Exercise testing with measurement of maximal oxygen uptake (VO2max) is increasingly used in the assessment of lung resection candidates, but its predictive value for postoperative complications remains controversial. We therefore sought to determine the prognostic value of VO2max compared with other pulmonary function tests. A consecutive group of 80 patients (mean age 61 yr; 57 males and 23 females) scheduled for lung resection (62 malignancies, 12 benign disorders, and 6 carcinoids) underwent pulmonary function tests and symptom-limited cycle ergometry. All patients underwent lung resections: 21 pneumonectomies, 45 lobectomies, and 14 segmental or wedge resections. Group A (64 patients, 80%) had an uneventful postoperative course, whereas Group B (16 patients, 20%) had complications; 3 of them died (4% overall mortality rate). In a stepwise logistic regression analysis used to determine independent risk factors for postoperative complications (within 30 d), VO2max expressed as a percentage of predicted (84 +/- 19 for Group A versus 61 +/- 11 for Group B) proved to be the best predictor (predictive value 85.5%). Although VO2max expressed in absolute values (ml/kg/min) was also highly predictive (79.5%), a ROC curve analysis proved the percentage predicted values to be significantly more sensitive. Of 9 patients with a VO2max < 60% of predicted, 8 had complications, including all 3 patients who died after resections of more than one lobe (sensitivity 50%, specificity 98%). The estimated probability (probit model SAS software package) of suffering no complication was 0.9 for VO2max > 75% of predicted and 0.1 for a VO2max < 43%.(ABSTRACT TRUNCATED AT 250 WORDS)