2022 Volume 45 Issue 7 Pages 895-903
A model-based, cost-effectiveness analysis was conducted to evaluate the difference in cost-effectiveness of nivolumab (NIVO), between first-line therapy in combination with chemotherapy and third-line or later monotherapy for patients with unresectable, advanced or recurrent gastric or gastro-esophageal junction cancer, with the aim of supporting the economic evaluation of healthcare in Japan. Data on overall survival and progression-free survival were obtained from the phase 3 clinical trials, ATTRACTION-4 and ATTRACTION-2. A partitioned survival model was developed to predict costs and outcomes. Direct medical costs were considered from the perspective of the Japanese national health insurance (NHI) payer. The model time horizon was set to 10 years. Health outcomes were defined as life years (LYs) and quality-adjusted life years (QALYs) gained. The incremental cost-effectiveness ratio (ICER) of NIVO compared to the control group was estimated. Sensitivity analyses were performed to assess the uncertainty of parameter setting. A willingness-to-pay threshold of 15 million JPY (Japanese yen) was used. Compared to each control group, the ICERs for NIVO treatment per 1 LY gained were 65745714 JPY for first-line treatment, 7420202 JPY for third-line or later treatment, and 74750097 JPY and 10496602 JPY per QALY gained, respectively. Probabilistic sensitivity analyses estimated that the probability of NIVO treatment being cost-effective for first-line and third-line treatment was 23.5 and 74.3%, respectively. From the perspective of the Japanese NHI payer, NIVO was cost-effective as third-line or later monotherapy for patients with advanced gastric cancer, but not in combination with first-line chemotherapy.
Gastric cancer (GC) is the fifth most prevalent cancer worldwide and the fourth leading cause of cancer deaths.1) Incidence rates are highest in Eastern Asian, and it is the third most common cause of cancer death in Japan.2,3) Systemic chemotherapy is indicated for GC patients with unresectable, advanced recurrence or noncurative resection who are in relatively good general condition and have preserved major organ function.4–6)
Nivolumab (NIVO) is a programmed cell death-1 (PD-1) immune checkpoint inhibitor that uses the body’s immune system to reactivate an anti-tumor immune response by inhibiting the PD-1 and PD-1 ligand pathways. Although two-drug combination chemotherapy with fluoropyrimidine and platinum was recommended for first-line treatment of Human Epidermal Growth Factor Receptor Type 2 (HER2) -negative patients, two recent trials have shown the benefit of NIVO in combination with chemotherapy for first-line treatment. CHECKMATE649 is a phase III trial evaluating NIVO combination therapy as first-line therapy versus chemotherapy alone in HER2-negative, advanced, unresectable gastric cancer and esophagogastric junction cancer, in which about 25% of the patients were Asian.7) In CHECKMATE649, the NIVO combination therapy showed a median overall survival (OS) of 13.8 months and a median progression-free survival (PFS) of 7.7 months. ATTRACTION-4 was a phase II/III clinical trial also conducted in Japan, Korea, and Taiwan.8,9) It confirmed a similar trend, although there was no significant difference in OS. Based on these results, NIVO combination chemotherapy is now recommended as first-line therapy in the guidelines.5,6) In December 2021, the Ministry of Health, Labour and Welfare approved NIVO for the first-line treatment of patients with metastatic or locally advanced GC in Japan. NIVO monotherapy was previously approved for late-line therapy after the ATTRACTION-2 trial reported its efficacy in third-line and later therapy. With this approval, it can now be used from the first treatment. Though NIVO is an effective drug for a variety of cancers with a revolutionary mechanism of action, it is also an expensive drug.10) There are several reports of the cost-effectiveness of NIVO for advanced GC. In China, the combination of NIVO and primary chemotherapy was reported to not be cost-effective.11) The data source for this analysis was the results of the CHECKMATE649 trial, which was not conducted mainly in an Asian population. However, considering the effects of race and region, an analysis using the results of the ATTRACTION-4 trial, which was conducted in an Asian population and