The survival benefit of combination chemotherapy to patients with advanced non-small-cell carcinoma of the lung (NSCLC) is controversial. To study this question, the National Cancer Institute of Canada (NCIC) Clinical Trials Group conducted a prospective randomized trial comparing best supportive care (BSC) to two chemotherapy regimens, vindesine and cisplatin (VP), and cyclophosphamide, doxorubicin, and cisplatin (CAP). Between February 1983 and January 1986, 23 centers across Canada entered 251 patients on study. Eighteen centers participated in the three-arm schema (150 patients); centers choosing not to participate in a study with a no-chemotherapy arm followed a two-arm schema comparing VP with CAP (101 additional patients). Altogether, 233 patients were eligible. Patients had measurable or evaluable disease, with either distant metastases (82.5%) or bulky limited disease considered inoperable or unsuitable for radical radiotherapy. The treatment groups were comparable in terms of age, sex, performance status, histology, disease extent, and weight loss. The overall response rates (complete response [CR] plus partial response [PR]) on the chemotherapy arms were CAP, 15.3%, and VP, 25.3% (P = .06). Patients on the three-arm portion of the trial had a median survival of 32.6 weeks when treated with VP, 24.7 weeks with CAP, and 17 weeks with BSC. The significance of the differences in survival, adjusted for prognostic factors, is as follows: chemotherapy v BSC, P = .02; VP v BSC, P = .01; and CAP v BSC, P = .05. Toxicity on the chemotherapy arms was significant, with leukopenia of severe or greater degree occurring in 37.8% (CAP) and 40.0% (VP), severe vomiting in 12.2% (CAP) and 23.3% (VP), and severe neurotoxicity in 15.6% (VP).