Potential Radiation Exposure in Military Operations: Protecting the Soldier Before, During, and After (1999)

Any employer has the ethical and legal obligation to provide care for employees who suffer harm as a result of their employment. This committee considers the government, with its management responsibility for military personnel, to be no different.

Although the law and associated regulations create a complex web of access to, provision of, and payment for health care, generally, the U.S. Department of Defense (DoD) provides medical care for active-duty personnel and retirees, whereas the U.S. Department of Veterans Affairs (VA) assumes the role for health care in certain circumstances after discharge from active duty. The current list of exceptions to complete coverage of health care responsibility, such as for the National Guard or the National Reserves when they are not actively deployed, is expected to grow as DoD moves to outsource much of its medical care and VA changes its eligibility criteria.

As a result, the locus of follow-up coordination—the focus of this chapter—is not uniform. The committee recognizes, with concern, that to actively ensure that the various government agencies provide appropriate care, including follow-up, for military personnel and veterans, a federal authority broader than that of either DoD or VA alone is required. A unified and comprehensive surveillance system that has access to and that uses preexposure and postdischarge outcome data is also necessary.

Some issues clearly affect the identification of long-term health effects. Some of these have already been discussed (e.g., good dosimetry and availability of records). A number of additional issues must also be recognized as important. For example, a bias may well occur when an active-duty soldier does not report an illness for fear of losing his or her military employment with a medical discharge. This will result in DoD assuming that there may not be a problem (or

that it is small) when the problem is actually significant. On the other hand, once a person is discharged and a pension is available because of a disability, there may be a tendency to overreport and overestimate potential problems.

Throughout this report, the committee has raised the issues of measurement and recordkeeping. These activities are aimed at identifying potentially hazardous situations, preventing or limiting exposure, measuring and documenting exposure, providing information for diagnosis and treatment, and learning from experiences. In Chapter 2, the committee described the factors that are known to determine the type and magnitude of health effects of radiation, such as dose, tissue, and sex. The immediate psychological impact and the subsequent psychological and psychosomatic effects of having been at risk of or exposed to radiation have a different set of determinants that are incompletely understood. These effects are not unique to radiation and may also pertain to exposures to other toxic substances.

It is important to identify the postexposure effects associated with radiation for two distinct purposes: (1) to help the individual medically or in claims of the individual or family members for compensation and (2) to form the basis of knowledge that might be used to prevent future harms to others. In this chapter, the committee considers the follow-up of persons after a known or suspected exposure to ionizing radiation while in military service for (1) medical purposes and (2) epidemiologic purposes.

Medical follow-up is intended to assist the exposed individual directly. It involves one or more of the following actions, depending on the situation and the needs of the individual at a given time after a known or suspected exposure. In this chapter, the committee uses the following definitions with respect to prior known or suspected exposures to radiation:

• medical assessment: early postexposure basic health evaluation;
• medical monitoring: the screening of asymptomatic populations;
• medical testing: the testing of an individual when judged to be necessary by a clinician on the basis of a clinical examination, history, and risk factors; and
• medical care: the management of clinically apparent injuries, diseases, or conditions.

Although there may be a concurrent or future benefit to the individuals involved, the purpose of epidemiologic follow-up is to identify deviations from normal health parameters among defined groups of people over the short and the long term. Using defined populations—such as all personnel deployed to a particular military operation or a subset of personnel who had been at risk of exposure to a specific agent (e.g., radiation or a vaccine)—epidemiologists try to identify or confirm and quantify associations between exposure (e.g., radiation dose, deployment, and personal behaviors) and health outcomes (e.g., specific illnesses, causes of death, or health care use).

Medical assessment may be defined generically as the early evaluation of an individual's basic health parameters in response to an acute health episode (e.g., severe chest pain or difficulty in breathing) or an unintentional exposure to a potential health hazard (e.g., a fall with a head injury or inhalation of toxic fumes). The purpose is to obtain data for comparison against previously recorded normal values for the individual or established norms for the general population and as a baseline against which the individual's future measurements can be evaluated for the assessment of progress. The assessment thereby provides a basis for diagnosing or ruling out acute conditions, patient management, establishing a prognosis, and future follow-up. Additional parameters are evaluated as indicated by the nature and circumstances of the exposure.

In the context of this report, medical assessment is defined as the evaluation of basic parameters of general and radiological health status after a known or suspected exposure to radiation or radioactive contaminants. Such an evaluation may be prompted by the development of nonspecific symptoms or trauma (e.g., nausea or blunt-instrument injury) or other detriments to an individual's performance during a military operation that carried a risk of exposure to radiation. Personnel are not likely to develop symptoms of acute radiation exposure at the dose range considered in this report (50 to 700 millisievert [mSv]); however, medical assessment is recommended after personnel exit areas of such potential exposure. The purpose of the assessment of asymptomatic individuals in these situations is (1) to rule out higher than expected doses, (2) to obtain baseline clinical data to assist in estimating the individual's radiation dose, and (3) to establish a basis for recommendations regarding the individual's need for medical care, periodic monitoring, or specific testing. This early postexposure assessment should be conducted by established protocols (Saenger, 1990; Voelz, 1990).

Medical monitoring, as defined in this report, is systematic screening of a population of asymptomatic individuals for preclinical disease with the purpose of preventing or delaying the development and progression of chronic disease in those individuals. Medical monitoring differs from both medical care of existing conditions and follow-up for purposes of epidemiological evaluation.

As early as 1922, the American Medical Association endorsed routine physical examinations for the general population to reveal current and prevent future illnesses. This approach, along with the use of multiphasic testing, yielded little new information or served to confirm already diagnosed illnesses. Therefore, in 1983, the American Medical Association issued a policy statement with-

drawing its support for the standard adult physical examination. Canadian and Australian authorities have reached similar conclusions.

Similarly, medical monitoring after radiation exposure is not routinely suggested or practiced for individuals with known or suspected exposures to radiation. An exposure or a presumed exposure to radiation is not by itself sufficient to justify a medical monitoring program. The decision about whether a medical monitoring program is appropriate and necessary in a given situation should be based on consideration of a number of factors including a rigorous cost-benefit analysis. This analysis should take into account the following characteristics: (1) the exposure of concern (e.g., its certainty, dose, and temporal relationship of exposure to observation), (2) the disease of interest (e.g., its natural history and prevalence in the population), (3) the characteristics of the available screening tests (e.g., their effectiveness, sensitivity, and specificity), (4) the potential for the tests used to themselves cause harm; (5) the potential for action when test results are positive (e.g., the availability of and risks from follow-up evaluation), and (6) whether there is evidence that an intervention can improve the clinical outcome. In this report, the committee considers these and other issues of concern associated with medical monitoring programs in general and, specifically, as they relate to persons exposed to radiation.

Because the effective dose range of interest for this report—50 to 700 mSv—is unlikely to cause delayed acute or chronic deterministic effects, the committee concentrates its discussion of medical monitoring or screening on malignant disease, which is the main long-term effect of radiation exposure.

Observations and research over the hundred years that radiation has been used and measured have identified certain malignant diseases that can be induced by radiation as well as by other known and unknown agents. Those malignant diseases that have been associated epidemiologically with prior radiation exposure are termed radiogenic; they include leukemia (all types except chronic lymphocytic leukemia); cancer of the female breast; cancers of the lung, stomach, thyroid, esophagus, small intestine, colon, liver, skeleton, central nervous system, and ovary; nonmelanoma skin cancer; non-Hodgkin's lymphoma; multiple myeloma; and cancer of the salivary glands (National Research Council, 1990). However, the government's approach to medical follow-up of potentially exposed individuals is based not only on scientific knowledge but also on the sociopolitical realities of veterans' concerns and congressional responses to them. Thus, it should be recognized that not all the health conditions identified as compensable under current laws and regulations have been associated scientifically with exposure to specific agents. Until September 1998, VA regulations

(CFR, 1998d) identified 22 conditions as radiogenic; most, but not all, of them are malignant diseases. They are

• all forms of leukemia except chronic lymphocytic leukemia
• cancer of the thyroid, breast, lung, bone, liver, skin, esophagus, stomach, colon, pancreas, kidney, urinary bladder, ovaries, salivary gland, rectum, brain, and central nervous system; multiple myeloma; and lymphomas other than Hodgkin's disease;
• posterior subcapsular cataracts;
• nonmalignant thyroid nodular disease; and
• parathyroid adenomas.

Effective September 24, 1998, VA added both the broad category of ''any other cancer" and the specific diagnosis of prostate cancer to this list of radiogenic conditions, despite the weakness of current scientific evidence for this conclusion. Under this regulation, any veteran who has a diagnosis of a disease identified in regulations as radiogenic and who can document a history of prior military exposure to radiation has access to VA medical care for the condition, provided that it clearly is not due to another (nonradiation) cause. This particular eligibility for care is not dependent on an officially adjudicated service-connected status. A more restricted list of radiogenic malignancies,* however, is defined in law for the designation of service-connection, which provides financial compensation and broader access to health care services (CFR, 1998c).

Separate from the consideration of government benefits, a number of investigators have discussed the principles for cancer screening in general. Taplin and Mandelson (1992) suggest a series of steps beginning with evaluation of the existing epidemiology literature in terms of the normal incidence of the disease of interest. It is not reasonable scientifically to screen for a disease that is extremely unlikely to occur as a result of a given exposure. If, for example, 100,000 people were exposed to a radiation dose that was estimated to increase the risk of developing cancer by one in a million, less than one additional case of cancer would be expected to result from that exposure. Screening of that population would not yield useful results. Screening is done for cervical cancer, which is diagnosed in 6 of 100,000 U.S. women annually, and for breast cancer, which is diagnosed in 85 of 100,000 U.S. women annually.

