Many cancers have established or suspected risk factors that are distributed unevenly in the environment. Exploring spatial patterns of cancer incidence and mortality can identify areas of significantly elevated risk and provide potential etiologic clues. Increasingly, residential addresses at time of diagnosis or study enrollment and even complete residential histories are collected in epidemiologic studies of cancer that enable a finer level of spatial analysis than in ecological studies. In spatial epidemiology studies of cancer risk the residential location is often used as a surrogate for the unknown environmental exposures that occur in and around the home. However, several methodological challenges arise in spatial analysis of cancer in epidemiologic studies when considering the temporal dimension of risk. Cumulative environmental exposures are not assessed and disease latency and population mobility are ignored when using the residential location at time of diagnosis as a marker for environmental exposures. There is a need for development and assessment of statistical methods to model cumulative spatial-temporal cancer risk in epidemiologic studies with residential histories while accounting for population mobility, disease latency, and known disease risk factors.
The specific aims of this research are 1) to develop cumulative spatial-temporal models of cancer risk and apply them to a case-control study of non-Hodgkin lymphoma (NHL) with residential histories, 2) to evaluate the accuracy of the statistical models for detecting areas of significant risk over time, and 3) to investigate possible exposures that may be associated with any detected areas of significantly elevated risk. NHL is suitable for a spatial pattern analysis because it is a cancer with an unclear etiology and established risk factors that account for only a small proportion of the total annual NHL cases in the United States. The expected outcomes of this research will be 1) new statistical approaches to model spatial-temporal risk that consider life-course environmental exposures, 2) a thorough assessment of the accuracy of the models, and 3) identification of areas of significant risk of NHL in space and time in four areas (Detroit, Iowa, Los Angeles, Seattle) of the United States. The significance of this research is two-fold. First, the development and evaluation of new approaches to spatial-temporal risk analysis that better consider life-course environmental exposures will advance the field of spatial analysis research. Second, this will be the first cumulative spatial-temporal risk analysis of a case-control study of NHL, which also adjusts for known environmental, genetic, and demographic risk factors. The work in this research application will serve as the foundation for a larger grant application to the National Cancer Institute to model spatial-temporal uncertainty in cancer risk. The assessment of the performance of new and existing methods for modeling spatial risk of cancer over time will be used to guide the selection and refinement of methods for inclusion in the later application and to demonstrate the benefits of the proposed approach to other cancers.
Many cancers have established or suspected risk factors that are distributed unevenly in the environment. When risk factors are unknown, studying the patterns in cancer risk over space and time may reveal important clues about what causes disease. This project seeks to develop and evaluate statistical methods that assess cumulative cancer risk over space and time and apply them to a public health study of non-Hodgkin lymphoma.
|Siangphoe, Umaporn; Wheeler, David C (2015) Evaluation of the performance of smoothing functions in generalized additive models for spatial variation in disease. Cancer Inform 14:107-16|