7
SCIENTIFIC FEASIBILITY OF EPIDEMIOLOGIC STUDY

IN THIS CHAPTER, the subcommittee addresses the utility and feasibility of conducting epidemiologic studies to assess whether adverse human health effects could be associated with exposure to ZnCdS as a result of the Army's dispersion tests. This assessment was performed by the subcommittee in response to citizens who requested that health studies be conducted in their communities.

The chapter has 3 sections. The first section introduces the field of epidemiology, providing an overview of what it can do and its limitations, its underlying assumptions, and the types of research designs that it uses. The second discusses key methodologic issues that need to be considered in deciding whether an epidemiologic study is feasible and likely to produce scientifically valid information about a particular exposure and health outcome. The last section discusses the types of epidemiologic studies that were considered by the subcommittee in light of available information, including that provided by members of the public at public hearings.



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 82
Toxicologic Assessment of the Army's Zinc Cadmium Sulfide Dispersion Tests 7 SCIENTIFIC FEASIBILITY OF EPIDEMIOLOGIC STUDY IN THIS CHAPTER, the subcommittee addresses the utility and feasibility of conducting epidemiologic studies to assess whether adverse human health effects could be associated with exposure to ZnCdS as a result of the Army's dispersion tests. This assessment was performed by the subcommittee in response to citizens who requested that health studies be conducted in their communities. The chapter has 3 sections. The first section introduces the field of epidemiology, providing an overview of what it can do and its limitations, its underlying assumptions, and the types of research designs that it uses. The second discusses key methodologic issues that need to be considered in deciding whether an epidemiologic study is feasible and likely to produce scientifically valid information about a particular exposure and health outcome. The last section discusses the types of epidemiologic studies that were considered by the subcommittee in light of available information, including that provided by members of the public at public hearings.

OCR for page 82
Toxicologic Assessment of the Army's Zinc Cadmium Sulfide Dispersion Tests NATURE OF EPIDEMIOLOGIC INVESTIGATIONS Epidemiology can be broadly defined as ''the study of the distribution and determinants of disease frequency'' in human populations (MacMahon and Pugh 1970). Implicit in that definition is the assumption that disease does not occur randomly and that systematic epidemiologic methods can be useful for identifying risk factors for various diseases. Epidemiology can be used to assess empirical associations between an exposure and disease, especially when the increase in risk of a disease attributed to an exposure is large, when the exposure can be measured with objective criteria, and when exposed and unexposed people can be clearly distinguished. Epidemiologic studies are less informative when the difference in risk between exposed and unexposed people is small, when exposures are poorly defined or measured, when exposed and unexposed people are not readily distinguished, or when confounding factors exert a much greater effect on the incidence of the disease than does the exposure of interest. An individual's risk of disease cannot be determined through epidemiologic study; rather, such study determines average risks for populations or at least for representative samples. In general, epidemiologic studies are designed to study risk factors for a specific disease or the possible health effects associated with a specific exposure. In recent years, environmental epidemiology has developed as a subspecialty of epidemiology. It focuses on the role of physical, biologic, chemical, and psychosocial factors—factors not necessarily under the individual's control—and human health (NRC 1991a; Rothman 1993). Environmental epidemiologists are increasingly asked about past exposures and their relation to present health problems. Various research designs are available to epidemiologists, depending on whether a study begins with a particular disease outcome or with an exposure. Retrospective or case-control studies categorize people on the basis of whether they have the disease of interest (cases) or are free of the disease (controls) at the time of study. Controls should be similar to cases except for the presence of the disease and need to be carefully selected for study; the two groups are compared with respect to the proportions that

