OBJECTIVES: The aim of the study was to identify and characterize long term survival population of advanced non–small-cell lung cancer patients treated with immunotherapy. METHODS: Data from 717 patients coming from two expanded used program and from two randomized trials evaluating the efficacy of CIMAvaxEGF and Vaxira in patients with advanced NSCLC, were used. Mixture models were fitted to Overall Survival with one or two population components. All analyzes were made using the NLMIXED procedure in SAS. We used the diagnostic tools provided by this procedure to check the models’ good of fit properties. The characterization of the two populations based in prognostic factors was done by classification tree models using RPART package in R. RESULTS: Two months of overall survival (OS) benefit were showed for CIMAvaxEGF and for Vaxira. The optimal mixture model with the fewest number of parameters that adequately describes the time survival data is a mixture model with 2 component distributions. Components represent short-term and long-term survival subpopulations. The proportions of the long term population increase with immunotherapy in 22% of patients for Vaxira and 18% for CIMAvaxEGF. The OS benefit was different for both subpopulation for both vaccines (vaxira: 2.08 months and 8.7 of OS benefit for short- and long- term survival populations respectively; CIMAvaxEGF: 1.96 months and 14.36 months of OS benefit for short- and long- term survival populations respectively). The performance status and the age were essential in the classification of the two populations. CONCLUSIONS: The results confirm that there are two subgroups among NLCLC patients. The separate analysis of subgroups can give more power to the evaluation of clinical trials. The use of mixture models in the analysis has implications for the design of new clinical trials. The use of classification trees allowed a good characterization of the two populations.