Primary results of a randomized two‐by‐two factorial phase II trial comparing neoadjuvant chemotherapy with two and four courses of cisplatin/S‐1 and docetaxel/cisplatin/S‐1 as neoadjuvant chemotherapy for advanced gastric cancer

Abstract Aim Neoadjuvant chemotherapy (NAC) is promising to improve the survival of resectable gastric cancer. However, suitable regimen and treatment duration for NAC have not yet been established. Methods We conducted a randomized phase II trial to compare two and four courses of neoadjuvant S‐1/cisplatin (SC) and S‐1/cisplatin/docetaxel(DCS) using a two‐by‐two factorial design for locally resectable advanced gastric cancer. Patients with M0 and either T4 or T3 in case of junctional cancer or scirrhous‐type cancer received two or four courses of SC or DCS. Then, patients underwent D2 gastrectomy and adjuvant S‐1 chemotherapy for 1 year. The primary endpoint was 3‐year overall survival. The planned sample size was 120 eligible patients. Results Between October 2011 and September 2014, 132 patients were assigned to CS (n = 66; 33 in 2‐courses and 33 in 4‐courses) and DCS (n = 66; 33 in 2‐courses and 33 in 4‐courses). The 3‐year OS was 58.1% in CS and 60.0% in DCS with hazard ratio of 0.80 (95% CI, 0.48‐1.34), while it was 53.1% in the two courses and 65.0% in the four courses with hazard ratio of 0.72 (95% CI, 0.43‐1.22). In the survival analysis by duration in each regimen, the 3‐year OS was 58.1% for both two and four courses in CS, while it was 48.5% for two courses of DCS and 71.9% for four courses of DCS. Conclusions Considering high 3‐year OS, four courses DCS has a value to be tested in a future phase III study to confirm superiority of neoadjuvant chemotherapy for locally advanced gastric cancer.

rior efficacy in phase III trial. 7 In addition to the development of regimen, duration of chemotherapy must be another key for NAC. A certain duration is required for reduction of the tumor cell; however, longer duration may miss the curative operation due to progression of the tumor. Previously, we demonstrated that four courses of CS did not show the survival benefit that two courses of CS did in phase II; 12 however, the sample size was only 60 patients and inconclusive. Moreover, it remains unclear whether the case for DCS is different from CS.
Based on these, we conducted a randomized phase II trial, COMPASS-D, to compare NAC using two and four courses of CS and DCS with a two-by-two factorial design for locally advanced gastric cancer. The primary endpoint was 3-year overall survival (OS).
In our previous report analyzing the early outcome, 16 the pathological response rate was found to be similar, regardless of the regimen or course, and the chemotherapy-related toxicities and surgical outcome of both the regimen and course were found to be feasible.
This report clarified the survival results of the COMPASS-D phase II study, which is intended to select a better regimen and course for the next phase III trial.

| ME THODS
The study enrolled patients diagnosed as clinical T4 disease or clinical T3 disease in cases of tumors invading the esophagus and/or of the scirrhous type, including giant type 3 with a maximum diameter of >8 cm. The T and N stage were precisely determined by the protocol. Details of the entry criteria are shown in Table 1.
The patients enrolled in this study received two or four courses of CS or DCS regimens as a neoadjuvant chemotherapy as following: Arm A, two courses of CS; Arm B, four courses of CS; Arm C, two courses of DCS; and Arm D, four courses of DCS. In the CS regimen, S-1 was given 80 mg/m 2 orally (p.o.) daily for 3 weeks of a 4-week cycle, and cisplatin was given as an intravenous infusion of 60 mg/ m 2 on day 8 of each cycle. In the DCS regimen, S-1 was given 80 mg/ m 2 p.o. daily for 2 weeks of a 4-week cycle, and 60 mg/m 2 cisplatin and 40 mg/m 2 docetaxel were given as an intravenous infusion on day 1 of each cycle. The details of the neoadjuvant chemotherapy treatments have been reported previously. 15 Following neoadjuvant chemotherapy, patients underwent gastrectomy plus standard D2 lymphadenectomy. D3 lymphadenectomy or combined resection of adjacent organs or minimum resection of the peritoneum is allowed for curative intent, but more extended radical surgery, for instance pancreaticoduodenectomy or Appleby's surgery, are not. After a Results: Between October 2011 and September 2014, 132 patients were assigned to CS (n = 66; 33 in 2-courses and 33 in 4-courses) and DCS (n = 66; 33 in 2-courses and 33 in 4-courses). The 3-year OS was 58.1% in CS and 60.0% in DCS with hazard ratio of 0.80 (95% CI, 0.48-1.34), while it was 53.1% in the two courses and 65.0% in the four courses with hazard ratio of 0.72 (95% CI, 0.43-1.22). In the survival analysis by duration in each regimen, the 3-year OS was 58.1% for both two and four courses in CS, while it was 48.5% for two courses of DCS and 71.9% for four courses of DCS.

