Epidemiology
EPIDEMIOLOGY FOR TESTING HYPOTHESES ABOUT TOXIC EXPOSURES
Science can be defined as a methodical approach to the acquisition of knowledge. The scientific method involves problem identification, hypothesis generation, and a study designed to test the initial hypothesis. Confirmation occurs when the results supporting or refuting the hypothesis are seen in repeated observations (Cohen, 1950; Irani, 1971; Feyerabend, 1981). An example of the application of scientific method is the investigation of a causal association between amphibole asbestos and mesothelioma. Wagner et al. (1960) wrote a preliminary publication describing 33 cases of diffuse pleural mesothelioma. Early in the investigation, the authors suspected that asbestos might be implicated, but this hypothesis was not supported at once from the original histories from the patients. After obtaining detailed occupational and residential histories, it was found that all but one case had a probable exposure to crocidolite asbestos called Cape blue. This landmark report opened a large area for epidemiological studies to test the initial hypothesis and replicate the finding (Newhouse, 1969; Wagner, 1991; McDonald and McDonald, 1998; Miller, 2004).
Epidemiology is the field of public health that studies the incidence, distribution, and etiologies of disease in human populations. It focuses on evaluating associations between exposures and disease in human populations. Well-performed epidemiological studies are the best way to determine potential risks and the effects of substances on humans. A U.S. Surgeon General's report describes the approach as a "direct measurement of association" (Bayne-Jones et al., 1964). The staff of the Office of Scientific Advisor writes, "EPA prefers high-quality human studies over animal studies because they provide the most relevant kind of information for human health identification" (U.S. EPA, 2004). Epidemiology requires a comparison group as opposed to case reports or case series. case series such as the one by Wagner et al. ( 1960) are descriptions of selected patients and as such are not analytical studies.
The aim of a body of epidemiological literature is to infer whether an association is causal and to derive an estimate of the magnitude of the excess risk, if one exists. An association is defined as a statistical dependence between two or more events, characteristics, or other variables. Association between two variables does not imply that one event causes the second. All associations reported in epidemiology studies can reflect varying degrees of bias (i.e., systematic errors), chance, and the reality of the situation under study. Errors may arise from biased selection of study participants, misinformation concerning the study or control groups, and confounding factors (Fraser, 1987; Rothman and Greenland, 2001; Savitz, 2003; U.S. Preventive Services Task Force, 2003). Confounding occurs whenever the effect of an exposure is distorted because of the association of the exposure with other factor(s) that influence the disease. Confounding can attenuate or exaggerate a relationship (Last, 2001) and pose obstacles to the interpretation of any epidemiological study (Savitz, 2003). Amphibole asbestos becomes a confounder when it is present in the air and/or tissues of study subjects when the intention is to test the hypothesis that chrysotile exposure causes a cancer risk.
Scientists use animal studies to study various toxicological aspects of substances to help assess human health risks. Although animal studies involve controlled exposures to wellcharacterized agents, many uncertainties are introduced when extrapolating the results of animal data to humans (Brent, 2004). With regard to modeling asbestos risk specifically, attempts to use animal data for human dose-response factors are not recommended (Berman and Crump, 2003). "In the end, if a choice has to be made between animal and human evidence as a basis for assessing human risk, adequate human data must be given a priority" (Hodgson and Darnton, 2000). The question of whether humans develop mesothelioma as a result of exposure to chrysotile asbestos fibers must be answered based on human experience rather than on animal experimentation (Elmes, 1994).
In a speech at the Royal Society of Medicine to occupational medicine physicians over 40 years ago, Sir Austin Bradford Hill, Professor Emeritus of Medical Statistics at the University of London, proposed a list of "nine different viewpoints" when interpreting observational and related studies as evidence of causation. Researchers and policymakers are still using the approach widely today. His decisive question was whether the frequency of undesirable event B will be influenced by a change in the environmental feature A. No formal tests of significance can determine cause and effect (Hill, 1965). Using the list as a "causation model" (see Lemen, 2004) without first establishing an association was not the original intent of Professor Hill. Discussed also in the section of the report called "Establishment of Association" for the U.S. Surgeon General by Bayne-Jones et al. (1964, pp. 179-182), this point was clearly reiterated by Hill when he introduced the list in his speech:
Disregarding then any such problem in semantics we have this situation. Our observations reveal an association bern-een two variables, perfectly clear-cut and beyond what we would care to attribute to chance. What aspects of that association should we especially consider before deciding that the most likely interpretation of it is causation? [italics added] (Hill, 1965)
If it he shown that an association exists, then the question is asked, 'Does the association have a causal significance?'. ..To judge or evaluate the causal significance of the association between the attribute or agent and the disease, or effect on health, a number of criteria must be utilized, [italics added] (Bayne-Jones, 1964)
Because the list of considerations was meant for studying positive associations "before we cry causation" according to Hill, the need to consider the credibility of an observed absence of association is not addressed (Savitz, 2003). Consideration of the Hill aspects should not be initiated at the current time because of the lack of a well-documented, "clear-cut" association with mesothelioma in the chrysotile epidemiological studies (as discussed later).