Articles. Interpretation S-1 for 1 year should remain as standard adjuvant chemotherapy for stage II gastric cancer.

Transcription

1 Four courses versus eight courses of adjuvant S-1 for patients with stage II gastric cancer (JCOG1104 [OPAS-1]): an open-label, phase 3, non-inferiority, randomised trial Takaki Yoshikawa, Masanori Terashima, Junki Mizusawa, Souya Nunobe, Yasunori Nishida, Takanobu Yamada, Masahide Kaji, Norimasa Fukushima, Shinji Hato, Yasuhiro Choda, Hiroshi Yabusaki, Kazuhiro Yoshida, Seiji Ito, Atsushi Takeno, Takashi Yasuda, Yasuyuki Kawachi, Hiroshi Katayama, Haruhiko Fukuda, Narikazu Boku, Takeshi Sano, Mitsuru Sasako Summary Background Postoperative adjuvant chemotherapy with S-1 for 1 year (corresponding to eight courses) is standard care for stage II gastric cancer. Whether the duration of S-1 could be shortened to 6 months (corresponding to four courses) without worsening survival is unclear. The aim of this study was to investigate the non-inferiority of four courses of S-1 compared with eight courses of S-1 for patients with stage II gastric cancer. Methods We did a phase 3, open-label, randomised controlled, non-inferiority trial at 59 hospitals in Japan. Patients aged years with stage II adenocarcinoma of the stomach were randomly assigned (1:1) to eight courses or four courses of S-1. Randomisation was done by the Japan Clinical Oncology Group Data Center website, using a minimisation method with a random component using institution, stage (IIA vs IIB), age (<70 years vs 70 years), and mode of operation (open gastrectomy with bursectomy vs open gastrectomy without bursectomy vs laparoscopic gastrectomy) as adjustment factors. One course was 80 mg/day per m² of S-1 administered for 4 weeks followed by a rest for 2 weeks. The primary endpoint was relapse-free survival, analysed by intention to treat, with a non-inferiority margin for the hazard ratio (HR) set at This study is registered at UMIN-Clinical Trial Registry, number UMIN Findings Between Feb 16, 2012, and March 19, 2017, 590 patients were enrolled (295 per group). 528 (89%) patients were analysed at the first planned interim analysis in March, 2017, at which time the point estimate of HR for the four-course group compared with the eight-course group was 2 52 (95% CI ), which exceeded 1 37 and met the prespecified criteria for early termination. Predictive probability for showing non-inferiority at the final analysis was calculated to be 2 9%. The study was stopped for futility. Updated 3-year relapse-free survival analysed in May, 2017, was 93 1% (95% CI ) for the eight-course group and 89 8% ( ) for the four-course group (HR 1 84, 95% CI ). The most common grade 3 4 adverse event was neutropenia, observed in 46 (16%) patients in the eight-course group and 51 (17%) patients in the four-course group. Interpretation S-1 for 1 year should remain as standard adjuvant chemotherapy for stage II gastric cancer. Funding Japan Agency for Medical Research and Development; the Ministry of Health, Labour and Welfare of Japan; the National Cancer Center Research and Development Fund, Japan. Copyright 2019 Elsevier Ltd. All rights reserved. Introduction Gastric cancer is the fifth most common neoplasm and the third leading cause of cancer-related deaths in the world. 1 Localised gastric cancer can be cured by complete resection or by multimodality treatment. The standard treatments for locally advanced gastric cancer are surgery followed by adjuvant chemotherapy, 2,3 surgery with perioperative chemotherapy, 4 and surgery with postoperative chemoradiation therapy. 5,6 Superior efficacy of adjuvant chemotherapy for stage II and III gastric cancer compared with surgery alone was shown by two phase 3 trials, the ACTS-GC trial 2 and the CLASSIC trial. 3 The ACTS-GC trial showed a survival benefit of S-1 for 1 year, 2 and the CLASSIC trial showed that postoperative doublet regimen with capecitabine and oxaliplatin for 6 months improved disease-free survival. 3 Notably, prognosis of patients with stage II gastric cancer was substantially improved by these regimens, with a 5-year survival of 84% with S-1 7 and 88% with capecitabine and oxaliplatin. 8 Considering severe adverse events caused by capecitabine and oxaliplatin and similar 5-year survival with both regimens, 1 year of adjuvant chemotherapy with S-1 is recognised as a standard care for stage II gastric cancer in Japan. However, the choice to administer adjuvant S-1 for 1 year in the ACTS-GC trial was not based on solid evidence, and capecitabine and oxaliplatin were ad ministered for 6 months in the CLASSIC study. 3 For colon cancer, a 6-month duration of adjuvant chemo therapy with fluoropyrimidine and leucovorin was estab lished on the basis of Lancet Gastroenterol Hepatol 2019 Published Online January 21, S (18) See Online/Comment S (18) Department of Gastric Surgery (T Yoshikawa MD), JCOG Data Center/Operations Office (J Mizusawa ME, H Katayama MD, H Fukuda MD), National Cancer Center Hospital, Chuo-Ku, Tokyo, Japan; Division of Gastric Surgery, Shizuoka Cancer Center, Nagaizumi-cho, Sunto-gun, Shizuoka Prefecture, Japan (M Terashima MD); Gastroenterological Surgery, Cancer Institute Ariake Hospital, Koto-Ku, Tokyo, Japan (S Nunobe MD, T Sano MD); Department of Surgery, Keiyukai Sapporo Hospital, Siroishi-Ku, Sapporo, Japan (Y Nishida MD); Department of Gastrointestinal Surgery, Kanagawa Cancer Center, Nakao, Asahi-Ku, Japan (T Yamada MD); Department of Surgery, Toyama Prefectural Central Hospital, Toyama, Japan (M Kaji MD); Department of Surgery, Yamagata Prefectural Central Hospital, Yamagata, Japan (N Fukushima MD); Department of Surgery, National Hospital Organization, Shikoku Cancer Center, Minami-umemotocho, Matsuyama Japan (S Hato MD); Department of Surgery, Hiroshima City Hiroshima Citizens Hospital, Naka-Ku, Hiroshima, Japan (Y Choda MD); Department of Digestive Surgery, Niigata Cancer Center Hospital, Kawagishimachi, Chuo-Ku, Niigata, Japan (H Yabusaki MD); Department of Surgical Published online January 21,

