A multicentre randomised phase III trial in chemotherapy-naive patients with advanced non-small-cell lung cancer (NSCLC) was undertaken to compare the therapeutic activity and toxicity of a cisplatin/carboplatin-etoposide-vinorelbine combination with that of a cisplatin-etoposide regimen. Patients with advanced (stage IIIB-IV) NSCLC were randomised, after stratification for stage (IIIB-IV) and performance status (0-1 and 2), to receive either (A) CDDP 40 mg m-2 + VP16 100 mg m-2 on days 1-3 as standard treatment or (B) CBDCA 250 mg m-2 on day 1 + CDDP 30 mg m-2 on days 2 and 3 + VP16 100 mg m-2 on days 1-3 + NVB 30 mg m-2 on day 1. Therapy was recycled on day 29 in both arms. We hypothesised a 15% minimum increment in the response rate with the experimental regimen over the 25% expected activity rate of the standard regimen. A two-stage design was chosen, which permitted the early termination of the trial (after the accrual of 52 patients in each arm) if the difference in response rates between the two regimens was less than 3% at the end of the first stage. A total of 112 patients (arm A = 57, arm B = 55) were enrolled in the study (53 with stage IIIB and 59 with stage IV), of which 105 eligible patients were evaluable for response on an "intention to treat' basis. Seven patients were excluded because they did not fulfil the inclusion criteria. Fifteen responses were observed in arm A (28%, 95% CI = 17-42) and 13 (one complete) in arm B (25%, 95% CI = 13-37). On multivariate logistic analysis, treatment did not affect the response rate, while stage IV and performance status 2 were significantly associated with a lower probability of response. Median survivals were similar in the two arms (31 vs 27 weeks). The experimental regimen was associated with an extremely poor median survival in patients with poor performance status (21 weeks). On Cox analysis, treatment failed to show a significant impact on survival: stage IV (relative risk = 1.6. CI = 1.0-2.6, P = 0.036) was the only prognostic variable significantly associated with a worse survival outcome and, although poor performance status adversely affected survival, this effect did not reach the level of statistical significance (relative risk = 1.6, CI = 0.98-2.5; P = 0.063). There were no significant differences in non-haematological toxicities between the two arms, although three patients in the control arm had to discontinue the treatment because of the persistence of severe nephrotoxicity (two patients) or neurotoxicity (one patient). In contrast, a significant increase in both neutropenia and thrombocytopenia was observed in the experimental arm. Four treatment-related deaths were registered in arm B (two due to neutropenic sepsis, one to myocardial failure and one to acute renal failure) compared with one toxic death (acute renal failure) in arm A. In view of these results, the trial was stopped and the null hypothesis (< 15% increase in response rate with the experimental regimen) has been accepted. Therefore, our combination does not deserve further evaluation as first-line treatment in advanced NSCLC patients. As our data suggest that an aggressive chemotherapy might have a negative impact on survival of patients with poor performance status, trials to evaluate the activity of new regimens should be conducted separately for each subset of patients with different performance status.