From September 1986 until December 1991, 139 patients with histologically-proven small cell lung cancer, age < 75 years, performance status > 40, absence of brain metastases and no previous treatment, were randomised to receive either CEV cyclophosphamide 1000 mg/m2 intravenous (i.v.), epirubicin 70 mg/m2 i.v., vincristine 1.2 mg/m2 i.v., every 3 weeks or PE (cisplatin 20 mg/m2 i.v. and etoposide 75 mg/m2 i.v. for 5 consecutive days, every 3 weeks) for six cycles. After three cycles, responding patients received radiotherapy to the chest (45 Gy/15 sessions) and to the brain (30 Gy/10 sessions--only in patients with limited disease achieving complete remission). 3 patients were ineligible. Patient characteristics included (CEV/PE) total number 66/70, median age 60/61 years, median performance status 80/80, extended disease 33/48 cases (P = 0.04). In evaluable patients, 42/62 (67.7%) responded to CEV while 42/58 (72.4%) responded to PE (P = non-significant); respective complete response rates were 16.1 and 29.3% (P = non-significant) and respective complete response rates in patients with extended disease were 9.4 and 28.9% (P = 0.03). Median survival was 10.5 months, without significant differences in the two treatment arms, even after adjustment for stage. PE was less well tolerated than CEV. Although PE is more active than CEV in certain subsets of patients, its apparent inability to improve survival in this and in other studies questions its routine use in small cell lung cancer.