinib and interleukin-2 shared a likelihood of nearly 50% when the threshold was $16,000. In China, local governments have the power to add additional therapies ino basic medical services supplied by central government according to their economic development level [30,31]. Although interleukin-2 achieved the majority of costeffective probability at the threshold of 36average per capita GDP of China, local governments are still expected to add new therapies. Table 5 listed the cost-effective probabilities of 5 alternative strategies for 32 Chinese provinces at 36local per capita GDP.

Discussion
New targeted therapies for the first-line treatment of mRCC have increased survival rates and improved quality of life. Nevertheless, the widespread use of new targeted therapies comes with a dramatic increase in health care costs. A costeffectiveness evaluation of the recommended first-line therapies in a health resource-limited setting is necessary to balance the economic burden with the health benefits. Using indirect comparison and decision-analytic modelling techniques, we estimated the cost-effectiveness of five first-line mRCC strategies over a lifetime horizon from the perspective of the Chinese healthcare system.
Our results suggest that targeted therapies as first-line treatment for mRCC provide more health benefits than do traditional cytokine therapies. Although the enhanced survival benefits resulting from the targeted therapies were not significant, the prolonged PFS survival improved the benefit of the targeted therapies over cytokines, which is an important reason why they are now covered by some developed countries [18]. Nevertheless, the gap between the costs of the targeted therapies and payment capacity in a health resource-limited setting is still large. The ICERs of sunitinib and bevacizumab plus interferon-alfa compared to cytokine therapies are far greater than the societal willingness-to-pay ($13,290for China). Interleukin-2 and interferon-alfa are still practical options in a resource-limited setting. An economic evaluation from the perspective of the UK NHS estimated that the cost per QALY ranged from $109,522 for sunitinib to $262,535 for bevacizumab plus interferon-alfa. The two therapies could not be considered cost-effective at a willingness-to-pay threshold of $46104/QALY, a result which is consistent with our findings [32]. Although sunitinib showed costeffectiveness as a first-line mRCC treatment from the perspective of the US, with a threshold of $50,000 to $100,000 per LY or QALY, the results are not applicable to the developing regions of the world because their thresholds are far less than $50,000.
If the sunitinib patient assistance program were available to patients from poor regions, the sunitinib strategy might be the optimal alternative option when the threshold of the willingnessto-pay is greater than $16,000( Figure 6B). However, our analysis showed the threshold of sunitinib strategy with the program is still higher than the value of thrice the average Chinese GDP per QALY, which indicates that adding sunitinib into the basic medical service would not be cost-effective for Chinese central government. The Chinese mainland has 32 provinces, among which the per capita GDP differs significantly. In 2010, for example, the per capita GDP ranged from $1,953 in Guizhou province to $10,828 in the city of Shanghai as showing Table 5 [27]. Local government could consider covering sunitinib in their local supplemental medical service according to local economic development level. The results from Table 5 could supply the decision information for local governments. At present, bevacizumab is also an alternative opion for mRCC in China, so the physicians, decision makers or patients would make the choice between sunitinib and bevacizumab. Our results showed sunitinib provides more health benefits for lower costs than bevacizumab plus interferon-alfa, regardless of the SPAP. Our findings are consistent with a recently published study based on the perspectives of the healthcare systems of the US and Sweden, which used indirect comparisons of survival data. The study showed that sunitinib as a first-line option for mRCC was a cost-effective alternative to bevacizumab plus interferon-alfa, because sunitinib was more effective and less costly [33].
The current study has some limitations. Using a Weibull distribution to extrapolate the outcomes beyond the time horizon of the trial follow-ups was an inevitable limitation of this analysis. The one-way sensitive analysis showed that the HR of the PFS and OS had important impacts on the final result. The short median followups of the pivotal clinical trials did not provide enough survival data for a comparison with the median survival that was estimated by the model. At the same time, we did not consider the potential benefits of many of the uncertainties surrounding long-term survival rates, such as new therapeutic agents for second-line therapy, which might improve survival and life quality. Our results could be updated when long-term data are available. However, no randomised controlled trial has yet determined the long-term mRCC outcomes of first-line, second-line, or third-line therapies to the point of death. Therefore, it may be difficult to accurately measure the benefits of first-line therapy in future analyses, suggesting that a modelling technique may be the only realistic alternative.
Another potential limitation was the choice the Chinese healthcare system as our baseline perspective, which led to only direct medical costs being included in the model. Considering a societal perspective, which adds the additional burden of disease on families and caregivers and other indirect costs, may increase the costs associated with mRCC. As such, oral medications (e.g., sunitinib), prolonged PFS (e.g., sunitinib and bevacizumab) and reduced toxicity may produce more favourable results. There is no well-established method for incorporating such societal perspectives when measuring the impact of cost-effectiveness of first-line therapies for mRCC. Because of the absence of head-to-head trials for the five firstline strategies for mRCC, an indirect comparison was used in the present analysis, which was another inevitable limitation. Similar cohort characteristics for the five strategies were assumed in our indirect comparison, and the results of the indirect comparison were imputed into the analytical model. Nevertheless, when no direct data is available, indirect comparisons using robust methods are accepted by many authors worldwide. Future research should directly compare the clinical efficacy of these strategies, especially those of different targeted therapies.
Other important limitations of the current economic analysis should be considered. In particular, we did not fully explore other therapeutic strategies for mRCC treatment, such as temsirolimus, everolimus and pazopanib, because they are still awaiting approval from the State Food and Drug Administration of China [10]. Targeted therapies have shown more favourable health benefits for certain subgroups [34]. Therefore, optimising the selection of the patients receiving targeted therapies could increase the costeffectiveness of more expensive strategies. However, we did not present economic outcomes for such subgroups, because we were unable to adjust the five strategies in an indirect comparison applicable to such subgroup cohorts. Finally, utility values were obtained from literature published abroad and thus may not reflect Chinese data. However, opinions from Chinese urologists and oncologists suggest, quality of life of mRCC patients in China should not be significantly different from external mRCC patients. Although utilities have some impact on the result, the results of sensitivity analysis indicated that the influence is limited. Nevertheless, we are confident that the model faithfully represented the common clinical conditions of mRCC in a health resource-limited setting. We believe this study has the potential to be an important reference point for decision makers.

Conclusion
In the Chinese healthcare setting, a representative health resource-limited region, traditional cytokine therapy is the costeffective option. When the threshold is higher than $16,000, sunitinib might be a cost-effective therapy option compared to bevacizumab plus interferon-alfa, interferon-alfa, interleukin-2 and interleukin-2 plus interferon-alfa, based on its superior PFS benefit and Patient Assistance Program.

Author Contributions
Conceived and designed the experiments: WX. Performed the experiments: BW. Analyzed the data: BW. Contributed reagents/materials/ analysis tools: BJD YJX. Wrote the paper: BW. Provided the clinical inputs and validated the model assumptions: QZ HFC. Defining the economic setting: JFS.