To attempt to improve the poor prognosis of extensive-stage small-cell lung cancer (SCLC) patients, we tried to administer late intensive combined modality therapy (LICMRX) to patients with good tumor regression after 12 weeks of conventional chemotherapy. Twenty-nine consecutive extensive-stage SCLC patients received 6 weeks of cyclophosphamide, methotrexate, and lomustine (CMC) induction therapy, followed by 6 weeks of vincristine, doxorubicin, and procarbazine (VAP). After restaging for assessment of tumor response, autologous bone marrow (ABM) was collected in patients in good medical condition with complete response (CR) or partial response (PR) and no tumor on marrow examination. LICMRX consisted of irradiation with 2,000 rad in five fractions for five days to sites of initial tumor involvement, followed by cyclophosphamide, 60 mg/kg for 2 days, and etoposide, 200 mg/m2 for 3 days and then by ABM infusion. Prophylactic cranial irradiation (PCI) was administered thereafter, but no further chemotherapy was used. Due to lack of tumor regression or poor medical condition, only ten of the original 29 patients were eligible for LICMRX; two refused, so only eight (28%) received therapy. Three patients who began LICMRX in CR developed recurrence of SCLC after an additional 4, 8, and 15 months. Of five patients with PR, one attained CR but relapsed at 3 months, two remained in PR and progressed at 2 and 4 months, and two died of infection without recovery from LICMRX. Mean time from ABM infusion to recovery of granulocyte count to 500/microL was 15.8 days in the six surviving patients (range, 12-22). The major non-hematologic toxicity of LICMRX was severe esophagitis. Among all 29 patients, there were six CRs (21%) and no 2-year survivors, compared with a CR rate of 36% and 10% 2-year survivors in 78 extensive-stage patients previously treated with CMC plus VAP without LICMRX. We conclude that the LICMRX given in this study can be administered to only a minority of extensive-stage SCLC patients and is very unlikely to yield substantial improvement in the fraction of 2-year survivors (95% confidence limits for 2-year survival 0% to 10%).