22 Background: Survival is the ultimate end-point in field of medical oncology. Although tumor response has been used as surrogate end-point in early phase clinical trial in patients with gastric cancer, clear correlation between tumor response and long-term survival has not been established yet. The objective of the current study is to determine whether early assessment could predict survival in patients with AGC. Methods: A total of 337 patients with AGC who had participated one of the three first-line, prospective, investigator-initiated, randomized trials carried out in Yonsei Cancer Center were included in the analysis. 238 had received fluoropyrimidine (F) + platinum (P), 61 had received F + taxane (T), 38 had received T + P. Response was assessed according to RECIST 1.1 at 8, 14 and 20 weeks after initiation of treatment. The correlation between response at each time point and the OS was evaluated using logistic regression test. Results: The median OS was 13.7 months. (95% CI 11.8 – 15.6) In terms of confirmed tumor response, 1 CR, 84 PRs, 104 SDs and 148 PDs were observed. At every landmark period, disease control rate (DCR, CR + PR + SD) was superior in predicting OS comparing with response rate. (RR, CR + PR) (Table). Among them, 8-week DCR was the most powerful predictor for OS. (HR 0.18, p< .001) There was no significant OS difference between patients with (13.7 months) vs. without (13.9 months) measurable lesions (p = .705) and between patients with 8-week response of PR (18.1 months) vs. SD (15.3 months) (p = .522). Conclusions: Disease control was a better predictor for OS in patients with AGC who had received first-line chemotherapy than response, with disease control at 8 week being the most powerful predictor. This could be used as a surrogate marker for long-term survival in early phase clinical trials. [Table: see text]