http://www.inel.gov/.research/.research_abstracts/.rb/index.html (World Wide Web Directory, ~04/1995)Leukemia has killed more than 500,000 in the United States since 1950; its etiology is generally unknown. Analysis and topographic mapping of all available U.S. mortality data revealed a shadow epidemic in the Midwest, with more than 7000 excess deaths observed in sixteen contiguous states. Distinct male and female patterns and long-term temporal trends were identified. Both the broad regional patterns and mortality statistics provide evidence of determinism. Low mortality was observed in five regions.
Analysis of leukemia mortality data for other nations offers an opportunity to confirm U.S. observations. If inferred associations with land-use patterns and occupational exposures are valid, the Canadian provinces of Alberta, Saskatchewan, and Manitoba should have elevated leukemia mortality rates. The major grain-producing regions of the former Soviet Union, i.e., the Ukraine and Belarus, should show similar effects. While the agricultural association may affect larger areas, localized effects of hydrocarbon-related activities within nations may also be observed.
One of the objectives of this study was the development of improved tools for analysis and representation of complex data sets. A by-product was improved understanding of cancer statistics; their deviation from standard reference distributions is such that common measures of significance testing may have little value. In small populations, standard analytical methods may find only statistical sheep in imaginary wolves' clothing. Geographical patterns do not create certainties. Perceptions and impressions do not yield quantitative results. But they can lead us in the right direction and help us ask better questions. The mortality patterns and statistical distributions for leukemia and other cancers establish a frame of reference for decision-making, identify promising areas for research, and provide, in part, Tufte's "revelation of the complex."
Leukemia mortality for white adults, 1950-80. 3-D topography represents combined, weighted mortality for both sexes; the upper map projects the computer-generated contours onto State boundaries. Red represents excesses of 7% or more, yellow, excesses of 4-7%, and green, at least 10% below normal.
Comparison of male and female mortality patterns relative to U.S. male and female reference rates for 1950-80. The region of male excess mortality is more widely developed than that of females; male excesses also tend to be higher. The number of SEA's with significant (Dx=0.01) male excesses is double the number with significant female excesses.
Time Trends. Temporal changes also appear significant, generally decreasing during the 1970s in the north-central region and increasing in Louisiana and Mississippi. If occupational exposures were the driving forces for the effects observed in the 50s and 60s, a change in both materials and exposures could explain the decline in the north.
Physiographic Associations. While mortality and demographic data have been recorded by governmental agencies, few other types of data are available on a county-by-county basis. Many of the data sets that are available have been published by the U.S. Geological Survey. Environmental monitoring and assessment programs have been started by EPA, but the variety of data required to begin to link causes to effects is difficult to conceive. An example of potential conceptual linkages is shown with a computer-projected contour of 4% excess leukemia superimposed on a map of the nation's physiography.
Both geographical patterns and mortality statistics offer compelling evidence of determinism for leukemia. Neither the effects of known causative factors nor any other single unknown agent can explain the mortality patterns observed in the U.S. during the last forty years. They exist in spite of the diluent effects of population migration.
WINCO-11892