The justification for a proposed screening or monitoring program can be assessed by considering the normal incidence rates and comparing these to the

excess number of cases expected as a result of some exposure. Consider the case of a disease that spontaneously occurs at an annual incidence of about 25 cases/100,000 population at age 50 but whose incidence rises to 50/100,000 at age 55. Excess cases induced by some toxic exposure would justify monitoring only if monitoring was also justified (and performed) for the increase (25 cases/100,000 population) that occurred spontaneously.

The latent period between radiation exposure and the development of a clinically detectable tumor may have an effect on the design of a screening program. In the case of military exposures, soldiers are usually between 20 and 40 years of age when they are exposed, and most radiation-induced tumors would be expected to begin to become clinical evident when they are older than age 40, and in most cases older than age 50. Since most cancers occur spontaneously at older ages, Berg (1991) has looked at cancer screening of a nonexposed general population over the age of 50. For such a population, he recommends periodic physical examination of the breast, mammography, a Pap test, physical examination of the skin, flexible sigmoidoscopy to 35 cm, and oral examination.

Recommended screening tests for cancer change with time as randomized clinical trials are completed and as technology develops. Probably the best comprehensive source of current information and guidance is the report of the U.S. Preventive Services Task Force (1996). It is of interest to note that most of the more than 50 screening interventions reviewed in the 1996 edition had insufficient evidence of effectiveness to warrant a U. S. Preventive Services Task Force recommendation.

Since actions are taken or are not taken on the basis of screening test results, that false-positive and false-negative results can and do occur must be considered when planning a test program. The U.S. Agency for Toxic Substances and Disease Registry (ATSDR), which is charged by statute with evaluating the need for medical monitoring programs at Superfund Sites (sites subject to cleanup of hazardous materials, including radioactivity), has developed criteria for the establishment of medical monitoring. These are designed in recognition of the serious consequences that can result from both false-positive and false-negative test results. ATSDR has also addressed the psychological consequences of false-positive results.

The prevalence of the disease of interest in the population has an effect on screening test accuracy. When a test is performed with a symptomatic population, the prevalence of the expected disease is reasonably high. However, in the screening of an asymptomatic population, the probability that the disease is actually present is low. As an example, if the test is being used with a population of 10,000 persons with a disease prevalence of I in 10,000 and the test has a 5 percent false-positive rate, there will be 501 positive results, of which I will repre-

sent true disease and 500 will be false-positive results (a positive predictive value of 1/501, or 0.2 percent). The use of more than one test further reduces the positive predictive value. Even if the prevalence of disease in the screened population is quite high (e.g., 1 percent), the positive predictive value of a one test screening program rises only to 16 percent.

Even with the availability of an accurate test, there must be a demonstration of the benefit of early detection. There must also be a lead time during which a tumor can be found as a result of monitoring before symptoms occur. If the patient presents with symptoms at the same time that the test becomes positive, then periodic testing will be of no benefit.

The availability of a sensitive and accurate test that detects a tumor before symptoms occur still is not sufficient reason to justify the use of such a test to monitor the health of a population. There must also be an intervention or a therapy that is effective, available, and acceptable to the patient. A number of screening programs have found smaller tumors in high-risk populations (e.g., chest x rays of smokers), but the mortality rate was unchanged, probably because the tumor had already spread to distant sites in the body. As a result, chest x rays are not recommended for monitoring or screening even of smokers, who are at 5 to 10 times higher risk for lung cancer than nonsmokers.

Randomized trials using the screening tests must show a benefit. The benefit can be measured in a number of ways. Commonly used parameters are the percentage of people who are cured or the percentage of fatalities that are averted. More difficult to measure—and therefore less desirable as study endpoints—are a decrease in years of life lost or an increase in quality of life remaining.

Finally, effective use of a test depends on the clinician's sufficient understanding of the test to know the appropriate interval for repeat testing, as well as the costs and risks of the test.

The International Agency for Research on Cancer (IARC, 1990) has pointed out that screening costs to be considered should include not only the financial cost of the initial medical actions but also the

• cost of intensive follow-up for false-positive results,
• emotional cost for false-positive results,
• cost of delayed diagnosis due to false-negative results,
• extension of period of morbidity for those in whom early detection does not improve survival,

• unpleasantness of screening test (e.g., colonoscopy), and
• risk from screening (e.g., mammography).

An example of the major psychological costs (Wardle and Pope, 1992) associated with screening programs involves mammography. Mammography has rates of false positivity of 70 to 80 percent, so three of every four women who test positive must have a biopsy or surgery—with the accompanying physical risk and psychological fear—before they learn that they do not have a malignancy.

