In a previous phase II randomized study, a cisplatin/gemcitabine/vinorelbine (PGV) regimen produced a 50-week median survival time (MST) in advanced non small-cell lung cancer (NSCLC) patients. The present trial was planned to randomly compare the outcome of patients treated with this new triplet regimen with those of patients receiving either cisplatin plus vinorelbine (PV) or cisplatin plus gemcitabine (PG) doublet combinations. One hundred eighty patients with stage IIIB (76) or IV (104) disease, aged <or= 70 years, and with Eastern Cooperative Oncology Group performance status (ECOG PS) <or= 1, were randomly allocated to receive: cisplatin, 50 mg/m2 plus gemcitabine, 1000 mg/m2 plus vinorelbine, 25 mg/m2 (PGV) on days 1 and 8 every 3 weeks; cisplatin, 100 mg/m2 on day 1 plus gemcitabine, 1000 mg/m2 (PG) on days 1, 8, and 15 every 4 weeks; cisplatin, 120 mg/m2 on days 1 and 29 plus vinorelbine, 30 mg/m2/week (PV). At the planned interim analysis, the MST of patients in the PGV, PG, and PV arms was 51, 42, and 35 weeks, respectively. The hazard of death (Cox analysis) for patients receiving PGV compared with those receiving PV was 0.35 (95% confidence index [CI], 0.16-0.77, P = 0.0058). The response rate was 47% in the PGV arm, 30% in the PG arm, and 25% in the PV arm. Severe neutropenia (75% vs. 45%), and vomiting (50% vs. 15%) significantly affected more patients in the PV than in the PGV arm. Since the difference in survival met early stopping rules, accrual to the PV arm was suspended. Enrollment still continues in the PGV and PG arms to ascertain whether the triplet regimen has a more significant effect on survival than that produced with the PG regimen.