Long-term health status as measured by EQ-5D among patients with metastatic breast cancer: comparison of first-line oral S-1 and taxane therapies in the randomized phase III SELECT BC trial

Breast cancer has the highest incidence among cancers in women worldwide, with a global age-standardized rate per 100,000 of 51.7 (75.0 in developed countries), corresponding to an increase of 16.6 % from 1990 [1]. Breast cancer is also a major factor that leads to death during the reproductive years [2]. Although diagnosis and treatment of breast cancer have advanced substantially in the past decade, breast cancer is still a common disease that affects the health-related quality of life (HRQOL) of patients and their families, and contributes significantly to healthcare costs.

The goal of chemotherapy for metastatic breast cancer (MBC) is to prolong survival and maintain HRQOL. Taxane and anthracycline are the first choice of chemotherapy for MBC, although these agents can reduce the HRQOL of some patients due to adverse events such as hair loss, peripheral neuropathy, and edema [3, 4]. Patients also frequently need to visit the hospital in order to receive intravenous chemotherapy, which leads to loss of productivity. S-1 [5] is an oral fluoropyrimidine drug for gastric and other cancers. While continuous infusions of 5-FU are known to be more effective than bolus injections for colorectal cancer patients [6], S-1 offers an advantageous and convenient alternative in that the maintenance of therapeutic blood levels of fluorouracil can be achieved with oral tablets alone, without requiring an injection pump.

SELECT BC is a phase III open-label randomized controlled trial (RCT) that compared S-1 with taxane for first-line MBC therapy [7]. The trial demonstrated non-inferiority of S-1 to taxane in overall survival (OS); median OS was 37.2 months in the taxane arm and 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.

SELECT BC also assessed HRQOL as a secondary endpoint using two instruments: the European Organisation for Research and Treatment of Cancer Core Quality of Life Questionnaire C30 (EORTC QLQ-C30) [8] and the three-level version of the EuroQol five-dimensional questionnaire (EQ-5D-3L) [9]. In terms of the primary endpoint, neither treatment is superior to the other; however, our hypothesis expected HRQOL in the S-1 group to be better than in the taxane group. This finding highlights the importance of considering HRQOL in the selection of chemotherapy. In some randomized phase III trials [10‐20], the HRQOL of MBC patients was reported using the Functional Assessment of Cancer Therapy (FACT) or EORTC, which are both standard cancer-specific instruments.

In this paper, we report our findings from the long-term data on EQ-5D-3L, which is the most commonly used preference-based measure [21, 22]. EQ-5D-3L index values can be also used to calculate quality-adjusted life years (QALYs) for the economic evaluation of healthcare technologies. As cost-effectiveness is included as a secondary endpoint of this trial, EQ-5D-3L index values will be also used for economic evaluation. In the SELECT BC trial, EQ-5D-3L measurements were continued over a prolonged period of time because measurements could be continued even when the disease progressed, whereas EORTC QLQ-C30 measurements were taken only during the first year. In many studies, measurements are stopped at the pre-determined time point or at disease progression. Thus, the long-term index value is often unknown. Against this backdrop, we report long-term EQ-5D-3L index values of MBC patients measured to the extent possible, i.e., until death.

HRQOL is an important outcome for patients with MBC. Our study surveyed long-term EQ-5D index values in the phase III randomized SELECT BC trial, which compared S-1 with taxane as first-line treatments for patients with MBC. We found that EQ-5D index values of respondents were higher in the S-1 arm than in the taxane arm, when the analysis was limited to the first 12 months during PFS. Grade 3 or higher adverse events (AE) including edema and sensory neuropathy occurred significantly more frequently in the taxane group, whereas diarrhea, mucositis, and nausea were more frequent in the S-1 group [7]. Different AE patterns might have influenced EQ-5D index values. EQ-5D index values did not differ between the two arms when observations were continued up to 60 months. These results may be due to crossover to S-1 at the time of disease progression or the influence of second-line or subsequent therapy. These results were supported by the results of the time to deterioration analysis. A previous SELECT BC trial report [7] showed that EQ-5D index values were significantly higher in the S-1 arm for 36 months after randomization. In contrast, the present study reports EQ-5D index values over a longer period (60 months), based on more detailed analyses.

EQ-5D index values after progression and before death were also assessed, although these index values did not significantly differ between groups. The SELECT BC trial demonstrated non-inferiority of S-1 to taxane in terms of OS. Considering our EQ-5D estimates, obtained QALYs may be greater in the S-1 arm if the EQ-5D index value of the S-1 arm during chemotherapy is higher.

The completion rate of the instrument was about 85 % at 3 months after randomization, which is the first time point after baseline. The rate at 12 months was about 75 %. The completion rate in the phase III CLEOPATRA trial [17], in which first-line MBC patients were randomly allocated to pertuzumab plus existing treatment (trastuzumab plus docetaxel) or placebo plus existing treatment, was 83.9 % in the placebo arm and 79.9 % in the pertuzumab arm at 9 weeks. Rates decreased to 76.5 and 81.1 % at 54 weeks (approximately 12 months), respectively. In the randomized JGOG3016 trial [29] for stage II–IV ovarian cancer conducted in Japan, completion rates were 74.5 % for conventional treatment and 73.0 % in the dose-dense chemotherapy arm at the third cycle (approximately 9 weeks) and 74.2 and 71.6 % at 12 months after randomization. When considering these rates, ours were similar to those of other randomized phase III studies.

We used both the MMRM and the time to deterioration analysis to compare EQ-5D index values between groups. Many recent studies have begun to use time to deterioration analysis. While the analysis may be more widely accepted, the results depend on various settings, such as the size of MID and the handling of other events. Indeed, as shown in Fig. 3, the size of MID clearly influenced the results. Results also differed depending on whether death or progression was considered an event or was censored, or neither. While our analysis treated death as an event, if it was censored in the sensitivity analysis, the Kaplan–Meier curves and median time to deterioration would likely have changed. Standardization of time to deterioration analysis is reportedly needed to improve transparency and comparability [30].

This study has some limitations. First, the SELECT BC trial was an open-label trial that compared oral and intravenous therapies. Thus, patients could readily determine which arm they were allocated to. Some patients participated in the trial in order to receive oral chemotherapy; if they were not allocated to the preferred arm, their disappointment might have led to decreased EQ-5D index values. Second, our analysis did not include missing index values at each time point. MMRM can be applied to data under the assumption of missing at random, which means that a missing index value can be predicted by the observed index value. However, if, for example, the missing index value depended on the value after drop out, the estimated index value by MMRM may be biased. Finally, in this study, EQ-5D was measured at intervals of 3–6 months, which might have been too long to capture a temporal deterioration of HRQOL (e.g., AE caused by chemotherapy), given that EQ-5D asks respondents about their momentary health state.

In conclusion, our study supports that the EQ-5D index value for the S-1 arm is higher during first-line chemotherapy. We also demonstrated that EQ-5D index values are similar between patients treated with S-1 and those treated with taxane over a prolonged period of time. In addition, we reported on EQ-5D index values after progression and before death. These data can contribute not only to economic evaluations, but also to the selection of treatment for first-line MBC patients based on individual preferences.