Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Feasibility and Design of an Epidemiologic Study Epidemiology is the art of the practical. Unlike the researcher in a labora- tory experiment, who has control over the relevant variables, including the ge- netic homogeneity of the animals and the precise exposures for an experiment, the epidemiologist must deal with events that have already occurred. Thus, the feasibility and design of an epidemiologic study are dictated by external, practi- cal matters as well as by the hypothesis under question. Some practical concerns are: 1. availability of an appropriate study population; 2. size and composition of the study population; 3. completeness (and lack of bias) with which study subjects can be en- rolled; 4. magnitude and distribution of exposure to the hazard being studied; 5. accuracy with which the exposure can be measured (measurement of ab- sorbed 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); 6. accuracy of disease identification (history of disease should be confirmed by hospital records, and causes of death should be determined by obtaining cop- ies of death certificates); 7. background rate of the disease being studied; 17
18 AD VERSE REPRODUCTIVE OUTCOMES 8. expected increase in disease among the exposed group; and 9. availability of information on other factors that might determine disease. STUDY COHORT One starting point for any epidemiologic study is a defined population. Ei- ther samples of this group of people or the entire group is studied. If only a sample is studied, it is important that the sample be representative of the whole, or at least that the differences between the sample and the total population be clearly identified. Problems arise when the study sample differs from the popu- lation in ways that are ill-defined or unknown. This may happen when people in the study population cannot be traced, have died, or decline to participate. The higher the percentage of the population lost for these reasons, the less certain are the conclusions that can be drawn from the people actually studied. MEASUREMENT OF EXPOSURE Hazardous exposures are difficult to measure under the best of circum- stances. When the study takes place years after the exposure has occurred, the difficulties are even greater. Few exposures can be measured accurately in retro- spect. Specific problems of radiation dose measurements are discussed in Chapter 8. The less precisely an exposure is measured, the harder it is to find a clear association between exposure and disease. When measurement of exposure is poor, the high-dose group may be so diluted by less exposed people that no ef- fCct can be seen. In addition, exposures to other disease-causing factors (such as cigarette smoking) may confound the connection between the suspected expo- sure and the disease. Information on these other factors (known as confounders) must be collected in an epidemiologic study. However, it can be as difficult to collect information on an individual's confounding factors ~ ~ periods as it is to measure the suspected exposure under study. DEFINING THE DISEASE AND ASCERTAINING THE CASES for previous time In an ideal study, the disease would be defined by examination of pathologic tissues or other direct diagnostic means. More often one must rely on medical records or self-reports. Medical records may be incomplete or inconsistent be- cause of the lack of standardized methods for examination and diagnosis. Phy- sicians usually do not arrive at a diagnosis by a well-standardized routine but by experience and judgment, which may vary widely among individuals. Also, medical records are so decentralized that it can be very difficult to find individ
FEASIBILI~AND DESIGN 19 ual records. Records are usually saved for a maximum of 20 years; this time period is often less if hospitals close or doctors end their practice. Reproductive problems are especially hard for epidemiologists to study, and the various types of reproductive outcomes require different kinds of ascertainment strategies. Hospital records may be good for identifying problems of delivery, but poor for finding fertility problems. No single epidemiologic design can identify all of the types of reproductive endpoints. This means that some choices regarding which endpoints are most important must be made at the onset of a study. The specific re- productive endpoints that should be included in a study of radiation-exposed men would depend on biologic plausibility, prior research results, and to some degree, anecdotal information. A detailed discussion of radiation and its most likely repro- ductive effects is provided in Chapter 6. A further difficulty in studying reproductive endpoints is that many are never detected or recorded in the medical system at all. Reproductive problems are not always obvious, and not all are diagnosed correctly. At least half of U.S. couples who are infertile never take their problem to a doctor (OTA, 1988~. Most spontaneous abortions occur so early in pregnancy that the woman does not know that she was pregnant (Wilcox et al., 1988~. Even serious birth defects can be underreported in routine records (Lie et al., 1994) When records are inade- quate, the couples themselves must be relied upon to provide information. Self- reports have their own limitations. People must sometimes be asked to recall events that occurred many years earlier. Study subjects are known to have trou- ble recalling even such major events as a spontaneous abortion, much less the clinical details of those events (Wilcox and Homey, 1984J, and men typically recall medical events relating to pregnancy, delivery, infancy, and childhood less well than women. Once the criteria for what defines a case are established, then the cases must be located or ascertained. In addition, every person in the reference population should be contacted personally by an investigator. This is usually impractical and may even be impossible. Existing medical records can be useful, but they are often difficult to track down, especially in the United States where people may get their medical care from many sources. Self-identification of cases may be possible, but then the concern is that people with a personal or even a finan- cial interest in an exposure to some hazard will selectively respond. When re- productive difficulties are the topic, there is the added problem that the most seriously affected individuals (such as a baby with a major birth defect) may no longer be alive, making ascertainment and diagnosis more difficult. Ultimately, however, the feasibility of an epidemiologic study depends on satisfying not one but a series of scientific requirements. It is necessary to evaluate the power of the proposed study (which will depend on the population size of those exposed and the magnitude of the expected risk differences) and to assess whether the necessary data on health effects, exposure, and potential con- founders can be obtained (refer to Chapter 95.
20 ADVERSE REPRODUCTIVE OUTCOMES Epidemiologic studies pose two closely related ethical issues: privacy and confidentiality. Privacy in this context refers to keeping particularly sensitive information about oneself a secret, whereas confidentiality refers more generally to keeping personal data out of the hands of others without the authorization of the subjects. Epidemiologic studies usually involve a review of various types of existing records, including medical records, and may also involve interviews with individuals regarding medical and other personal information. It has be- come standard ethical practice in the United States to have epidemiologic proto- cols reviewed by federally mandated Institutional Review Boards (IRB) to en- sure that researchers take adequate steps to preserve the confidentiality of the data they collect, re- quiring 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 infonned consent of the subjects will be obtained be- fore interviews are conducted . . . (Wallace, 1982; OPRR, 19931. There is some discretion on the part of IRBs to authorize record review by accountable individuals who agree to protect confidentiality, at least when no information that can be linked to the individual is kept. However, when confi- dential information can be linked to the individual, each subject must consent to the study under most circumstances (OPRR, 1993~. It is also important to have a mechanism to inform participants of the results of the study and to counsel them when appropriate, especially if the results might influence their health care, medical future, or other important life decisions.