Combining cohort and period methods for retrospective time trend analyses of long-term cancer patient survival rates.

Assessing trends in long-term cancer patient survival is an essential component of monitoring progress against cancer by cancer registries. Traditional assessment of long-term survival ('cohort analysis') is very useful to disclose trends in long-term survival rates of patients diagnosed many years ago, but it does not allow the disclosure of recent trends in long-term survival rates. The latter can be achieved by an alternative method of survival analysis ('period analysis'), which has been proposed a few years ago. On the other hand, unlike cohort analysis, period analysis does not provide estimates of long-term survival rates for patients diagnosed in the early years after initiation of cancer registration. In this paper, a method of retrospective analysis of time trends in long-term survival rates is introduced, which combines the advantages of both cohort and period analysis ('mixed analysis'). This method thereby allows for a comprehensive monitoring of trends in long-term survival over an extended time span from the earliest to the most recent years of cancer registration. The use of the method is illustrated for retrospective time trend analyses of long-term survival of cancer patients in the United States with the 1973-1999 database of the Surveillance, Epidemiology, and End Results Program of the National Cancer Institute.