2 Oncology, Gifu University, Graduate School of Medicine, Gifu City, Japan (Prof K Yoshida MD); Department of Gastroenterological Surgery, Aichi Cancer Center Hospital, Chikusa-ku, Nagoya (S Ito MD); Department of Surgery, Kansai Rosai Hospital, Amagasaki, Japan (A Takeno MD); Department of Gastroenterological Surgery, Hyogo Cancer Center, Akashi, Japan (T Yasuda MD); Department of Surgery, Nagaoka Chuo General Hospital, Nagaoka, Japan (Y Kawachi MD); Division of Gastrointestinal Medical Oncology, National Cancer Center Hospital, Chuo-Ku, Tokyo, Japan (N Boku MD); and Department of Multidisciplinary Surgical Oncology, Hyogo College of Medicine, Nishinomiya, Japan (Prof M Sasako MD) Correspondence to: Dr Takaki Yoshikawa, Department of Gastric Surgery, National Cancer Center Hospital, Chuo-Ku, Tokyo , Japan tayoshik@ncc.go.jp See Online for appendix Research in context Evidence before this study We searched PubMed without language restrictions for studies published before February 27, 2018, with the terms S-1 AND gastric cancer AND adjuvant. The phase 3 ACTS-GC study showed that adjuvant chemotherapy with S-1 for 1 year was effective for patients with stage II or stage III gastric cancer. However, the 1-year duration was not determined on the basis of solid evidence. There has been much interest in adjuvant chemotherapy, including S-1, in patients with gastric cancer. However, most studies were retrospective and no randomised studies have focused on the duration of adjuvant S-1 chemotherapy. A 6-month duration of adjuvant chemotherapy has been established for colon cancer on the basis of several phase 3 studies that compared different treatment durations. Whether adjuvant chemotherapy with S-1 for 1 year could be shortened to 6 months without worsening survival is unclear. several phase 3 studies comparing different treatment durations Moreover, 3 months of adjuvant chemotherapy has been introduced for colon cancer on the basis of the international IDEA collaboration, a large scale combined analyses of phase 3 studies. 12 Furthermore, 6 months of adjuvant chemotherapy with S-1 was shown to be superior to gemcitabine for resected pancreatic cancer. 13 Thus, whether a 1-year duration of adjuvant S-1 chemotherapy (corresponding to eight courses) is optimal for gastric cancer is unclear. We report the primary results of a phase 3, open-label, multicentre, randomised controlled trial the JCOG1104 (optimal period of adjuvant chemotherapy S-1 [OPAS-1] study) to assess the non-inferiority of four courses of S-1 (corresponding to 6 months) compared with eight courses of S-1 (corresponding to 1 year) for patients with stage II gastric cancer. Methods Study design and participants OPAS-1 was a multicentre, prospective, randomised, phase 3, open-label trial, done in 59 Japanese hospitals by the Stomach Cancer Study Group of the Japan Clinical Oncology Group. The appendix (p 4) lists each site from which patients were recruited, the names of principal investigators responsible for each site, and the number of patients which were recruited from each site. Eligibility criteria were histologically proven adeno carcinoma of the stomach; stage II disease, excluding T1N2 3 and T3N0; Eastern Cooperative Oncology Group (ECOG) performance status 0 or 1; R0 resection achieved by gastrectomy with D1 plus or D2 dissection for clinical stage IA, or D2 dissection for clinical stage IB or more (surgical approach by laparotomy was recommended in all cases and the laparoscopic approach was allowed for clinical stage I); enrolment within 7 weeks after surgery; Added value of this study To our knowledge, our study is the first randomised controlled trial to investigate non-inferiority of four courses of S-1 to eight-courses of S-1 in patients with stage II gastric cancer. The trial was stopped early for futility after 528 patients were enrolled because of a lower relapse-free survival in the group receiving four courses of S-1, although the scientific relevance of the study is limited because of the low number of events. Implications of all the available evidence This study provides the first evidence regarding duration of S-1 adjuvant chemotherapy in patients with stage II gastric cancer. S-1 for 1 year should remain as standard adjuvant chemotherapy for stage II gastric cancer. Our findings warn against the shortening of adjuvant chemotherapy and suggest that adjuvant chemotherapy should be continued for eight courses without early termination. age years; no remnant gastric cancer; no previous chemotherapy, radiation therapy, or molecularly targeted therapy for any other malignancies; sufficient oral intake; and adequate organ function including neutrophil count ⁹/L or more, haemoglobin 80 g/l or more, platelet count 75 10⁹/L or more, aspartate aminotransferase 100 IU/L or lower, alanine aminotransferase 100 IU/L or lower, total bilirubin concentration 20 mg/l or less, and creat inine clearance 60 ml/min or more if body surface area was above 1 25 m² or 40 ml/min or more if body surface area was less than 1 25 m². Tumours were staged in accordance with the Japanese Classification of Gastric Carcinoma. 