COLLEGE STATION, November 5 -- Texas A&M University statisticians help medical researchers analyze epidemiological data, mapping and assessing geographic clustering of cancer and leukemia.
"There have been many celebrated cases in the nation where a few people in one area, all of a sudden, have a similar type of cancer. People are worried about cancer clusters and want to know if the disease in these cases occurred by chance, or if there are certain underlying causes for the disease," said Michael Sherman, an associate professor in the statistics department at Texas A&M. "This is one typical study that my spatial statistics research can address."
Sherman, along with Shirley Bame of the College of Architecture, has studied whether cases of renal failure across the state of Texas occur in a natural way or are concentrated in a few hotspots.
"We found out the distribution of renal failure cases in Texas is consistent with random phenomena," Sherman said. "There could be some reasons to explain 'hotspots' of cancer cases, but we cannot conclusively say this is a persistent pattern across the whole state of Texas."
Sherman is beginning work on another epidemiological study, addressing whether a pregnant woman's exposure to pesticides might cause leukemia in her children. Specifically, Sherman and his collaborator Susan Carroza in the School of Rural Public Health want to know whether children born to women exposed to pesticides during pregnancy have leukemia at higher rates than those born to women not exposed to those pesticides.
A current popular method in epidemiology is called a case-control study. The cases here are children that have leukemia, Sherman explained, and the controls are children that are similar to those cases in terms of age and sex but who do not have leukemia. One basic question then is whether the children with leukemia tend to come from areas with higher levels of pesticides than those children without leukemia. This is largely a non-spatial approach to analyzing these data.
"An approach that uses the actual locations would compare the spatial distributions of the cases with those of the controls, and see if they look similar," Sherman said. "If the cases that have leukemia look like they are more clustered than the controls that do not have leukemia, we then believe there might be some reasons that cause those leukemia cases to be clustered.
"This could refine the non-spatial search, for example, by determining which pesticides are typically elevated where leukemia cases tend to cluster."
Specific pesticide information from satellite images can tell researchers the typical pesticide exposure for each leukemia case cases and for the non-leukemia control.
Sometimes, the statistical analysis might support the researchers' work; sometimes, it might not. The most important thing for statisticians, Sherman emphasized, is to be honest and fair in their evaluation of data.
"Statisticians should use proper and creative methods to tease out what we can learn from the data," Sherman said. "We cannot be influenced by what other people may want to see in the data; we are interested in finding out what the data are truly telling us. I think that is the most important thing a statistician should do."
Materials provided by Texas A&M University. Note: Content may be edited for style and length.
Cite This Page: