Cost-effectiveness analysis of apatinib treatment for chemotherapy-refractory advanced gastric cancer

Abstract

Background

Apatinib, a third-line or later treatment for advanced gastric cancer (aGC), was shown to improve overall survival and progression-free survival (PFS) compared with placebo in the phase III trial. Given the modest benefit with high costs, we further evaluated the cost-effectiveness of apatinib for patients with chemotherapy-refractory aGC.

Methods

A Markov model was developed to simulate the disease process of aGC (PFS, progressive disease, and death) and estimate the incremental cost-effectiveness ratio (ICER) of apatinib to placebo. The health outcomes and utility scores were derived from the phase III trial and previously published sources, respectively. Total costs were calculated from the perspective of the Chinese health-care payer. Sensitivity analysis was used to explore model uncertainties.

Results

Treatment with apatinib was estimated to provide an incremental 0.09 quality-adjusted life years (QALYs) at an incremental cost of $8113.86 compared with placebo, which resulted in an ICER of $90,154.00 per QALY. Sensitivity analysis showed that across the wide variation of parameters, the ICER exceeded the willingness-to-pay threshold of $23,700.00 per QALY which was three times the Gross Domestic Product per Capita in China.

Conclusions

Apatinib is not a cost-effective option for patients with aGC who experienced failure of at least two lines chemotherapy in China. However, for its positive clinical value and subliminal demand, apatinib can provide a new therapeutic option.

Introduction

According to the GLOBOCAN estimates for 2012, gastric cancer (GC) ranked as the fifth most common malignancy and the third leading cause of cancer-related death worldwide (Ferlay et al. 2015). Over 70% of new cases with gastric cancer occurred in developing countries, and half of the world total were in eastern Asia, mainly in China. GC is often diagnosed at the middle or advanced stage. Even after treatment by gastrectomy with D2 lymphadenectomy and adjuvant therapy, patients in stages II and III are still at high risk of recurrence and metastasis, and the 3-year disease-free survival rate is about 72–78% (Sakuramoto et al. 2007; Noh et al. 2014; Lee et al. 2012). Once recurrence or metastasis occurs, the overall 5-year survival rate for advanced GC (aGC) is only 5–20% and median survival time is about 1 year (Kamangar et al. 2006; SEER Cancer Statistics Review 2016; Cunningham et al. 2005).

Both first-line and second-line chemotherapy (Ross et al. 2002; Van Cutsem et al. 2006; Ajani et al. 1994; Hironaka et al. 2013) have been demonstrated to improve survival of patients with aGC. In addition, two target drugs are available for aGC, including trastuzumab combined with chemotherapy in the first-line setting (Bang et al. 2010) and ramucirumab monotherapy or combined with chemotherapy in the second-line setting (Fuchs et al. 2014; Wilke et al. 2014). However, after the failure of these standard regimens, there was no approved treatment available for aGC prior to the introduction of apatinib.

Apatinib, a small-molecule tyrosine kinase inhibitor targeting vascular endothelial growth factor receptor 2 (VEGFR-2), is the first targeted drug demonstrated to be efficient in third-line or later treatment for patients with aGC in China in a phase III study. The results showed that apatinib could significantly improve median overall survival (mOS) compared with placebo (6.5 vs 4.7 months; hazard ratio, 0.709; P = 0.0156) with an acceptable safety profile (Li et al. 2016). As for the most common grade 3/4 adverse events (AEs) of antiangiogenic agents, hand–foot skin reaction (HFSR), hypertension, and proteinuria occurred in 8.5, 4.5, and 2.3% of patients in the apatinib group, respectively. Moreover, common grade 3/4 hematologic AEs related to apatinib included anemia (6.3%) and neutropenia (5.7%). Based on data from the phase III study, apatinib was approved for aGC after second-line chemotherapy by the China Food and Drug Administration (CFDA) in 2014 (Apatinib got CFDA approval 2015).

