Introduction Cancer information obtained by suitable method becomes a basis of planning the effective cancer control program. Age-Period-Cohort (APC) model is widely used in the analysis of longitudinal cancer data or prediction. This is constructed by regression model whose explanatory variables are age, period and cohort. However, this model includes the problem that these three elements are not estimated uniquely, because there is a linear dependent relationship among these elements, cf., “cohort = period—age”. So there are many kinds of improvements in order to divide these three effects.
Methods To avoid difficulties, we propose to use an interaction model instead of APC model, and visualise cancer risk on the age-period space. We compare three models, that is, traditional APC model, geographically weighted regression (GW) model and interaction model. Those models can be described on Poisson regression model. To help understanding those mathematical models, we also show some scripts, which are executable in statistical software, R.
Results and conclusion Applying those regression models to the data of lung cancer in Denmark which is available as lungDK of Epi package in R and liver cancer in Japan, which is said to have strong cohort effect, cohort effects were visualised in sex effect of Denmark data and in Japanese male data.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.