Printer Friendly

Challenges of practicing environmental epidemiology in the US military.

When service members return to their normal routines following the end of a war, some of them develop illnesses. These may be associated with heredity, personal risk factors such as smoking, normal aging, or the result of their occupational exposures. When someone develops a serious illness, it is common for them to review their past to look for a potential cause. This is human nature and part of the "why did this happen to me?" assessment. For many, their military deployments are likely to be considered. Deployments represent changes to routines such as eating and sleeping and lack of control over daily life, and may also include changes in hygiene and unpleasant exposures to things such as trash burning; old, abandoned warehouses and industrial sites; and sources of radiation. For many, a deployment may be the most unusual and noteworthy experience in their memory. Therefore, it is not uncommon for returning service members to ask their health care providers whether or not the deployment was a factor in their illness. Associations between exposures to burning trash, blowing sand (particulate matter), old chemical weapons, and subsequent illnesses have been considered following Operations Iraqi Freedom, Enduring Freedom, and New Dawn. Generally, few clear associations have emerged. The challenges of practicing environmental epidemiology in a deployed military force are identified and reviewed.


Epidemiological studies may be conducted to assess potential relationships between exposures during deployment and health outcomes. Epidemiology is a field concerned with the identification of factors associated with disease and development of public health measures for disease reduction. Environmental epidemiology is the study of the effect on human health of physical, biologic, and chemical factors in the external environment, broadly conceived. (1) It examines associations between exposures and outcomes to determine if an association places a segment of the population at increased risk of certain outcomes, usually a disease. The magnitude of the association measures how much increased risk of an outcome can be associated with the exposure. Avoidance of the exposure should then result in a proportionally lowered risk for the outcome. Thus, one aim of public health is to optimize health by limiting hazardous exposures. There are several different types of studies which may be conducted, the selection of which depends on the study question(s) being addressed, the frequency of the exposure(s), and outcome(s) of interest, as well as the expected strength of association between them, the current state of knowledge related to the topic at hand, issues regarding efficiency and validity, and matters of practicality and ethical considerations.

Cross-sectional Studies

A cross-sectional study attempts to assess exposure and outcome at the same time. It provides a snapshot view (cross-section) of disease patterns in the population of interest while assessing the degree of presumed or documented exposure. (2) The main difficulty with this type of study is that the "snapshot" captures only one moment in time, and by looking at exposure and outcome at the same time, it is not always possible to say with certainty that the exposure preceded the outcome. This is a noteworthy limitation of cross-sectional studies due to the fact that in order to draw conclusions about exposure-outcome causation, epidemiologists must be able to first establish correct temporal sequencing. Also, because cross-sectional studies are a snapshot, cases of illness can be counted only if present at the time of the look, which presents another disadvantage of this study design: it tends to identify prevalent cases of long duration and may miss capturing individuals who are lost to follow-up shortly after the outcome is established or acute outcomes among individuals who recover quickly. Advantages of cross-sectional studies include high generalizability and minimal resource requirements, relatively speaking, due to the fact that they can be conducted quickly.

An example of a cross-sectional study would be testing pulmonary function in service members exposed to particulate matter levels in the desert at the time of exposure. The aforementioned limitation would be highlighted if a unit demonstrated some impairment in pulmonary function, as it would not be possible to say for certain that it was due to in-theater particulate matter exposure. It is possible that some unit members had another type of exposure prior to deployment that influenced their pulmonary function. Individuals who were so affected that they had been evacuated from theater, or who were not onsite at the time of the testing would not be included.

Cohort Studies

Cohort studies assemble an exposed group and an unexposed group, follow them forward in time for the occurrence of health effects, and compare the rates in the 2 groups. This study type is suitable when the at-risk population is well-defined, the disease is common, or the exposure is rare or specific. When the at-risk population is known and there is a specific or rare exposure, it is easy to identify the exposed versus the unexposed. The advantages of this method are that accurate measures of exposure are possible before the outcome occurs, and that multiple outcomes can be monitored. A source of bias in these studies is that the knowledge of exposure status may alter the follow-up and subsequent diagnosis. Other limitations include the fact that many years may pass between exposure and outcome, which makes the study costly, increases the likelihood of losing people to follow-up, and is particularly impractical when looking at rare diseases or diseases with long latency periods, such as cancer. An unexposed population is used to provide a comparison rate for the health effects of interest. However, it is critical that the exposed and unexposed populations differ only with respect to exposure status, something that is not always easily accomplished. To alleviate some of the limitations, nonconcurrent or retrospective cohort studies are often performed. This type of study reconstructs exposed and unexposed groups based on exposure in the past. These groups are then compared with respect to the rate of the outcomes of interest. The chief advantage is that since exposure occurred in the past, sufficient time for the outcome to develop may have already passed. The disadvantage occurs when exposure is not well quantified at the time of occurrence. (3) For example, to study the effects of burn pit exposure on the respiratory system, health outcomes for individuals identified as having been at base camps with burn pits were compared to individuals who were at base camps without burn pits. One key limitation was that exposures associated with the burn pits had not been well characterized. (4)