There are situations when risk is low (and monitoring the general population is not warranted) but a monitoring program might be justified for selected subgroups (Fearon, 1997; Perera, 1997). Such subgroups might include those who are genetically susceptible to a particular disease such as cancer. Relative to radiation exposure, the predominant general factor that appears to affect radiation sensitivity to a number of cancers is age at the time of exposure (with more risk per unit dose at the very much younger ages, which is not a factor for military personnel). Sex is also related to the incidence of cancer following radiation exposure: females have a slightly higher risk per unit dose than men due to the occurrence of breast and thyroid cancers.

At present, genetic testing is only beginning to be used, and its ramifications are not clear (Ponder, 1997). The issues of efficacy of intervention, test cost and accuracy, and disease prevalence considered throughout this chapter also apply to genetic testing. At present genetic testing is used only in the clinical management of families with well-defined inherited cancer syndromes.

Although certain types of leukemia and some cancers are generally accepted as having a scientific basis for their designation as "radiogenic cancers," to date screening programs have been shown to effectively reduce mortality only for cancers of the female breast and colon among this group of potentially radiogenic tumors. Although the Pap smear for the early detection of cervical cancer has proven to be highly successful in reducing the rate of mortality due to this cancer among women, the cancer's association with exposure to radiation is equivocal. The Pap smear is therefore unlikely to be useful for the detection of potentially radiogenic cervical cancers. The same may be said for prostate cancer, but prostate cancer is mentioned here because it has been added to regulations governing the VA's list of radiation-related conditions.

A medical monitoring program for asymptomatic persons exposed to radiation must take into account a wide variety of major factors before it is instituted. The major long-term effect that one might find after exposure to radiation in the dose range of 50 to 700 mSv is cancer. The risk of cancer is high even in nonexposed populations, and few tests have been shown to be of benefit in terms of improving either survival or quality of life. Those that have been endorsed include the Pap smear and mammography. However, the incidence of radiation-induced tumors among those exposed to the dose range of interest would almost always be less than the normal spontaneous incidence. If a monitoring or screening test is developed and an effective therapy is available, it is the spontaneous cancer risk (not the radiogenic cancer risk) that should drive a decision to do monitoring. It is theoretically possible that a test may be developed that could assess radiation-induced genetic damage likely to lead to malignancy. If such a test were developed it could prove useful. None of the above should prevent symptomatic persons from receiving appropriate diagnostic tests.

Although a particular diagnostic test may not be indicated when it is applied to an asymptomatic group of persons, in select situations the value of the test can be improved significantly in terms of specificity and sensitivity by clinical examination, history, and evaluation of risk factors. A familiar example is that of testing for human immunodeficiency virus (HIV) infection. It is unreasonable to test the general population for HIV. It would be useful, however, to test an asymptomatic medical worker who was stuck with a needle that had been used on a patient diagnosed with AIDS. All of these situations need to be assessed individually.

The radiation situation is more complex, but examples can be given. If a 35-year-old female presented with a solid palpable lump in her breast and the examining physician knew that she had received a high radiation dose in a military operation, a mammogram or an aspiration needle biopsy may be ordered. Without the high-dose radiation history, the physician may have elected to do an ultrasound or wait and not do any diagnostic procedure. If the clinical information was that the lump appeared within 5 years of the radiation exposure, the physician would also not have ordered the tests since the risk of radiogenic breast cancer is very small or zero at only 5 years since the exposure.

On the basis of the risk from the dose range considered in this report (50 to 700 mSv) and the lack of effective screening tests for neoplasms such as leukemia, radiation exposure should not play a significant role in the decision to test individuals.

Medical care following exposure to ionizing radiation concerns the management of the early and delayed deterministic effects resulting from doses above threshold levels, such as radiation-induced injuries to skin and bone marrow depression, and the management of stochastic effects, primarily nonspecific tumors that may become clinically evident years after exposure to radiation (see Chapter 2). With the dose range that the committee is considering, the greatest risk is of the appearance of benign and malignant tumors years later. However. because of the uncertainty of the dose that may be encountered in hostile situations, brief consideration of the deterministic effects that will appear within months of certain types of acute exposures to radiation above the threshold levels is required.

As described in Chapter 2, it is unlikely that symptoms of deterministic effects will appear in the absence of acute whole- or partial-body doses of less than I Sv (100 rem) of penetrating radiation. Early evidence of acute radiation-induced cellular injury, for example, structural changes in the chromosomes of some circulating lymphocytes, and falls in the absolute lymphocyte and sperm counts, is, however, clinically detectable in asymptomatic individuals who received lower doses.

Examples of scenarios in which soldiers may become involved with a risk of exposure to radiation within the 50- to 700-mSv range include (1) responding to a nuclear reactor accident, (2) securing a negligently or deliberately abandoned sealed radiation source, or (3) containing radioactive materials exposed to the environment, as may occur if a nuclear waste dump is disturbed. Such events could occur in the course of normal peacetime duty on friendly territory, on hostile or nonhostile peacekeeping missions, or as the result of terrorist actions. In these instances, exposures may be acute or chronic, they may involve nonuniform irradiation resulting in high doses to specific areas of the body, they may occur alone or with radioactive contamination, and they may occur with or without trauma or other injuries or illnesses.