Although apatinib can be considered as a new option for chemotherapy-refractory aGC, given its modest benefit with high costs, it is very important to consider the value of this treatment relative to its benefit (Ilson 2016). Such consideration of cost-effectiveness may be critical for clinicians or patients as well as from a social perspective. In the current study, we developed an economic model to evaluate the cost-effectiveness of apatinib as third-line or later treatment for aGC from the perspective of Chinese society.

Materials and methods

Patients and regimens

The population analyzed in the current study was derived from the phase III, randomized, controlled study that compared apatinib with placebo in patients with aGC who had experienced failure of at least two lines of chemotherapy (Li et al. 2016). As a result, 176 patients were randomly assigned to receive 850 mg oral apatinib once per day until development of intolerable toxicity or disease progression, or withdrawal of consent from the study. There were 91 patients assigned to the placebo group. The percentage of patients who received more than two lines of chemotherapy before enrollment was 34.8% in the study. The demographics and baseline characteristics of both groups were well balanced.

Model structure

A Markov model was established by TreeAge Pro Suite 2011 (TreeAge Software Inc., Williamstown, MA) to simulate the disease process of aGC and estimate the incremental cost-effectiveness ratio (ICER) of apatinib to placebo. A life-span time for patients was divided into three health states including progression-free survival (PFS), progressive disease (PD), and death according to disease turnover. All patients initially entered the model in the PFS state and could remain in the original state or move to the other states at the end of each model cycle, based on the transition probability between the two health states (Fig. 1). Each model cycle represented 1 month, and the time horizon was 1 year. Transition probabilities of health states per model cycle were calculated based on the equation: P (1 month) = 1 − (0.5) ^{(1/median time to event)}, which was a descendant of the equations: P = 1−e^−R and R = −ln[0.5]/(time to event/number of treatment cycles) (Sonnenberg and Beck 1993; Purmonen et al. 2008).

Fig. 1.

Markov model for advanced gastric cancer. Markov decision tree model comprised three health states (progression-free survival, progressive disease and death) and simulated the disease process

Cost estimates

Total costs were calculated from the perspective of the Chinese health-care payer, which meant all the costs were included in cost estimate regardless of whether the costs were paid by the patients or by the social medical insurance. Total costs included two parts: direct medical costs and societal costs. Direct medical costs were related to drugs, tests, and treatments for grade 3/4 AEs as well as outpatient fees (registration fees), while societal costs referred to travel fees and time costs (absenteeism fees). Expenses for drugs and tests consulted the 2015 fee standards of West China Hospital, Sichuan University, respectively. On the condition that average societal costs were assumed to be identical, travel costs were estimated at $10.20 per patient each trip to the hospital based on referring to a published cost-effectiveness analysis from the same hospital as ours (He et al. 2013) and the taxi fare per kilometer in Sichuan, China, in 2016. Time costs were estimated at $31.58 per day according to the average monthly salary in China in the first half of 2016. Travel costs and absenteeism fees were applied to the scheduled outpatient visits of 2 times a month, 1 day each time. Costs after PD were calculated according to the placebo group costs without considering the expense of subsequent supportive care. All costs were converted to US dollars according to the exchange rate of $1 = RMB 6.5590 in January 2016 (CNY Central Historical Parity Rate 2016).

Effectiveness estimates

The survival data were extracted from the results of the phase III study and converted into quality-adjusted life years (QALYs) gained. Utility scores of Markov states were derived from previously published studies, with 0.797 for the PFS state and 0.577 for the PD state (Shiroiwa et al. 2011). Since the upper limits of the 95% confidence interval of OS in the two groups were significantly <1 year, the annual discount of costs and effectiveness was not considered.

Sensitivity analysis

One-way sensitivity analysis was performed to explore the influence of factors on our model. The parameters analyzed were varied over a range of ± 30% of their baseline values. Probabilistic sensitivity analysis using a Monte Carlo simulation of 1000 patients was performed to quantify the uncertainty because of the variation of factors and estimate probabilities of different treatments being considered as optimal strategies at varying willingness-to-pay (WTP) thresholds, and the results of which were summarized by cost-effectiveness acceptability curves. Based on the WHO Guide to Generalized Cost-Effectiveness Analysis, the WTP threshold in our analysis was set to three times the Gross Domestic Product per Capita (GDP) of China in 2015, which was $23,700/QALY. (WHO Guide to Generalized Cost-Effectiveness Analysis 2016; List of Chinese administrative divisions by GDP per capita 2016).