Case-control Studies

The case-control method consists of identifying people with the disease of interest (cases), and people without the disease (controls). The unknown variable to be ascertained is the type, frequency, and duration of past exposure(s). Critical to the validity of a case-control study is careful selection of controls, in that they come from the same population as the cases and exposed and unexposed controls have the same likelihood of being selected for inclusion. (5) This type of study is useful when the population at risk is not well-defined, when the disease is rare and the exposure common, and although only one disease can be investigated at a time, multiple risk factors can be assessed. The major weakness of this method is that exposure is investigated after disease status is established. This leads to recall bias, where those who are diseased are more likely to remember and report exposures than are those in good health. Furthermore, accuracy can be particularly challenging when conducting case-control studies due to the fact that relevant exposures may have occurred many years prior to data collection, which can be difficult to ascertain and define correctly. To assess the potential association between deployment and the development of specific cancers, one approach would be to compare cancer cases with controls who have no such diagnosis. The main variable to be assessed would be whether or not the cases had a stronger history of deployment.


Given that these study designs each have strengths and weaknesses, other factors affect the ability of any given study to accomplish what it sets out to do--support or refute an association between an exposure and an outcome. Unfortunately, these limitations are common in environmental epidemiology studies. Illustrating this point, the US Department of Veterans Affairs asked the Institute of Medicine (IOM) to form a committee, its main charge being a determination of the association between exposure to burn pits during deployment in support of Operations Enduring and Iraqi Freedom and subsequent long-term health effects. (6) Although the IOM task was to respond to a specific exposure, the committee's conclusions regarding feasibility and design issues apply on a much broader scale. They identified the major challenges of any epidemiologic study to be both exposure assessment and outcome ascertainment. Regardless of type of study selected, the committee reiterated the elements characteristic of any well-designed epidemiological study, including selection of a relevant study population of adequate size, comprehensive assessment of exposure, careful evaluation of health outcomes, reasonable methods for controlling confounding and minimizing bias, appropriate statistical analyses, and adequate follow-up time.

Magnitude of the Disease Risk

If an exposure is strongly associated with an outcome, following a group of people who are exposed and comparing their incidence of illness with that of a group without the exposure allows one to calculate the excess risk which can be attributed to the exposure. This is the cohort study design. For example, nonsmoking physicians in England were compared with smoking physicians, and followed over time to count the number of lung cancers in each group. It became clear that cigarette smoking was associated with a risk of lung cancer, perhaps 10 times greater than that seen in nonsmokers. (7) In this study design, accurate information about the magnitude of exposure is collected for all individuals and all individuals should be followed for a period of time sufficient to allow the outcome to occur. If an exposure produces a risk which is 4, 5, or even 10 times greater than in people without the exposure, it is relatively straightforward to sort this out and quantify the risk. In contrast, lung cancer risk from passive smoke exposure is thought to be elevated to less than 2 times that of nonexposed individuals. (8) This small difference is difficult to detect. When an exposure is associated with only a slightly elevated risk, that risk sometimes "blurs into the baseline" upon investigation. This is due to the fact that in order to detect a slightly elevated risk, one must be able to distinguish it from the baseline, or "usual incidence" of the outcome without any exposure. This is the role of control groups. However, small risks require that the studies include large numbers of people so that the study has sufficient statistical power. Since large numbers of service members deployed in the most recent conflict, this may not be a problem, but might be if the outcome is particularly rare. Typically, the increased risk associated with environmental exposures is estimated to be small. It has been estimated that the probable range of increased risk for an outcome due to chronic, low level exposure is less than 2 times that of someone without the exposure. (8) This has implications for the number of people necessary in each group in order to detect a difference. This is especially so when competing exposures such as smoking may lead to a much higher elevation of risk, so the role of smoking would have to be thoroughly assessed.