14 Patients were excluded if they had any of the following criteria: active coexisting cancer; active infection; body temperature of 38 C or more; persisting surgical complications; severe mental disorder; use of flucytosine, phenytoin, or warfarin treatment; systemic administration of corticosteroids; uncontrolled diabetes; unstable hypertension; unstable angina or myocardial infarction within 6 months; positive for hepatitis B antigen or hepatitis C antibody; or severe respiratory disease. The study protocol was approved by the JCOG Protocol Review Committee and the institutional review board of each participating hospital before initiation of the study. The Data and Safety Monitoring Committee monitored data and operation of the study. This study was done in accordance with the international ethical recommendations stated in the Declaration of Helsinki and the Ethical Guidelines for Medical and Health Research Involving Human Subjects. 15 All patients were required to provide written informed consent prior to enrolment. Randomisation and masking After the researcher had obtained patient consent, the researcher or support staff registered the patient into the 2 Published online January 21,

3 JCOG Data Center s web entry system. Eligible patients were randomly assigned (1:1) to receive eight courses of S-1 or four courses of S-1 by a minimisation method with a random component using institution, stage (IIA vs IIB), age (<70 years vs 70 years), and mode of operation (open gastrectomy with bursectomy vs open gastrectomy without bursectomy vs laparoscopic gastrectomy) as adjustment factors. Patients and all investigators were unmasked to treatment assignment. The JCOG Data Center did central monitoring to ensure patient eligibility, data submission, protocol compliance, safety, and on-schedule study progress. Monitoring reports were independently reviewed by the JCOG Data and Safety Monitoring Committee. Procedures All patients had curative gastrectomy at a participating hospital in the Stomach Cancer Study Group of the JCOG. Within 50 days after surgery, both groups initiated oral S-1 80 mg/m² per day ( mg/day total dose depending on the patient s body surface area: 80 mg if <1 25 m²; 100 mg if m²; and 120 mg if >1 5 m²) on days 1 28, followed by 2 weeks rest (one course), repeated every 6 weeks. Patients allocated to the eightcourse group received S-1 for eight courses or until completion of the last full course at 1 year after surgery. Patients assigned to the four-course group received S-1 for four courses or until completion of the last full course at 7 months after surgery. The criteria for initiating each course were: body temperature less than 38 C; neutrophil count 1200 cells per µl or more, or white blood cell count 2500 cells per µl or more; haemoglobin 80 g/l or more; platelet count ⁴ cells per µl or more; aspartate aminotransferase 100 IU/L or less; alanine aminotransferase 100 IU/L or less; total bilirubin concentration 0 02 g/l or less; creatinine g/l or less; reduction to grade 1 or less of any grade 2 diarrhoea, nausea, vomiting, appetite loss, or stomatitis observed in the previous course, or of any grade 3 or higher toxicities observed in the previous course. S-1 was suspended when a patient experienced any of the following toxicities: neutrophil count less than 1000 cells per µl (or white blood cell count less than 2000 cells per µl); platelet count less than ⁴ cells per µl; aspartate aminotransferase more than 100 IU/L; alanine aminotransferase more than 100 IU/L; total bilirubin concentration more than 0 02 g/l; creatinine more than g/l; grade 2 or 3 diarrhoea, nausea, vomiting, appetite loss, stomatitis, or fever; and any other grade 3 toxicity. When the investigator judged that these adverse events were attributed to S-1, dose reduction or switch to a 3-week schedule was required. In the 3-week schedule, a single course comprised S-1 administered on days 1 14, followed by a 1-week rest, repeated once. S-1 was stopped in the case of relapse, grade 4 nonhaematological toxicities, adverse events requiring rest of more than 28 days, adverse events requiring dose reduction of S-1 beyond the allowed level, patient refusal to continue treatment for any reason, or on the basis of the attending physician s judgment. During treatment, patients in both treatment groups were assessed at least once every 2 weeks in the first course and at least once every 3 weeks in the second and further courses. Patients who received two continuous courses of S-1 without changing the schedule could be assessed at least once every 6 weeks. Patients were assessed at least every 6 weeks for the first year after surgery, every 3 months for the next 2 years, and every 6 months thereafter until 5 years after randomisation. Assessment consisted of physical examination, ECOG performance status, blood count, blood chemistry, and toxicity according to the Common Terminology Criteria for Adverse Events (CTCAE version 4.0). Carcinoembryonic antigen and carbohydrate antigen 19 9 were checked every 6 weeks during treatment, then every 3 months for the first 3 years after randomisation, and every 6 months thereafter for the next 2 years. Abdominal and pelvic CT scanning was done every 6 months for the first 3 years after randomisation and every year thereafter for the next 2 years. Chest X-ray or thoracic CT scanning was done every year for 5 years. Upper gastrointestinal endoscopy was done every year after surgery in patients who received distal gastrectomy, proximal gastrectomy, or pylorus-preserving gastrectomy. Outcomes The primary endpoint was relapse-free survival, defined as the time from random assignment to the date of relapse confirmed by the imaging or clinically diagnosed by the physician, to the date of death from any cause, whichever came earlier, and censored on the last contact for a surviving patient without recurrence. Secondary endpoints