Cancer, Geography of

Cancer Maps

Maps are important tools in the geographic study of cancer; however, when producing and interpreting [Page 321]cancer maps, it is essential to consider how cancer is measured and displayed. Incidence, mortality, survival, prevalence, and stage of diagnosis are common measures of cancer disease outcomes. Typically, these measures are presented as rates, standardized by population size, age, sex, and race, as a means of accounting for the underlying demographic differences between regions. Since cancers are relatively rare diseases, data are often grouped across many years to provide sufficient stability to the rates. Some maps are descriptive (e.g., for purposes of surveillance), while others are used in spatial analyses undertaken to detect statistically significant patterns or possible associations with potential risk factors. An important issue is the spatial resolution of the cancer data. Depending on the purpose of the map, data may either be aggregated to some geographic level, for example, national, census units, postal districts, and so on, or the data may be mapped using the locations of individual persons (e.g., location of residence). In most applications the resolution is often determined by the availability of data, which is driven by confidentiality concerns, resulting in very few spatial studies of individual-level cancer. As a consequence, the modifiable areal unit problem (MAUP) must be considered, necessitating tests of whether results are consistent across different scales and levels of aggregation; without considering the MAUP, results should be interpreted cautiously.

Spatial Analysis of Cancer

Beyond descriptive maps of cancer, researchers use spatial analyses to quantify and explain spatial and temporal patterns of cancer. First, it is essential to determine if observed patterns are simply random or if there are true geographic differences. Common spatial analytic algorithms include Kulldorff's Spatial Scan and Cuzick and Edwards's Nearest Neighbour Test, although others also exist. If the patterns are statistically significant, researchers may try to identify some underlying factors that could explain the observed patterns. Such factors include smoking, diet, lifestyle, genes, environment, health care resources, and differences in cancer registration. If the spatial analyses point to certain possible risk factors, a new hypothesis may be developed and tested, for example, by using spatial regression analysis such as Geographically Weighted Regression (GWR) or in a more structured epidemiologic study. One example would be the relationship between large concentrations of arsenic in drinking water and the high cancer mortality rates in Taiwan. This association was first suggested based on aggregated data on cancer mortality, which coincided with aggregated data on arsenic concentrations. Later, several case-control and cohort studies supported this hypothesis at the individual level. It should be noted that it is essential that any hypothesis originating from aggregated data be tested at the individual level due to the ecological fallacy.

Geographic studies of cancer, by and large, have generated little etiologic insight. This may be attributable to the neglect of individual mobility information. In the analysis of chronic diseases such as cancer, causative exposures may occur over a long time, and the disease may be manifested only after a lengthy latency period. During this latency period, individuals may move from one place of residence to another. This can make it difficult to detect patterns of cases in relation to the spatial distribution of their causative exposures without having information on residential mobility. Building on Torsten Hägerstrand's concepts, space-time mobility paths can now be visualized using temporal GIS technology. Statistical algorithms have recently been developed for evaluating space-time clustering in residential histories. This approach is theoretically appealing but largely untested, and it is not yet known whether it will reveal insights into disease etiology.

...