Survival and risk model for stage IB non-small cell lung cancer

The aim of this work is to estimate the prognostic value of a set of clinical-pathological factors in patients resected for non-small cell lung cancer (NSCLC) and classified as stage IB, in order to create a prognostic model for establishing risk groups, and to validate that model.

Among 637 patients resected and classified as stage IB, we analyzed sex, age, symptoms, location, type of resection, cell type, histology, and tumor size. The Kaplan-Meier method was used to estimate the survival. The results were compared using the log-rank test. All the significant variables from this univariable method were then included in a multivariable method of estimation of the proportional risk for survival data developed by Cox, using the variables selected, a regression model was developed for accurately predicting survival. To validate the predictive capability of the regression model, we randomly divided our patients into training and test subsets, containing 322 and 315 cases, respectively.

The overall 5-year survival rate of the series was 60%. The cell type, the squamous or non-squamous and the tumor size showed a significant influence on survival in the univariable analysis, while, according to the Cox model, only the tumor size and the squamous or non-squamous type entered into regression. Hazard rates were calculated for each patient. The mean risk was 0.87 +/- 0.25 (range 30-1.94). The series was divided into three risk groups (low, intermediate, and high risk) according to the fitted hazard rates, using cut-off points (one standard deviation from the mean). The 5-year survival rates were 85, 59, and 44%, respectively. To validate the model, we repeated the analysis for training and test subsets. Only the tumor size had a significant influence on survival in the univariable analysis. Using the Cox model, also the tumor size entered into regression. The mean risk was 0.79 +/- 0.29 (range 0.09-2.12). Cut-off points were 0.50 and 1.08 for the low, intermediate, and high-risk groups. The 5-year survival rates were 83, 58, and 40%, respectively. We validated the regression model obtained in the training subset by demonstrating its capacity in identifying risk groups in the test subset. The 5-year survival rates were 83, 61, and 49.5% for the low, intermediate, and high-risk groups, respectively (P = 0.0104).

Stage IB does not succeed in configuring a group of patients with a homogeneous prognosis, as there is a wide variability in a 5-year survival. The estimation of prognosis derived from a multivariable analysis can obviate the limitations of the actual staging system for NSCLC.