Meta-analysis of three randomized trials of capecitabine plus cisplatin (XP) versus S-1 plus cisplatin (SP) as first-line treatment for advanced gastric cancer

S-1 plus cisplatin (SP) and capecitabine plus cisplatin (XP) are standard first-line regimens for advanced gastric cancer (AGC) worldwide. We conducted a meta-analysis using individual participant data (IPD) to investigate which is more suitable. IPD from three randomized trials were collected. In these trials, patients with AGC were randomly allocated to SP (S-1 80–120 mg for 21 days plus cisplatin 60 mg/m2 (q5w)) or XP (capecitabine 2000 mg/m2 for 14 days plus cisplatin 80 mg/m2 (q3w)). In 211 eligible patients, median overall survival (OS) for SP versus XP was 13.5 and 11.7 months (hazard ratio [HR], 0.787; p = 0.114), progression-free survival (PFS) was 6.2 and 5.1 months (HR, 0.767; P = 0.076), and TTF was 5.1 and 4.0 months (HR, 0.611; P = 0.001). The most common grade ≥ 3 adverse events with SP or XP were neutropenia (18% vs. 29%) and anorexia (16% vs.18%). Subgroup analysis demonstrated significant interaction between treatment effect and performance status > 1 (HR, 0.685; P = 0.036), measurable lesion (HR, 0.709; P = 0.049), primary upper third tumor (HR, 0.539; P = 0.040), and differentiated type (HR, 0.549; interaction, 0.236; P = 0.019). For the differentiated type, OS was significantly longer in the SP group (13.2 months) than in the XP group (11.1 months) (HR, 0.549; P = 0.019). For the undifferentiated type, OS was similar in the SP group (14.2 months) and in the XP group (12.4 months) (HR, 0.868; P = 0.476). SP and XP were both effective and well tolerated. SP might be suitable for the pathological differentiated subtype of AGC. Clinical Trial Registration: The HERBIS-2, HERBIS-4A, and XParTS II trials were registered with UMIN-CTR as UMIN000006105, UMIN000006755, and UMIN000006045, respectively.


Introduction
Gastric cancer remains one of the leading causes of cancerrelated deaths worldwide [1].Combination chemotherapy regimens provide better response rates (RRs) and modest survival benefits compared with single-agent therapies [2].A combination of fluoropyrimidine and platinum is the most commonly used [3,4].S-1 or capecitabine can be used as an alternative to infusional 5-fluorouracil (5-FU) in doublet regimens for advanced gastric cancer (AGC).
Based on results from the SPIRITS trials, S-1 combined with cisplatin (SP) has become the accepted standard treatment regimen for AGC in Japan [5].Conversely, capecitabine plus cisplatin (XP) was reported to have significantly non-inferior median progression-free survival (PFS) compared with 5-FU plus cisplatin (5.6 months vs. 5.0 months; hazard ratio [HR] = 0.81) [6].More recently, XP therapy was the active comparator in three multinational phase III studies designed to evaluate the potential of molecular targeted Kazuhiro Nishikawa and Hisato Kawakami have contributed equally to this study.
Extended author information available on the last page of the article therapeutics for AGC [7][8][9].Thus, both SP and XP are listed as first-line treatment in the Japanese guidelines [10].
Regarding S-1 versus capecitabine not in combination with cisplatin, two randomized trials in elderly patients with AGC have been reported.One involved S-1 or capecitabine monotherapy and the other involved S-1 plus oxaliplatin (SOX) and capecitabine plus oxaliplatin (CAPOX) [14][15][16].In addition, several meta-analyses reported in abstracts have compared S-1-based and capecitabine-based chemotherapy [17][18][19].All these studies showed that S-1 and capecitabine are almost equally effective.However, there is scant evidence to suggest whether S-1 or capecitabine would be preferable for precision medicine.
Thus, it remains unclear whether SP or XP is more suitable as first-line treatment for AGC.It is also unclear whether clinical and molecular characteristics of patients could assist in deciding whether to use SP or XP.Therefore, we conducted a meta-analysis with individual participant data (IPD).