Magnitude of the Exposure

Detection of effects associated with exposure is also greatly simplified when the magnitude of the exposure is great. Excess numbers of leukemias were easily detected following the exposure to radiation from atomic bombs dropped on Hiroshima and Nagasaki because the magnitude of exposure was so large. (9) High doses of radiation are strongly associated with leukemia as an outcome. In contrast, environmental exposures are classically small--doses lower than those known to produce effects in animal studies. Often the exposures may be chronic low-level exposures, occurring over weeks or months, but which may not have been quantified since they were not recognized. (10) Although some studies ask individuals about whether or not they were exposed, they do not always know. This can lead to misclassification bias, where it becomes difficult to categorize individuals as exposed or nonexposed, and thus difficult to make causal interpretations about their outcomes. One study evaluated health outcomes for service members who were present at the time of a sulfur mine fire that burned for several weeks in Iraq. (11) The "exposed" group included individuals within a 50 km radius of the site, based on satellite images of the plume and rosters of units in the area. However, some individuals may not have been present, and it was difficult to assess the level of exposure.

Dose-response Relationship

An axiom of toxicology states, "The dose makes the poison." The implication of this is that the larger the dose, the larger the effect. Associations between exposures and outcomes are strengthened when this relationship can be shown to be logical and uniform. This does not necessarily mean that the greater the exposure, the sicker the individual, but that more outcomes (eg, more cases of cancer) occur in the higher exposure groups as compared with those less exposed. For example, more lung cancers would be expected in groups of individuals who smoke 3 packs of cigarettes a day for 20 years versus one pack a day for 5 years. This introduces again the requirement for accurate, quantitative exposure data. Unfortunately, we can rarely reconstruct individual exposures in deployed settings. For example, the study looking at health outcomes associated with having deployed to a base camp with a burn pit was reviewed by the IOM for their report, Long Term Health Consequences of Exposure to Burn Pits in Iraq and Afghanistan. (4) One of their criticisms was that the study did not identify those with the highest exposures, such as those who guarded the burn pits, compared to those with lesser exposures. (6) However, this information on individuals was not available. In another example, the study of the health outcomes associated with those who were "under the plume" of the sulfur mine fire was not able to distinguish individuals who were at locations with higher levels of exposure versus those at locations which were briefly under the plume. (11) This is because the information on exact daily locations of individuals was not available. Efforts must be made to accurately measure contaminants at the individual level, but having sufficient sampling capabilities at the location of an exposure, particularly an unplanned one, is a daunting task.

Controlling for Confounding

Confounding occurs when an association between exposure and outcome is over or underestimated due to a third factor. It is best explained by example. Studies have shown that smoking causes an increase in pancreatic cancer. An investigator attempted to ascertain the relationship between coffee consumption and pancreatic cancer. He questioned cases with cancer and controls about coffee consumption and found that cases had a greater history of exposure to coffee. He concluded that coffee consumption was associated with an increased risk of pancreatic cancer. Critics challenged his conclusion by pointing out that those who drink coffee are more likely to smoke as well, and that by not questioning subjects about smoking behaviors and adjusting for it, he was in effect seeing the association between cigarettes and pancreatic cancer, not coffee and pancreatic cancer. (12) Cigarette smoking, the unmeasured "true" risk exposure, was the confounder. In many studies of deployment and health outcomes, individual smoking status is unavailable. Furthermore, the rate of smoking has been shown to increase on deployments. Since outcomes such as respiratory conditions and cancer are strongly associated with smoking, it is an important variable to consider.

Latency Period

The job of the epidemiologist is easier when the time from exposure to outcome is short. This time is called the "latent period." An exposure to high levels of chlorine gas leads quite quickly to respiratory distress. Inferring the association between exposure and outcome is relatively clear. The same is true in that exposure to someone with measles is quickly followed by measles in a susceptible person, and ingestion of contaminated food quickly leads to gastrointestinal illness in those who ingest it. An outbreak or cluster of outcomes is clearly defined, and searches for "exposures" and associations ensue.

Unfortunately, in environmental scenarios, most of these conditions are not met. Cancer outcomes typically have latent periods on the order of decades. This presents difficulty in that the exposure can go unrecognized or forgotten because it does not lead to an acute outcome; and if it eventually leads to an outcome, sufficient time may have elapsed that the individual no longer recalls the exposure, or that the past "dose" or magnitude of the exposure can only be estimated by memory. (10) This, as mentioned, leads to recall bias.

Specificity of the Health Effect

Phocomelia, the congenital absence of the proximal part of a limb, is an uncommon defect in infants. When infants were seen to have this defect in greater than expected numbers, mothers were questioned about the use of medications during pregnancy. It became apparent that infants with the defect were more likely to be born to women who had taken Thalidomide, compared with women who had not. (13) In this instance, the outcome was specific enough to allow for the assembly of a group of these mothers to compare with mothers who had normal outcomes, and compare their prenatal exposures. Another classic example was when occupational health physicians noticed an increased incidence of an unusual liver cancer called angiosarcoma in certain workers and determined that they had an increased history of exposure to vinyl chloride. The key was that the outcome was one with few other causes. Legionnaire's disease was discovered and described because an unusual pneumonia developed with high frequency in a group of men at a convention in a single hotel, and was linked to the air conditioning system. (14) Hantavirus illness in the southwestern United States was characterized after a physician became alarmed that 2 cases of fatal respiratory illness occurred one week apart. Suspicion was raised that there was some common exposure. (15)

When the effect of a low-level hazardous exposure is undefined, those exposed may attribute multiple health effects from varied causes to the exposure. Individuals who have registered for the burn pit registry have reported diverse medical conditions. (16) This is likely because they are reporting all health conditions that they may have. Without a comparison group, it is difficult to determine which effects might be associated with an exposure. The lack of a comparison group made it difficult to draw conclusions from clinical evaluations of redeployed service members who were veterans of the first Gulf War, when there was a concern about potential health effects resulting from their exposures in 1991. (17)

Evaluating the relationship between health effects that are known but nonspecific and environmental exposures can also be challenging. Exposures to solvents, which are metabolized in the liver, can lead to elevations of liver enzymes in the blood, but so can a host of other chemicals including alcohol and acetaminophen (confounding). Respiratory symptoms may follow exposure to burn pit smoke, dust storms, or other factors. The investigations which attempt to relate exposures to effects must take into account other possible exposures and question the individuals about these and control for these in the data analysis.


When studying health effects, it is important to consider how the effect is measured. In some studies, health effects were measured as symptoms reported by the study participants. Reporting bias may occur when people with concerns report their symptoms. (1) Studies which use databases that report disease by diagnostic codes are subject to problems with miscoding. Analyses of healthcare encounter diagnostic codes following redeployment are subject to unique limitations. For example, acute changes in health status during deployment due to exposures in theater could be missed. Likewise, changes in health status that occur over the long-term could be delayed beyond the available follow-up period and are therefore unobservable. This would be demonstrated, and particularly problematic, in evaluations of cancers with longer latency periods. Using medical encounter diagnostic codes to define cases may result in false positives, capturing as cases those individuals for whom health care providers entered codes for conditions that were being considered but not yet rejected or confirmed.

Multiple Hypothesis Testing

Since health effects are not always known, a common strategy employed in investigations is to question patients about multiple health effects. Questionnaires may ask about respiratory symptoms, gastrointestinal symptoms, neurologic symptoms, and vague, nonspecific symptoms. This can result in the calculation of a significantly high disease rate due to chance alone. Statistical "rules" for study precision generally allow for 5% error due to chance. This means that if 100 symptoms were asked about, 5 of the 100 could appear to be increased due to chance alone. (3)

Power and Sample Size

The power of a study is its statistical ability to detect a difference between 2 groups if it truly exists (for example, an elevated cancer rate) between the exposed and the unexposed. The power of a study is intimately related to the sample size, which is the number of people in each of the 2 groups. It is also related to the prevalence of the disease of interest in the population. In order to detect significant patterns, rare diseases such as cancer in relatively young populations, are studied using a case-control approach, or require large populations. More common diseases can be studied in smaller populations. However, to the extent that multiple causes (or exposures) are involved, as they are with most chronic diseases, larger populations are generally required in order to obtain significant results in studies of more common diseases as well. Refining the measures of diseases and the assessment of exposure can improve the power of a study to detect an association.


Some situations lend themselves to an epidemiologic investigation of the strength of association. The strength of an association, while not proof of causation between the exposure and the outcome, gives one confidence that the two may be related. These situations occur when the exposure is large and leads to a greatly increased risk of a specific outcome. The exposure must have occurred before the outcome, and ideally, the outcome follows soon after the exposure. Ideally, current and past exposures to chemicals would be precisely quantified, and testable hypotheses of specific health effects would be generated. An accurate assessment of adverse health effects in exposed and unexposed populations would be made, and account would be taken of all potential confounders. The sample size would be adequate, and irrelevant hypothesis would not be included. Because all studies have limitations, these conditions are never achieved. However, practitioners of military preventive medicine must continue to work to quickly identify potentially hazardous exposures, precisely measure the exposures, and document those who were exposed and the extent of their exposures to facilitate these evaluations. * 1


(1.) Schneider D, Lilienfeld DE. Lilienfeld's Foundations of Epidemiology. 4th ed. New York, NY: Oxford University Press. 2015.