In evaluating the effects on health of radiation released by the detonation of nuclear weapons or the dispersion of nuclear materials, DoD has concentrated its preparedness planning and extensive research efforts on the acute deterministic effects of radiation, including the acute radiation syndrome and the associated bone marrow depression. Events in which these types of radiation-induced injuries have occurred have been documented and reviewed extensively and have been presented together with the current recommendations for evaluation, medical care, and follow-up of exposed individuals (Mettler et al., 1990; Reeves et al., 1998).

Expression of acute radiation injury in some cell systems is delayed for weeks or months after an acute exposure to penetrating radiation above threshold levels (see Chapter 2). A period of transient male infertility may follow exposure to doses beginning at the upper end of the 50- to 700-mSv range. After higher but sublethal whole-body doses, males will experience a low sperm count with a nadir at about 45 days postexposure or an absence of sperm postexposure for a period that is directly proportional to the dose. Females may experience a period of amenorrhea after acute radiation exposure. Several sievert to the gonads are required to cause permanent sterility in previously fertile males and females of reproductive age; thus, sterility will not be a problem for individuals at risk of doses in the 50- to 700-mSv range (IOM, 1995). The threshold doses for the typically delayed (for weeks or months) expression of acute radiation injury to other tissues—such as the endothelial cells lining the blood vessels and connective tissue and their replacement by fibrous tissue (i.e., fibroatrophy), the optic lens (cataract), and the thyroid gland (thyroid hypofunction)—also are considerably higher than the range of the 50 to 700 mSv that is of interest for this report; thus, there will be no indication for specific medical care for soldiers with such exposures.

As noted previously, the primary stochastic or late effect of exposure to radiation is the development of radiation-induced tumors of types that are not caused only by exposure to radiation; they may be benign or malignant. It is assumed that the probability that such tumors will become clinically evident among a population some years after exposure to radiation above background levels is directly related to the dose. Their occurrence in an exposed population may be observed as an increase in the rates of occurrence of specific tumors above the rate for the spontaneous occurrence of tumors among the nonexposed population, beginning at ages at which the rate of occurrence of spontaneous tumors begins to increase. Radiation-induced or radiogenic tumors are histologically and clinically indistinguishable from spontaneously occurring tumors. Their diagnosis, treatment, and management are the same as those for spontaneously occurring cancers of the same type.

When the military knows that its soldiers have been or might have been exposed to agents that could produce long-term effects, it has an ethical obligation to notify them of this fact and to inform them of any new information concerning their exposure or ways to minimize its health effects. Once the military complies with these two obligations, an ethically responsible follow-up program would

provide for reasonable health screening to try to detect damage, and if an effective treatment is available, that treatment should be provided as well. If medical monitoring would be of net benefit, it should be done. However, the committee agrees that routine monitoring programs established specifically for persons with known or suspected exposure to radiation would not be useful at this time, given the limitations of current cancer screening programs. The follow-up obligation is directly applicable to soldiers who were at risk of exposure, and for whom the increased risk of long-term adverse health effects is known or reasonably suspected. There is no legal or ethical obligation to follow-up nonexposed soldiers for exposure-related health effects since there is no reason to assume they would be at special risk of harm.

In those instances when government-initiated follow-up is appropriate, the current organization of health care services in the United States significantly complicates adequate follow-up. In recent years, only about 10 percent of current veterans receive care from VA, in part because of eligibility requirements and access to other sources of care. Were the government to uncover information that would be of interest or of importance to veterans, how would it communicate this to the relevant veterans? The Departments of Defense, Veterans Affairs, and others with related responsibilities may want to develop policies and procedures to govern proactive contact of veterans. These might cover logistics and ethical issues such as consent and secondary uses of available data.

All the medical follow-up processes described in the preceding sections involve direct contact with individuals who may have been exposed to radiation. Epidemiologic follow-up is based primarily on the records for groups of those individuals. Epidemiologic studies seek to identify the distribution and determinants of disease among human populations by comparing groups that have some experience or exposure, such as radiation, in common. Although such research may benefit the individuals studied, it contributes primarily by increasing scientific understanding of the relationships between exposure and subsequent health outcomes.

Epidemiologic follow-up of a group of persons known or presumed to have been exposed to a potentially hazardous agent may be implemented to

• identify adverse health effects in an at-risk group and to determine whether the risk of such effects is greater than that for a comparable but nonexposed group of individuals,
• determine whether the increased risks that may be identified are associated statistically with the exposure,

• determine whether the increased observed risk is related to or influenced by other factors associated with or independent of the exposure, such as tobacco smoking and radon, and
• add to the scientific knowledge base, which can then be used to derive and refine risk estimates and to develop interventions.

