Studies of time trends in cancer incidence and mortality have been useful for understanding disease etiology, as well as for health care planning. This project has made use of the age-period-cohort model as a method of understanding time trends for cancer. It has developed a number of ways of incorporating external information into the model in order to deal with the nonidentifiability problem, including the use of theoretical models for carcinogenesis and the inclusion of trends in known cancer risk factors. Existing methods of analysis for population based registry data will be extended by developing a comprehensive two-compartment model that will simultaneously describe the effects of cancer incidence, survival from cancer, and cancer mortality. The model development will first give separate consideration to incidence and survival, making adjustment for relevant demographic covariates. It will also analyze trends in tumor characteristics, such as stage and histology, along with the effect of those characteristics on survival. These separate analyses will provide information required for the specification of a comprehensive model that describes the entire process from cancer diagnosis to cancer death. Once a good fitting model has been found, this project will develop practical ways for using information contained in the model, including a back-calculation method for estimating cancer incidence and prevalence using mortality data. This methodology could prove to be valuable for purposes of health care planning in parts of the U.S. that are not covered by population based cancer registries. The methodology will be applied to the following cancer sites: lung, female breast, colon/rectum, and Hodgkin's disease. It will also be used to study ethnic differences in cancer incidence, survival and mortality. The methods will first be developed using data from Connecticut, and then tested further using data from the other SEER Registries.