Study design
This meta-analysis was planned in 2020 in accordance with Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA)-IPD [20].We included all completed and published peer-reviewed randomized clinical trials that investigated the effect of XP or SP as first-line chemotherapy for gastric or gastroesophageal junction adenocarcinoma.We searched PubMed, Embase and the Cochrane Central Register of Controlled Trials up to 30 September 2020.The search terms were "gastric cancer OR gastric carcinoma OR gastroesophageal junction adenocarcinoma" AND "capecitabine" AND "S-1" AND "CDDP OR cisplatin" AND "metastatic OR recurrent OR first-line."The inclusion criteria were as follows: (1) prospective randomized controlled clinical trial; (2) pathological diagnosis of AGC or gastroesophageal junction adenocarcinoma; (3) XP or SP as firstline treatment regimen.Based on PRISMA-IPD, three randomized trials, the ECRIN XParTS II trial, OGSG HEBIS-2 trial, and OGSG HERBIS-4A trial, were included (Fig. 1).All three trials compared SP and XP at the same dose and the treatment methods were identical, and patients were randomly assigned to receive either SP (S-1 40-60 mg twice daily for 21 days plus cisplatin 60 mg/m 2 on day 8, every 5 weeks) or XP (capecitabine 1000 mg/m 2 twice daily for 14 days plus cisplatin 80 mg/m 2 on day 1, every 3 weeks).In terms of patient enrollment, inclusion and exclusion criteria were reported previously [11][12][13].
We first verified the integrity of the IPD from the three trials.All clinical data were extracted and held centrally at the OGSG data center.This study was conducted in compliance with the ethical principles of the Declaration of

Patients
Briefly, patients with histologically confirmed HER2-negative AGC were eligible for all three trials.For XParTS II, no previous chemotherapy or radiotherapy was allowed, but prior adjuvant chemotherapy was allowed if more than 6 months have passed since the end of such treatment [11].For HERBIS-4A, patients were required to be naïve to systemic chemotherapy [13].Patients who had recurrence within 6 months after the completion of S-1 adjuvant therapy were eligible for HERBIS-2 [12].Other eligibility criteria were as follows: age ≥ 20 years; written informed consent; Eastern Cooperative Oncology Group performance status (PS) of ≤ 2; adequate organ function (white blood cell count ≥ 3000/mm 3 for HERBIS-2 and HERBIS-4A, neutrophil count ≥ 1,500/mm 3 , platelet count ≥ 100,000/ mm 3 , hemoglobin level ≥ 8.0 g/dL, aspartate aminotransferase and alanine aminotransferase levels ≤ 100 IU/L for HERBIS-2 and HERBIS-4A or ≤ 2.5 × upper limit of normal (ULN) at each institution (≤ 5 × ULN if there were metastases) for XParTS II, total bilirubin level ≤ 1.50 mg/ dL for HERBIS-2 and HERBIS-4A or ≤ 1.5 × ULN at each institution for XParTS II, serum creatinine ≤ 1.20 mg/dL for HERBIS-2 and HERBIS-4A, creatinine clearance ≥ 60 mL/ min estimated with Cockcroft-Gault equation.Exclusion criteria consisted of additional malignancies or significant comorbidities.

