Regional Variations in Cancer Mortality
For some types of cancer, the maps show regional variations in mortality that are mainly attributable to differences in cigarette smoking (lung cancer), smoking and alcohol consumption (cancer of the oral cavity, esophagus, and larynx), or dietary habits (stomach cancer). The maps also show geographic associations between lung cancer mortality and regions with unusually high levels of occupational exposure (Wismut). In certain regions of western Germany as well (Lower Rhine and Ruhr, Saarland), it may be assumed that occupational exposures and perhaps environmental air pollutants contribute to the increased risks that are shown on the maps. The preventive strategies outlined above and mentioned in the various sections of "Cancer Mortality by Site" have particular relevance to this situation. For many types of cancer, however, the maps do not indicate definite frequency patterns.
In years past, many efforts have been made to analyze regional clusters of cancer cases, and considerable experience in this area has been acquired. There have often been basic misunderstandings regarding the nature and purpose of "cancer mapping." To clarify this issue, we shall review several key points in the interpretation of cluster data. Further information can be found in other sources such as Becker and Wahrendorf (1991) and Becker (1995).
- The maps show cancer mortality for the period from 1981 to 1990. Because cancers have latent periods (interval between exposure/induction and diagnosis) that in some cases are very long, the risk patterns must date back at least 10 years (e.g., for tobacco-related lung cancer) and perhaps as much as 40 years (e.g., certain risk factors for stomach cancer). Consequently, the patterns that appear on the maps cannot be associated with risks that were incurred even as late as the 1980s. Moreover, finding an increased cancer mortality in certain areas does not necessarily mean that the risk pattern underlying this finding still exists today (e.g., stomach cancer in Bavaria). Consequently, an increase in cancer mortality does not necessarily imply an immediate health concern.
- In cases where no conspicuous special influences are known to exist, and subject to the proviso noted above, basically the same risk factors are responsible for regional differences in cancer incidence that are generally known to be associated with the cancer in question, and they are operative in their usual order of importance. For example, regional mortality differences for lung cancer are due mainly to differences in smoking habits - by far the dominant risk factor. Occupational or environmental factors should be considered only if they increase the regional risk to such a degree that they surpass the dominant effect of smoking. The Wismut uranium mines are proof that such situations can occur, and they serve as an example of the "conspicuous special influences" that must be present in order to produce a detectable change in cancer mortality.
- If relatively little is known about the causes of a particular cancer (e.g., prostate cancer, leukemias), any increased mortality from that cancer noted in a particular region cannot, for the time being, be explained. One of the traditional goals of cancer atlases is to identify high-risk areas for etiologically obscure cancers, on the theory that these areas will provide an excellent setting for etiologic research. But given the limitations that are imposed on epidemiology as a nonexperimental science, the conditions for such research are not favorable in every region. The question of whether an epidemiologic study appears feasible is decided not by the degree of the risk increase that is observed in a region but by determining whether the data available for the region create favorable prospects for making discoveries that go beyond the existing body of knowledge.
- It is important to note that the data on which these maps are based are subject to random variations that may raise values into the "statistically significant" range or may results in markedly reduced values, which generally receive less attention. These random fluctuations are superimposed upon any real regional differences in underlying risk patterns that may be present. For methodological reasons, it is often impossible to differentiate real from random (for further information see the publications cited above).
Usually, the primary purpose of maps showing regional differences in cancer mortality rates is to identify areas of increased mortality and seek the causes of the increase. From the standpoint of preventive medicine, however, at least equal attention should be given to regions that show persistent low rates of cancer mortality. It may be rewarding to study the potential for cancer morbidity and mortality prevention that already exists under the socioeconomic conditions that prevail within the region.