(2.) Morgenstern H, Thomas D. Principles of study design in environmental epidemiology. Environ Health Persp. 1993; 101(4):23-38.

(3.) Aschengrau A, Seage, GR: Essentials of Epidemiology in Public Health. London, UK: Jones and Bartlett Publishers International; 2003.

(4.) Abraham JH, Eick-Cost A, Clark LL, et al. A retrospective cohort study of military deployment and postdeployment medical encounters for respiratory conditions. Mil Med. 2014; 179:540-546.

(5.) Rothman KJ, Greenland S, Lash TL. Case-control studies. Encyclopedia of Quantitative Risk Analysis and Assessment. Vol I. West Sussex, England: John Wiley & Sons; 2008:192-204.

(6.) Institute of Medicine. Long Term Health Consequences of Exposure to Burn Pits in Iraq and Afghanistan. Washington DC: The National Academies Press; 2011:1-9,117-128. Available at: https:// Accessed April 6, 2017.

(7.) Doll R, Hill AB. The mortality of doctors in relation to their smoking habits. BMJ. 1954; 328(7455):1529-1533.

(8.) Hori M, Tanaka H, Wakai K, Sasazuki S, Katanoda K. Secondhand smoke exposure and risk of lung cancer in Japan: a systematic review and metaanalysis of epidemiologic studies. Jpn J Clin Oncol. 2016; 46(10):942-951.

(9.) National Research Council. Health Risks from Exposure to Low Levels of Ionizing Radiation. BEIR VII Phase 2. Washington DC: The National Academies Press; 2006:141-154. Available at: https:// Accessed April 6, 2017.

(10.) National Research Council. Environmental Epidemiology, Volume 1: Public Health and Hazardous Wastes. Washington, DC: National Academies Press; 1991.

(11.) Baird CP, DeBakey S, Reid L, Hauschild V, Petruccelli B, Abraham JH. Respiratory health status of US Army personnel potentially exposed to smoke from 2003 Al-Mishraq Sulfur Plant fire. J Occup Environ Med. 2012; 54 (6):717-723.

(12.) MacMahon B. Concepts of Multiple Factors. In: Lee DHK, Kotin P, eds. Multiple Factors in the Causation of Environmentally Induced Disease. New York, NY: Academic Press: 1972:chap1.

(13.) Taussig HB. Thalidomide and phocomelia. Pediatrics 1962; 30(4):654-659. Available at: http://pedi Accessed April 6, 2017.

(14.) Chin J, ed. Control of Communicable Diseases Manual. 17th ed. Washington, DC: American Public Health Association; 2000.

(15.) Centers for Disease Control and Prevention. Outbreak of acute illness-southwestern United States, 1993. MMWR Morb Mortal Wkly Rep. 1993; 42(22):421-424.

(16.) National Academies of Sciences, Engineering, and Medicine. Assessment of the Department of Veterans Affairs Airborne Hazards and Burn Pit Registry. Washington, DC: The National Academies Press; February 28, 2017. Available at: https://www. ment-of-veterans-affairs-airborne-hazards-andopen-burn-pit-registry. Accessed April 6, 2017.

(17.) Institute of Medicine. Adequacy of the Comprehensive Clinical Evaluation Program: A Focused Assessment. Washington, DC: The National Academies Press; 1997. DOI: https://doi.org10.17226/6004.

Coleen Baird, MD, MPH

Jessica Sharkey, MPH

Joel Gaydos, MD, MPH


Dr Baird is the Program Manager for the Environmental Medicine Program of the Occupational and Environmental Portfolio, US Army Public Health Center, Aberdeen Proving Ground, Maryland.

Ms Sharkey is an Epidemiologist for the Environmental Medicine Program of the Occupational and Environmental Portfolio, US Army Public Health Center, Aberdeen Proving Ground, Maryland.

Dr Gaydos is an Occupational Medicine Physician for the Clinical Public Health and Epidemiology Directorate, US Army Public Health Center, Aberdeen Proving Ground, Maryland.
COPYRIGHT 2017 U.S. Army Medical Department Center & School
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2017 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Baird, Coleen; Sharkey, Jessica; Gaydos, Joel
Publication:U.S. Army Medical Department Journal
Article Type:Report
Geographic Code:1USA
Date:Jan 1, 2017
Previous Article:Blast-associated traumatic brain injury in the military as a potential trigger for dementia and chronic traumatic encephalopathy.
Next Article:Indirect military occupational lead exposure to children at home: a case report.

Terms of use | Privacy policy | Copyright © 2021 Farlex, Inc. | Feedback | For webmasters |