Each year the National Cancer Institute, the American Cancer Society, the Centers for Disease Control, and the North American Association of Central Cancer Registries produce an Annual Report to the Nation on the Status of Cancer in the United States. These reports contain estimates of overall annual cancer incidence and mortality, as well as incidence/mortality within population subgroups defined by sex, race, ethnicity, and other characteristics. The findings in these Annual Reports influence cancer research priorities, cancer prevention programs, and public health expenditures. In order to accurately determine the existence and magnitude of year-to-year changes in cancer rates, the estimates of these rates must have two properties: 1) the estimates for the subgroup of interest (for instance, black females), must not be influenced by between-year differences that exist in other risk factors (for instance, age and economic status);2) the differences in the between-year estimates for the subgroup of interest must be unbiased for the difference in cancer rates that would have been seen if all the other risk factors were identical in each year. The Annual Reports produces estimates of cancer rates using a procedure called direct age-standardization. We have proven that the estimates produced by direct age-standardization have neither of the required properties [1]. We have developed a new method of standardization that has both of these properties [1]. We have also shown that the conclusions currently made from the Annual Reports have a one-to-one correspondence with the factors that give rise to differences in standardized rates produced by this new method [1]. In the analyses of SEER data that we have so far examined, the standardized rates from the direct age-standardization method and our new method are quite different. The biases in the estimates of cancer rates using the current method are greatest for minority groups. Our work to date allows us to use the new method of standardization to produce estimates of cancer rates with the desired properties. However, we have not yet developed the statistical methods and software necessary to make formal inferences from analyses of SEER data. Ultimately our goal is to produce freely available, user-friendly software that will allow any researcher to utilize the new standardization procedures to analyze SEER data. In this application we request funding for the initial stage of software development.
Each year the National Cancer Institute, the American Cancer Society, the Centers for Disease Control, and the North American Association of Central Cancer Registries produce an Annual Report to the Nation on the Status of Cancer in the United States. Based on the year-to-year changes in the incidence of specific cancers, these findings influence cancer research priorities, cancer prevention programs, and public health expenditures. The current method of estimating these rates produces biased estimates of cancer rates, with the magnitude of the bias largest in minority groups. We have developed a method for producing unbiased estimates of these cancer rates.