were overall survival, time to treatment failure, proportion of treatment continuation at each time, adverse events, the incidence of early mortality, and treatment-related death. Overall survival was defined as the time from random assignment to the date of death from any cause and censored on the last date of contact for a surviving patient. Time to treatment failure was defined as the time from random assignment to the date of treatment failure, except if due to early termination of the study, censored on the last date of contact for a surviving patient. Adverse events were assessed according to the CTCAE. Early mortality was defined as deaths that occurred either during the protocol treatment period or within 30 days after the final date on protocol treatment. Statistical analysis This study was designed to confirm the non-inferiority of four courses of S-1 compared with eight courses of S-1 in terms of relapse-free survival. Expecting 3-year relapse-free survival of 85% in both groups with a noninferiority margin for the hazard ratio (HR) set at 1 37, 1000 patients (246 events) were required to achieve statistical power of 80% with one-sided α of 5%, with an Published online January 21,

4 295 assigned to receive eight courses of S started treatment 172 completed eight courses 590 randomly assigned 6 post-hoc non-eligible 1 laparoscopic surgery for stage II 1 laparoscopic proximal gastrectomy 3 laparoscopic total gastrectomy 1 lower creatinine clearance 123 did not complete eight courses 5 did not start treatment 3 relapse 31 adverse events 29 patient refusals related to adverse event 3 patient refusals unrelated to adverse event 50 due to early termination of the study 7 other 295 assigned to receive four courses of S started treatment 231 completed four courses 18 post-hoc non-eligible 6 laparoscopic surgery for stage II 3 violations for extent of nodal dissection 1 laparoscopic surgery for stage II and violation for extent of nodal dissection 4 laparoscopic total gastrectomy 1 previous radiation therapy for prostate cancer 1 stage IIIA due to N3a 1 neutrophil unknown 1 remnant gastric cancer 64 did not complete eight courses 3 did not start treatment 4 relapse 13 adverse events 15 patient refusals related to adverse event 27 due to early termination of the study 5 other early for efficacy or futility were planned at 50% and 100% enrolment, respectively. The prespecified stopping criteria of the study were (1) for efficacy, when relapse-free survival for four courses of S-1 was significantly superior to that of eight courses of S-1; or (2) for futility, when relapse-free survival of four-courses of S-1 was below that for eight courses of S-1 (ie, HR >1 37). The Data and Safety Monitoring Committee of the JCOG independently reviewed the interim analysis report and could recommend early termination of the study. Data from all randomly assigned patients were analysed for relapse-free survival and overall survival on an intention-to-treat basis. Overall survival, relapse-free survival, and time to treatment failure curves were calculated by the Kaplan-Meier method. HRs for relapsefree survival were estimated with a stratified Cox regression model stratified by pathological stage, and HRs for overall survival and time to treatment failure were estimated with an unstratified Cox regression model. Safety was assessed on a per-protocol basis. The p value for the primary analysis of the relapse-free survival is one sided and all other p values are two sided. Interactions between treatment and nine prespecified sub groups were investigated in terms of relapse-free survival and overall survival: sex, surgical procedure (open gastrectomy with bursectomy vs open gastrectomy without burusectomy vs laparoscopic gastrectomy), age (<70 years vs 70 years), histological type (differentiated vs undifferentiated), ECOG per formance status (0 vs 1), stage (IIA vs IIB), T status (T2 vs T3), N status (N0 vs N1 vs N2), and lymph node dissection (D1 or D2, or both). In a posthoc analysis, time to recurrence was estimated by a cumulative incidence function approach with death as a competing risk. HR for time to recurrence was estim ated with the Fine and Gray model. Statistical analyses were done by the JCOG Data Center with SAS software, version 9.4. This study is registered with UMIN-CTR, number UMIN included in efficacy analysis (intention-totreat) 290 included in safety analysis Figure 1: Trial profile 295 included in efficacy analysis (intention-totreat) 292 included in safety analysis expected accrual period of 5 years, and a follow-up period of 3 years. Because of slow patient accrual, the protocol was amended on Aug 16, 2016, to prolong the total accrual period from 5 years to 10 years. Two interim analyses were planned, with adjustments for repeated comparisons using the Lan and DeMets method and the O Brien-Fleming type α spending function. 16 The first and second interim analyses for stopping Role of the funding source The funders of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data in the study and final responsibility for the decision to submit for publication. Results Between Feb 16, 2012, and March 19, 2017, 590 patients from 59 hospitals in Japan were enrolled and randomly assigned to receive eight courses of S-1 (n=295) or four courses of S-1 (n=295). Among them, 528 patients enrolled between Feb 16, 2012, and Oct 31, 2016, were included in the first planned interim analysis with a data cutoff date of Nov 29, 2016, when there were 27 observed events of the total 246 expected, and information time was (27/246). At this time, the point estimate of HR for relapse-free survival in the four-course group 4 Published online January 21,

