Analysis of vital statistics data.

Population based cause-specific incidence and mortality data have been routinely collected for surveillance purposes in many parts of the world. Cohort or age-period-cohort analyses of such data can provide an initial screen of etiologic hypotheses and may provide the only data available on long-term disease trends. The atlases of cancer mortality produced by the Environmental Epidemiology Branch of the US National Cancer Institute is a good example of initial screening. This paper focuses upon a number of statistical models for analyzing temporal trends in chronic disease incidence. In particular the authors are interested in distinguishing the background hazard and representing what is biologic and inescapable the effect of being born in a certain period and the effect of having survived until a given time period. The cohort effects model age-period model age-period-cohort model and models which incorporate the effects of a specific risk factor are considered. Smoking and carcinomas of the bladder and pancreas in England and Wales and smoking and lung cancer among males in England and Wales are examined. When appropriate data on important risk factors are available the models described in this paper can be extended to incorporate that data.