The combination of vindesine and cisplatin is considered a reference regimen in advanced NSCLC which has yielded a significant improvement in the duration of survival. A phase II study of a new semi-synthetic vinca alkaloid, Navelbine, reported an unusually high 29% response rate in stage III-IV NSCLC and a phase I-II study established the feasibility of the combination of Navelbine and cisplatin. We, therefore, designed a prospective randomized trial to compare Navelbine and cisplatin (NVB-P) to vindesine and cisplatin (VDS-P) and to evaluate whether the best of these regimens affords a survival benefit compared to Navelbine alone (NVB), an outpatient regimen. Forty-five centers included 612 patients in this study: 206 in NVB-P, 200 in VDS-P and 206 in NVB. Navelbine was given at a dose of 30 mg/m2 weekly, cisplatin at 120 mg/m2 on day 1, day 29 and then every 6 weeks and vindesine at 3 mg/m2 weekly for 6 weeks and then every other week. Treatment was continued until progression or toxicity. Patients' characteristics were similar in the three groups with 59% of patients presenting with metastatic disease. An objective response rate was observed in 30% of patients in NVB-P versus 19% in VDS-P (P = .02) and 14% in NVB (P < .001). The median duration of survival was 40 weeks in NVB-P compared to 32 weeks in VDS-P and 31 weeks in NVB. The comparison of survival between the three groups demonstrated an advantage for NVB-P compared to VDS-P (P = .04) and NVB (P = .02). Neutropenia was significantly higher in the NVB-P group (P < .001) and neurotoxicity more frequent with VDS-P (P < .004). Since our results have demonstrated that NVB-P yields a longer survival duration and a higher response rate than VDS-P or NVB alone, with acceptable toxicity, this combination should be considered a reference regimen in advanced NSCLC.