MODULE 4
Basics of Radiation Epidemiology
Department of Epidemiology, School of Public Health, University of North Carolina
OBJECTIVES
After studying this module, the reader will be able to
- describe the explanatory approach generally used in radiation epidemiology
- understand and discuss conceptual and technical issues in radiation epidemiology
- identify the practical and theoretical limitations of the approaches utilized in radiation epidemiology
Introduction
Epidemiology is the study of health and disease in populations. Prior to World War II the field was primarily concerned with infectious diseases, and it is from the study of epidemics that the field draws its name. Epidemiological research now addresses a broader range of topics, including environmental agents such as ionizing radiation.
The goals of this module are to describe: (1) the explanatory approach generally used in radiation epidemiology and (2) the practical and theoretical limitations to this approach. The focus on the logic of explanation is intended to encourage critical evaluation of potentials and limitations of epidemiological studies.
The Logic Of Explanation In Radiation Epidemiology
Most investigation in radiation epidemiology addresses questions about the association between radiation exposure, or dose to certain tissues, and an outcome. For example, the relationship between a specific form of ionizing radiation and occurrence of a particular medical outcome is studied. The method of epidemiology is to observe whether disease occurs more or less commonly among individuals who have the exposure or factor than among those who do not. Risk-factor epidemiology explains disease in populations by enumerating all the risk and protective factors, the independent variables or causes, and their relationships with a list of disease outcomes derived from clinical practice, the dependent variables, or effects.
The randomized experiment serves as a model or ideal design for the evaluation of specific agents. First, subjects with specific characteristics, including absence of a disease or outcome of interest, are chosen for study. Next, they are randomized to be exposed or unexposed to a factor, a process that tends to produce an even distribution of the heterogeneous study subjects between exposure groups over the course of many trials. During a period of exposure or non-exposure, all other conditions affecting the subjects can be held constant. Finally, the researcher determines the outcome characteristics in members of each group using a standardized protocol.
The analysis of such a study amounts to a comparison of the frequency of the outcomes of interest between the groups. Differences in frequency that persist over many trials, or that are obtained in a small number of large trials, are attributed to the action of the experimental agent. This method is essentially the same approach that is used in toxicological studies of animals. However, humans should not be experimentally exposed to potentially harmful agents for research purposes. Consequently, most epidemiological studies are observational in nature.
Observational studies attempt to imitate the controlled experiment by making exposed and unexposed groups as similar as possible, except that one group has the exposure itself. This is accomplished both through the design of the study and through statistical analysis of the data. An occupational study of the effects of whole-body exposure to gamma radiation on cancer rates might be designed to include only workers of a certain type; for example, males who worked longer than six months at a specific facility. This avoids some initial differences between exposure groups. The study then might compare workers who had received different cumulative radiation exposures within strata of age, other occupational exposures, and behavioral attributes of interest. This provides a summary estimate of the exposure-disease relationship "adjusted" for differences in those other factors. In attempting to yield results that would have been obtained in an experiment, the observational study attempts to control "extraneous" factors.
It is assumed that well-designed studies can provide an estimate of the radiation-cancer dose-response relationship that characterizes the change in cancer rates for each unit change in radiation. It is often assumed that this is a universal dose response law that could be identified by experimental studies if those were possible [Wing, 1994]. However, many observational studies lack measurements or estimates of individual doses. The disease experience of an entire group of potentially-exposed individuals, such as workers or downwinders, must be compared to some "standard" or "expected" disease rates. Although such studies may identify an excess or deficit of disease, they cannot generally provide information about dose response [Shleien et al., 1991].
Identification of a risk factor or disease agent using the approach based on experimental logic is accomplished by noting a higher disease rate (or excess of observed compared to some predicted number of cases) among an exposed group compared to an unexposed group. However, within the exposed group there is no way to distinguish cases that would not have occurred in the absence of exposure from cases that would have occurred anyway. Thus, risk-factor epidemiology is about factors associated with excess disease in groups. It cannot specify the cause of any particular case of disease.
Some factors associated with disease rates are viewed as causal while others are viewed as spurious. Causal means that the factor acts to create disease. Spurious means that the association occurs either because of another factor or because disease leads to the presence of the factor. This is an overly simplistic view of complex causation of disease within an organism or of disease rates in a population. Nonetheless, this focus on finding exposure-disease associations and then distinguishing causal from non-causal exposures has dominated thinking in modern risk-factor epidemiology.