Results

Health outcomes

With reference to the results of the phase III study, the median PFS and median OS were 2.6 and 6.5 months in the apatinib group, respectively. For patients treated with placebo, the median PFS was 1.8 months and median OS was 4.7 months. Efficacy and grade 3/4 AEs are shown in Table 1. Based on the aforementioned equation, the monthly transition probability from the PFS state to the PD state (P _PFS–PD1) was 0.234 for the apatinib group, from the PD state to death (P _PD–D1) 0.163, and from the PFS state to death (P _PFS–D1) 0.101. Similarly, the transition probability from the PFS state to the PD state (P _PFS–PD2), PD state to death (P _PD–D2), and PFS state to death (P _PFS–D2) for the placebo group was 0.320, 0.213, and 0.137, respectively. When the Markov process ended, apatinib treatment provided 0.36 QALYs compared with 0.27 QALYs for the placebo.

Table 1.

Clinical efficacy and adverse events

Variable	Base-case value
	Apatinib	Placebo
Clinical efficacy—months (95% CI)
Median OS	6.5 (4.8–7.6)	4.7 (3.6–5.4)
Median PFS	2.6 (2.0–2.9)	1.8 (1.4–1.9)
Probability of grade 3/4 AEs—%
Leukopenia	1.7	0
Neutropenia	5.7	1.1
Anemia	6.3	4.4
Thrombocytopenia	2.8	1.1
Elevated transaminase	8.0	4.4
Proteinuria	2.3	0
Hypertension	4.5	0
HFSR	8.5	0
Fatigue	2.8	2.2
Decreased appetite	2.8	1.1
Diarrhea	1.1	1.1

OS overall survival, PFS progression-free survival, AEs adverse events, HFSR hand–foot skin reaction

Cost

The estimated monthly costs of apatinib, tests, grade 3/4 AE treatments, and outpatient as well as societal costs are shown in Table 2. To summarize, the total monthly costs in the apatinib group were $3391.40 for the PFS state, of which apatinib expenses accounted for 89.0% ($3018.33/$3391.40) and $366.08 for the PD state, while the values in the placebo group were $366.08 for both the PFS state and PD state. After running the Markov model to the estimated time horizon, the cumulative costs were $9915.48 for the apatinib group, which was significantly higher than that of $1801.62 for the placebo group (Table 3).

Table 2.

Costs and utility scores

Variable	Base-case value
	Apatinib	Placebo
Direct medical costs for PFS state—$/month
Apatinib	3018.33	0.00
Tests	223.74	223.74
Outpatient fees	9.12	9.12
Grade 3/4 AEs	36.25	29.26
Total direct medical costs for PFS state—$/month	3287.44	262.12
Societal costs—$/month
Absenteeism	63.16	63.16
Travel	40.80	40.80
Total societal costs—$/month	103.96	103.96
Total costs for PFS state—$/month	3391.40	366.08
Total costs for PD state—$/month	366.08	366.08
Utility scores
Utility for PFS state	0.797	0.797
Utility for PD state	0.577	0.577

AE adverse event, PFS progression-free survival, PD progressive disease

Table 3.

Results of cost-effectiveness analysis

Parameters	Apatinib	Placebo
Costs—$
Costs for the PFS state	8378.02	615.64
Costs for the PD state	1537.46	1185.98
Total costs	9915.48	1801.62
Effectiveness—QALYs
Effectiveness for the PFS state	0.16	0.11
Effectiveness for the PD state	0.20	0.16
Total effectiveness	0.36	0.27
C/E ratio—$/QALY	27,543.00	6672.67
ICER—$/QALY	90,154.00

QALY quality-adjusted life year, C/E cost/effectiveness, ICER incremental cost-effectiveness ratio

Cost-effectiveness

Taking the above cumulative costs and effects into account, the apatinib strategy was more costly, with a cost of $27,543.00 per QALY compared with $6672.67 per QALY for the placebo. Treatment with apatinib was estimated to provide an incremental 0.09 QALYs at an incremental cost of $8113.86 compared with the placebo, resulting in the incremental cost-effective ratio (ICER) of $90,154.00 per QALY, which was significantly higher than the WTP threshold of $23,700.00. Detailed information is shown in Table 3.

