From April 1985 to September 1987, 92 patients with advanced NSCLC were randomized to receive cytotoxic chemotherapy, Arm A (treated), or supportive care, Arm B (control). Chemotherapy consisted of the CEP combination (cyclophosphamide 500 mg/m2 i.v. day 1; epirubicin 50 mg/m2 i.v. day 1; cisplatin 80 mg/m2 i.v. day 1) alternated every 4 weeks with the MEC' combination (methotrexate 30 mg/m2 i.v. day 1; etoposide 200 mg/m2 i.v. day 1; CCNU 70 mg/m2 per os, day 1) until progression. Eight-nine patients (44 treated and 45 controls) were eligible for survival and 77 evaluable for response (38 treated and 39 controls). Response rate was: in Arm A, 8/38 (21%) partial response, 20/38 (53%) stable disease and 10/38 (26%) progressive disease; in Arm B, 18/39 (46%) stable disease and 21/39 (54%) progressive disease. Median time to progression was 4 months (range = 1-14) for treated and 2 months (range = 1-9) for controls (P = 0.001). Median survival was 8.5 months (range = 1+ to 25) for Arm A versus 5 months (range = 1+ to 28+) for Arm B; this difference was not statistically significant (Breslow test: chi-square = 2.75, P = 0.097; Mantel-Cox: chi-square = 0.32, P = 0.56). Treatment related toxicity was gastrointestinal WHO grade 3 in 22/102 (22%) CEP courses and in 10/91 (11%) MEC' courses respectively. Other observed side-effects were not clinically important. From these data our treatment was not clearly superior to supportive care in prolonging survival. This suggests the need for the inclusion of a control group in future chemotherapeutic trials of NSCLC.