5 compared with the eight-course group was 2 52 (95% CI ), which was greater than 1 37 and met the prespecified criteria for early termination of the study. The predictive probability for showing non-inferiority at the final analysis was calcu lated to be 2 9%. The JCOG Data and Safety Monitoring Committee recommended that the trial be terminated early, and the JCOG Stomach Cancer Study Group accepted this recommendation. The study was termin ated on March 19, After randomisation, six (2%) of 295 patients in the eight-course group and 18 (6%) of 295 patients in the four-course group were deemed ineligible mainly because of laparoscopic surgery for clinical stage II disease (n=8) or for laparoscopic total gastrectomy (n=7; figure 1). One patient allocated to the four-course S-1 group was deemed ineligible because of stage IIIA disease. Efficacy analysis was done in all 590 randomly assigned patients and safety analysis was done in all patients who initiated treatment (n=582; 290 in the eight-course group and 292 in the four-course group); eight patients did not initiate treatment. Patient demographics, tumour characteristics, and surgical procedures were well balanced between the groups (table 1). With a data cutoff date of May 31, 2017, the median follow-up in the intention-to-treat group was 1 75 years (IQR ). The updated relapse-free survival at 3 years was 93 1% (95% CI ) for the eight-course group and 89 8% (95% CI ) for the four-course group (HR 1 84, 95% CI ) (figure 2A). The updated overall survival at 3 years was 96 1% (95% CI ) for the eight-course group and 92 6% (95% CI ) for the four-course group (HR 3 37, 95% CI ; figure 2B). All subgroups had an HR of more than 1 0 in relapse-free survival and over all survival except N2 in relapse-free survival (appendix, p 1). Cumulative risk of relapse at 3 years was 5 5% (95% CI ) in the eight-course group and 7 7% (95% CI ) in the four-course group (HR 1 59, 95% CI ; appendix p 3). Relapse was observed in 11 (4%) of 295 patients in the eight-course group and 17 (6%) of 295 patients in the four-course group. Relapses at the peritoneum and lymph nodes were more frequently observed in the four-course group than in the eight-course group, and relapses at liver and distant sites were similar between the two groups (table 2). At 3 months, 91 8% (95% CI ) of patients in the eight-course group and 92 9% ( ) of patients in the four-course group had not reached treat ment failure. At 6 months, 83 7% (95% CI ) in the eight-course group and 87 2% ( ) in the four-course group had not reached treatment failure, and 77 2% (95% CI ) and 72 7% (95% CI ) at 9 and 12 months in the eight-course group. Reasons for treatment termination were relapse (three [1%] patients in the eight-course group vs four [1%] patients in the four-course group), adverse events (60 [20%] vs 28 [9%]), Eight-course group (n=295) Four-course group (n=295) Age (years) 64 (58 71) 65 (59 70) < (73%) 215 (73%) (27%) 80 (27%) Sex Men 184 (62%) 193 (65%) Women 111 (38%) 102 (35%) EGOG performance status (92%) 268 (91%) 1 23 (8%) 27 (9%) T status T2 130 (44%) 119 (40%) T3 83 (28%) 102 (35%) T4 82 (28%) 74 (25%) N status N0 82 (28%) 74 (25%) N1 163 (55%) 181 (61%) N2 50 (17%) 39 (13%) N3 0 (0%) 1 (<1%) Stage IIA 80 (27%) 79 (27%) IIB 215 (73%) 215 (73%) IIIA 0 (0%) 1 (<1%) Surgical procedure Open bursectomy 44 (15%) 45 (15%) Open non-bursectomy 176 (60%) 180 (61%) Laparoscopic surgery 75 (25%) 70 (24%) Operation time (min) 236 ( ) 243 ( ) Blood loss (ml) 176 (40 330) 165 (50 330) Extent of gastrectomy Proximal 4 (1%) 3 (1%) PPG 1 (<1%) 4 (1%) Distal 215 (73%) 210 (71%) Total 75 (25%) 78 (26%) Lymph node dissection D1 0 0 D1+ 26 (9%) 32 (11%) D2 269 (91%) 262 (89%) Other* 0 1 (<1%) Pathology Well differentiated 127 (43%) 127 (43%) Poorly differentiated 163 (55%) 168 (57%) Other 5 (2%) 0 Data are n (%) or median (IQR). ECOG=Eastern Cooperative Oncology Group. PPG=pylorus-preserving gastrectomy. *Nodal dissection for remnant gastric cancer. Lymphoid infiltrating carcinoma. Table 1: Baseline characteristics patient refusal (three [1%] vs zero [0%]), early termination of the trial (50 [17%] vs 27 [9%]), and other reasons including protocol deviations because of physicians misunderstanding and occurrence of secondary malignancies (seven [2%] vs five [2%]). Among all patients who Published online January 21,

6 Relapse-free survival (%) Number at risk (number censored) A n Events 3-year relapse-free survival (95% CI) HR 1 84 (95% CI ) Eight courses % ( ) Four courses % ( ) Eight courses 295 (0) Four courses 295 (0) Eight courses Four courses 295 (0) 295 (0) 270 (23) 262 (25) 272 (23) 266 (25) 217 (49) 215 (43) 221 (50) 220 (44) 176 (38) 169 (41) 181 (40) 175 (43) Figure 2: Kaplan-Meier curves (A) Relapse-free survival (n=590). (B) Overall survival (n=590). 132 (42) 126 (41) 136 (44) 131 (42) 106 (26) 98 (28) 108 (27) 102 (28) 78 (26) 74 (22) 80 (26) 77 (23) 56 (22) 47 (26) 57 (23) 50 (26) 33 (23) 23 (23) B n Events 3-year overall survival (95% CI) 20 Eight courses % ( ) 10 Four courses % ( ) HR 3 37 (95% CI ) Years after randomisation Number at risk (number censored) Overall survival (%) 33 (24) 23 (25) 14 (19) 15 (8) 14 (19) 15 (8) 2 (12) 2 (13) 2 (12) 2 (13) 0 (2) 0 (2) 0 (2) 0 (2) 0 (0) 0 (0) 0 (0) 0 (0) Eight-course group (n=295) Any Peritoneum 3 12 Lymph nodes 1 5 Liver 6 4 Other distant sites 4 4 Table 2: Relapse site Four-course group (n=295) were treated, dose reduction of S-1 was required in 132 (46%) of 290 patients in the eight-course group and in 114 (39%) of 292 patients in the four-course group. The most common adverse events of any grade were leukopenia, neutropenia, anaemia, thrombocytopenia, ele