Limitations In The Epidemiological Approach
Studies of radiation and cancer conducted among Japanese atomic bomb survivors and nuclear workers show that generic difficulties related to comparability, measurement, and knowledge of what to measure pose conceptual and technical problems for radiation epidemiology.
Comparability
Comparability refers to the similarity of individuals with different degrees of exposure. When exposure groups are comparable in other respects, exposure becomes a more plausible explanation of differences in disease rates than when groups differ in other respects. If groups differ in potential to develop cancer, or in exposures to other carcinogens or susceptibility factors, then the absence or presence of a radiation-cancer association of a given form could be due to these other factors.
For many scientists, studies of survivors of the atomic bombing of Hiroshima and Nagasaki have played a dominant role in the assessment of radiation health effects [BEIR V, 1990]. The special circumstances of their exposure in 1945, and survival for inclusion in the population assembled for epidemiological study five years later, may make radiation-disease associations look quite different than they would in other situations. Biased conclusions about the radiation-cancer association in other populations could occur if groups with different degrees of radiation exposure are not comparable in other respects.
Stewart and Kneale [1993a] have presented evidence that differential mortality from the time of the bombing in 1945 until the assembly of the population for epidemiological study in 1950 produced more highly selected groups of robust individuals at higher than at lower exposures. The more robust survivors at higher dose groups would reduce the apparent radiation effect. Further differences among survivors of different levels of exposure may relate to long-term effects of radiation on immune function. This situation raises questions about the applicability of radiation-cancer associations among atomic bomb survivors to other populations. A critical review of the use of atomic bomb survivors as a standard for evaluating radiation health effects is found in Nussbaum and K�hnlein [1994].
Different issues of comparability arise in studies of workers exposed to low-level radiation. To begin with, workers must be healthy enough to be employed. Some studies compare workers with each other rather than with the general population in order to avoid problems of comparability between workers and non-workers. However, workers that enter and remain in jobs involving radiation exposure differ from workers in other jobs; for example, they may have different work skills and preparation. Employers use medical exams to screen workers for dangerous or high-security jobs, and prohibit smoking in some work areas. This could lead to health differences between unexposed workers and those employed in jobs that entail greater radiation exposure.
In one study, workers in jobs with potential for internal contamination with radionuclides had lower mortality from cardiovascular disease and all causes of death combined (but not cancer) than workers who had never been monitored. This suggests general health differences [Wing et al., 1991]. Selective occupational exposure also occurs because occupational exposures accumulate gradually over many years or decades. Only workers healthy enough to remain employed for many years generally reach higher dose levels, while workers who leave employment early due to illness or other reasons generally have lower doses. Workers exposed to radiation may also be exposed to chemical carcinogens, and they may have different smoking or dietary patterns.
Measuring Exposure and Disease
A second major problem in the technical practice of radiation epidemiology is dose measurement. It is important to correctly categorize individuals into groups based on their dose in order to avoid under- or over-estimating an association. Even if disease rates increase with dose, the increase cannot be detected if enough people are incorrectly categorized. This situation leads to "false negative" studies.
Dose estimates for A-bomb survivors were derived from models which consider the amount and energy of radiation released from the bombs and interview data on location and shielding collected five or more years after the bombing. The accuracy of physical models is called into question by several recent changes in estimates of the amount and types of radiation released [BEIR V, 1990; Straume et al., 1992]. Dose misclassification would also result from the use of retrospective survey data used to locate individuals at the time of the bombing.
Workers at some Department of Energy facilities have been issued personal dosimeters to monitor external penetrating radiation exposures, a seemingly ideal measurement situation. However, changes over time in who was monitored, the sensitivity of dosimeters, and the frequency of reading dosimeters can affect the reliability of recorded doses. In the early years, dosimeters were read daily or weekly to help quickly identify workers with higher exposures. But frequent reading may not allow dosimeters to be sufficiently exposed to reach the detection threshold. Doses well below exposure standards have not been of regulatory concern, but are of epidemiological interest, especially when they are accumulated over many years. Other errors occur because of difficulties in matching hundreds of thousands of dosimeter readings collected over many decades to thousands of workers, failure of workers to wear the correct badges, equipment errors, and variation in reading instruments [Wing et al., 1994].