OCR for page 82
Toxicologic Assessment of the Army's Zinc Cadmium Sulfide Dispersion Tests have had a particular exposure. Most information is collected retrospectively for case-control studies. Prospective or cohort studies identify populations or samples for study with respect to an exposure and follows them over time to identify the onset of disease. Information collected prospectively is considered by some to be more reliable than information collected retrospectively. Regardless of design, epidemiologic studies need to conform to the scientific principles of research methodology if their findings are to be considered scientifically valid and useful. KEY METHODOLOGIC ISSUES Three key methodologic issues would have to be considered together in designing a scientifically valid epidemiologic study that would be capable of answering the concerns of people who were exposed to ZnCdS: (1) Is a well-defined population comprising clearly defined exposed and unexposed individuals available for study? (2) Is the exposure capable of causing the suspected health outcomes? (3) Is the other information needed for a well-designed epidemiologic study available? These issues are discussed below as they relate to a possible study of health effects of exposure to ZnCdS. 1.Is a well-defined-population comprising clearly defined exposed and unexposed individuals available for study? Accurate measurement of an exposure is a key aspect of any epidemiologic study. In environmental epidemiology, the crudeness with which an exposure is measured often renders the results inconclusive. A somewhat different set of circumstances surrounds the Army's testing with ZnCdS. Extensive data on ZnCdS dispersion patterns are available for several geographic areas—including Minneapolis, Fort Wayne, and Corpus Christi-because the Army was interested in measuring the dispersion of the compound and monitored it closely starting soon after its release. However, there are two major limitations to the exposure data. First, data are not available regarding the actual exposures or doses that people

OCR for page 82
Toxicologic Assessment of the Army's Zinc Cadmium Sulfide Dispersion Tests (as opposed to geographic areas) received as a result of the Army's tests. Exposure and dose are important concepts for environmental epidemiology, as discussed by the NRC (1989, 1991b). "Exposure" refers to the concentration of (and duration of exposure to) an agent that a person receives from the environment, whereas "dose" refers to the amount deposited in or absorbed by the body (Hatch and Thomas 1993). The subcommittee considered whether geographic residence could be used to estimate individual exposure. This approach was considered a crude means of measuring exposure, because of the number of assumptions that underlie it. For example, it assumes that all people who lived in a geographic area were there at the time of the exposure and that all people received the same exposure regardless of their exact location (for instance, inside or outside). Furthermore, ZnCdS exposure occurred over 35 years ago, so it is likely that many people could not be found or would not recall where they were at the time of the Army's testing. A second limitation of the exposure data involves the inability to distinguish between cadmium from the Army's tests and cadmium derived from other sources. If the exposure of interest for determining health effects is exposure to cadmium, the exposure to cadmium from ZnCdS would be very low and indistinguishable from background environmental and industrial sources of cadmium (see Chapter 5). Cadmium is naturally present in soil and groundwater and has wide industrial uses, for example, in paints, plastics, and nickel-cadmium batteries. In addition to the difficulty in knowing how much cadmium might be available from ZnCdS (see Chapter 5), measurement of the cadmium exposure that would have resulted from ZnCdS would yield a highly imprecise estimate of total exposure to cadmium. If it is assumed that exposure to ZnCdS is equivalent to exposure to cadmium, almost all persons received a total cadmium exposure from the Army tests of less than 1 µg and 24.4 µg is the highest estimated total potential dose received by any person in a populated area. The average daily intake of cadmium from ambient air ranges from 0.1 to 0.8 µg in urban areas. The average daily intake of cadmium from all sources (air, water, food, and so on) ranges from 12 to 84 µg in industrial areas. Therefore, most people received a total cadmium exposure from the Army tests that was no greater than that associated with living in a typical urban area for less than a month. Such a small difference between cadmium from the

OCR for page 82
Toxicologic Assessment of the Army's Zinc Cadmium Sulfide Dispersion Tests Army's ZnCdS tests and that from background sources makes it very unlikely to detect health effects in an epidemiologic study. Measurement of internal dose is needed for analyzing potential dose-response relationships, although often this information is not available in community-wide situations. When a dose-response relationship can be observed, researchers are more likely to attribute the observed health effects to the exposure. To assess dose-response relationships, people need to be categorized as to whether they have been exposed and, if so, by the amount (dose) of the substance that was deposited or absorbed. Dose is more difficult to assess in environmental studies than in occupational studies in which workers are followed and monitored directly. No data are available for assessing individual exposure to ZnCdS. Not all exposed persons received the same dose, nor will all exposed persons manifest adverse health effects. Several characteristics will determine whether exposed people exhibit disease, such as age, genetic predisposition, immune function, and health status. In addition to dose, the duration and timing of the exposure might affect the magnitude and severity of human health effects, as discussed below. In summary, the subcommittee recognizes that people could be classified with respect to ZnCdS exposure on the basis of residence at the time of the Army's tests. However, this approach would yield only a crude approximation of exposure and would be subject to bias stemming from difficulties in finding residents of exposed areas and from misclassification of people with respect to exposure status. Accurate measurement of dose for individuals is not possible, given the very small contribution of ZnCdS to the total exposure to cadmium. 2.Is the exposure capable of causing the suspected health outcomes? As discussed in previous chapters, few, if any, human-health data are available on ZnCdS exposure from previous studies. In considering the feasibility of conducting an epidemiologic study, the subcommittee evaluated the contribution of cadmium from ZnCdS in a worst-case scenario. Much of the previous work on cadmium toxicity has focused on occupational groups in which exposures were very high and continued for long