In the circumstances considered in this report, base and field commanders could use the information obtained from epidemiologic follow-up studies to weigh the costs of different potential exposure scenarios.

Epidemiologic follow-up studies may describe a disease situation in a defined group at a specific point in time (cross-sectional prevalence studies) or may collect information about group members over an extended time period (longitudinal studies). In prospective longitudinal studies, a defined population (or cohort) that has a common experience or exposure is followed forward in time to determine if there is an increased risk of disease among this cohort relative to that among a comparable nonexposed cohort. Alternatively, groups of individuals with and without a specific disease, condition, or cause of death can be compared retrospectively, using recorded data, to determine if the risk of exposure was greater in the diseased than in the nondiseased group.

The planning and implementation of epidemiologic research involve many practical concerns (IOM, 1995), including the

• availability of a clearly defined and appropriate study population with unique individual identifiers;
• size and composition of the study population;
• completeness (and lack of bias) with which study subjects can be enrolled;
• magnitude and distribution of exposure to the hazard being studied;
• accuracy—including the unbiased collection of data and adherence to a defined time frame—with which the exposure can be measured (measurement of absorbed dose, as in the atomic bomb survivors, is extremely important since the most compelling evidence of causality is the demonstration of a dose-response relationship);
• accuracy—including the unbiased collection of data and adherence to a defined time frame—of disease identification (history of disease should be confirmed from hospital records, and causes of death should be determined by obtaining copies of death certificates);
• background rate of the disease being studied;
• expected increase in the incidence of disease among the exposed group;
• availability of information on other factors that might determine disease; and

• procedures to ensure valid consent for those research settings in which it is appropriate.

An epidemiologic follow-up study typically begins by identifying two groups of people—those exposed and those unexposed to the agent, treatment, or characteristic being studied—and then seeks to determine whether the groups experience different health outcomes. The choice of an outcome—the measure of health—affects study design, complexity, and feasibility.

Mortality is the outcome most conducive to an epidemiologic study because the occurrence is clearly definable, happens at most once per person, and relatively complete records are available. Mortality is not, however, always the health outcome of interest. Many questions involve diseases and conditions that affect the quality of life but that do not kill the individual. Physical and emotional health are often grouped under morbidity, yet concomitant employment, economic, and social well-being outcomes are increasingly being used as measures of effect. Finally, although not a direct measure of an individual's health, health care use—and its cost to the individual, the military, and other government agencies—is a reasonable choice of outcome for some epidemiologic follow-up studies. The study of each of these outcomes—death, illness, and cost—poses substantial challenges to the epidemiologist.

A robust study design includes a clearly defined and identified study population and assurances that adequate data (in terms of completeness and lack of bias) regarding those individuals can be acquired. All of the products of the exposure monitoring and recordkeeping activities that the committee discussed in earlier chapters of this report are available for use in epidemiologic follow-up studies. Assessing whether and to what extent potentially hazardous exposures (e.g., ionizing radiation) are present is complicated by the demanding conditions arising from the hazard itself, as well as by limitations associated with the devices used to quantify the exposure. Problems specific to the measurement of radiation exposure are discussed in Chapter 2 of this report. The quality of the exposure data tremendously influences the feasibility and usefulness of such studies.

In part because of the very limited existence of prospectively designed and funded epidemiologic studies, researchers often turn to available databases. Administrative databases and registries are prime examples. They can be very useful in the consideration of some questions, but they have severe limitations in many epidemiologic applications.

Because of the need for unbiased sample selection and exposure and outcome measurements, it is a great advantage for epidemiologists to chose the population to be studied and to measure prospectively the baseline characteristics (demographic, clinical, and risk factors) of the individuals in that population. Also, outcome information must be sought in ways that make all members of the population equally likely to be identified. Registry data are therefore not necessarily well-suited for use in epidemiologic analyses. The drawbacks of registry data include the following: (1) the individuals with adverse outcomes may be more likely to register; (2) even among those with adverse outcomes, only a subset will register; (3) the outcome is influenced by more than the putative exposure, and those confounding factors—such as access to medical care (diagnosis and treatment) or health-related behaviors such as tobacco use—are usually not recorded in registry databases; and (4) reports to registries are often associated with compensation claims.

Confounding factors, including disease-causing behaviors such as smoking and alcoholism, may obscure the relationship under study. This often is exacerbated when studies begin many years after the occurrence of an exposure and require many years to complete.

