Gastric cancer (GC) is the fifth most common malignancy and the third most common cause of cancer-related deaths worldwide [1]. In localized GC, surgery is the curative treatment modality; however, postoperative recurrence is common and observed in situations where attempting a curative treatment approach is difficult. In most cases of metastatic GC, the metastasis is usually widespread at the time of initial diagnosis. Hence, a curative treatment approach is almost impossible. Therefore, treatment for metastatic or recurrent GC (MRGC) remains a significant challenge.

Palliative chemotherapy is a cornerstone of treatment for patients with MRGC and is known to prolong overall survival (OS) and improve the quality of life (QOL) compared with best supportive care only [2]. The standard first-line chemotherapy for MRGC is based on fluoropyrimidine plus platinum combination, with trastuzumab added for HER2 + tumors [2, 3]. Among platinum agents, oxaliplatin has proven to be as effective as cisplatin when combined with 5-fluorouracil or capecitabine [4, 5]; as oxaliplatin causes less emesis and does not require large amounts of fluid infusion for renal protection compared with cisplatin, oxaliplatin has been replacing cisplatin in many cases with GC. Among fluoropyrimidine agents, both oral agents, capecitabine and S-1, are non-inferior to intravenous 5-fluorouracil and thus can replace 5-fluorouracil [5,6,7].

S-1 is an oral agent consisting of tegafur (5-fluorouracil prodrug), 5-chloro-2,4-dihydroxypyridine (an inhibitor of dihydropyrimidine dehydrogenase), and potassium oxonate (an inhibitor of the phosphorylation of 5-fluorouracil in the gastrointestinal tract). In a Japanese phase III study (JCOG9912), S-1 was proven to be non-inferior to 5-fluorouracil as monotherapy in MRGC [7]. Additionally, in another Japanese phase III study (SPIRITS), the combination of S-1 and cisplatin (SP) demonstrated a superior OS than S-1 monotherapy [8]. Based on these studies, SP has been widely used as one of the standard first-line therapies for patients with MRGC, especially in East Asian countries. However, in Korea, a 3-weekly SP regimen (SP3) is more commonly used than the Japanese 5-weekly SP (SP5) regimen [9]. Recently, the efficacy of S-1 plus oxaliplatin (SOX) regimens have been actively investigated. Several phase II studies have tested various doses and schedules of SOX regimens as first-line therapy in MRGC and have reported an objective response rate (ORR) of 30–59%, progression-free survival (PFS) of 4.4–7.0 months, and OS of 7.8–16.5 months [10,11,12,13,14,15]. Therefore, a phase III SOPP study was performed to verify non-inferiority or superiority of SOX compared to SP3 in terms of PFS, using the 3-weekly schedule of both regimens, in Korean patients with MRGC.


Study design

This study was a multicenter, open-label, randomized, non-inferiority phase III trial, conducted at 10 Korean institutions. Eligible subjects were randomly assigned to receive either SOX or SP3 in a 1: 1 ratio using permuted block randomization via interactive web response services. Three stratification factors were used as follows: (1) disease status (recurrence after curative gastrectomy versus initially unresectable or metastatic disease without gastrectomy versus initially curatively unresectable or metastatic disease with palliative gastrectomy; (2) presence of measurable disease; and (3) institution. The randomization results of the allocated study treatments were open to the investigators and patients.

This study was conducted in compliance with the Declaration of Helsinki and Good Clinical practice guidelines. The study was approved by the institutional review boards at each center. All patients gave their written informed consent and this trial was registered with (NCT01671449).

Patient eligibility criteria

The inclusion criteria were as follows: patients with unresectable locally advanced, metastatic or recurrent adenocarcinoma of the stomach; age of ≥ 18 years; Eastern Cooperative Oncology Group (ECOG) performance status (PS) 0–2; the existence of at least one measurable or evaluable lesion (according to Response Evaluation Criteria In Solid Tumors (RECIST, version 1.1); adequate bone marrow, liver, and kidney function; life expectancy of > 3 months; and no previous chemotherapy for MRGC. If a patient with resectable GC had received non-platinum-based neoadjuvant or adjuvant chemotherapy and the tumor recurred > 6 months after the completion of neoadjuvant or adjuvant chemotherapy, the patient could be enrolled in this study. Patients with recent major surgery or radiotherapy within 4 weeks of randomization, central nervous system metastasis, or other serious comorbidities were excluded.

