The impact of screening on cancer incidence and mortality rates is one of important issues in cancer prevention and control, and many researchers conducted studies on the link of screening, for example, prostate-specific antigen (PSA) screening for prostate cancer and mammography for breast cancer, to cancer incidence, mortality and survival rates. The proposed research seeks to develop a method to investigate how cancer mortality responds to incidence history, which would help one to assess the impact of screening. We propose an autoregressive joinpoint model for mortality trends, which is expressed as an autoregressive model in terms of the preceding incidence and mortality rates and a joinpoint regression model in terms of the time covariate. First, we plan to develop a procedure to select the model, including the number of joinpoints and the numbers of significant incidence and mortality terms, and to derive asymptotic confidence intervals for the parameters. In order to select the model, we will use a modified Bayesian Information Criterion as well as a multi-stage permutation procedure. We will first select the number of joinpoints with all possible preceding incidence terms included, and then use traditional regression model selection procedures to determine significant preceding incidence and mortality terms, and finally run the permutation test to finalize the model. To construct confidence intervals for the model parameters, asymptotic results available in literature will be extended and applied. Second, we propose a recursive residual test for a change in the causal relationship between incidence and mortality trends. Since one of main causes for a change in such a relationship is a change in the incidence pattern, the proposed test is expected to provide information on if and how changes in the incidence pattern in0uence mortality trends. Finally, we will apply the developed method to cancer incidence and mortality data avail- able at Surveillance, Epidemiology and End Results (SEER) data base at the National Cancer Institute (NCI). We plan to conduct autoregressive joinpoint analysis for prostate, breast and brain cancer data. Prostate and breast cancer data for which screening is known to have made a significant contribution to the increase in incidence would help one to quantify the nature of screening effects and the analysis of brain cancer data is expected to provide insights on rather generic changes. By applying the proposed method to mortality rates categorized by disease stage and other baseline factors, we plan to investigate how to handle biases such as lead time, length bias, and over-diagnosis.
Public Health Relevant Statement The purpose of the proposed research is to investigate the impact of cancer screening on cancer incidence and mortality rates using population based data available at the SEER data base. The proposed research would help one to quantify effects of screening tests in reducing mortality, to understand the nature of screening effects according to baseline factors, and thus to develop more e1cient cancer control programs.
