RESULTS: The training cohort included 431 patients with a mean age of 67.7 years who underwent wedge resection (n=49), segmentectomy (n=118), lobectomy (n=264), and other procedures (n=6) for clinical stage IA (n=325) and stage IB (n=112) NSCLC. Postoperative major complications (Grade IVa or more; life-threatening complication requiring intensive care unit management) occurred in 8 patients (2%), and the postoperative 30-day mortality rate was 0.2% (n=1; acute exacerbation of idiopathic pulmonary fibrosis). The optimal cutoff value for prediction of postoperative complications determined using ROC curve analysis was 45.7% of DLCO (area under the curve [AUC], 0.954; 95% confidence interval [CI], 0.931-0.976; P< 0.001), which indicated 92.3% sensitivity and 100% specificity. The OS and disease-free survival (DFS) decreased to 68.3% and 63.7% at 5 years in patients with 50% or less of DLCO, compared with 92.3% and 85.2% in patients with more than 50% of DLCO (P<0.001, P<0.001, log-lank, respectively). Multivariate survival analysis using Cox's regression model revealed that less than 50% of DLCO was an independent predictor for OS (hazard ratio 4.21, P=0.001) and DFS (hazard ratio 3.44, P=0.001). In a validation cohort consisting of 784 patients, similar results were obtained showing that less than 50% of DLCO was an independent predictor for postoperative complications and OS (odds ratio 4.28, 95% CI, 1.40-12.96, P=0.010; hazard ratio 2.32, 95% CI, 1.37-3.91, P=0.002, respectively).