Conclusions:
Considering high 3-year OS, four courses DCS has a value to be tested in a future phase III study to confirm superiority of neoadjuvant chemotherapy for locally advanced gastric cancer.

K E Y W O R D S
cisplatin/S-1, docetaxel/cisplatin/S-1, gastric cancer, neoadjuvant therapy macroscopic curative resection was achieved, postoperative chemotherapy using S-1 of 80 mg/m 2 p.o. daily for 28 days, every 6 weeks, is initiated within 6 weeks after surgery and continued for 1 year.
After surgery, the patients received a physical examination, laboratory test, and abdominal computed tomography scan at least once every 6 months until 3 years after the accrual. Recurrence was confirmed by imaging studies, including computed tomography, ultrasonography, laparoscopy, gastrointestinal radiography, and endoscopy.
Individual patients were followed up for at least 5 years after the accrual in accordance with the protocol.
The present study was open-label, randomized phase II trial of the selection design proposed by Simon et al. This study was designed to evaluate the effectiveness of four courses of NAC compared with two courses, and the effectiveness of the DCS regimen compared with CS. Primary endpoint was 3-year OS and the key secondary endpoint was progression-free survival (PFS). The regimen showing a higher 3-year OS rate was considered more promising for a following phase III trial. Initially, we predicted that the 3-year OS rate of reference arm would be around 50% as previously reported. 10 The DCS and four-course regimens were expected to be 10% better in the 3-year OS rate than the CS and two-course regimens. Thus, we assumed that the 3-year OS rate of DCS and four-course regimens would be 60%. The sample size required to ensure 85% probability of the correct selection of a more effective regimen was calculated to be 110 patients, with 55 patients per arm. Considering the likelihood of dropouts and ineligible patients, the number of patients to be accrued was set at 120 in total. The primary analysis in this study aimed to estimate the 3-year OS.
OS and PFS were summarized by the Kaplan-Meier method in a full analysis set including the randomized patients with start of allocated treatments. The survival curves were compared by the log-rank test and hazard ratios (HRs) were estimated by the Cox regression models. Subgroup analysis were carried out to assess whether the relative effect from regimens and duration of NAC varies according to baseline characteristics, and treatment effect in each subgroup was presented using a forest plot. All clinical data were held centrally at the data center and analyzed using SAS version 9.4 (SAS Institute Inc).
Eligible patients were registered and subsequently randomised using a centralised dynamic randomization method with the following stratification factors: scirrhous type including giant type 3 (yes/ no), tumors invading the esophagus (yes/no), cT3-4a/T4b disease, lymph node metastasis (yes/no), and institution.
Although no significant differences were observed among these four arms (P = .2377), the 3y-OS of four courses DCS (arm D) was 14% better than the that of two courses CS (arm A) as a reference The forest plots presenting effects of treatment on OS are shown in Figure 4 by regimen and in Figure 5 by number of courses.
Remarkably, four courses were better in type 4 and giant type 3 than in other macroscopic types.

| DISCUSS IONS
The major finding of the present trial was that the four-course neo- Another interesting result was that four-course arm had superior survival than two-course arm, which was apparent only in a shorter follow-up period is more favorable because our aim is not a definitive result but is to explore a better regimen and course for future phase III study. Based on these, we set 3-year OS as the primary endpoint in this phase II study.
In conclusion, four-course DCS regimen of neoadjuvant chemotherapy is recommended as a promising regimen in a future phase III study for locally advanced gastric cancer.