Measurement of exposures, outcomes, and possible confounding factors is further complicated by the availability and quality of event records (e.g., medical and dose records). Records may be poorly maintained, stored in decentralized locations, or discarded after a set time period. For example, in conducting a mortality study of military participants in Operation CROSSROADS, a nuclear test series done in 1946, researchers (Johnson et al., 1996) required records maintained by the VA's health and benefits components; DoD ship logs and morning reports, which are now stored in paper files in cartons at numerous facilities of the National Archives and Records Administration; National Personnel Records Center in St. Louis, Missouri (Army and Air Force records in that facility sustained heavy damage in a 1973 fire); vital records departments in 50 states and additional territories; and the National Center for Health Statistics National Death Index; among others. Record systems may also inconsistently document events in situations in which examiners, such as pathologists and physicians, do not use standardized diagnostic routines.

A major consideration in the design of a study and the analysis and interpretation of its data is statistical power. To determine whether there is a difference in outcome between an exposed and an unexposed population, each population must be large enough (sample size) so that normal variation does not dwarf any real differences. That sample size is determined by the prevalence of the outcome in the unexposed population and the level of uncertainty that the researcher is willing to take in terms of false-negative results (not finding an exposure-outcome relationship in the data when there actually is one).

Attempts to study the late health effects of exposure to radiation—assuming that malignant disease is the major concern—include other specific difficulties:

• no unique disease: all malignancies are pathological and clinically the same regardless of cause;
• population size and dose need to be large enough to be able to detect a statistically significant difference in risk between exposed and nonexposed populations;
• long interval between time of exposure to and occurrence of disease (latent period);
• dose uncertainties that may overwhelm the true dose at low dose levels; and
• confounding factors that may mask a radiation effect if there is one.

A complicating factor in studies of military-related exposures is that the authorities of the involved government agencies overlap. DoD and VA, along with the Department of Health and Human Services,* have some common responsibilities for the health of veterans. Federal legislation and the budget process, however, leave each of these agencies without the authority or funding to perform other necessary activities in support of veterans' health. Because ascertainment of health outcomes must be equally likely for all individuals in a study, this diffuse authority for follow-up can affect an epidemiologic study's time line, expense, complexity, and, ultimately, validity.

DoD and VA are steadily improving their automated records systems to allow sufficient follow-up over time. These will be a valuable source of data for those service members and veterans who seek all of their health care through those agencies. Most veterans, however, do not go to VA for health care, and those who do are predominantly those who have service-connected disabilities or who are eligible for coverage because of low income. Hence, any study findings limited to VA health care databases could not easily be generalized beyond those groups. Furthermore, it may be that both service members and veterans seek care for personally sensitive health care needs outside of the government systems for privacy reasons.

Administrative obstacles arise because a single agency is not responsible for all care provided by or paid for by the government. This relates to the choice of study population, which was mentioned earlier. An epidemiologic follow-up study poses a question and attempts to answer it by using data from the study

population. Study cohorts that have been selected according to different sets of criteria may yield different answers to the study question. Examples of different study populations include veterans (or their records) who actively respond to active VA or DoD requests for volunteers, veterans whom VA or DoD identifies passively through administrative records, and scientifically designed samples or groups of veterans with either a common exposure or a common adverse health outcome who are systematically followed for appropriate data collection.

Epidemiologic studies necessitate consideration of the privacy and confidentiality concerns associated with the use of personal records. Privacy refers to keeping sensitive information about oneself secret. Confidentiality refers more generally to keeping personal data from being used by others without informed consent (IOM, 1995). In the United States, federally mandated institutional review boards (IRBs) serve to ensure that researchers take adequate steps to preserve the confidentiality of the data they collect, requiring that they specify who will have access to the data, how and at what point in the research personal information will be separated from the data, and whether the data will be retained at the conclusion of the study. IRB reviewers also make sure that the informed consent of the subjects will be obtained before interviews are conducted (Wallace [1982] and OPRR [1993], as cited in IOM [1995, p. 20]).

There are two types of epidemiologic investigations. One is an experiment, in which the researcher exposes one group of individuals to a hazard or a vaccine and does not expose another group and then measures and compares the outcome in both groups. The other type of epidemiologic research is an observational study, in which researchers use data that are available from an operationally caused exposure not planned or influenced by the researchers. Both of these investigations require IRB approval. Different ethical rules may apply for certain kinds of observational studies, when, for example, anonymous or unidentifiable data are used. Whether adhering to the "common rule" or developing its own policies and procedures, the Army should follow ethically appropriate rules in all research. Despite the specialized context of military service, set privacy and confidentiality protections should be maintained.

The Institute of Medicine's Medical Follow-up Agency (MFUA) has evaluated a number of military veteran populations for potential late health effects as a result of exposures during military service. These include a published study of the mortality of veterans who participated in Operation CROSSROADS in 1946, the first postwar U.S. atmospheric test of nuclear weapons (Johnson et al., 1996),

and a study, which is now underway, of participants at five other test series in the 1950s.

