The medical criteria for inoperability have been difficult to define in patients with lung cancer. Sixty-six patients with non-small cell lung cancer and radiographically resectable lesions were evaluated prospectively in a clinical trial. The patients were considered by cardiac or pulmonary criteria to be high risk for pulmonary resection. If exercise testing revealed a peak oxygen uptake of 15 mL.kg-1.min-1 or greater, the patient was offered surgical treatment. Of the 20 procedures performed, nine were lobectomies, two were bilobectomies, and nine were wedge or segmental resections. All patients were extubated within 24 hours and discharged within 22 days after operation (median time to discharge, 8 days). There were no deaths, and complications occurred in 8 (40%) of the 20 patients. Five patients whose peak oxygen uptake was lower than 15 mL.kg-1.min-1 also underwent surgical intervention; there was one death. Thirty-four patients whose peak oxygen uptake was less than 15 mL.kg-1.min-1 and 7 who declined operation underwent radiation therapy alone (35 patients) or radiation therapy and chemotherapy (6 patients). There were no treatment-related deaths, and the morbidity rate was 12% (5/41). The median duration of survival was 48 +/- 4.3 months for the patients treated surgically and 17 +/- 2.7 months for those treated medically (p = 0.0014). We conclude that a subgroup of patients who would be considered to have inoperable disease by traditional medical criteria can be selected for operation on the basis of oxygen consumption exercise testing. There is a striking survival benefit to an aggressive surgical approach in these patients.