Cost-effectiveness analysis of the introduction of S-1 therapy for first-line metastatic breast cancer treatment in Japan: results from the randomized phase III SELECT BC trial

A number of novel anticancer drugs have been developed this decade, leading to a gradual improvement in the outcomes of cancer patients. However, the economic influence of these drugs on current public medical expenditures has become substantial due to their relatively high prices. Under these circumstances, and with the present budget constraints in healthcare, it is important to consider not only the safety and efficacy, but also the cost-effectiveness of anticancer drugs. In fact, many health technology assessment (HTA) organizations focus on new innovative anticancer drugs. For example, the National Institute for Health and Care Excellence (NICE) in the UK began evaluating all anticancer drugs in 2016 [1, 2] with the reform of cancer drugs fund. Some HTA agencies (e.g., NICE, the pan-Canadian Oncology Drug Review (pCODR) in Canada, and the Pharmaceutical Benefits Advisory Committee (PBAC) in Australia) concluded that certain chemotherapy regimens are not cost-effective and should not be recommended for routine use under the public healthcare system [3‐5].

S-1 [6] (tegafur with gimeracil and oteracil, Teysuno®/TS-1®) is an oral fluoropyrimidine anticancer drug that does not require intravenous administration, unlike many other chemotherapy agents. Thus, patients receiving oral S-1 therapy do not need to bear long hours of intravenous administration and adverse events (e.g. phlebitis) associated with intravenous administration. In addition, a hospital visit is required to receive chemotherapy whenever intravenous anticancer drugs are administered. Therefore, S-1 may not only provide a convenient option for metastatic breast cancer (MBC) therapy, but may also improve the efficiency of treatment. S-1 has been approved in some Asian countries (Japan, Korea, Mainland China, Singapore, Taiwan, etc.) and European countries (UK, Germany, Sweden, etc.) for gastric cancer. However, Japan is the first country to avail S-1 to MBC patients.

SELECT BC [7] is a phase III, open-label, randomized controlled trial (RCT) that compared S-1 with taxanes (paclitaxel or docetaxel) for first-line MBC therapy. According to treatment algorithms (Hortobagyi [8] and NCCN guidelines [9]), patients irresponsive to endocrine therapy receive cytotoxic chemotherapy in standard cases. Taxanes are among the first-choice chemotherapy agents for MBC patients. However, the trial demonstrated non-inferiority of S-1 to taxane in overall survival (OS), with a median OS of 37.2 months in the taxane arm vs. 35.0 months in the S-1 arm (hazard ratio (HR) 1.05, 95% CI 0.86–1.27, p = 0.015), at a median follow-up of 34.6 months. The SELECT BC trial also evaluated cost-effectiveness as a secondary endpoint in addition to some clinical endpoints, including health-related quality of life (HRQOL).

In the SELECT BC trial, EuroQol 5-dimension (EQ-5D) measurements [10, 11] and claims (receipt) data collection for economic evaluation were also performed. These longitudinal patient-level EQ-5D and claims data can be used to calculate quality-adjusted life years (QALYs) and medical costs for evaluation of long-term cost-effectiveness. Such a trial-based [12, 13] cost-effectiveness analysis could improve the robustness of analysis and validity of internal comparison compared to a model-based approach [14] (e.g., using Markov model [15]). In this paper, we report on a cost-effectiveness analysis of the introduction of S-1 therapy to first-line MBC treatment using data from the SELECT BC trial.

We performed a cost-effectiveness analysis of oral S-1 therapy for MBC patients. The analysis was mainly based on information (survival, QOL, and treatment costs) collected from the Phase III randomized SELECT BC trial. Our results suggest that S-1 is cost-saving and the probability of being dominant (i.e., superior in both effectiveness and costs) is high compared with standard taxane therapy. A number of economic evaluations concluded that some anticancer drugs are either not cost-effective or have increased treatment costs even if they are cost-effective. However, our results revealed a high probability that S-1 therapy is cost-saving or dominant with high probability. Considering these results, S-1 may become one of the standard therapies used to treat MBC patients.

This study used Japanese unit costs (e.g., acquisition costs and drug prices) to estimate expected costs of chemotherapy, and our results cannot be simply extrapolated to other countries. However, in Europe, although S-1 has not yet been approved for MBC, the introduction of S-1 therapy for MBC patients may have larger economic effects because the difference in drug costs between the S-1 and taxane group was larger in both the UK and Germany than Japan. According to the British National Formulary (BNF) and Rote Liste, a 20 mg capsule of S-1 is JPY 564.7 (USD 5.1) in Japan, GBP 2.96 (USD 3.7, GBP 1 = USD 1.26) in UK and EUR 6.01 (USD 6.5, 1 EUR = USD 1.07) in Germany. For Docetaxel (Taxotel®), an 80 mg vial is JPY 52,835 (USD 480) in Japan, GBP 504.27 (USD 635) and EUR 783.17 (USD 838) in Germany. A 100 mg vial of Paclitaxel is JPY 22,071 (USD 201) in Japan, GBP 200.35 (USD 252) in UK and EUR 400.57 (USD 428.6) [the lowest price] in Germany. If the drug prices of S-1 and taxane in the UK were applied, the cost would be GBP 6200 (USD 7810) for S-1 and GBP 9310 (USD 11,700) for taxane. Similarly, drug cost as calculated by German pricing was EUR 11,900 (USD 12,700) for S-1 and EUR 16200 (USD 17,300) for taxane. Differences in drug costs between groups in the UK and Germany were larger than those in Japan, because the list price of taxane in the UK and Germany is higher than Japan; conversely, the cost of S-1 is similar or lower.