Assessment
The primary objective of this study was to evaluate prospectively collected IPD from the XParTS II, HERBIS-2, and HERBIS-4A trials to compare SP with XP and to determine the optimal first-line chemotherapy for patients with HER2negative unresectable advanced or recurrent gastric cancer.The primary endpoint of this study was OS.The secondary endpoints were PFS, time to treatment failure (TTF), post-protocol survival (PPS), overall response rate (ORR), disease control rate (DCR), and adverse events (AEs).
OS was defined as time from randomization to death.PFS was defined as time from randomization to radiographic progression or death.TTF was defined as time from treatment to resistance or progression of disease, transformation to another malignancy, or death from any cause.PPS was defined as time from the end of therapy as per protocol to death.Tumor responses were assessed using Response Evaluation Criteria In Solid Tumors version 1.1 and classified as complete response (CR), partial response (PR), stable disease (SD), or progressive disease (PD) based on central review.ORR was defined as the proportion of patients with CR or PR.DCR was defined as the proportion of patients with CR, PR, or SD.The DCR of patients without measurable lesions was defined as the proportion of patients with CR or non-CR and non-PD.PFS and RR were monitored with abdominal Computed Tomography or Magnetic Resonance Imaging every 6 weeks and measuring levels of the tumor markers carcinoembryonic antigen and cancer antigen 19-9.AEs were evaluated using the Common Terminology Criteria for Adverse Events version 4.0.
Tumor histology was based on the Japanese classification of gastric carcinoma [10].Differentiated-type (DIFF) tumor was defined as papillary or tubular adenocarcinoma.Undifferentiated-type (UNDI) tumor was defined as poorly differentiated adenocarcinoma, signet ring cell carcinoma, or mucinous adenocarcinoma.Other histology types were designated as other.

Statistical analysis
Survival curves were estimated using the Kaplan-Meier method.The log-rank test with stratification of allocation factors was used to compare survival curves between arms.HRs and their 95% confidence intervals (CIs) were calculated with Cox proportional hazards models; 95% CIs were used for median OS, PFS, and TTF.ORR was compared between arms with Fisher's exact test.All statistical analysis was performed with R version 3.3.1 (R Foundation for Statistical Computing, Vienna, Austria) and SAS version 9.4 (SAS Institute, Cary, NC).A P value of < 0.05 was considered statistically significant.

Patient characteristics
We extracted data from 211 patients eligible for the HER-BIS-2 (N = 17), HERBIS-4A trial (N = 84), and XParTS II (N = 110) trials.There were 105 patients (50%) allocated to the SP group and 106 (50%) allocated to the XP group.Both arms were well balanced in terms of baseline clinical characteristics (Table 1).

Adverse events
Toxicity profiles are summarized in Table 2.The incidence of grade ≥ 3 AEs was low in both groups.There were no significant differences in the incidence of grade ≥ 3 AEs between treatment arms, except for diarrhea, which occurred more frequently in the SP group (7% vs. 0%; P = 0.006).There were no deaths resulting from AEs in either arm.The incidence of all-grade neutropenia (52% vs. 67%), hyponatremia (22% vs. 40%) and peripheral sensory neuropathy (4% vs. 18%) was higher in the XP group.The incidence of all-grade abdominal pain (21% vs. 8%) was higher in the SP group.