is thus a more appropriate data source, is needed. Cost-effectiveness analysis tailored to each country’s healthcare system and standard of care will allow international comparisons. In Japan, the Center for Outcomes Research and Economic Evaluation for Health analyzed late-line NIVO monotherapy. The details of this analysis have not been not published, but based on this result, the drug price of NIVO was reduced through expanded approval. In addition, a cost-effectiveness analysis comparing Trifluridine/tipiracil hydrochloride and NIVO for third-line treatment of advanced gastric cancer in Japan has been reported, and an incremental cost-effectiveness ratio (ICER) has been estimated.12) In this study, applying a willingness-to-pay (WTP) threshold of 5 million Japanese yen (JPY) per quality-adjusted life year (QALY), Trifluridine/tipiracil hydrochloride therapy appeared to be more cost-effective than NIVO therapy. However, no study to date has examined the cost-effectiveness of NIVO combination first-line therapy for HER2-negative, advanced, recurrent gastric cancer from the perspective of the Japanese national health insurance (NHI) payer. There has also been no report of examining difference in the cost-effectiveness between first-line and late-line NIVO treatments. From 2019 onwards, the results of health economic evaluations will be used to adjust the prices of drugs in Japan. In order to make universal health insurance sustainable, economic evaluation from the perspective of the NHI payer is important, especially in the field of oncology, where treatment is becoming increasingly expensive. Extreme differences in the cost-effectiveness of the same drug for different targets may pose a serious challenge for the drug pricing system, but to the best of our knowledge, there has been no quantitative evaluation to date. Therefore, a model-based economic evaluation was performed to assess the cost-effectiveness of NIVO therapy for patients with HER2-negative, advanced or recurrent, gastric cancer comparing first-line and late-line treatment with the aim of supporting the economic evaluation of healthcare in Japan.
For first-line therapy, the target population was a cohort of Japanese patients with previously treated, unresectable, non-HER2-positive, gastric, gastroesophageal junction, or esophageal adenocarcinoma. For third-line or later-line treatment, the cohort consisted of patients who had received at least two prior chemotherapy regimens for advanced or recurrent gastric cancer. The cost-effectiveness of first-line and third-line or later NIVO therapy was compared for advanced gastric cancer from the perspective of the Japanese NHI payer.
First-Line Treatment StrategiesNivolumab CombinationNIVO plus chemotherapy (SOX or CapeOX)
Patients were administered intravenous oxaliplatin 130 mg/m2 on day 1 plus either oral S-1 40 mg/m2 (tegafur–gimeracil–oteracil potassium; 40 mg/dose for body surface area (BSA) <1·25 m2, 50 mg/dose for BSA between ≥1·25 and <1·5 m2, and 60 mg/dose for BSA ≥1·5 m2) twice daily on days 1–14 (SOX), or oral capecitabine 1000 mg/m2 twice daily on days 1–14 (CapeOX), every 3 weeks, in addition to either 360 mg NIVO intravenously every 3 weeks.
Control TherapyPlacebo plus chemotherapy (SOX or CapeOX).
Late-Line StrategiesNivolumab MonotherapyPatients received 240 mg/body NIVO intravenously every 2 weeks.
Control TherapyPatients received placebo intravenously every 2 weeks.
Clinical DataThe published results of a randomized, multicenter, double-blind, placebo-controlled, phase 3 trial, ATTRACTION-4, conducted in Japan, Korea, and Taiwan as a first-line chemotherapy were used as the data source for the analysis.9) Eligible patients were aged 20 years and older with previously untreated (except for neoadjuvant or adjuvant chemotherapy completed ≥180 d before recurrence), HER2-negative, unresectable, advanced or recurrent gastric or gastro-esophageal junction cancer (regardless of PD-L1 expression), at least one measurable lesion per the Response Evaluation Criteria in Solid Tumors guidelines (version 1.1), and a baseline Eastern Cooperative Oncology Group (ECOG) performance status of 0 or 1. Patients were randomly assigned to chemotherapy every 3 weeks (intravenous oxaliplatin plus either oral S-1 (SOX) or oral capecitabine (CapeOX), in addition to either 360 mg NIVO intravenously every 3 weeks or placebo.