vated aspartate aminotransferase and alanine aminotransferase, nausea, diarrhoea, and anorexia in both groups, whose incidences were similar between the two groups (table 3). The most common grade 3 4 adverse events were neutropenia, observed in more than 15% of patients in both groups. Other grade 3 adverse events were uncommon, occurring in 6% of patients or less in the both groups. One treatment-related death occurred in the four-course group due to sudden death with unknown cause. Discussion This study was a phase 3 trial to examine the duration of adjuvant chemotherapy with S-1 for resected stage II gastric cancer. Unexpectedly, the trial was stopped early for futility at the first interim analysis, because the HR for relapse-free survival of the four-course group compared with the eight-course group was found to exceed the prespecified stopping criteria for futility, with a low possibility to show non-inferiority even if the study had been completed. Moreover, the HR for overall survival and subgroup analyses showed a similar trend of unfavourable survival in the four-course group. Thus, 6 Published online January 21,

7 Eight-course group (n=290) Four-course group (n=292) Grade 1 2 Grade 3 Grade 4 Grade 1 2 Grade 3 Grade 4 Leukopenia 136 (47%) 7 (2%) 0 (0%) 120 (41%) 7 (2%) 0 (0%) Anaemia 250 (86%) 8 (3%) 1 (<1%) 256 (88%) 7 (2%) 0 (0%) Thrombocytopenia 144 (50%) 4 (1%) 0 (0%) 165 (57%) 3 (1%) 0 (0%) Neutropenia 184 (63%) 45 (16%) 1 (<1%) 161 (55%) 46 (16%) 5 (2%) Hypoalbuminaemia* 224 (78%) 3 (1%) 222 (76%) 1 (<1%) Elevated serum bilirubin 73 (25%) 0 (0%) 0 (0%) 59 (20%) 1 (<1%) 0 (0%) Serum glutamic oxaloacetic transaminase 173 (60%) 3 (1%) 1 (<1%) 162 (55%) 4 (1%) 0 (0%) Serum glutamic pyruvic transaminase 119 (41%) 3 (1%) 1 (<1%) 126 (43%) 4 (1%) 0 (0%) Elevated creatinine 28 (10%) 0 (0%) 0 (0%) 24 (8%) 0 (0%) 0 (0%) Fever 39 (13%) 1 (<1%) 0 (0%) 15 (5%) 0 (0%) 0 (0%) Fatigue 118 (41%) 5 (2%) 98 (34%) 7 (2%) Abdominal pain 47 (16%) 6 (2%) 35 (12%) 3 (1%) Diarrhoea 139 (48%) 16 (6%) 0 (0%) 119 (41%) 13 (4%) 0 (0%) Nausea 88 (30%) 3 (1%) 100 (34%) 2 (<1%) Vomiting 30 (10%) 3 (1%) 0 (0%) 41 (14%) 5 (2%) 0 (0%) Oral mucositis 75 (26%) 4 (1%) 0 (0%) 54 (18%) 2 (<1%) 0 (0%) Alopecia 10 (3%) 5 (2%) Skin hyperpigmentation 88 (30%) 73 (25%) Palmar-plantar erythrodysesthesia syndrome 19 (7%) 1 (<1%) 16 (5%) 1 (<1%) Maculopapular rash 41 (14%) 5 (2%) 55 (19%) 2 (<1%) Anorexia 140 (48%) 13 (4%) 0 (0%) 135 (46%) 17 (6%) 0 (0%) Watering eyes 60 (21%) 5 (2%) 49 (17%) 1 (<1%) Data are n (%). *Data for two patients were missing in the eight-course group. For some adverse events, grade 4 is not defined in the Common Terminology Criteria for Adverse Events (version 4.0). Table 3: Adverse events although underpowered, the study did not support the conclusion that survival was non-inferior after four courses of S-1 compared with eight courses of S-1. Our conclusion is that S-1 for 1 year should remain as standard adjuvant chemotherapy for stage II gastric cancer. The reason why the study was negative is unclear. Although the number of relapses was small, betweengroup differences in the pattern of relapse were interesting. Relapses at the peritoneum and lymph nodes were high in the four-course group, and relapses at the liver and other distant sites were similar between the two groups. The ACTS-GC trial 7 suggested that adjuvant S-1 for 1 year after surgery was effective especially for preventing relapse at the peritoneum and lymph nodes compared with surgery alone.the most frequent relapse site was the peritoneum in the S-1 group of the ACTS-GC trial, 7 whereas it was the liver in the eight-course group in the present study. This finding could be explained by differences in the study populations. T4 disease was dominant in the ACTS-GC trial for stages II and III gastric cancer, 2 but this was not the case for patients with stage II disease in our trial. The most frequent relapse site in the four-course group of our study was the peritoneum. In the CLASSIC trial, comparing surgery alone and surgery plus chemotherapy with capecitabine and oxaliplatin for 6 months, haematogenous relapse was less frequent in the chemotherapy group but peritoneal relapse was similar. 8 Gastric cancer is characterised by frequent peritoneal metastases, by contrast with other neoplasms such as colon cancer, in which adjuvant chemotherapy for 6 months is established and haematogenous relapse predominates. Future basic or clinical research might clarify whether longer adjuvant chemotherapy can prevent peritoneal recurrence of gastric cancer. Previously, there were two phase 3 studies comparing duration of chemotherapy for resectable gastric cancer. One was the AMC trial comparing combination chemotherapy with mitomycin C and short doxifluridine (Mf) with mitomycin C, long doxifluridine, and cisplatin (MFP) in the adjuvant setting. 17 In the Mf regimen, patients received mitomycin C for 3 6 weeks followed by doxifluridine for 3 months. 17 In the MFP regimen, patients received mitomycin C for 3 6 weeks followed by cisplatin for 6 months and doxifluridine for 12 months. 17 The AMC trial did not show a benefit of MFP for overall survival and relapse-free survival as compared with Mf, 17 suggest ing that increased duration of treatment with an oral pyrimidine conferred no survival benefit, by contrast with our results with S-1. However, unlike the case for S-1, there is no solid evidence that the Mf or MFP regimen improves survival over surgery alone, nor is there clear evidence that adjuvant chemotherapy using old pyrimidines such as Published online January 21,