Personal dosimeters do not detect radiation dose from internally-deposited radionuclides. Estimates of doses from internally-deposited radionuclides are made based on information about the solubility of the compounds, the amount excreted in urine and feces, and values for internal transport and residence times derived from models based on clinical studies. Estimates can also be made using whole-body counters that detect the penetrating radiation emitted by the internally-deposited particles. Still, most epidemiological studies of nuclear workers have not quantified internal doses. In any case, the estimates used to classify individuals' doses, as is necessary in an epidemiological study of the exposure-response relationship, are based on a variety of assumptions. This means that quantification of dose-response relationships is speculative and leaves many unanswered questions.
The ability of an epidemiological study to quantify radiation risks also depends on measurement of the outcome. Radiation epidemiology has focused on outcomes such as cancer and major birth defects which are easier to count than some other illnesses. Studies using cancer as an outcome measure usually rely on cancer mortality rather than cancer incidence. Reporting of deaths is legally required and death certificates are gathered in a central registry, but not all states have tumor registries. Death certificates often fail to yield information on cancers that are in remission, unrelated to the primary cause of death, or undetected at the time of death from other causes. The poor quality of death certificate diagnoses remains a problem [Jablon et al., 1990].
What to Measure?
A more fundamental problem in dose-response assessment is the inadequacy of the theoretical basis for knowing what to measure. It is uncertain which aspects of a dose need to be quantified in order for a study to be sensitive to complex radiobiological effects. Among workers exposed to penetrating ionizing radiation over long periods, for example, the total cumulative dose over a worker's employment history is typically studied. Sometimes only the doses received up to a certain number of years in the past are considered in forming exposure groups. These "lag" or "latency" analyses are based on the assumption that cancers take time to develop and that recent exposures are not relevant to disease.
Alternatively, doses received in the distant past might not be etiologically important. The doses that should be counted might be those accumulated around the time of the emergence of the hypothetical mutations leading to radiogenic cancer [Pearce, 1988; Stewart and Kneale, 1993b]. Then again, it might not be the cumulative dose that is critical, but whether the dose is delivered in one or a few short time periods, or is drawn out slowly. Chronic exposures might have a greater opportunity to impact an organism during especially susceptible states, or, in some systems, could allow defense mechanisms to operate. Other aspects of dose that might be important to measure are the peak dose or the coincidence of radiation with other carcinogens or susceptibility states.
Another difficulty in interpreting radiation-cancer associations is that the mechanisms of radiocarcinogenesis are not well understood. Also, there is increasing evidence to suggest that there is variation in the extent to which different cancers are radiogenic. Unlike difficulties of lack of comparability and measurement that may be substantially reduced in experimental settings, the problem of measuring the right thing affects the controlled experiment just as seriously as the observational study.
Epidemiologists know well the problems discussed above of comparability, sometimes called confounding and selection factors, and measurement errors that distort dose-disease relationships. The solutions are to improve measurement, to select study subjects in a way that makes them more comparable, and to statistically adjust for remaining sources of non-comparability that can be identified and quantified. Refinement of epidemiologic method has occurred, and the field has contributed to knowledge about many pathogenic agents, including ionizing radiation. Epidemiological techniques are well-suited to documenting strong risk factors, such as regular cigarette smoking or high-dose ionizing radiation, that show little or minor variation in impact in various population subgroups.
Epidemiological methods, however, are not well-suited for assessing radiation health effects when doses are low and measurements are poor. Relatively small differences in disease occurrence, such as those that are suspected in the case of many environmental radiation exposures, are difficult to detect [McMichael, 1989]. But small increments in disease incidence can have a great population impact when many people are exposed [Rose, 1992]. These are the very situations that are often of most concern to the public and most commonly seen by clinicians.
Other modules in this monograph summarize the research findings of many radiation epidemiology studies. Ironically, studies of downwinders, whose exposures are the focus of the Hanford Health Information Network, are among the most ambiguous of radiation epidemiology studies. The ambiguity derives from all the generic problems reviewed above. Factors contributing to ambiguity include: migration; the long delays between time of exposure and manifestation of a health effect such as cancer; and lack of measurements of the exposures of interest.
Conclusion
Epidemiology's major contribution to the understanding of radiation health effects is the identification of "late effects" of radiation exposure in human populations. Epidemiological studies have observed long-term differences in disease, primarily cancers, associated with radiation exposures. Epidemiological studies are important because of the uncertainties involved in extrapolating health effects from animals to humans, especially when the latency period between exposure and clinical appearance of disease exceeds the life span of most experimental animals.