For first-line treatment, apart from the ATTRACTION-4 trial, there was also another open-label, phase III trial, CHECKMATE649, which was conducted in 29 countries in Asia, Australia, Europe, North America, and South America.7) In this analysis, a scenario analysis using the published results of this CHECKMATE649 trial as another data source was also conducted. Eligible patients were aged 18 years or older. Other inclusion criteria were similar to those of the ATTRACTION-4 trial. Patients were randomly assigned to NIVO (360 mg every 3 weeks or 240 mg every 2 weeks) plus chemotherapy (CapeOX every 3 weeks or leucovorin, fluorouracil, and oxaliplatin [FOLFOX] every 2 weeks), NIVO plus ipilimumab, or chemotherapy alone.
In a review report by the Japanese Pharmaceuticals and Medical Devices Agency (PMDA), the OS of CHECKMATE649 was inconsistent between the overall population and the Japanese population. However, based on the overall judgment together with the consistency of the PFS efficacy of ATTRACTION-4, which was conducted in Asian subjects and included a large number of Japanese subjects, the efficacy in the Japanese population was judged to be promising, leading to its approval. Therefore, it was decided to extrapolate the results of the overall population of ATTRACTION-4 to the Japanese population in this analysis.
For third-line or later-line treatment, the data source was a double-blind, phase III trial, ATTRACTION-2, conducted in Japan, Korea, and Taiwan. Eligible patients were aged 20 years or older with unresectable advanced or recurrent gastric or gastro-esophageal junction cancer histologically confirmed to be adenocarcinoma refractory to, or intolerant of, standard therapy.10) Patients treated with two or more previous chemotherapy regimens in the advanced or recurrent setting were eligible, and they had to have an ECOG performance status of 0 or 1 and a life expectancy of at least 3 months.
Disease ModelingA partitioned survival analysis model was developed to examine expected costs and QALYs of each therapy.13) A partitioned survival model is a type of economic model commonly used to track a theoretical cohort over time, moving through a series of comprehensive and mutually exclusive health states. The area under the survival curve can be used to estimate effectiveness and cost, using common functions such as Exponential and Weibull to represent survival. The partitioned survival model was constructed using 3 health states: PFS; post-progression survival (PPS); and death. Kaplan–Meier curves were extracted and digitized from clinical trials, using WebPlotDigitizer (https://automeris.io/WebPlotDigitizer/). Parametric functions were examined according to a previously published method.13) The parametric functions were determined to be curve-fitted functions based on the Akaike information criterion, Bayesian information criterion, and visual adequacy (Fig. 1). For OS in the first-line treatment, a log-logistic function was selected for the NIVO combination arm and a log-normal function for the control arm; for PFS, a generalized gamma function was selected for the NIVO combination arm, and a log-normal function for the control arm. In the third and later-line treatment, for both the NIVO and control arms, log-normal functions for OS and log-logistic functions for PFS were selected. OS is a clinical parameter that includes death unrelated to advanced gastric cancer. As a result, background mortality was not included in the simulation. A 10-year time horizon was used in the model (i.e., costs and outcomes for patients were considered up to 10 years after the initiation of treatment). This time horizon was assumed to reflect the lifetime of the patients. This was because the model predicted that more than 99% of patients in each treatment group would have died by this point, and there was no need to further extrapolate the model results for decision-making. In this model, patients were assumed to be treated with each strategy until PFS.
OS = Overall survival; PFS = Progression-free survival.
In ATTRACTION-4 trial, there was a difference in the percentage of nivolumab use among the subsequent therapies, and its impact on cost was considered in the model. Based on the results of the post-treatment, ramucirumab plus paclitaxel, nivolumab, and irinotecan were selected as the standard subsequent therapy. According to the Japanese Society of Nephrology, the standard height and weight of Japanese people are 170 cm and 63 kg, respectively.14) Therefore, in this analysis, a hypothetical cohort of population with a weight of 63 kg and a body surface area of 1.73 m2 was used as a base case.