Study treatment

In SOX therapy, patients received intravenous oxaliplatin (130 mg/m2 over 2 h; day 1) followed by oral administration of S-1 (40 mg/m2, twice daily; days 1–14). In SP3, cisplatin (60 mg/m2) was administered intravenously for 1 h on day 1 with adequate hydration, and S-1 (40 mg/m2; twice daily) was administered on days 1–14. In both arms, the calculation of S-1 dosage was based on body surface area (BSA) and the range of the S-1 dosage (80 mg/m2/day) was 100–160 mg/day (Supplementary Table 1). The study treatment was repeated every 3 weeks and continued until progressive disease (PD) according to the local investigator’s assessment, unacceptable adverse effects (AEs), or withdrawal of consent.

In both arms, the dose of each anticancer drug was appropriately reduced based on the degree of AEs that developed during the previous cycle. If a severe hematologic AE developed, the administration of S-1 and cisplatin (or oxaliplatin) was delayed until recovery to absolute neutrophil count > 1.5 × 109/L and platelet count > 100 × 109/L, and then S-1 and cisplatin (or oxaliplatin) were administered at a one-dose-level reduction. If an investigator deemed that the AE was only caused by one drug (for example, nephrotoxicity or neuropathy due to cisplatin), no dose modifications for the other drug were required. Even if cisplatin or oxaliplatin was permanently discontinued due to AEs, S-1 was allowed to continue until PD. However, the administration of only cisplatin or oxaliplatin without S-1 was not permitted under any circumstances. Guidelines for dose reduction according to the hematologic or non-hematologic AEs are detailed in Supplementary Table 1.


The primary endpoint of the study was PFS, defined as the time from the date of randomization to the date of documented PD or death from any cause, whichever occurred first. According to the RECIST, tumor assessments using computed tomography (CT) of the abdomen (± chest) were performed every 2 cycles (6 weeks: window, ± 7 days) from randomization until PD. Patients whose study treatment was terminated for reasons other than PD underwent a tumor assessment every 6 weeks before PD or subsequent second-line chemotherapy. All CT images for tumor responses and PFS were reviewed by the investigators. OS was defined as the time from randomization to death from any cause. AEs were classified according to the Common Terminology Criteria for Adverse Events (CTCAE; version 4.03). QOL was measured using the 3-level version of EuroQoL 5-dimensional instrument (EQ-5D-3L) at the baseline, every 2 cycles (at the time of tumor assessment), and at the end of study treatment. The EQ-5D-3L comprises a short descriptive system questionnaire on 5 distinct dimensions (mobility, self-care, usual activities, pain/discomfort, and anxiety/depression) and a visual analog scale (EQ VAS); each dimension in the questionnaire has 3 levels (indicating no problem, some or moderate problem, and extreme problem) [16].

Statistical considerations

Efficacy and safety analyses were conducted in patients included in the full analysis set (FAS), which is a group that could meet the intention-to-treat principle. The FAS included all randomized patients except those who met the following criteria: subjects who violated inclusion or exclusion criteria, subjects without any information after randomization, or subjects who did not receive any study treatment after randomization. Among patients included in the FAS, those who had completed at least one cycle of treatment and had at least one tumor response evaluation were included in the per-protocol set (PPS).

The primary endpoint of this study was PFS and secondary endpoints were OS, ORR, QOL, and safety. The primary objective of this study was to demonstrate the non-inferiority of 3-weekly SOX in terms of PFS when compared with SP3. A non-inferiority margin of 1.4, power = 80%, one-sided α = 0.025, accrual period of 24 months, follow-up period of 12 months, expected median PFS in SP3 = 6 months were expected. Based on these conditions, a total of 277 events were needed and a total of 304 subjects were required. Considering a 10% loss to follow-up, the enrollment of 338 subjects was finally planned.

