UICC World Cancer Congress 2006

The relationships between risk factors and cancer occurrence at a geographical level are not easily understood, interpretation being greatly dependent on scale. Maps of cancer incidence on a global scale show clear spatial patterns between different parts of the world (http://www-dep.iarc.fr) (Ferlay et al 2004). The overall pattern shows that total cancer incidence is higher in affluent parts of the world, most markedly for cancer types closely related to socio-economic status (SES).

At a continental scale, the United States (US) National Cancer Institute (NCI) publishes interactive maps of cancer mortality at US County or State Economic Area levels (one or more socio-economically similar counties within a state) (www.nci.nih.gov) (Devesa et al 2001). These maps show (for example) that the higher the County poverty rate, the greater the lung cancer mortality rate among men, whereas patterns are less clear for other cancers. NCI notes that ‘‘the study of geographic patterns of cancer may provide important clues to the causes of cancer and improvements in cancer control'', but also that it ‘‘does not provide information about why death rates may be higher in certain localities than in others''.

At a small area level (such as a municipality), spatial patterns of disease are often far more difficult to interpret. Using modern geographical information systems (GIS), data are usually aggregated up to a practicable level, depending on which geographical units have relevant data available (Jarup and Best 2003).

An overview of cancer mapping at different scales (global, national, small area level) will be given. Interpretation of small area disease patterns as well as strengths and weaknesses of ecological epidemiological study designs will be discussed.

A range of GIS methods for the analysis and mapping of cancer will be discussed with an emphasis on how a spatial approach can complement epidemiological studies for risk assessment. The role of maps and visualisation of the results of such spatial analysis is also important in conveying information and must be carefully considered to ensure effective communication of data.

The session will also motivate the use of Bayesian hierarchical models for mapping cancer risk at small area level. An attractive feature of these methods is their ability to smooth spurious variation in the raw data, leading to more interpretable estimates of cancer risk. The estimates can also be summarized to give probabilities of the risk exceeding a particular level for the purposes of risk assessment. These features will be illustrated using US cancer data. Extension of the Bayesian models to the joint mapping of several cancers with potentially similar etiology will also be discussed (Knorr-Held and Best 2001).

Finally, a software tool for disease mapping and risk analysis, the Rapid Inquiry Facility (RIF), will be demonstrated, using US cancer data. The RIF has been developed by SAHSU in collaboration with the US CDC Environmental Public Health Tracking Program. The RIF produces rates and relative risks for any specified disease endpoint in relation to environmental sources of pollution, adjusted for covariates such as socioeconomic status. The RIF can also produce disease maps at a small area level, using Bayesian smoothing techniques to minimize problems of sparse case data in small areas (Jarup 2004).

References

Devesa SS, Grauman DG, Blot WJ, Pennello G, Hoover, N, Fraumeni JF Jr. Atlas of Cancer Mortality in the United States,1950–94 NIH Publ No. (NIH) 99-4564. NIH,1999.

Ferlay J, Bray F, Pisani P, Parkin DM. GLOBOCAN 2002: Cancer Incidence, Mortality and Prevalence Worldwide IARC CancerBase No. 5. version 2.0, IARCPress, Lyon, 2004.

Jarup L, Best N. Editorial comment on Geographical differences in cancer incidence in the Belgian Province of Limburg by Bruntinx and colleagues. Eur J Cancer. 2003;39:1973-5.

Jarup L. Health And Environment Information Systems for Exposure and Disease Mapping, and Risk Assessment. Environ Health Perspect 2004;112:995-7.

Knorr-Held L, Best NG. A shared component model for detecting joint and selective clustering of two diseases. Journal of the Royal Statistical Society A 2001;164:73–85.

Singh GK, Miller BA, Hankey BF, Edwards BK. Area Socioeconomic Variations in U.S. Cancer Incidence, Mortality, Stage, Treatment, and Survival, 1975–1999. NCI Cancer Surveillance Monograph Series, Number 4. Bethesda, MD: National Cancer Institute, 2003. NIH Publication No. 03-5417.