CostsCosts were estimated from the perspective of the NHI payer, and, therefore, included only direct medical costs (Table 1). The direct medical costs considered in this analysis included drugs, outpatient chemotherapy, and palliative care. Costs associated with outpatient chemotherapy include outpatient service fees, intravenous infusion fees, laboratory testing, and diagnostic imaging. There have been no reports of whether the costs of best supportive care, end-of-life hospitalization, outpatient care, and home care in Japan depend on the chemotherapy administered. Based on the results of the Cost of End-of-Life Care Survey in Japan, the estimated cost of end-of-life care was defined at 17481 U.S. dollars (USD).15) The costs for cases who received palliative care in the hospital and died in the hospital were calculated using insurance claims data. The subsequent therapy cost was estimated based on the proportion of drugs used in subsequent therapy in ATTRACTION-4, and the expected cost was estimated using the PFS of each drug in clinical trials or a valid cohort study.9,10,16,17)
| Parameter | Unit cost, JPY (USD) | Range (%) | Distribution | Reference |
|---|---|---|---|---|
| Outpatient chemotherapy (monthly) | ||||
| CT scan diagnostic fee (once in 3 months) | 3330 (31.2) | 70–130 | Gamma | 19) |
| Contrast medium (once in 3 months) | 1670 (15.6) | 70–130 | Gamma | 19) |
| Peripheral blood tests (>10) | 2180 (20.4) | 70–130 | Gamma | 19) |
| Plasma protein immunological test | 320 (3.00) | 70–130 | Gamma | 19) |
| Blood drawing fee | 500 (31.6) | 70–130 | Gamma | 19) |
| Peripheral blood tests diagnostic fee | 2500 (23.4) | 70–130 | Gamma | 19) |
| Biochemical tests fee | 2880 (27.0) | 70–130 | Gamma | 19) |
| Immunological test fee | 2880 (27.0) | 70–130 | Gamma | 19) |
| Outpatient reexamination fee | 1440 (13.5) | 70–130 | Gamma | 19) |
| Tumor maker tests fee | 8000 (74.9) | 70–130 | Gamma | 19) |
| Prescription fee | 840 (7.87) | 70–130 | Gamma | 19) |
| Dispensing fee | 220 (2.06) | 70–130 | Gamma | 19) |
| Dispensing Technology Basic fee | 280 (2.62) | 70–130 | Gamma | 19) |
| Examination in urine | 520 (4.87) | 70–130 | Gamma | 19) |
| Outpatient chemotherapy fee (once per administration) | 12000 (112) | 70–130 | Gamma | 19) |
| Drug cost | ||||
| Palonosetron | 14764 (138) | 70–130 | Gamma | 18) |
| Dexamethason 1.65 mg i. v. | 98 (0.92) | 70–130 | Gamma | 18) |
| Dexamethason 4 mg p. o. | 30 (0.28) | 70–130 | Gamma | 18) |
| Saline solution 100 mL | 140 (1.31) | 70–130 | Gamma | 18) |
| Saline solution 50 mL | 140 (1.31) | 70–130 | Gamma | 18) |
| 5% glucose solution 250 mL | 177 (1.66) | 70–130 | Gamma | 18) |
| Oxaliplatin 100 mg | 58630 (549) | 70–130 | Gamma | 18) |
| Capecitabine 300 mg | 223 (2.09) | 70–130 | Gamma | 18) |
| Nivolumab 240 mg | 366405 (3,430) | 70–130 | Gamma | 18) |
| l-Leucovorin 100 mg | 7107 (66.5) | 70–130 | Gamma | 18) |
| S-1 20 mg | 445 (4.16) | 70–130 | Gamma | 18) |
| 5-FU 1000 mg | 1008 (9.44) | 70–130 | Gamma | 18) |
| Subsequent therapy in first-line NIVO combination therapy (monthly) | 209364 (1960) | 8,10,16,17,18,19) | ||
| Subsequent therapy in first-line control therapy (monthly) | 279167 (2614) | 8,10,16,17,18,19) | ||
| Paliative care | 11775 (110) | 70–130 | Gamma | 15) |
Exchange rate in 2020, 1 USD = 106.8 JPY. JPY = Japanese yen; USD = U.S. dollars.
