Standard adult cancer patients populations are derived in this paper as a tool for the calculation of age-standardised cancer survival figures. Previously used standards in survival analysis have been site- and/or study-specific. Here, multivariate methods have been used to define the smallest possible number of general standard cancer patient populations which are simple to use and provide standardised survival values close to the raw ones for the largest possible number of cancer sites. The analysis was based on data for over 1.1 million cancer patients included in the EUROCARE-2 study. The proposed standard populations consist of three age distributions, appropriate for cancers with incidence patterns: (1) increasing with age - the vast majority of cancers; (2) broadly constant with age and (3) mainly affecting young adults. The three standard distributions are presented by both broad and five-year age classes. The latter can be used to determine which of the three standards would be used for sites not included in the cluster analysis because their survival is generally calculated in unusual age groups. Overall, standard I is appropriate for over 91% of cases, standard 2 for just over 7%, and standard 3 for less than 2%. The proposed standards were tested on European (EUROCARE-2 and EUROCARE-3) and US (Surveillance, Epidemiology and End Results Program, SEER) relative survival data. There was very good correspondence between the raw (population weighted) and age-standardised survival figures. (C) 2004 Elsevier Ltd. All rights reserved.