Sensitivity analysis

The results of the one-way sensitivity analysis are shown in the tornado diagram (Fig. 2). Duration of the PFS state for the apatinib group and cost of apatinib played a key role in the model. When PFS in the apatinib group varied from 1.82 to 3.38 months, the ICER decreased from $112,202.57 to $59,174.41 per QALY. If cost of apatinib increased from $2112.83 to $3923.83 per month, the ICER increased from $56,423.81 to $99,044.73 per QALY. In addition, utilities of different states, duration of PFS state for the placebo group, duration of PD state, costs after PD, and costs of tests in the two groups were important influential factors on ICER. Societal costs and outpatient fees had minor influence on ICER. However, across the wide variation of each parameter, the ICER still exceeded the WTP threshold of $23,700.00 per QALY. After the Monte Carlo simulation of 1000 patients, the mean costs in apatinib group were $9952.570 [95% confidence interval (CI) 9969.289–9935.851], with 0.368 QALYs (95% CI 0.367–0.369). The results of the probabilistic sensitivity analysis further explored no probability that apatinib was considered to be a cost-effective strategy when WTP value was <$70,000.00 per QALY. If the threshold of cost-effectiveness analysis increased to $81,870.00 per QALY, apatinib may be considered acceptable (Fig. 3).

Fig. 2.

Tornado diagrams of one-way sensitivity analyses. Tornado diagrams show the influence of factors on the Markov model. The factors are listed in descending order of the influence on ICER with variation of factor values. (ICER incremental cost-effectiveness ratio, PFS progression-free survival, PD progressive disease, AE adverse event)

Fig. 3.

Cost-effectiveness acceptability curves. The cost-effectiveness acceptability curves summarize the results of probabilistic sensitivity analysis by estimating probabilities of different treatments being considered as optimal strategies at varying WTP thresholds. (WTP willingness to pay, CE cost-effectiveness, QALY quality-adjusted life year)

Discussion

To our knowledge, for the patients with aGC experience progression after two or more lines of chemotherapy, apatinib was the first and the only efficient regimen with an acceptable safety demonstrated in a phase III study. Apatinib prolonged median OS for only 1.8 months accompanied with significantly increased costs. It remains unclear whether apatinib is a cost-effective strategy (Ilson 2016). The current study is the first to evaluate the health and economic outcomes of apatinib as third-line or later treatment for patients with aGC. According to our analysis, apatinib treatment provided an incremental 0.09 QALYs at an incremental cost of $8113.86 compared with placebo, resulting in an ICER of $90,154.00 per QALY, which was significantly higher than the WTP threshold set in our study ($23,700.00/QALY). In one-way sensitivity analyses using varied parameters over a range of  ± 30 , the ICER exceeded $23,700.00 per QALY. Subsequent probabilistic sensitivity analysis revealed that it was highly unlikely that apatinib would be cost-effective at any WTP threshold <$600,000.00 per QALY. Thus, on the basis of our model, apatinib is not a cost-effective choice for patients with chemotherapy-refractory aGC in China.