In this analysis, the effects of serious adverse events with an incidence of 5% or higher and grade 3 or higher were included. Although bone marrow suppression, anemia and anorexia were included, it was assumed there was no adverse event cost associated with the analysis because the basic measure was withdrawal. Costs were calculated according to the social insurance reimbursement schedule and drug tariff of the fee-for-service system in Japan.18,19) Using the exchange rate for the year 2020 (1 USD = 106.8 JPY), costs calculated in JPY were converted to USD.20)
Health-Related UtilityIn this analysis, LYs and QALYs were obtained as measures of efficacy. Total QALYs were estimated by adjusting survival time for health-related QOL. The utility values for each strategy for first-line treatment were obtained from the ATTRACTION-4 trial and were 0.81 for the NIVO combination therapy and 0.83 for the control therapy (Table 2). Because this difference in utility values includes the effects of adverse events and other factors, it was not necessary to consider disutility due to adverse events. In the ATTRACTION-4 trial, EQ-5D was not evaluated after disease progression, so utility values from alternative sources were required. Therefore, for the base-case analysis, the score of 0.6, which was used in a previous Japanese study based on the National Institute for Health and Clinical Excellence (NICE) technology appraisal, was used.21,22) For the third-line or later, no utility value was reported in the ATTRACTION-2 trial. Therefore, the utility values for third-line treatment were determined based on the results of the QOL survey of the INTERGRATE trial, a randomized, placebo-controlled, phase 2 trial of regorafenib in advanced gastric cancer.23) These utility values have been used in a previous cost-effectiveness analysis of NIVO in Japan.12)
| Parameter | Base case | Range | Distribution | Reference |
|---|---|---|---|---|
| Utility value | ||||
| Progression-free survival of Nivolmab combination arm (first-line) | 0.810 | 90–110% | Beta | 8) |
| Progression-free survival of placebo arm (first-line) | 0.830 | 90–110% | Beta | 8) |
| Post-progression survival (first-line) | 0.600 | 90–110% | Beta | 21,22) |
| Progression-free survival (third- or later-line) | 0.730 | 90–110% | Beta | 12,23) |
| Post-progression survival (third- or later-line) | 0.690 | 90–110% | Beta | 12,23) |
| Median relative dose intensity in NIVO combination arm | ||||
| Nibvolumab (SOX) | 0.9377 | 70–130% | Beta | 8) |
| Nibvolumab (CapeOX) | 0.9545 | 70–130% | Beta | 8) |
| Oxalplatin (SOX) | 0.7318 | 70–130% | Beta | 8) |
| Oxaliplatin (CaoeOX) | 0.8242 | 70–130% | Beta | 8) |
| S-1 (SOX) | 0.7811 | 70–130% | Beta | 8) |
| Capecitabine (CapeOX) | 0.7482 | 70–130% | Beta | 8) |
| Median relative dose intensity in placebo arm | ||||
| Oxalplatin (SOX) | 0.7593 | 70–130% | Beta | 8) |
| Oxaliplatin (CaoeOX) | 0.8282 | 70–130% | Beta | 8) |
| S-1 (SOX) | 0.8086 | 70–130% | Beta | 8) |
| Capecitabine (CapeOX) | 0.7711 | 70–130% | Beta | 8) |
| Rate of subsequent therapy in NIVO combination arm | ||||
| Ramucirumab + paclitaxel | 43.6 | 70–130% | Beta | 8) |
| Nivolumab | 15.7 | 70–130% | Beta | 8) |
| Irinotecan | 20.4 | 70–130% | Beta | 8) |
| Rate of subsequent therapy in placebo arm | ||||
| Ramucirumab + paclitaxel | 51.7 | 70–130% | Beta | 8) |
| Nivolumab | 42.1 | 70–130% | Beta | 8) |
| Irinotecan | 21.0 | 70–130% | Beta | 8) |
| Duration of Subsequent therapy (month) | ||||
| Ramucirumab + paclitaxel | 4.40 | 4.2–5.3 | Uniform | 16) |
| Nivolumab | 1.61 | 1.54–2.3 | Uniform | 10) |
| Irinotecan | 3.19 | 2.3–4.08 | Uniform | 17) |
| Discount rate | 0.02 | 0–0.05 | 24) | |
Cost-effectiveness was assessed using the ratio of incremental cost per incremental QALY as the ICER in the cost-utility analysis. The threshold for price adjustment of anticancer drugs in the cost-effectiveness evaluation system introduced in Japan in 2019 was set with reference to the gross domestic product per capita and thresholds in other countries, with three levels: 7.5, 11.25, and 15 million JPY per QALY.24) The experts will scientifically determine the validity of which threshold to apply. In this analysis, the threshold of WTP for the base case analysis was set at 15 million JPY per QALY, based on the maximum threshold. A sensitivity analysis was conducted to examine the variation in the probability of acceptance based on changes in WTP.24,25) A base-case analysis incorporating baseline parameters was performed. In the base-case analysis, costs and QALYs were discounted at an annual rate of 2% based on the Guideline for the Economic Evaluation of Medical Technology in Japan.24)
