OBJECTIVES: Traditionally, patients who survive 5 years from diagnosis of cancer are considered to be cured. More recent analysis of adjusted mortality rates suggests this assumption may not be valid. We sought to identify data on relative survival at 10 to 20 years for common cancers to determine which showed a continuing increase in mortality beyond 5 years. METHODS: National or regional cancer databases reporting net or relative survival beyond 5 years from diagnosis were identified, with 5- and 10-year data available from England and Wales, Scotland, the USA, Switzerland and Slovakia; 5-, 10- and 15-year data from Norway;  and 5-, 10-, 15- and 20-year data from Germany and Sweden. The percentage decrease in net survival at each 5-year step was calculated at 10, 15 and 20 years to identify where there was a persistent increase in all-cause mortality above expected for the age- and sex-adjusted cancer-free population. RESULTS: Regional setting had a large impact on relative survival for each type of cancer. However, oropharyngeal, head and neck, liver, lung, pancreatic and ovarian cancers, chronic lymphocytic leukaemia, mesothelioma, multiple myeloma, and Kaposi’s sarcoma consistently showed between a 10% and 70% further decrease in relative survival at 10 compared with 5 years. A further 10% or greater decrease in relative survival was seen between 10 and 20 years from diagnosis for head and neck, lung, laryngeal and prostate cancer, and was around 40% lower than population norms at 20 years for people with multiple myeloma.  CONCLUSIONS: When modelling the impact of treatments for cancer, it is important to consider the whole period of increased risk of mortality. For certain cancers, in particular multiple myeloma, lung, prostate and laryngeal cancer, this may require a time horizon of 20 years or longer.