Tom Webster grew up in San Diego County, California, loving mountains, deserts, the ocean and Mexican food (he still does). As a kid, he wanted to be a marine biologist but was wooed away by physics and math as an undergraduate student at MIT. His interest in environmental science was also kindled in college and became a passion while working with Dr. Barry Commoner at the City University of New York. Tom got his doctoral degree in environmental health at Boston University and then joined the faculty. Tom thinks one of the exciting aspects about public health and the Superfund program is the interdisciplinary nature of the research. Maybe he’ll even get to do some marine biology some day!
Research Summary
Risk of disease typically varies from place to place. Understanding the reasons for this variation may provide clues to the causes of a disease and methods of prevention. Our group is developing new, improved ways to map disease risk. We are particularly interested in several important aspects of this problem. First, since many diseases such as cancer take years or decades to develop, where a person lives at the time of diagnosis is less important than where they lived earlier in life. Second, we want to create maps that take into account known causes of disease (for example the effect of smoking on lung cancer) so that the map shows the unexplained risk. Amazingly, it turns out that some of these mapping techniques are useful for analyzing the toxicity of mixtures of toxic chemicals. Understanding mixtures is a difficult but important problem in toxicology because while scientists typically study one chemical at a time, we are exposed to many simultaneously.
Q. Your work includes analyses of epidemiologic and toxicological data. How do you use similar methods to look at these two different types of data?
A. Perhaps you have seen three dimensional models of mountain ranges. Imagine connecting points of equal height on this model with lines. Such lines are called contour lines. By drawing contour lines on a map, we can show elevation at each point (maps like this are called topographic maps). For example, the 500’ contour line shows places that are 500’ above sea level. Isolated hills thus appear as concentric rings. In our epidemiology work, we use fancy statistics to estimate the risk of disease at different locations. Imagine creating a three dimensional model of this information where height at each point represents disease risk. To display the information on a flat map, we can use contour lines or colors, with shades of color representing different amounts of risk.How can we apply this idea to toxicology? Toxicologists typically study one chemical at a time, but we are exposed to many in real life. Understanding the toxicity of mixtures of chemicals is an important but hard problem. But in one way it’s similar to mapping elevation or disease risk. Suppose we do a bunch of experiments examining the toxicity of different concentrations of two different chemicals. We can think of the results as a three dimensional model with latitude and longitude representing concentrations of the two chemicals and height representing toxicity. We can again produce a map of this information using contour lines. It turns out that the shape of the contour lines tells us a lot about the behavior of the mixture, for example whether it is acting in a “synergistic” (greater than expected) way.
Q. In your analysis of cancer on Cape Cod, how did you tell if there was a cancer cluster (or hot spot) in a particular area?
A. We begin with carefully collected epidemiologic data on the cancer of interest. We use statistical techniques to construct a map of the risk of the cancer at each point on the Cape. In doing so, we take into account known causes, such as smoking for lung cancer, so that the maps represent the unexplained risk. Cancer clusters (“hot spots”) are those areas of the resulting map with high unexplained risk, a “hill” in our cancer map. We use statistics to determine how likely it is that the hill might be due to chance.Q. What is ecological bias, and how does it affect the validity of epidemiology studies?
A. Standard epidemiological studies collect information for each person in a study: exposure, disease outcomes, and other information (such as known causes of the disease). For example, to study the effect of radon on lung cancer, an epidemiologist would collect information on radon exposure, lung cancer and smoking for each individual in the study. This is a lot of work! In contrast, in an ecological study--also called a group-level study--one only has information about groups of people. For example, one might know the average radon exposure, average lung cancer risk and average rate of smoking for a number of counties. Such data might be available from standard references, making it much cheaper to obtain than individual level information. The problem is that serious errors can occur when you try to use group-level data to estimate disease risk. One cause of the problem is that it often much more difficult to properly take into account the known disease causes. Epidemiologists call these problems ecologic bias.
Q. What is the advantage to using Open Source software? How do you hope the techniques and methods that you developed will be used?
A. With Open Source software, anyone can obtain the computer programs that we use to do our statistical analysis for mapping. This makes it much easier for other researchers to use the software as well as to improve it or modify it for their purposes. We hope that researchers and health departments will use our methods to better understand the distribution of disease in human populations, providing clues to causes.
If you have a question for Tom, email it to
This e-mail address is being protected from spambots, you need JavaScript enabled to view it
