A total of 154 patients were recruited: 79 G17DT and 75 placebo. A final analysis of the intention-to-treat population, using a proportional hazards model, stratifying by disease stage and adjusting for interim analysis, gave a hazard ratio for mortality of 0.75 (95% confidence interval, 0.51-1.10, P = 0.138; G17DT/placebo). A conventional analysis without adjustment for disease stage or interim analysis, censoring for chemotherapy and excluding protocol violators, gave median survival periods of 151 (G17DT) and 82 days (placebo) (log-rank test, P = 0.03).Patients developing anti-G17DT responses (73.8%) survived longer than nonresponders or those on placebo (median survival, 176 vs 63 vs 83; log-rank test, P = 0.003). G17DT was well tolerated.