8 doxifluridine confers a survival benefit. The phase 3 OE05 trial compared two cycles of preoperative cisplatin and fluorouracil with four cycles of preoperative epirubicin, cisplatin, and capecitabine for oesophageal adenocarcinoma 18 and found no difference in survival between the groups. 18 The aim of neoadjuvant chemotherapy is early eradication of micrometastatic tumour cells, but a risk of this approach is delayed surgical resection, which might reduce survival. In the OE05 trial, pathological response was achieved in less than 20% of patients in both groups. 18 Among patients receiving S-1 in the ACTS-GC study, treatment was continued for at least 3 months in 452 (87 4%) of 517 patients, 2 at least 6 months in 403 (77 9%) patients, at least 9 months in 366 (70 8%) patients, and 12 months in 340 (65 8%) patients. 2 In our study, the proportion of patients continuing treatment was slightly higher at all time points in both groups. Although information time in the interim analysis was about 11%, the JCOG Data and Safety Monitoring Committee recommended early termination and the JCOG Stomach Cancer Study Group accepted this recommendation, despite the low number of events that had accrued at this time. When stopping early for futility was discussed during the interim analysis, the balance between risk to the patients and robustness of the scientific results was considered. JCOG policy states that the risk to patients must be more important than robustness of data. We therefore prespecified the time for the interim analyses by the number of patients enrolled rather than the number of events. Although the number of events was small, we decided that the trial should be stopped considering the risk to the patients who were assigned to the four-course group. Both 3-year relapse-free survival and overall survival were over 90% in the eight-course group, which was better than expected; in the ACTS-GC study, 3-year relapse-free survival was 84 3% and overall survival was 91 1%. 7 Possible reasons for this discrepancy include the short follow-up and the lack of power due to the low number of events in our study. The proportion of patients continuing treatment with S-1 was also slightly higher in our study than in the ACTS-GC trial. 2 Frequently observed adverse events were similar in the both groups and were concordant with the ACTS-GC phase 3 study. 2 Moreover, few grade 3 or grade 4 adverse events were observed in the ACTS-GC study, 2 also consistent with our study. We observed no new toxicities. Thus, S-1 adjuvant chemotherapy is feasible and safe after D2 gastrectomy. Our study has several limitations. First, the planned sample size was not achieved because the trial was stopped early, restricting the statistical power to support conclusions. Second, the timing of the interim analysis was not based on the number of events, but rather on the number of enrolled patients. Therefore, although this criterion was prespecified, the scientific relevance of the study might be limited as a result. Although these findings cannot be considered scientifically conclusive, powered trials to identify the optimal duration of S-1 would not be ethical without strong rationale that supercedes our results. To confirm the findings of this trial, we plan to follow up the patients for longer, which is anticipated to lead to an increased number of events. Third, relapse-free survival was judged at each institution; there was no central review of imaging for the assessment of relapse. As such, there is a potential risk for lack of concordance with investigator-judged relapse. Therefore, we plan to do a central review of imaging. Finally, this study was an open-label, unblinded, randomised controlled trial, and the absence of placebo after finishing S-1 in the fourcourse group might also affect the results. In conclusion, S-1 for 1 year should remain as standard adjuvant chemotherapy for stage II gastric cancer. Although time to treatment failure associated with S-1 was relatively high in this study as compared with the ACTS-GC trial, increased compliance of S-1 might improve effectiveness. Whether the optimal period of adjuvant S-1 treatment is 1 year or whether this could be shortened by increasing compliance is unclear and requires further studies. Contributors TYo, MS, and TS had the original idea for the study, and MS then MT chaired the study group. TYo wrote the protocol, assisted by MT, JM, HK, HF, TS, and MS. Statistical analyses were done by JM. All authors, except JM, HK, and HF, recruited patients into the study. JM, HK, and HF were responsible for data management, statistical analysis, and interpretation. TYo drafted the paper. MT, JM, HK, HF, NB, TS, and MS revised the paper. SN, YN, TYam, MK, NF, SH, YC, HY, KY, SI, AT, TYas, YK, HK, HF, TS, and MS reviewed the paper. All authors approved the final version. Declaration of interests TYo reports grants from the Japan Agency for Medical Research and Development (AMED) during the conduct of the study, lecture fees from Taiho Pharmaceutical, Chugai Pharmaceutical, Ono Pharmaceutical, Bristol-Myers Squibb, Merck Sharp and Dohme (MSD), Daiichi Sankyo, Yakult Honsha, Takeda Pharmaceutical, Nippon Kayaku, Kaken Pharmaceutical, Abbott Japan, Medtronic, Johnson & Johnson, and Olympus, and grant support from Taiho Pharmaceutical, Chugai Pharmaceutical, and Yakult Honsha, outside of the submitted work. JM reports grants from Ministry of Health, Labour and Welfare, Japan, and the Japan AMED during the conduct of the study. KY reports grants, personal fees, and non-financial support from Taiho Pharm Co during the conduct of the study, and grants, personal fees, and non-financial support from Sanofi, Chugai Pharm Co Ltd, Yakult Honsha Co Ltd, Eli Lilly Japan KK, Daiichi Sankyo Co Ltd, Merck