Discussion
This meta-analysis with IPD is the first report to compare SP and XP regimens in a large sample.With PRISMA-IPD methods, we were able to evaluate efficacy in detail, especially with subgroup analyses.We showed that SP is associated with superior TTF compared to XP as well as non-significantly better OS and PFS.SP and XP achieved comparable ORR with acceptable toxicity levels.The overall DCR for SP was significantly higher than for XP.Various factors could account for the discrepancy between OS and TTF results.In Japan, more than 70% of patients with AGC receive second-line chemotherapy.Demonstrating the superiority of a first-line regimen in terms of OS is difficult because OS is strongly affected by subsequent treatment.Takashima et al. reported strong correlations between the proportion of patients receiving with second-line chemotherapy and post-progression survival and OS, respectively [23].Although the SP group was able to undergo primary treatment longer due to maintenance treatment with S-1 as a single agent after SP therapy, the XP group might have had more effective treatment after the shift to second-line treatment.Our subgroup analysis of OS revealed a significant interaction between treatment effect and PS > 1 (P = 0.036), primary U lesion (P = 0.040), measurable lesion (P = 0.049), and DIFF type (P = 0.019) (Fig. 3).
It is unknown why SP resulted in superior OS in patients with poor performance status.The major difference between SP and XP regimens other than S-1 and capecitabine is the initial dose and cisplatin schedule (SP, 60 mg/m 2 , every 5 weeks [12 mg/m 2 /week] vs. XP, 80 mg/m 2 , every 3 weeks [26.7 mg/m 2 /week]).In the SOS study, cisplatin 60 mg/m 2 every 3 weeks in combination with S-1 (SP3) was compared to standard every 5 weeks S-1 plus cisplatin (60 mg/m 2 ) [22].In this trial, SP3 was slightly superior to standard SP in terms of PFS, but not in OS.However, there were significantly more ≥ grade 3 AEs in the SP3 arm than in the standard SP arm (73% vs. 51%).In our study, although AEs with SP and XP were comparable, better tolerability of cisplatin 60 mg/m 2 every 5 weeks might have led to more treatment cycles and longer TTF than cisplatin 60 mg/m 2 every 3 weeks.In addition, SP is more convenient because it can be administered less frequently in hospitals and partially on an outpatient basis given the lower dose of cisplatin.Thus, one possible reason for the favorable prognosis of patients with SP was the better tolerability against cisplatin.
Regarding primary U tumors, gastric cancer arising from the upper portion of the stomach, including the gastroesophageal junction and cardia, is associated with chromosomal instability [24], suggesting that precise biomarker analysis might facilitate the selection of patients most likely to benefit from SP.Further study of the clinical and molecular characteristics of AGC is needed to guide decisions on using SP or XP therapy.
We also showed that OS was significantly better with SP than with XP in patients with pathological DIFF tumors.There are few reports suggesting the use of S-1 or capecitabine based on histological type.Initially, in the subset analysis of the FLAGS trial, S-1 appeared to be superior to 5-FU in the diffuse gastric cancer subgroup [25], but the DIGEST trial following the FLAGS trial failed to demonstrate that S-1 was superior to 5-FU for diffuse-type gastric cancer [26].A previous study showed that compared to diffuse-type tumors, intestinal-type tumors tend to have lower expression levels of excision repair crosscomplementation group 1 (ERCC1), a nucleotide excision repair pathway gene that provides protection against platinum-based chemotherapy-induced DNA damage [27].The better OS of SP compared with XP for differentiatedtype tumors might be due to the fact that relatively low doses of platinum (SP: 12 mg/m 2 /week vs. XP: 26.7 mg/ m 2 /week) are enough for tumors with low ERCC1 expression and SP is better tolerated than XP.Further molecular biology studies are needed to determine whether tumor differentiation can be used to select S-1 or capecitabine for AGC.SP yielded superior OS than XP in patients with measurable lesions.We have worked on elucidating as a separate study, and found that in DIFF tumors, the depth of tumor shrinkage in SP is deeper than that in XP, which is why there is no difference in PFS or TTF, but a large difference opens up in OS [21].On the other hand, in the UNDI tumors, we found differences in PFS and TTF.
This study has several limitations.This study only included Japanese patients and ethnic differences between Asian and Western patients could have affected the overall results; the results should be interpreted with caution.Second, the planned doses of concomitant cisplatin were different, so it is not possible to determine whether S-1 or capecitabine should be used.Lastly, as the platinum agent used in this study was cisplatin, it is unclear what the results would be with other platinum products.Recently, SOX or CAPOX has become more commonly used than SP or XP [28,29].Furthermore, the standard of care for HER2-negative AGC is an anti-programmed death receptor-1 antibody with oxaliplatin and fluoropyrimidine [29][30][31].Whether SOX or CAPOX should be used as the base combination chemotherapy is debated.
In conclusion, the efficacies of XP were similar to those of SP.SP might be suitable in the differentiated subtype of AGC, although histological subtyping is not adequately sensitive for selecting S-1 or capecitabine.Further research that classifies AGC based on other biomarkers is necessary to enable individualized treatment.

Fig. 1
Fig. 1 PRISMA flow diagram of the literature search

Fig. 2 Fig. 3 Fig. 4
Fig. 2 Kaplan-Meier curves for OS a and PFS b by treatment group.OS overall survival, PFS progression-free survival

Table 1
Patient characteristicsECOG Eastern cooperative oncology group, PS performance status, HER2 human epidermal growth factor receptor 2, SP S-1 plus cisplatin, XP capecitabine plus cisplatin a Patients can be included in more than one category