In view of poor prognosis of aGC and limited survival improvement by current targeted therapy, it is difficult for expensive targeted drugs to become cost-effective, let alone in the third-line setting. It is well established that the ToGA trial indicated that for aGC with HER2 over expression or amplification, the median OS was 16.0 months in the trastuzumab plus chemotherapy group, compared with 11.8 months in the chemotherapy alone group. Despite median survival prolongation of 4.2 months being considered a breakthrough for aGC treatment, a cost-effectiveness analysis based on the ToGA trial in Japan demonstrated that when the threshold of WTP was JPY 6.0 million (€55,000.00) per QALY, the probability of trastuzumab being considered cost-effective was only 23.4% for the population with IHC2+/FISH+ or IHC3+ tumors and 47.9% for the population with IHC3+ tumors, of which median OS reached 17.5–17.9 months. Finally, the study concluded that trastuzumab treatment may be cost-effective for IHC3+ populations, whereas the cost-effectiveness for IHC2+/FISH+ or IHC3+ populations is debatable (Shiroiwa et al. 2011). Ramucirumab, a monoclonal antibody VEGFR-2 antagonist, is a standard option for aGC in the second-line setting. Compared with placebo, ramucirumab improved median OS by 1.4–5.2 months in the REGARD study (Fuchs et al. 2014). There has been no health economic evaluation of ramucirumab for aGC. From the view of improved survival, apatinib did not appear to be inferior to ramucirumab and was similar in safety. The former data are 1.8 months and derived from third-line or later settings. Because of racial heterogeneity and different data sources, differences in efficiency between the two angiogenesis antagonists need to be further verified. However, apatinib as an oral drug is more convenient, while reducing the costs.

Similarly, regorafenib, an oral inhibitor of angiogenesis targeting VEGFR-2, was approved by the US Food and Drug Administration to become a standard care regimen for metastatic colorectal cancer (mCRC) refractory to standard treatments, which improved median OS by 6 weeks compared with placebo (6.4 vs 5.0 months) in the CORRECT trial (Grothey et al. 2013). However, a recent study of cost-effectiveness analysis from the US payer perspective indicated that regorafenib provided minimal incremental benefit at high ICER of $900,000 per QALY versus placebo in the third-line treatment of mCRC and exceeded the generally accepted values for cost-effectiveness (Goldstein et al. 2015).

In one-way sensitivity analysis, duration of the PFS state in the apatinib group and cost of apatinib were the most important factors affecting ICER. Efficacy is uncontrollable unless validated biomarkers predicting clinical outcomes are found. Fortunately, the China Pharmaceutical Innovation and Research Development Association (PhIRDA) initiated a charitable assistance program in May 2015 to provide donations of apatinib to patients with low income who paid for apatinib for the initial 3 months and had an effective outcome. Under the assistance program, apatinib is still not an economical drug, but the economic burden will be decreased from the patients’ perspective. In addition, the cost-effectiveness of apatinib could be improved by using value-based pricing. According to our calculation, if the cost of apatinib reduces by 77.6% ($675.38 per month), it may become a cost-effective option.

There are limitations to this study. First, our analysis was based on the data of the phase III trial, which was not patient-level data in clinical practice. Second, because of the lack of precise utility scores for the PFS and PD states during third-line treatment for aGC, we estimated these values based on published articles and performed further sensitivity analysis to evaluate the impact of these uncertain parameters on our results. Third, dose modifications owing to toxicity occurred in 21.0% of patients treated with apatinib. However, it was difficult to estimate an average dose without providing relative information in the article. Considering that the drug is available in 425-, 375-, and 250-mg tablets, we calculated ICER according to the permissible doses of 800, 675, and 500 mg once per day, and the corresponding ICER values were $86,828.33, $76,660.67, and $63,189.78 per QALY, respectively, which all exceeded the threshold of WTP in the current analysis. An additional limitation is that so far, clinical research on apatinib has all been carried out in China. The heterogeneity between Asian and Western populations may lead to differences in efficacy and toxicity which may weaken the generalizability of our results.

In conclusion, our study demonstrated that apatinib provided modest additional benefit at high ICER in the third-line or later treatment for aGC in China. However, for its positive clinical value and subliminal demand, it can provide a new therapeutic option for the population. Value-based pricing may allow apatinib to become cost-effective from the social perspective. Additionally, new assistance programs or medical insurance can also decrease the economic burden from the perspective of patients with aGC. Presently, there were many studies exploring apatinib in combination with chemotherapy as first- or second-line therapy (Roviello et al. 2016; ClinicalTrials.gov 2015; AHEAD-301 2016). We look forward to the results of these studies contributing to guide high-value medical interventions of patients with aGC.

Compliance with ethical standards

Conflict of interest

The authors declare no conflict of interest.

Ethical approval

This article does not contain any studies with human participants or animals performed by any of the authors.