Evidence against the proposition that "UK cancer survival statistics are misleading": simulation study with National Cancer Registry data.

摘要: Objectives To simulate each of two hypothesised errors in the National Cancer Registry (recording date recurrence cancer, instead diagnosis, for registrations initiated from a death certificate; long term survivors who are never notified to registry), estimate their possible effect on relative survival, and establish whether lower survival UK might be due one or both these errors. Design Simulation study. Setting England Wales. Population Patients diagnosed as having breast (women), lung, colorectal cancer during 1995-2007 Wales, with follow-up 31 December 2007. Main outcome measure Mean absolute percentage change year five associated simulated error. Results explain differences after between Sweden, under first hypothesis, diagnosis would have been incorrectly recorded by an average more than 70% women known dead. Alternatively, second failure register even 40% less half difference survival. were similar lung cancers. Conclusions Even implausibly extreme levels registry data could not international observed other European countries.