OCR for page 82
Toxicologic Assessment of the Army's Zinc Cadmium Sulfide Dispersion Tests periods. Occupational exposure to cadmium is not the same as exposure to cadmium in the form of ZnCdS. That is an important distinction because it appears that when cadmium is ingested in the form of ZnCdS, it is not readily soluble in the body (Leighton 1955; Weast and others 1986). A review of the scientific literature suggests that only a few health outcomes might be linked to cadmium exposure: lung cancer, dysfunction of the proximal renal tubules, and possibly infertility and low birth weight. The spectrum of health effects reported in the subcommittee's public hearings is much wider—infertility, mental retardation, autoimmune disease, melanoma, lymphoma, leukemia, chronic lung disease, breast cancer, joint pain, skin problems, and many others. The subcommittee recognizes several serious limitations of research on health effects of ZnCdS. The first is the lack of previous human studies on this compound, although cadmium has been studied extensively and is known to be associated with health risks. Second, there is a lack of health information systematically collected before, during, and after the periods of exposure in the affected communities; it is essential to have data on the temporal relationship of exposure to health outcomes to establish a cause-effect relationship (Hill 1965). Third, although health outcomes can be determined in individuals who identify themselves to investigators, it would be difficult to link specific health outcomes to the ZnCdS exposures. In general, it is possible to confirm the presence of disease in individuals. It is not possible to attribute illnesses to ZnCdS exposures that occurred over 35 years ago. The long latency period between exposure and disease increases the difficulty in collecting the necessary information and makes interpretation of results in establishing causality more difficult. Residents who voluntarily came forward in public meetings might not constitute a representative sample for an epidemiologic study. 3.Is the other information needed for a well-designed epidemiologic study available? A well-designed epidemiologic study requires information on attrition, statistical power, and confounders. It can be hard to find people, especially if they have moved or changed names and especially in the absence of personal identifiers, such as social-security numbers. The absence of

OCR for page 82
Toxicologic Assessment of the Army's Zinc Cadmium Sulfide Dispersion Tests that information in communities exposed to ZnCdS could complicate follow-up. The subcommittee considered the issue of statistical power in assessing the feasibility of an epidemiologic study of cancer or dysfunction of the proximal renal tubules in relation to ZnCdS. Statistical power is the ability to detect a true difference in health status, given the size of a population, the prevalence of exposure, and the expected size of the effect. We specifically chose those 2 health outcomes on the basis of scientific evidence and public concern. Our maximal estimates of increased risk of cancer attributed to ZnCdS for most communities is approximately 1 additional cancer case per million people exposed (see Chapter 6). As an example, assume that even our conservative estimate of 1 extra cancer death per million people exposed is too low by a factor of 100. Could we carry out an epidemiologic study if in fact ZnCdS exposure causes 1 extra cancer death among 10,000 exposed people? Our power calculations indicate that to detect an effect of that magnitude, we would need to study more than 62 million people, half of whom are in a high-exposure category. Similarly, the subcommittee performed power calculations for lung cancer and proximal renal tubular dysfunction, which are reported to affect about 2% of the general U.S. population (American Cancer Society 1996; OSHA 1992). If we assume that exposure to ZnCdS results in 1 extra case of lung cancer or renal disease per 10,000 people, we would need to study more than 30 million people, half of whom are in a high-exposure category. That figure is several times larger than the present combined populations of Minneapolis, Fort Wayne, and Corpus Christi, where some estimates of population exposure are available. Assessment of confounding is a major consideration for any epidemiologic study. Information on known or potential confounders—factors that are associated with both the study exposure and the outcome—needs to be collected. Potential confounders of particular relevance to any ZnCdS studies include industrial exposures in Fort Wayne, refinery exposures in Corpus Christi, and indoor air pollution in Minneapolis. Without proper control in either the design or analytic phase of research, confounders can distort the true relation between an exposure and an outcome (Kleinbaum 1994). Simply stated, uncontrolled confounders can overestimate or underestimate the magnitude of health risk associated with an exposure. For

