Dose Response Modeling
All chemical and physical agents can induce injury or illness under specific conditions of dose and exposure duration. For example, the short-term effects of alcohol consumption (with ethanol as the toxic component) increase and intensify as consumption increases, with slurred speech occurring at a lower blood-alcohol concentration than coma. Thus, all agents have characteristic dose response relationships, and an understanding of dose response is central to the assessment of risk of harm from an exposure. An agent's dose response is studied experimentally or through observation of uncontrolled exposures, and can be described quantitatively as data permit.
We use our expertise in toxicology, mathematics, and statistics to develop dose response models to inform the assessment of exposures to potentially hazardous substances. Examples of such projects follow.
Sample Projects
Toxicity of methyl mercury in sushi
An individual who ate sushi daily had an abnormally high concentration of mercury in his blood, and claimed that he suffered from methyl mercury-poisoning. We reviewed and analyzed the qualitative and quantitative evidence on methyl mercury-induced health effects, and found that his symptoms were not consistent with his likely exposures to methyl mercury.
Derivation of emergency planning guidelines
For a manufacturer of specialty chemicals, we collected and assessed toxicologic and other data regarding certain process chemicals that might be released during an accident. In keeping with procedures developed by U.S. EPA and industrial hygiene organizations, we derived limits for these chemicals in air based on specified exposure durations and effects that might delay or prevent one's escape.
Uncertainty distributions for cancer potency factors: combining epidemiological studies with laboratory bioassays - the example of acrylonitrile
For most of the materials known through epidemiologic study to be carcinogenic to humans, cancer bioassays performed on laboratory animals are also available. If, as happens, the procedures used for estimating human carcinogenic potencies from laboratory animal bioassays are to be relied on in cases where there are no human epidemiological data, such methods should be also used where there is epidemiological evidence. We developed consistent methods of incorporating the results of both epidemiological studies and laboratory animal bioassays into a single probability distribution for cancer potency-values, using acrylonitrile as an example. The methods are sufficiently general to allow the incorporation of any combination of positive and negative bioassay and epidemiological data. Details are provided in the original publication:
Crouch, E.A.C. (1996). Uncertainty distributions for cancer potency factors: combining epidemiological studies with laboratory bioassays - the example of acrylonitrile. Human and Ecological Risk Assessment 2:130Ð149.
Analysis of dioxin cancer bioassays
For a project involving a large area of soil contaminated with polychlorinated dibenzo-p-dioxins and furans, we reviewed carcinogenicity bioassays for 2,3,7,8-tetrachloro dibenzo-p-dioxin (TCDD). In particular, we evaluated ten large, long-term, carcinogenicity bioassays using 2,3,7,8-TCDD, in male and female hamsters, rats, and mice. We obtained carcinogenic potency estimates and their approximate uncertainty distributions from each experiment using likelihood methods applied to linearized multistage dose-response curves. We then investigated various methods for extrapolating the animal data to humans including (1) constraining the between-experiment variability to have the geometric standard deviation (a factor of 10) previously observed in animal-to-human extrapolations, and (2) using the U.S. EPA approach of extrapolation using the 1/4 power of body weight ratios. Because a better dose metric for extrapolation to humans may be the lifetime average concentration in the body, this metric was estimated for three experiments (two in female rats, one in mice) based on a dosimetric model. Finally, we performed an analysis of the most informative bioassay that allows for the possibility of a threshold and concluded that such a threshold does appear to exist, below which there is no carcinogenic response at all.
Epidemiologic analysis of asbestos exposure, cigarette smoking, and lung cancer risk
We investigated the carcinogenicity to the lung of asbestos exposure and cigarette smoking, the relationship between exposure to both materials, and a means to attribute the increase in lung cancer risk to asbestos only, smoking only, or the interaction between the two. In epidemiologic studies of exposure to an agent that may cause a particular disease, it is typical to estimate the relative risk (RR) of disease for those exposed compared to those unexposed. Ideally, the RR is a direct measure of the probability of causation in those with the disease. With sufficient data, the variation of relative risk with exposure variables, such as dose and period of exposure, may be obtained. We quantitatively evaluated the relative risk of lung cancer as a function of age, age at termination of exposure, age at first exposure, and average intensity of exposure for each of asbestos exposure and smoking, and developed a model to reflect the interaction of these exposures. We wrote a computer program that allows, for a specific individual, allocation of the relative risks for lung cancer attributable to these two agents, given a particular quantitative history of cigarette smoking and asbestos exposure. In the context of this work, we also performed a meta-analysis of the relationship between exposure to asbestos and risk of lung cancer. Details are provided in the publication:
Lash, T.L., Crouch, E.A., and Green, L.C. (1997). A meta-analysis of the relation between cumulative exposure to asbestos and relative risk of lung cancer. Occupational and Environmental Medicine 54:254Ð263.