From April 1985 to September 1988, 128 patients with advanced non-small-cell lung cancer (NSCLC) were enrolled in a prospective randomized trial evaluating chemotherapy (arm A) versus best supportive care (arm B). Chemotherapy consisted of cyclophosphamide 500 mg/m2 intravenously (IV) day 1, epirubicin 50 mg/m2 IV day 1, and cisplatin 80 mg/m2 IV day 1 (CE'P regimen) alternating every 4 weeks with methotrexate 30 mg/m2 IV day 1, etoposide 200 mg/m2 IV day 1, and lomustine (CCNU) 70 mg/m2 orally day 1 (MEC' regimen) until progression. Of the 123 patients (62 treated and 61 controls) eligible for survival, 115 were fully evaluable for response (58 treated and 57 controls). Response rates were 21% partial response, 53% stable disease, and 26% progressive disease in arm A, and 47% stable disease and 53% progressive disease in arm B. Median survival was 34.3 weeks (range, 4.3 to 218.6+ weeks) in arm A versus 21.1 weeks (range, 4.3 to 188.6 weeks) in arm B; the difference was not significant at P = .153 (Mantel-Cox). Subgroups of patients retrospectively analyzed by age, performance status, stage M0/M1, and weight loss or not showed no significant difference in survival. Poor-risk patients (at least two of the following: poor performance status, stage M1, weight loss) of arm A survived significantly longer than poor-risk patients of arm B (23.6 weeks v 12.4 weeks, Mantel-Cox P = .008); a significant difference in survival was also observed between nonsquamous cell patients of arm A and those of arm B (median survival, 38.6 weeks v 16.7 weeks; Mantel-Cox P = .041). Toxicity on the chemotherapy arm was hematologic (World Health Organization [WHO] grade greater than 3) in 12% of CE'P and in 13% of MEC' courses and gastroenteric (WHO grade greater than 3) in 24% of CE'P courses and in 8% of MEC' courses. Our alternating treatment was not significantly superior to supportive care. It is likely that certain subgroups of the NSCLC category may have an advantage with chemotherapy.