OBJECTIVES: To ascertain the existence of several populations regarding overall survival (OS) in patients with advanced non-small-cell lung cancer (NSCLC). METHODS: Data of OS from the Cuban National Cancer Registry (CNCR) and from Cuban multicentre trials of immunotherapy were analysed with a lognormal mixed model assuming 1 to 6 underlying populations. The Bayesian Information Criterion (BIC) was used to select the best model fitted to the data in all cases.  RESULTS: The CNCR provided data for 31133 patients diagnosed with lung cancer since January 1998 until December 2008. Of those, 7286 patients presented stages IIIb-IV of NSCLC at diagnosis and were selected for analysis. The immunotherapy Cuban trials provided data for more than 750 patients enrolled in 8 trials conducted since 1997 until 2010. The mixed model applied to CNCR data separated 4 populations: very high risk (OS mean time = 0.62 months, 23% of the sample); high risk (OS mean time = 3.1 months, 34%); medium risk (OS mean time = 9.2 months, 35%); and low risk (OS mean time = 29.1 months, 8%). Results for clinical trials separated 2 populations for controls and 3 populations for the immunotherapy groups. For controls a population of medium risk (OS mean time = 11.6 months, 61%) and other of low risk were obtained (OS mean time = 31.7 months, 38%). For NSCLC patients with immunotherapy a population of medium risk (OS mean time = 11.2 months, 55%); a population of low risk (OS mean time = 23.8 months, 33%); and another of very low risk or long-term survival were obtained (OS mean time = 55.5 months, 12%).  CONCLUSIONS: Our analyses support the existence of several populations regarding OS among advanced stage lung cancer.