Sequential tests for non-inferiority were planned with a non-inferiority margin of 1.4 and 1.25. In the first step, a test for non-inferiority margin of 1.4 was conducted, and if the null hypothesis was rejected, then it was planned to conduct the second test to verify whether it would satisfy the non-inferiority margin of 1.25. If non-inferiority was confirmed with the non-inferiority margin of 1.25, then a test for superiority of SOX to SP3 was planned. The Kaplan–Meier method was used to estimate the PFS distribution, and the Cox proportional hazards regression model was used to calculate the hazard ratio (HR) and adjust the influence of prognostic and confounding parameters on PFS. Regarding sub-group analysis, univariable Cox regression model was used to estimate the HR in each group with respect to PFS and OS. Student’s t-test was performed to compare means, and Chi-square test was used to compare percentages in cross tabulations. P-values < 0.05 were considered statistically significant. Statistical analyses were conducted using SAS (version 9.3; SAS Institute, Inc., Cary, NC).


Patient characteristics

From October 2012 to October 2014, 338 patients were randomized. One patient who was assigned to the SP3 group immediately withdrew consent and did not receive the study treatment; therefore, the FAS for efficacy and safety analyses included 337 patients (Fig. 1). Overall, the clinical characteristics were well balanced between SOX and SP3 (Table 1). Although the age of patients in the SOX group tended to be higher than that in the SP3 group, this was not statistically significant. Of the 337 patients in FAS, 304 were included in the PPS.

Fig. 1
figure 1

CONSORT diagram (SOX, S-1 plus oxaliplatin; SP3, S-1 plus cisplatin)

Table 1 Patient characteristics

Delivery of study treatments

The analyses of treatment delivery and dose intensities (DIs) were conducted in PPS. In the SOX regimen, the median treatment cycles of S-1 and oxaliplatin were 6 (range 1–45) and 5 (1–30), respectively. In the SP3 regimen, median treatment cycles of S-1 and cisplatin were 5 (range 1–35) and 5 (1–31), respectively. In SOX, dose reductions of S-1 and oxaliplatin were conducted in 70% and 70% of patients, respectively. In SP3, dose reductions in S-1 and cisplatin were conducted in 72% and 71%, respectively. In SOX, the relative DIs (RDIs) of S-1 and oxaliplatin were 80% (± 14) (mean ± standard deviation) and 84% (± 12), respectively. In SP3, the RDIs of S-1 and cisplatin were 77% (± 16) and 82% (± 12), respectively (Supplementary Table 2).

Of the 173 patients in SOX (FAS), 100 patients (58%) received second-line therapy after the discontinuation of SOX [taxane-based (N = 50), irinotecan-based (N = 36), and others (N = 14)]. In SP3, 101 (62%) of the 164 patients received second-line chemotherapy [taxane-based (N = 39), irinotecan-based (N = 38), and others (N = 24)].


The duration of the median follow-up was 15.6 months (range 0.1–33.3) in surviving patients as of the data cut-off date (Dec 23, 2015). The median PFS with the SOX and SP3 regimens were 5.6 months [95% confidence interval (CI), 4.4–6.9] and 5.7 months [95% CI 4.9–6.7], respectively (Fig. 2a). The unadjusted HR of SOX over SP3 was 0.85 (95% CI 0.67–1.07). Since the upper margin of the HR was 1.07, which was below the non-inferiority margin of 1.25, the non-inferiority of SOX was confirmed with both non-inferiority margins of 1.4 and 1.25. When the Cox regression model was adjusted for stratification factors (disease status at enrollment, presence of measurable disease, and institution), the adjusted HR was 0.90 (95% CI 0.70–1.16). In the final step, a superiority test was conducted; however, the superiority of SOX over SP3 was not proven.

Fig. 2
figure 2

Survival curves for a progression-free survival and b overall survival (SOX, S-1 plus oxaliplatin; SP3, S-1 plus cisplatin)

The median OS with SOX and SP3 treatments was 12.9 months (95% CI 10.3–14.6) and 11.4 months (95% CI 9.9–12.4) (Fig. 2b). The unadjusted and adjusted HRs of SOX over SP3 was 0.86 (95% CI 0.66–1.11; P = 0.242) and 0.84 (95% CI 0.63–1.11; P = 0.223), respectively. Tumor response evaluation was performed in patients with measurable lesions (N = 173). ORRs with SOX and SP3 treatments were 58% and 60% (P = 0.700), respectively. Disease control rates with SOX and SP3 regimens were 79% and 81% (P = 0.725), respectively (Supplementary Table 3).