Sensitivity AnalysisA sensitivity analysis was performed to assess the uncertainty and robustness of the model. In the sensitivity analysis, parameters were selected to cover all areas of uncertainty, including survival curve parameters for PFS and OS, drug costs, and health-related utility values. A one-way sensitivity analysis assessed the impact of changes in each parameter on the ICER. Variations of 10% were considered for utility values to the range that PFS and PPS utilities were not reversed, 95% confidence intervals for curve parameters (but only 30% for the scale parameter because the generalized gamma distribution maintains its shape), and 30% as sufficiently reasonable wide range for other parameters, and discount rates were varied from 0 to 5%. A specific sampling distribution was also assigned to each input parameter, and a probabilistic sensitivity analysis (PSA) was performed with Monte Carlo simulation of 10000 samples. The distributions were chosen according to the nature of the parameters. A normal distribution was used for the curve parameter, a gamma distribution was used for the cost parameter, and a beta distribution was used for the utility parameter and percentage. The results of clinical trials and real-world evidence were used to estimate the cost of subsequent treatment, and uniform distributions based on confidence intervals for PFS were adopted.10,16,17) The parameters of the PFS and OS curves were generated using the variance covariance matrix and multivariate normal distribution. The change of the cost-effective probability with the WTP threshold fluctuation was examined in the cost-effectiveness acceptability curve. All analyses were conducted using TreeAge Pro software (version 2022; TreeAge, Williamstown, MA, U.S.A.).
The results of the base-case model are shown in Table 3. In first-line treatment, NIVO combination therapy yielded benefits of 0.150 LYs and 0.132 QALYs compared with the control strategy. The 10-year incremental cost was 9845389 JPY (92185 USD), with an ICER per LY of 65745714 JPY (615597 USD) and an ICER per QALY of 74750097 JPY (699907 USD). In the third-line or later treatment, NIVO monotherapy yielded benefits of 0.343 LYs and 0.243 QALYs compared with the control therapy. The 10-year incremental cost was 2548509 JPY (23862 USD), with an ICER per LY of 7420202 JPY (69478 USD) and an ICER per QALY of 10496602 JPY (98283 USD). A scenario analysis using published data from the CHECKMATE649 trial showed that the NIVO combination therapy provided a benefit of 0.331 LY and 0.213 QALY compared with the control strategy (Table 4). The 10-year incremental cost was 9632469 JPY (90192 USD), with an ICER per LY of 29134731 JPY (272797 USD) and an ICER per QALY of 45309920 JPY (424250 USD).
| Strategy | Cost, JPY | Incremental Cost, JPY | LYs | Incremental LYs | ICER, JPY/LY (USD/LY) | QALYs | Incremental QALYs | ICER, JPY/QALY (USD/QALY) | |
|---|---|---|---|---|---|---|---|---|---|
| First-line | NIVO combination | 17085501 | 9845389 | 2.162 | 0.150 | 65745714 | 1.564 | 0.132 | 74750097 |
| Control | 7240112 | 2.012 | (615597) | 1.432 | (699907) | ||||
| Third-line or later | NIVO combination | 4761366 | 2548509 | 0.850 | 0.343 | 7420202 | 0.599 | 0.243 | 10496602 |
| Control | 2212857 | 0.506 | (69478) | 0.356 | (98283) |
Exchange rate in 2020, 1 USD = 106.8 JPY. JPY = Japanese yen; USD = U.S. dollars; LY = Life-year; ICER = Incremental cost-effectiveness ratio; QALY = Quality adjusted life-year; NIVO = Nivolumab.
| Strategy | Cost, JPY | Incremental Cost, JPY | LYs | Incremental LYs | ICER, JPY/LY (USD/LY) | QALYs | Incremental QALYs | ICER, JPY/QALY (USD/QALY) | |
|---|---|---|---|---|---|---|---|---|---|
| First-line (CHECKMATE649) | NIVO combination | 15213911 | 9632469 | 1.406 | 1.737 | 29134731 | 1.264 | 0.213 | 45309920 |
| Control | 5581442 | 1.737 | 1.406 | (272797) | 1.051 | (424250) |
Exchange rate in 2020, 1 USD = 106.8 JPY. JPY = Japanese yen; USD = U.S. dollars; LY = Life-year; ICER = Incremental cost-effectiveness ratio; QALY = Quality adjusted life-year; NIVO = Nivolumab.
