Temporal-spatial Analysis of Breast Cancer Risk on Upper Cape Cod, Massachusetts, USA

Abstract

P-665 Introduction: The reason for elevated breast cancer rates on upper Cape Cod, Massachusetts remains unknown despite epidemiological studies to investigate possible environmental risk factors. Two existing population-based breast cancer case-control studies of upper Cape Cod (1983–1993) provide a geocoded residential history and information on confounders for all subjects. Combining the two studies creates an invaluable dataset for temporal-spatial analysis of subjects’ residences from 1943 to 1993. Methods: Geographic Information Systems (GIS) have allowed the use of spatial analytic techniques in public health studies previously not feasible. Generalized Additive Models (GAMs) are an effective approach to mapping spatial distributions of data that combines: smoothing of geographical location and residency duration, inclusion of risk factors, selection of optimum degree of smoothing (bandwidth), hypothesis testing, ability to handle various types of outcome data, and use of standard software. The combination of mapping and statistical modeling provides a powerful tool for visualizing disease risk. Case-control data were modeled with GAMs and GIS to create breast cancer risk maps that are smoothed over time and space using a bivariate smooth term for location and a univariate smooth term time for residency duration. We also examined years of residency by applying the GAM model to the dataset restricted to residences for given time periods. A series of maps were used to create movies of changing breast cancer risk over time. The movies allow us to visualize changes in magnitude, geographic size, and location of elevated risk for the fifty years of residential history. Results: Maps show a large area of elevated breast cancer risk near the Massachusetts Military Reservation (MMR) beginning in the late 1940s and remaining consistent in size and location until early 1960s. Throughout much of the 1970s and 1980s, the risk map is relatively flat. In the late 1980s, an area of elevated risk is present in the northeast region of the study area. In the maps of residency duration, an area of elevated breast cancer risk north of the MMR appears after ten years of duration and increases in magnitude and size with longer duration. Maps of promoter periods showed very little spatial variation. Discussion and Conclusions: Temporal-spatial analysis of the breast cancer data may help identify new exposure hypotheses that warrant a future epidemiologic investigation. By determining the location, residency years, and duration associated with elevated breast cancer risk, we may gain a better understanding of any environmental factors involved.