Subgroup analyses of PFS and OS are shown in Supplementary Fig. 1. Overall, between SOX and SP3, no difference in PFS or OS were observed, except subgroups of body mass index (BMI) ≥ 25 and age ≥ 60 years. In patients with BMI ≥ 25, SOX demonstrated significantly longer PFS and OS than SP3. In patients aged ≥ 60 years, the PFS was longer with SOX treatment than SP3; however, the OS was not statistically different between SOX and SP3.


The proportion of patients responding to the QoL questionnaire has steadily decreased because of tumor progression or discontinuation of study treatment; the percentage of patients who responded to the questionnaire after the end of the 8th cycle of SOX or SP3 was less than 50% of the baseline. Therefore, the analysis on the QoL was conducted at the time point of the baseline, after the completion of 2nd, 4th, and 6th cycles of study treatment and at the end of study therapy. During the study treatment, there was no significant difference in all EQ-5D-3L dimensions and VAS between 2 treatment groups (Fig. 3 and Supplementary Table 4).

Fig. 3
figure 3figure 3

Changes in quality of life [3-level version of EuroQoL 5-dimensional instrument (EQ-5D-3L)] during study treatments: a mobility, b self-care, c usual activities, d pain/discomfort, e anxiety/depression and f EQ-5D visual analog scale


Treatment-related AEs are described in Table 2. Among hematologic AEs (all grades), anemia (30% versus 22%), leucopenia (66% versus 54%), and neutropenia (71% versus 57%) developed more commonly with SP3 than with SOX. Thrombocytopenia was more frequent with SOX than with SP3 (71% versus 58%). Febrile neutropenia developed in 5% of SP3-treated patients compared with 0.6% of SOX-treated patients.

Table 2 Treatment-related adverse events (per patient analysis; N = 337)

Among non-hematologic AEs (all grades), oral mucositis developed more commonly with SP3 than with SOX. Nausea and vomiting were more common with SOX than with SP3. As expected, peripheral sensory neuropathy was much common with SOX than with SP3 [all grades (59% versus 35%) and grade ≥ 3 (9% versus 4%)].


The efficacy, safety, and QOL data of this phase III study (SOPP) were reported. In patients with MRGC, SOX was significantly non-inferior but not superior to SP3 in terms of PFS (median 5.6 versus 5.7 months). Additionally, the OS in both groups did not differ (median 12.9 versus 11.4 months). ORRs were similar between SOX and SP3 (58% versus 60%). The two regimens were well tolerated with different toxicity profiles. The influence of both regimens on QOL showed no significant difference during the treatment period.

For MRGC, palliative chemotherapy improves OS compared with the best supportive care alone. The combination of intravenous 5-fluorouracil and cisplatin (± epirubicin) has been considered the standard first-line chemotherapy in MRGC [2, 5, 6]. Recent trials have shown that oral fluoropyrimidine (capecitabine and S-1) can replace intravenous 5-fluorouracil [5,6,7] and oxaliplatin can replace cisplatin [4, 5]. Considering convenience and safety profiles, oral fluoropyrimidine and oxaliplatin are being extensively used in MRGC.

