Spatial epidemiology of cancer: data types, spatial aggregation and geographical patterns at mainland Portugal

The importance of research on the spatial epidemiology of cancer has been progressively recognized and the number of studies in this field has been increasing steadily, having experienced significant growth over the last decade. Overall, spatial epidemiology is a comprehensive research area that covers a variety of subjects. Among those subjects, and particularly in the case of spatial epidemiology of cancer, one important topic is the level of spatial data aggregation. The analysis of individual data may present ethical data protection issues, so geographically-aggregated data are often used in epidemiological research. Conversely, the analysis of aggregated data also raises certain methodological issues, such as: a) the importance of calculating standardized rates (usually by sex and age) to allow for the comparison of different geographical units; b) the need to take into account the fact that the results obtained at a given level of geographical aggregation cannot be extrapolated as true for other levels of aggregation; and c) the possible existence of the “small numbers problem” and the relevance of using smoothed rates to try to solve it. This thesis takes a first look at this subject through a systematic review of the literature on the spatial epidemiology of cancer. The findings show that the questions which are considered most frequently in the literature consist of: a) the methodology used; b) the spatial distribution of cancer and its temporal evolution; and c) risk factors. The thesis then seeks to address these three aspects. In order to do that, several methodological issues related to the research of the spatial epidemiology of cancer are evaluated and the mortality distribution of all types of cancer is analyzed. Given the high rates of incidence and mortality due to colorectal cancer among the Portuguese population, this cancer type is taken as a case-study, and the distribution of its incidence and mortality is studied and compared with the distribution of the prevalence of the main risk factors. With regard to methodology, it is find that data aggregation at the municipality level makes it possible to obtain detailed information on the geographical distribution of cancer in mainland Portugal, and simultaneously address the small numbers problem. The use of statistical methods to calculate smoothed rates based on the spatial vicinity of the units under analysis also contributes to solving this latter problem. As far as cancer’s spatial distribution patterns are concerned, it is find significant heterogeneity across the municipalities of mainland Portugal, both when examining overall cancer mortality and when specifically looking at colorectal cancer incidence and mortality. Regarding the mortality distribution of all cancer types, it is find that the highest mortality rates tend to cluster around Oporto (on the north coast), Lisbon (on the west coast) and in the municipalities of southeast mainland Portugal. With respect to colorectal cancer, it are detected different patterns of incidence and mortality. The distribution of colorectal cancer incidence is more heterogeneous than that of colorectal cancer mortality. In addition, both colorectal cancer incidence and mortality exhibit especially high values for men in two municipalities in the center of mainland Portugal, which seems to call for a more detailed analysis of this area. In terms of risk factors, the following main risks, which have been identified as having a possible association with colorectal cancer, were considered: tobacco smoking; alcohol intake; and dietary behavior. Screening, as a secondary prevention measure aimed at the early detection of cancer and mortality reduction, was also considered in the analysis. However, data on these risk factors are representative only at much more aggregated geographic units, such as NUTS II. Therefore, a detailed analysis of association could not be performed. Finally, the natural background radiation in mainland Portugal was also analyzed, computing a mean by municipalities, and the conclusion is that its association with colorectal cancer is generally weak. This work points to a variety of interesting topics to study in future research, such as: a) to apply the same methodologies to other cancer types; b) to study the geographical association of risk factors, like socio-demographic and environmental risk factors, with cancer incidence and mortality; c) to test other spatial methods in the research of spatial epidemiology of cancer in Portugal, like, for an instance, kriging or multilevel models; and d) to establish suitable methodologies to study spatial epidemiology of cancer in the Portuguese islands.