Our previous study examined longer-term (60 months) EQ-5D index values [26] and reported that the values were higher in the S-1 arm than the taxane arm when the analysis was limited to the first 12 months during progression-free survival (PFS). However, the values did not differ between arms when observations were continued up to 60 months. In the present evaluation, a 48-month analysis was performed to conform to the time horizon of cost-effectiveness analysis, although the above descriptions are also applied to the results of the EQ-5D in this analysis. This suggests that EQ-5D values of S-1 might be higher than those of taxane when patients receive chemotherapy. However, the difference was not statistically significant due to variation in EQ-5D values after chemotherapy, which are longer and more influential toward the results. In fact, the scores of EORTC QLQ-C30 were higher in the S-1 arm than in the taxane arm during 12 months from randomization for global health state (by 4.5; p = 0.039), as well as for all five functional domains including physical functioning, role functioning, emotional functioning, cognitive functioning, and social functioning [7].

In this study, the HRQOL and costs population comprised only a portion of the whole population. Only patients from contacted institutions completed the survey on HRQOL and costs. This design came about after considering the feasibility that some organization could not collect these data because of human resource restraints (e.g. lack of a clinical research coordinator at small institutions). Of course, while this design may have also caused potential selection bias of patients, institution was one of the adjusted factors for allocation in the SELECT BC trial. As shown in Table 1, the background factors of QOL and cost population were comparable to those of the whole FAS population. While the study design may be one limitation of the present investigation, we also feel that our results maintained high internal validity.

There are some limitations to claims data collection in randomized phase III trials. First, expenditures in a clinical trial and daily medical practice may not be the same. This may affect the generalizability of results [27]. However, we believe that the influence was similar in both arms. Second, in this trial, claims data were received from each institution with patient approval. As such, it was difficult to collect data if patients had received treatment from other clinics or hospitals, or changed their hospitals. For example, some patients might have been transferred to another institution to receive terminal care, but such data could not be collected. Although costs of terminal care may differ between the two groups, in many cases with cancer, most procedures are provided by experts; therefore, costs of cancer treatment provided by non-experts (i.e., other clinics and hospitals) can be regarded as unrelated medical costs [28]. Although there remains controversy about the handling of unrelated medical costs, the Japanese economic evaluation guideline [20] recommends that these costs should not be included in treatment costs. Lastly, claims data from pharmacies could not be collected for the same reason. Instead, we recorded prescribed drugs, which were then included in the costs of drugs. Calculating pharmacy fees in Japan is complicated (e.g., it depends on the type of pharmacy), and it is difficult to predict exact fees based only on the information provided by claims and prescription data. In this analysis, pharmacy fees were not included, although the standard pharmacy fee for 28 days of S-1 use ranges approximately from JPY 2000 to JPY 2500 based on a simple calculation. This was not reflected in our results.

The time horizon of our analysis was limited to 4 years. We believe this period is long enough to evaluate the cost-effectiveness of S-1 therapy. Normally, in an economic evaluation, a survival curve is estimated parametrically and extrapolated to obtain an estimated curve; the expected survival time or other measures are calculated using this curve. In the present analysis, the results were not changed even when the survival curve was extrapolated. Therefore, we used a more robust non-parametric Kaplan-Meier method without extrapolation.

The SELECT BC trial is one of the first oncology studies in Japan that collected EQ-5D and claims data continuously over a long period. The present analysis mainly used data from this trial, which enabled a robust analysis, and demonstrated that it is highly likely that oral S-1 therapy is cost-effective. In the area of outcomes research, attention is focused on real-world data (e.g., registry, claims database), although results sometimes have internal validity issues (even if external validity is high) when compared between two different treatment groups. We believe that trial- and real-world-based methods are complementary to each other, and even if studies based on real-world data increase due to improved availability of such data, the importance of trial-based analysis, such as the present study, should not be underestimated.

Our results show that the introduction of oral S-1 therapy for metastatic breast cancer is cost-effective with a high probability. S-1 demonstrates potential for becoming a standard therapy for first-line metastatic breast cancer treatment in comparison with taxenes from the perspective of cost-effectiveness.