S-1 has demonstrated different safety profiles between East Asian and Caucasian patients [17]. According to ethnicity and geographical area, diverse dosing schedules of SP have been used. Based on the result of the SPIRITS trial [10], 5-weekly SP [SP5; S-1 80–120 mg/day (days 1–21) and cisplatin 60 mg/m2 (day 8)] is widely used as first-line chemotherapy for MRGC in Japan. However, 3-weekly SP [SP3; S-1 80 mg/m2/day (days 1–14) and cisplatin 60 mg/m2 (day 1)] is popular in Korea [9]. Interestingly, in the Japanese SP5 regimen, the S-1 dose is capped at 120 mg/day if BSA ≥ 1.5 m2; however, in the Korean SP3 regimen, the S-1 dose (80 mg/m2/day) is not capped and can be increased to 160 mg/day if BSA ≥ 2.0 m2 (Supplementary Table 1). In the Korea–Japan collaborative phase III S-1 Optimal Schedule Study (SOS), SP3 was not only non-inferior but was superior to SP5 in terms of PFS [median 5.5 versus 4.9 months; HR 0.82 (95% CI, 0.68–0.99)]. However, because the PFS improvement was small and no difference in OS was observed, the Korean and Japanese investigators concluded that both SP3 and SP5 can be recommended as first-line chemotherapy for MRGC [9]. In contrast, considering inter-ethnic variabilities in S-1 tolerability, a 4-weekly regimen was established [S-1 50 mg/m2/day (days 1–21) and cisplatin 75 mg/m2 (day 1)] in Western countries [18].

As S-1 has been popularly used in East Asian patients rather than Western patients, the safety and efficacy of SOX have been actively investigated in East Asian countries. However, the tested dosages and schedules of SOX were diverse in several phase II trials for MRGC [10, 12,13,14,15]. In the Japanese G-SOX phase III study (N = 685), the oxaliplatin dose was chosen as 100 mg/m2 (day 1) in SOX and the SP5 regimen was used in the reference arm. S-1 was used with the dosage cap at 120 mg/day if BSA ≥ 1.5 m2 in both SOX and SP5. The investigators demonstrated that SOX is non-inferior to SP5 in terms of PFS [5.5 versus 5.4 months (median); HR 1.004, 95% CI 0.840–1.199; predefined non-inferiority margin 1.30] in the per-protocol population (N = 642). The OS [14.1 versus 13.1 months (median)] and ORRs (56% versus 52%) did not significantly differ between SOX and SP5 [19]. In this SOPP study, different dosage of oxaliplatin (130 mg/m2; day 1) in the SOX was used and SP3 was chosen as the reference arm. In both SOX and SP3 arms, the S-1 dose (80 mg/m2/day) was not capped and could be increased to 160 mg/day according to the subject’s BSA. This study also demonstrated that SOX is non-inferior to SP3 in terms of PFS [5.6 versus 5.7 months (median); HR 0.85, 95% CI 0.67–1.07]; when both predefined non-inferiority margins of 1.4 and 1.25 were used, SOX was proven to be non-inferior to SP3. OS and ORR did not differ between SOX and SP3 [OS (median), 12.9 versus 11.4 months; ORR, 58% versus 60%]. Therefore, regarding efficacy, although different dosages and treatment schedules of SOX and SP regimens were evaluated, both SOPP and G-SOX studies arrived at the same conclusion.

In terms of severe AEs (≥ grade 3), anemia (11% versus 5%), leucopenia (10% versus 2%), neutropenia (40% versus 16%), and febrile neutropenia (5% versus 0.6%) were more common in SP3 than in SOX; conversely, peripheral sensory neuropathy was more common with SOX than SP3 (9% versus 4%). Although thrombocytopenia of all grades was more frequent with SOX than SP3 (71% versus 58%), the frequency of severe thrombocytopenia (≥ grade 3) was similar between the 2 arms (8% versus 5%). In this study, nausea and vomiting (all grades) were more common with SOX than SP3. However, in the G-SOX study, nausea (all grades) was more frequent with SP5 than SOX. This differential pattern in the incidence of nausea and vomiting could be attributed to the different oxaliplatin doses used in the two studies. Although some differences exist in the pattern of developing treatment-related AEs between the SOPP and G-SOX studies, the overall differences in the AE profiles observed with SOX and SP in this SOPP study were consistent with those observed in the G-SOX study [19].