OCR for page 82
Toxicologic Assessment of the Army's Zinc Cadmium Sulfide Dispersion Tests example, very high cadmium exposures have been linked to lung cancer (IARC 1993; Thun and others 1991) and, in 1 study, diminished infant size (Laudanski and others 1991). Both those outcomes also are adversely affected by cigarette smoke, which contains cadmium, as well as nicotine, tars, carbon monoxide, and other toxic substances. After weighing various methodologic considerations, the subcommittee concluded that it is not possible to gather supporting health information, including that pertaining to possible confounders, and that an extremely large sample would be required to detect any statistically significant differences between exposed and unexposed groups. Those limitations, coupled with difficulties in identifying exposed and unexposed populations other than by crude estimates based on residence, argue against an epidemiologic study. TYPES OF EPIDEMIOLOGIC STUDIES The subcommittee considered several epidemiologic approaches in determining the feasibility of conducting an epidemiologic study of the health effects from ZnCdS exposure. We considered various study designs in light of the information presented by members of the public attending the public hearings. Possible study designs are discussed separately below and in relation to the three key research issues addressed above. The subcommittee considered strategies for the retrospective follow-up of populations exposed to the Army's dispersion tests, despite the absence of individual-exposure data. Several barriers to a retrospective follow-up study were identified. Mortality or cause-of-death data will be available for communities only if death certificates were filed with particular states' health departments for people who died. Population-based tumor registries for assessing cancer deaths in Texas, Indiana, and Minnesota were not established until 1949, 1987, and 1988, respectively, and cannot be used in all areas for cancer-mortality studies covering much of the period of interest. For some diseases, excess-mortality studies can be carried out with information on death certificates for persons in exposed and unexposed communities. However, death certificates are not reliable for detecting

OCR for page 82
Toxicologic Assessment of the Army's Zinc Cadmium Sulfide Dispersion Tests many of the diseases mentioned in the public meetings. Excess-mortality studies have other serious limitations, including lack of information on place of residence during the time of exposures, lack of information on the background characteristics of study populations that affect disease likelihood (National Center for Health Statistics 1994) and lack of information on other exposures that could cause the diseases being studied. For other nonfatal health outcomes, there are no standardized reporting mechanisms in place that would permit a scientific analysis. The long time since exposure occurred, the limitations of health information, the diversity of reported health concerns, and the variations in survival and residence of individuals argue against conducting an excess-mortality or excess-morbidity study. Cancer was a frequently mentioned health concern at all public meetings. That could in part reflect the importance and prevalence of cancer in the U.S. population. Cancer (of all types) is the second leading cause of death and illness in the United States (American Cancer Society 1996). About one in 1 in 2 males and 1 in 3 females will develop cancer. Lung cancer will occur 1 in 12 males and 1 in 19 females. Smokers are 10-20 times more likely than nonsmokers to develop lung cancer. Breast cancer is a leading cause of cancer in women, affecting 1 in 8 women at some point in their lifetime. Some people who spoke at the public hearings felt that cancer was unusually common in their communities and wondered whether there had been a cluster of cancer deaths in their areas. Epidemiologists have investigated numerous reports of cancer clusters throughout the world. In most instances, if a higher than expected rate of cancer is observed, no cause can be established. The Centers for Disease Control reviewed 108 cancer clusters in 29 states and 5 other countries and noted that no clear cause could be found for any of the clusters (Caldwell 1990). In fact, cluster-based studies work best for outbreaks of infectious diseases, such as Legionnaire's disease or Salmonella infections, which have an acute onset in a well-defined population soon after exposure. The subcommittee has concluded that a cluster-based epidemiologic design would not be a sound approach to study the chronic-disease consequences of ZnCdS exposure. The subcommittee also considered a prospective cohort study of children who attended the Clinton Street School in Minneapolis. Any study

