Multicenter phase III comparison of cisplatin/S-1 (CS) with cisplatin/5-FU (CF) as first-line therapy in patients with advanced gastric cancer (FLAGS): Secondary and subset analyses

4511 Background: The primary analysis of FLAGS (ASCO-GI-2009) showed that CS and CF had similar overall survival (OS) but CS had a significantly superior safety profile. Methods: 1,053 (1,029 treated; CS=521/CF=508) patients with untreated, advanced gastric/gastroesophageal adenocarcinoma were randomized to either S-1 (25 mg/m2 bid, d 1–21)/cisplatin (75 mg/m2 d 1) q 28 d or 5-FU (1,000 mg/m2/d 5-d infusion)/cisplatin (100 mg/m2 d 1) q 28 d. OS analyses for non-inferiority (NI), by pre-specified stratifications, and by the largest histologic subset (diffuse type histology) were performed. Results: OS for NI: OS from CS compared to CF had a HR=0.92 (two-sided 95% CI, 0.80–1.05). HR=1.05 being much lower than HR=1.22 derived from the literature. Using a stringent HR non-inferiority margin of 1.10, CS remains statistically significantly non-inferior (p=0.0068) to CF. The 74% preserved control effect by CS is well above the suggested 50% by Rothmann et al. (Statist-Med2003;22:239–264). OS by stratifications: Of 12 stratification sub-categories, CS produced OS HR=<1.0 in 9 and HR=>1.0 in 3. Subset analysis: OS analysis for diffuse type histology (n=590) showed that CS (median survival=9.0 months) resulted in a superior OS (Log rank p=0.0413; HR, 0.83 [95% CI, 0.70 to 0.99]) than CF (median survival=7.1 months). Conclusions: CS is statistically non-inferior to CF while proving much safer for the patients. CS resulted in a HR=<1.0 in the majority of pre-specified stratifications and CS produced statistically superior OS for patients with diffuse type histology (needs prospective studies). CS is an optimum substitute for CF. Supported by Taiho Pharma, USA. No significant financial relationships to disclose.