The top 10 most influential parameters from the one-way sensitivity analysis are shown in Fig. 2. In one-way sensitivity analysis for first-line treatment, the most influential parameters for ICER were OS and PFS survival curve parameters, PFS and PPS utility values, and NIVO price. ICER was over 15 million JPY per QALY, even if the value of each parameter changed greatly. Other parameters had even less effect on ICER. On the other hand, for the third-line or later treatment, the most influential parameters for ICER were NIVO price, OS and PFS survival curve parameters, PFS and PPS utility values, and the discount rate. ICER was below 15 million JPY per QALY, even if the value of each parameter changed greatly.
NIVO = Nivolumab; OS = Overall survival; PFS = Progression-free survival; PPS = Post-progression-free survival; ICER = Incremental cost-effectiveness ratio; JPY = Japanese yen; QALY = Quality-adjusted life year.
PSA results (1000 points from 10000 samples) are presented in the cost-effectiveness planes (Fig. 3). The cost-effectiveness plane plotted the incremental costs against the incremental effects of the NIVO treatment strategy compared to the control. Each point of the cost-effectiveness plane was obtained from the random selection by PSA. The WTP line of the plot had a slope of 15 million JPY per QALY threshold. Points below the WTP line indicate high cost-effectiveness, while points above the line indicate low cost-effectiveness. In the cost-effectiveness acceptability curve, the cost-effective probability of PSA results when the WTP changed (Fig. 4). The probability that NIVO treatment was cost-effective versus placebo was 22.5 for first-line treatment and 74.4% for third-line or later treatment when the WTP threshold was 15 million JPY per QALY.
JPY = Japanese yen; QALY = Quality-adjusted life year; WTP = Willingness-to-pay.
JPY = Japanese yen; QALY = Quality-adjusted life year.
NIVO has been reported to be effective both as first-line and third-line or later therapy in unresectable advanced or recurrent gastric or gastro-esophageal junction cancer.7,9,10) Although the efficacy of NIVO has been reported in various types of cancer other than advanced gastric cancer, there are many cost-effectiveness analyses because of concerns about its economic feasibility due to its high price.26,27) Although the economic impact of NIVO on NHI is large for gastric cancer patients due to the large number of cases, there are no reports comparing the cost-effectiveness between different treatment lines.3) This is the first study examining comparative efficiency between different treatment lines. This analysis examined the economic impact of NIVO treatment using model analyses that extrapolated the results of clinical trials on which approval was based. It was found that NIVO treatment in third-line or later was cost-effective from the perspective of the Japanese NHI payer, whereas NIVO added to chemotherapy in first-line was not cost-effective. Various sensitivity analyses further validated the robustness of the results. Extrapolation of the survival curve contributed significantly, but did not overturn the results. These results provide meaningful information and serve as important evidence for value-based medicine.
The ATTRACTION-4 trial, which examined first-line treatment with NIVO for unresectable advanced or recurrent gastric or gastro-esophageal junction cancer, did not show an OS advantage, but the CHECKMATE649 trial showed an OS advantage, as well as a PFS advantage.7,9) The higher post-treatment rates after first-line treatment in the Asian population are unlikely to result in differences in OS. A scenario analysis using the results of the CHECKMATE649 trial as the data source was also performed, but even including the OS advantage of the CHECKMATE649 trial, the results were not as cost-effective. Even when using the higher benefit first-line evidence available at the time of this analysis, this study showed that there were serious challenges with efficiency. In order to improve efficiency, it is necessary to identify the target population with high benefit. The CHECKMATE649 trial examined the Combined Positive Score (CPS) as a pathological diagnosis for predicting response, suggesting a benefit in response rates and hazard ratios for OS and PFS at a cutoff value of PD-L1 CPS 5.7) Because the ATTRACTION-4 trial did not examine subgroups by CPS, its impact as a prognostic factor in Japanese and Asian populations is unclear.9) If future studies identify prognostic factors such as PD-L 1 CPS in the Japanese population, this analysis could be updated.