However, epidemiological studies have important limitations. They generally do not address the reasons for any specific case of disease. Many serious conditions have received little or no attention. Epidemiology's ability to identify an association between exposure and disease is sensitive to the knowledge about which aspects of exposure to measure, the quality and completeness of the exposure and disease measures, and the comparability of the groups being studied. Low-level effects are especially difficult to detect, and there has been relatively little attention to possible subgroups that may be more sensitive to radiation than other groups. Although there has been much excitement about new molecular methods in epidemiology, these new techniques have promised much more than they have delivered [Pearce et al., 1995].
There is often a demand for epidemiological studies when populations have been exposed and there is public concern about such problems. Such studies can be helpful in documenting relationships between specific agents and disease rates when good measurements of both can be made and when sufficiently large numbers of people can be studied. However, when exposures are low and the disease of interest can arise in the absence of exposure, measurement quality must be high and large numbers of people must be studied. In the case of Hanford downwinders, individual doses are especially difficult to document due to uncertainties about what was released; chaotic and complex environmental distribution; and variability in individuals' diets, home, work, physical activity, and biological processing.
Large and expensive historical epidemiological studies may compete for funding with environmental clean-up or clinical services, and such studies may not have the power to detect existing associations. In some cases it has been argued that such studies are funded for the very reason that they are unlikely to detect effects [Sterling, 1980].
As in all areas of science, the construction and interpretation of evidence from epidemiological studies reflects the social context in which the research is produced [Wing, 1994]. Political and economic interests in nuclear industries, including military, energy, and medical uses, have created especially obvious social influences on studies of radiation health effects. However, ionizing radiation exposures are only one of the many important ways that radiation-producing industries affect public health. Health professionals, scientists, policy makers, and activists should consider radiation exposures in the context of these more global issues.
REFERENCES
-
Committee on the Biological Effects of Ionizing Radiations. Health Effects of Exposure to Low Levels of Ionizing Radiation (BEIR V). Washington, DC: National Academy Press, 1990.
Jablon S, Thompson D, McConney M, Mabuchi K. Accuracy of Cause-of-Death Certification in Hiroshima and Nagasaki, Japan. Ann NY Acad Sci. 1990;609:100-109.
McMichael AJ. Setting Environmental Exposure Standards: The Role of the Epidemiologist. Int J Epidemiol. 1989;18:10-16.
Nussbaum R, K�hnlein W. Current Perspectives on Low-Dose Health Effects of Ionizing Radiation: Consistencies, Discrepancies and Open Questions. Env Health Persp. 1994;102(8):656-667.
Pearce N. Multistage Modeling of Lung Cancer Mortality in Asbestos Textile Workers. Int J Epidemiol. 1989;17:747-752.
Pearce N, Sanjose S, Boffetta P, Kogevinas M, Saracci R, Savitz D. Limitations of Biomarkers of Exposure in Cancer Epidemiology. Epidemiol. 1995;6(2):190-194.
Rose G. The Strategy of Preventive Medicine. NY: Oxford Press, 1992.
Schleien B, Ruttenber AJ, and Sage M. Epidemiologic Studies of Cancer in Populations Near Nuclear Facilities. Health Physics. 1991;61:699-713.
Sterling TD. The Health Effects of Low-Dose Radiation on Atomic Workers: A Case of Employer-Directed Research. International Journal of Health Services. 1980;10:37-46.
Stewart A, Kneale GW. A-bomb Survivors: Further Evidence of Late Effects of Early Deaths. Health Phys 1993a;64:467-472.
Stewart A, Kneale GW. The Hanford Data: Issues of Age at Exposure and Dose Recording. PSR Quarterly. 1993b;3:101-111.
Straume T, Egbert SD, Woolson WA, Finkel RC, Kublik PW, Gove HE, Sharma P, Hoshi M. Neutron Discrepancies in the DS86 Hiroshima Dosimetry System. Health Physics. 1992;63:421-26.
Wing S. Limits of Epidemiology. Medicine and Global Survival. 1994;1:74-86.
Wing S, Shy C, Wood J, Wolf S, Cragle D, Frome E. Mortality among Workers at Oak Ridge National Laboratory: Evidence of Radiation Effects in Follow-up through 1984. JAMA. 1991;265:1397-1402.
Wing S, West CM, Wood JL, Tankersley W. Recording of External Radiation at Oak Ridge National Laboratory: Implications for Epidemiological Studies. J Expos Assess Environ Epidemiol. 1994;4:83-93.TOP OF PAGE