Interestingly, PFS and OS curves of SOX and SP3 tended to separate at about 7 and 10 months after randomization, respectively. This trend was not observed in the G-SOX study. At the end of this study, the study treatment was still ongoing in 16 subjects (9%) in SOX and 5 (3%) in SP3 (Fig. 1). Between the two regimens, there was no difference in the delivery of platinum agents. However, although there was no statistical difference, S-1 in SOX was delivered longer than S-1 in SP3 [median treatment cycles of S-1, 6 (range 1–45) in SOX versus 5 (range 1–35) in SP3; mean treatment cycles of S-1, 8.6 in SOX versus 7.2 in SP3 (Supplementary Table 2)]. When looked into the causes of study treatment discontinuation (Fig. 1), the proportion of subjects who were dropped out of study treatment owing to adverse events or refusal of treatment due to toxicity was numerically higher in SOX (9%) than in SP3 (5%). However, unexpectedly, subjects in SP3 rejected the study treatment more frequently for reasons other than identified toxicities (9%) than those in SOX (3%). Although the exact reasons of more treatment refusal in SP3 than in SOX was not clear, it is thought that there is still a possibility that treatment compliance decreased more in SP3 than in SOX due to unrecognized adverse events or other reasons. In addition, it was found that 13 subjects received conversion gastrectomy (9 in SOX and 4 in SP3) after study treatment (Supplementary Table 5); all these subjects had favorable OS outcomes and the median OS of these subjects did not reach. It is thought that the role of conversion surgery after palliative chemotherapy is not yet established in MRGC, so more researches are necessary. Although these explanations may not be enough, more conversion gastrectomy and more prolonged administration of S-1 in SOX group than in SP3 group may have contributed to the separation of PFS and OS curves after the observed time points between SOX and SP3.

In this study, subgroup analyses indicated no overall differences in PFS and OS between SOX and SP3. However, patients with a BMI ≥ 25 or age ≥ 60 years demonstrated longer PFS or OS with SOX than with SP3 (Supplementary Fig. 1). In the G-SOX study, the investigators have reported that patients with peritoneal metastasis had significantly longer OS with SOX than with SP5; the investigators could not observe any other subgroups with significantly improved OS with either SOX or SP5. The G-SOX investigators did not include BMI as one of the parameters for subgroup analyses of survival. In contrast, organs (such as the liver, lungs, or peritoneum) involved by metastasis were not included as parameters for subgroup analyses of PFS or OS in our study. Therefore, a direct comparison of subgroup analyses is not feasible between G-SOX and the current studies.

This study has some limitations. First, the study treatments were not blinded and 49% of enrolled subjects (164/337) had no measurable target lesion. The fact that only investigators, without the confirmation of independent central reviewers, determined tumor progression might be a limitation. Second, some experts may criticize the non-inferiority margin 1.4 as the basis for sample size calculation in this study was too wide. However, since non-inferiority margins of both 1.4 and 1.25 were satisfied through the two-step tests, it is thought that SOX was sufficiently proven non-inferior in terms of PFS compared to SP3. Third, some experts may question whether it is appropriate to set the primary endpoint to PFS in a phase III clinical trial of first-line chemotherapy in MRGC. In a meta-analysis of the GASTRIC group, the investigators showed that chemotherapy effects on PFS and OS were only moderately correlated, and thus they could not confirm the validity of PFS as a surrogate endpoint for OS in MRGC [20]. It is agreed that PFS is not a good surrogate endpoint for OS in superiority trials for MRGC. However, as in other tumors, OS is greatly influenced by salvage chemotherapy after first-line chemotherapy, and the salvage chemotherapy is influenced by regional difference and practice pattern of the attending physicians [21]. In addition, as the efficacy of salvage chemotherapy has been recently demonstrated in MRGC, the frequency of using second-line or later-line chemotherapy is increasing [22,23,24]. Therefore, it is believed that PFS as a primary endpoint is practically acceptable in non-inferiority trials of first-line chemotherapy in MRGC. For example, PFS was previously used as a primary endpoint in a phase III trial (ML17032), which had shown the non-inferiority of capecitabine/cisplatin compared to 5-fluorouracil/cisplatin in MRGC [6].

In conclusion, this SOPP study demonstrated that SOX is non-inferior to SP3. The two regimens were well tolerated with different toxicity profiles. Considering the SOPP and Japanese G-SOX studies, the SOX regimen can be recommended as a first-line treatment of MRGC in East Asian countries.