OCR for page 82
Toxicologic Assessment of the Army's Zinc Cadmium Sulfide Dispersion Tests based on the Clinton Street School would have several methodological shortcomings. First, exposure status would need to be assigned on the basis of geographic residence, assuming that all children in a geographic area had the same exposure. Second, records that could be used to identify all the children who attended the school are not available. No records regarding the children's health status before and after the Army's tests are available. As previously noted, documenting changes over time is important to establish a temporal relationship between exposure and outcome. Furthermore, a small number of students are available for study; the subcommittee's power calculations highlight the large number of people needed for valid studies of selected health outcomes. Finally, information on potential confounders in this population (such as second-hand exposure to cigarette smoke or exposures related to indoor and outdoor air pollution) cannot be reliably assessed, because too much time has elapsed. Collectively, those serious limitations argue against further research involving this special population. CONCLUSIONS The subcommittee evaluated the feasibility of an epidemiologic study in relation to various methodologic issues. The evaluation is summarized in Table 7-1 with the subcommittee's rationale for determining the feasibility of future epidemiologic studies. As Table 7-1 reflects, only a crude estimate of exposure based on people's residence at the time of the Army's tests can be ascertained for affected communities. Although current health status could be determined for some, an epidemiologic study would not be able to determine scientifically whether ZnCdS exposure caused a disease, given the absence of reliable information on exposure, individual dose, or other potential confounders. Small increases in cadmium exposure or cancer risk attributable to cadmium from ZnCdS exposure would be indistinguishable from those related to background sources of cadmium and other risk factors for cancer. And a very large number of people would be needed for a study to determine statistically significant differences attributable to ZnCdS exposure. In summary, the subcommittee has concluded that there are three major

OCR for page 82
Toxicologic Assessment of the Army's Zinc Cadmium Sulfide Dispersion Tests TABLE 7-1 Methodologic Considerations in Assessing the Feasibility of an Epidemiologic Study of the Health Effects Associated with ZnCdS Exposure Methodologic Consideration Feasibility Rationale 1. Is a well-defined population comprising clearly defined exposed and unexposed individuals available for study? Yes The Army monitored the dispersion of ZnCdS. Exposure could crudely be estimated on the basis of residence at the time of the tests; however, this ecologic approach is subject to misclassification bias with respect to exposure status. 2. Is the exposure capable of causing the suspected health outcomes? No Although lung cancer and kidney disease have been linked to occupational cadmium exposure, no known health effects have been associated with ZnCdS. Exposures (to cadmium from ZnCdS in a worse-case scenario) resulting from the Army's tests are believed to be too low to cause health effects. No comprehensive health records exist for people affected for the periods of interest. Even if health status were evaluated currently, individual health complaints could not be linked specifically to past ZnCdS exposure. 3. Is the other information needed for a well-designed epidemiologic study available?     a. Can confounders be adequately addressed in the study? No No information is available on people with respect to known and potential confounders, such as cigarette-smoking, air pollution, diet, and other occupational exposures that might result in erroneous findings.   No   b. What sample size would be required to conduct a valid epidemiologic study?   At least 15 million exposed and 15 million unexposed individuals would be needed for valid assessment of such health concerns as lung cancer or kidney disease.

OCR for page 82
Toxicologic Assessment of the Army's Zinc Cadmium Sulfide Dispersion Tests barriers to carrying out an epidemiologic study of the health effects of ZnCdS: lack of complete and accurate exposure data on individuals; inadequacies in data on health outcomes before, during, and after the periods of exposure; and, because of the low exposures, the requirement of a huge sample to detect any small increase in adverse health effects. Accurate measurement of individuals' doses is not possible, given the very small contribution of ZnCdS to the concentration of cadmium in the environment. Information on potential confounders also would be lacking. The subcommittee concludes that an epidemiologic study of the affected populations is not feasible.