Serono Co Ltd, Novartis Pharma, EA Pharma Co Ltd, Johnson & Johnson KK, Covidien Japan, Takeda Pharmaceutical Co Ltd, and Ono Pharm Co Ltd, grants and personal fees from Nippon Kayaku Co Ltd amd Otsuka Pharma Co Ltd, grants from Sumitomo Dainippon Pharma, Bristol-Myers Japan, Tsumura, Kyowa Hakko Kirin, Astellas, Toyama Chemical, Kinetic Concepts Inc, Abbott Japan, and Asahi Kasei, personal fees from Olympus, Terumo, Denka, MSD, and Bayer Yakuhin Ltd, outside of the submitted work. SI reports grants from MSD and Ono Pharmaceutical, personal fees from Chugai Pharmaceutical, and Otsuka Pharmaceutical Factory outside of the submitted work. HK reports grants from Ministry of Health, Labour and Welfare, Japan, and Japan AMED during the conduct of the study, and personal fees from Johnson & Johnson, outside of the submitted work. HF reports grants from Ministry of Health, Labour and Welfare, Japan, and National Cancer Center, Japan during the conduct of the study, personal fees from Taiho Pharma and 8 Published online January 21,

9 Chugai Pharmaceutical, outside of the submitted work. NB reports honoraria from Taiho, Ono, Chugai, Shionogi, Yakult, Bristol-Myers Squibb, Eli Lilly, and Merck Serono, and study grants from Taiho, Ono, and Bristol-Myers Squibb, outside of the submitted work. TS reports personal fees from Taiho Pharma, outside of the submitted work. MS reports personal fees from Taiho Pharmaceutical during the conduct of the study. All other authors declare no competing interests. Data sharing Individual participant data that underlie the results reported in this Article will not be shared because the follow-up of the patients is continued until March, After publication of the final follow-up, using data as of March, 2022, individual participant data that underlie the results after deidentification will be shared if investigators whose proposed use of the data has been approved by the investigators from JCOG Stomach Cancer Study Group identified for this purpose. Proposals should be directed to Acknowledgments We thank Megumi Miyazawa for data management. The study was supported in part by the National Cancer Center Research and Development Funds (23-A-16, 23-A-19, 26-A-4, 29-A-3) and by AMED under Grant Number JP17ck References 1 GLOBOCAN2012. Estimated cancer incidence, mortality and prevalence worldwide in factsheets/cancers/7-stomach-fact-sheet.pdf (accessed Oct 5, 2018). 2 Sakuramoto S, Sasako M, Yamaguchi T, et al. Adjuvant chemotherapy for gastric cancer with S-1, an oral fluoropyrimidine. N Engl J Med 2007; 357: Bang YJ, Kim YW, Yang HK, et al. Adjuvant capecitabine and oxaliplatin for gastric cancer after D2 gastrectomy (CLASSIC): a phase 3 open-label, randomised controlled trial. Lancet 2012; 379: Cunningham D, Allum WH, Stenning SP, et al. Perioperative chemotherapy versus surgery alone for resectable gastroesophageal cancer. N Engl J Med 2006; 355: Macdonald JS, Smalley SR, Benedetti J, et al. Chemoradiotherapy after surgery compared with surgery alone for adenocarcinoma of the stomach or gastroesophageal junction. N Engl J Med 2001; 345: Al-Batran SE, Homann N, Pauligk C, et al. Effect of neoadjuvant chemotherapy followed by surgical resection on survival in patients with limited metastatic gastric or gastroesophageal junction cancer: the AIO-FLOT3 trial. JAMA Oncol 2017; 3: Sasako M, Sakuramoto S, Katai H, et al. Five-year outcomes of a randomized phase III trial comparing adjuvant chemotherapy with S-1 versus surgery alone in stage II or III gastric cancer. J Clin Oncol 2011; 29: Noh SH, Park SR, Yang HK, et al. Adjuvant capecitabine plus oxaliplatin for gastric cancer after D2 gastrectomy (CLASSIC): 5-year follow-up of an open-label, randomised phase 3 trial. Lancet Oncol 2014; 15: O Connell MJ, Laurie JA, Kahn M, et al. Prospectively randomized trial of postoperative adjuvant chemotherapy in patients with high-risk colon cancer. J Clin Oncol 1998; 16: Andre T, Quinaux E, Louvet C, et al. Phase III study comparing a semimonthly with a monthly regimen of fluorouracil and leucovorin as adjuvant treatment for stage II and III colon cancer patients: final results of GERCOR C96.1. J Clin Oncol 2007; 25: Chau I, Norman AR, Cunningham D, et al. A randomised comparison between 6 months of bolus fluorouracil/leucovorin and 12 weeks of protracted venous infusion fluorouracil as adjuvant treatment in colorectal cancer. Ann Oncol 2005; 16: Grothey A, Sobrero AF, Shields AF, et al. Duration of adjuvant chemotherapy for stage III colon cancer. N Engl J Med 2018; 378: Uesaka K, Boku N, Fukutomi A, et al. Adjuvant chemotherapy of S-1 versus gemcitabine for resected pancreatic cancer: a phase 3, open-label, randomised, non-inferiority trial (JASPAC 01). Lancet 2016; 388: Japanese Gastric Cancer Association. Japanese classification of gastric carcinoma: 3rd English edition. Gastric Cancer 2011; 14: Ethical Guidelines for Medical and Health Research Involving Human Subjects. Provisional Translation (as of March 2015). oukouseikagakuka/ pdf (accessed Oct 5, 2018). 16 Lan KKG, DeMets DL. Discrete sequential boundaries for clinical trials. Biometrika 1983; 70: Kang YK, Chang HM, Yook JH, et al. Adjuvant chemotherapy for gastric cancer: a randomised phase 3 trial of mitomycin-c plus either short-term doxifluridine or long-term doxifluridine plus cisplatin after curative D2 gastrectomy (AMC0201). Br J Cancer 2013; 108: Alderson D, Cunningham D, Nankivell M, et al. Neoadjuvant cisplatin and fluorouracil versus epirubicin, cisplatin, and capecitabine followed by resection in patients with oesophageal adenocarcinoma (UK MRC OE05): an open-label, randomised phase 3 trial. Lancet Oncol 2017; 18: Published online January 21,