When designing this trial, there was no evidence that salvage chemotherapy (SLC) in advanced gastric cancer (AGC) resulted in substantial prolongation of survival when compared with best supportive care (BSC). However, SLC is often offered to pretreated patients with AGC for anecdotal reasons. Patients with AGC with one or two prior chemotherapy regimens involving both fluoropyrimidines and platinum and with an Eastern Cooperative Oncology Group performance status (PS) 0 or 1 were randomly assigned in a ratio of 2:1 to SLC plus BSC or BSC alone. Choice of SLC-either docetaxel 60 mg/m(2) every 3 weeks or irinotecan 150 mg/m(2) every 2 weeks-was left to the discretion of investigators. Primary end point was overall survival (OS). Median OS was 5.3 months among 133 patients in the SLC arm and 3.8 months among 69 patients in the BSC arm (hazard ratio, 0.657; 95% CI, 0.485 to 0.891; one-sided P = .007). OS benefit for SLC was consistent in most of the prospectively defined subgroups, including age, PS, number of prior treatments, metastatic sites, hemoglobin levels, and response to prior chemotherapy. SLC was generally well tolerated, and adverse events were similar in the SLC and BSC arms. We found no median OS difference between docetaxel and irinotecan (5.2 v 6.5 months; P = .116). To our knowledge, this is the largest phase III trial comparing SLC plus BSC with BSC alone in AGC. In pretreated patients, SLC is tolerated and significantly improves OS when added to BSC.