Seventy-three patients with regional, inoperable non-small-cell lung cancer received treatment with initial chemotherapy for two cycles (vinblastine-mitomycin followed in 3 weeks by vinblastine-cisplatin), with planned subsequent neutron irradiation to the primary site and concurrent, elective whole-brain irradiation using photons, followed by two more cycles of identical chemotherapy. Histology was reported as adenocarcinoma or large cell in 75%, and 60% had Radiation Therapy Oncology Group (RTOG) stage 3 disease; the remainder had stage 4. The response rate to chemotherapy induction was 51%. There were 58 patients in a second phase of the study who were potentially eligible for treatment with a medically dedicated cyclotron having more favorable characteristics with regard to treatment planning and dose delivery (neutrons "B"). The overall response rate in this group was 79%. Chemotherapy toxicity included four fatalities (5%), with three related to mitomycin C induced bilateral pneumonitis, and an additional five patients (7%) with life-threatening events that required hospitalization. Two fatalities were attributed to combined effects of chemotherapy and radiation, and six more to chest radiation therapy, for an overall treatment-related death incidence of 12 of 73 (16%). Four of the six deaths related to chest irradiation occurred after treatment with a "physics-based" neutron generator (neutrons "A"). Among the 45 who received neutrons in the B group, two (4%) had radiation-related deaths, and another four (10%) had clinically evident radiation pneumonitis. Pretreatment performance status (PS) and response to chemotherapy, but not RTOG stage or weight loss, were significantly associated with survival. Among patients who actually received chest irradiation, only initial response to chemotherapy remained as a significant predictor of survival in univariate analysis, with a median survival of 20 months in responders v 9 months in chemotherapy nonresponders. The patterns of first relapse observed in B group patients revealed that 28% were distant, while 64% were locoregional. This represents a reversal of the usual pattern in studies of chest irradiation alone. It probably reflects elimination of brain relapse by the use of elective whole-brain irradiation, impact of systemic chemotherapy on micrometastases elsewhere, and conservative treatment volumes employed for the chest irradiation in an attempt to minimize its toxicity. Further exploration of combined modality therapy is indicated for regional non-small-cell disease, with a real potential for survival impact if the therapeutic index can be improved.