The difference in ICERs between first-line treatment and third-line or later treatment was significant. The divided results suggest that problems with the NHI pricing system existed for advanced gastric cancer, but the problems will become even more apparent when a variety of cancer types are included. Under the current Japanese system, NHI prices can be adjusted according to the results of Health technology assessment and market size. However, for the sustainability of the NHI, it may be necessary to consider a more efficient system, such as introducing the current Health technology assessment into insurance reimbursement availability. It is necessary to challenge the coexistence of efficiency and promotion of drug development.
The variability in the plot of PSA results for first-line NIVO combination therapy demonstrated the magnitude of uncertainty. In particular, the results of one-way sensitivity analysis suggested that ICER may not decrease much, but it may increase significantly due to variation of model curve parameters. In the partitioned survival model, the choice of parametric model was made carefully, but extrapolation of the survival curve had a large impact on the results, and the validity of the model might remain a challenge. However, it shows a trend consistent with the results of cost-effectiveness of first-line NIVO treatment in the Chinese healthcare system, although the medical environment is different.11) In addition, the present ICER results for NIVO treatment in late-line advanced gastric cancer showed consistency with that of a previous study.12) It can be said that the validity of the method used in this model analysis was supported. The long-term effects of immune-related side effects are also still unknown. For more accurate extrapolation, future studies are expected to update data sources and re-evaluate efficiency with long-term treatment prognosis, including the impact of side effects.
The ATTRACTION-2 trial included HER2-positive patients, whereas the ATTRACTION-4 trial included only HER2-negative patients, a difference in patient background because the use of the anti-HER2 antibody trastuzumab is recommended as primary therapy for HER2-positive patients. However, post-analysis of the ATTRACTION-2 trial showed that survival outcomes were consistent regardless of whether trastuzumab was used.28) Therefore, this cost-effectiveness comparison appears to be well-justified.
There are several limitations to this study. First, there is still limited information on the long-term prognosis of NIVO-treated patients in primary treatment, so extrapolation beyond the clinical trial period of the partitioned survival model is a challenge. The results of the sensitivity analysis showed that the parameters of the parametric curve had a significant impact, but extrapolation did not seem to have affected the results against WTP, since an increasing trend in ICER was obtained when the curve parameters were changed. Second, because EQ-5D data were not collected in the ATTRACTION-2 trial, the utility value obtained in the INTEGRATE trial, which was also used in the previous study, was used. In the ATTRACTION-2 trial, all patients received two or more lines of chemotherapy, whereas in the INTEGRATE trial, about 60% of patients received two or more lines of chemotherapy.12,23) This may result in an overestimation of the actual utility value. The variability in utility values was examined in the one-way sensitivity analysis, but the results were not reversed. Third, various assumptions about patient background, medical costs, and palliative care costs based on study conditions and prior reports were made, which may lead to different results for specific subgroups. However, from the perspective of the NHI payer, it is necessary to assume an average target population. As for uncertainty, distributions were set for the parameters based on the available and valid evidence, and they were examined in the PSA, which did not affect the results. Furthermore, the results of CHECKMATE649 were used for scenario analysis, which showed that the ICERs were still far from the WTP and that the additional benefit gained by the additional cost of NIVO was marginal. Future studies are needed to identify effective clinical conditions for survival outcomes. Forth, details on the regimens used in the subsequent treatment of each arm of the Attraction-4 trial, the duration of treatment, and other details were not disclosed. We estimated the cost of subsequent treatment, including later-line NIVO, based on the proportion of drugs used in subsequent treatment. The sensitivity analysis also examined the impact of subsequent treatment cost, but the results were not affected. The accuracy of this model could be improved if further data on subsequent treatment after the initial NIVO treatment is made available.
The present results are useful evidence for countries, payers, and decision-makers where NIVO is reimbursed for the treatment of advanced GC and raise a serious issue for the drug pricing system.
From the perspective of the NHI payer in Japan, NIVO is cost-effective as late-line monotherapy for patients with unresectable advanced or recurrent gastric or gastro-esophageal junction cancer, but not as first-line combination therapy with chemotherapy.
The authors declare no conflict of interest.