In considering moving beyond mortality endpoints in studies of Atomic Veterans, MFUA convened an expert committee to explore the feasibility and potential design of studies of reproductive outcomes, a concern of many veterans. That committee's report (IOM, 1995) stated that it will be extremely difficult, if not impossible, to find and contact a sufficiently high and representative percentage of veterans' families, to establish a good measure of dose for each veteran, to identify and accurately document reproductive problems that occurred over a 50-year interval, and to measure other factors that cause reproductive problems and therefore might confound any observed relationship between radiation exposure and reproductive problems (IOM, 1995, p. 79).

In addition to the physical effect of radiation on tissue, psychological effects occur following real or perceived radiation exposure, and these have been studied in a number of populations (IAC, 1991). Most of these studies, such as those concerning Chernobyl, have been related to accidental exposures. Numerous other studies relate to both the measurement and the perception of risk in general (National Research Council, 1996). In this brief section of its report, the committee raises the issue of psychological effects, including stress; this report is not the setting for a complete discussion of the complex components of that subject.

Usually, the perception of the risk from radiation exposure is much greater than the actual risks described in the scientific literature. Much of the concern about radiation exposure is because of its unfamiliarity, the fact that it is related to a dreaded illness (cancer), and, depending upon the situation, the fact that exposure is nonvoluntary. Media coverage of exposure situations can amplify the psychological effects. When, for example, the media report exaggerated or false claims, the potentially exposed population becomes even more worried than they were initially, resulting in even greater media attention (Lee, 1996).

Since the accident at Chernobyl the psychological effects of the accident have been studied quite extensively. In a paper presented at the 1996 International Atomic Energy Agency Conference "One Decade After Chernobyl: Summing up the Consequences of the Accident," Lee writes that "[T]he main human legacy of the accident has been anxiety about health and a social disruption that has manifested in widespread health disorders not induced by radiation" (Lee, 1996, p. 285). The accident presented an unfortunate but unique test situation. Hundreds of villages were exposed to fallout, with the absorbed doses being at the lower end of those considered in this report (about 50 mSv). Due to the nonuniform nature of the atmospheric dispersion of the radioactivity, however, interspersed among those exposed villages were many villages that were not exposed to radiation. Although the two populations were significantly different in terms

of stress and anxiety, the absolute levels of stress and anxiety were high in both populations compared to what would be expected in general populations. Persons residing in unexposed villages reported a very high incidence of health complaints that they believed to result from the radiation exposure. For example, 45 percent of the people in contaminated villages indicated that they were sure they had an illness due to radiation, whereas 30 percent of persons in clean villages reported illnesses due to radiation (Lee [1996] describing an unpublished report by Drottz-Sjoberg et al. [1994]). Thus, a severe obstacle in studying health effects in Chernobyl was the lack of a clear definition of either contaminated areas versus noncontaminated areas or exposed persons versus nonexposed persons.

Initial reports from the former Soviet Union described a number of ill-defined entities including radiophobia, chronic radiation sickness, and vegetative dystonia (IAC, 1991). There was also the issue of whether these persons suffered from posttraumatic stress disorder (PTSD). PTSD is usually the result of witnessing a sudden catastrophic event (e.g., a battle, earthquake, or fire) that is over in a short period of time. It is manifested by intrusive recollections of the event and avoidance symptoms. Although Chernobyl firemen may have had PTSD, persons distant from the accident had symptoms inconsistent with the diagnostic criteria for PTSD, leading Lee (Lee, 1996) and others to propose a related but distinct entity—chronic environmental stress disorder.

Although the factors contributing to symptoms are complex (e.g., food restrictions, relocation, and financial incentives), the major international psychological studies invoke chronic environmental stress as the major etiology of these reported symptoms (IAC, 1991). Stress can generally be defined as adverse mental experiences that have negative effects on bodily functions; it can be measured by physiological indices (Lee, 1996).

Delayed and incomplete transfer of information from responsible authorities to potentially exposed persons has been a major cause of psychological stress in many radiation exposure situations (Lee, 1996). Transfer of information on the extent and magnitude of the risk or potential risks should occur before persons are exposed, but if this is not practical (as sometimes occurs in accidents or military situations) it should be done as soon as possible, depending on the nature of the circumstance (mission).

Stress can be alleviated in a number of ways (Lee, 1996). A straightforward strategy is to remove the stressor (e.g., decontaminate the area). This, however, is of little help to people who have already been exposed to radiation. A second alternative is to increase people's sense of control. This may include the implementation of specific medical procedures to help eliminate internally deposited radioactive materials, the institution of voluntary food controls, or the formation of community action groups. The third way to alleviate stress is diffusion of knowledge that changes the way that the radiation source and risks are perceived. This is important not only for the exposed persons and their families but also for the medical community, media, and other groups that are involved.

Particular attention should be paid not only to relieving the stress of the individuals involved but also to the social task of reestablishing support and mutual understanding between individuals. The social stigmatization of exposed persons—such as after the atomic bombings of Hiroshima and Nagasaki as well as after Chernobyl—results in prejudice and cuts off social contact and communication (Lee, 1996).