Patients with inoperable non-small-cell lung cancer (NSCLC) were randomly assigned to receive one of three dosage regimens: (1) vindesine and cisplatin (VP); (2) mitomycin, vindesine, and cisplatin (MVP); or (3) etoposide and cisplatin alternating with vindesine and mitomycin (EP/VM). In 199 assessable patients, the response rates were VP, 33%; MVP, 43%; and EP/VM, 19%. The addition of mitomycin to the VP regimen did not significantly improve the response rate. The response rate was significantly lower with the EP/VM regimen than with the MVP regimen (P less than .01). The median survival times were VP, 50 weeks; MVP, 42 weeks; and EP/VM, 40 weeks. These differences were not significant. Grade III or IV thrombocytopenia was significantly greater (P less than .01) in MVP patients (22%) than in the VP (5%). Other toxicities were similar in the three groups. Analyses of prognostic factors showed that treatment with MVP, sex, and histologic classification (squamous cell carcinoma) were predictive of improved response. Important factors for improved survival, according to the Cox regression analysis, were the stage of disease, performance status, sex, weight loss before diagnosis, and hemoglobin concentration.