Cancer registries provide valuable information on trends for cancer incidence in a specified population and the Connecticut Tumor Registry is the oldest registry of its kind. An appropriate understanding of these trends often provides valuable clues for the etiology of cancer, demonstrates the success of the health care system in maintaining the health of the population, and assists in planning for future health care requirements of a population. However, valid interpretations of these trends require a good understanding of the interplay of different perspectives of time, including the effects of age, time period of diagnosis, and birth cohort. These effects are best appreciated through the active collaboration of researchers with a good understanding of disease etiology, as well as the statistical complications that arise from these analyses. The investigator's previous work in this area has concentrated on the development and application of methods for using age-period-cohort models to understand cancer incidence trends. These methods have been used in an extensive analysis of all the major cancer sites, and the investigators have recently used similar approaches to better understand trends by histologic type, stage, and tumor subsite. In addition, they have developed models for lung cancer incidence that incorporate external information about the carcinogenesis process, as well as population-based information on cigarette smoking. These methods have helped many epidemiologists to better understand incidence and mortality trends, although they also can point out fundamental limitations to our understanding that arise from the unidentifiability of some effects. A rich array of tools is now available for descriptive epidemiology and these need to be made more accessible while, at the same time, there are promising new areas that require further study before they can be recommended for general use. This is a proposal to build on work that has developed and applied a variety of statistical methods to the study of the descriptive epidemiology of cancer with the goal of obtaining a better understanding of the reasons for existing trends in cancer incidence and mortality.
The specific aims of this work are: 1. Study the importance of histology, stage of disease, and tumor subsite on time trends for the following malignancies: cervix, prostate, brain, ovary, thyroid, endometrium, bladder, and testis. 2. Develop a comprehensive model for breast cancer incidence and mortality in order to better understand the role of trends in detecting earlier stage tumors on the discrepancy between incidence and mortality trends; and, 3. Develop a Bayes model for incorporating estimates of trends in known risk factors for disease on incidence in order to quantify the extent to which existing knowledge about etiology can explain observed trends in cancer incidence using information on the effect of cigarette smoking on lung cancer incidence as an example to illustrate the method.
