Toxicological Principles for the Safety Assessment of Food Ingredients
Return to Redbook 2000 table of contents
- Descriptive Epidemiology Studies
- Analytic Epidemiology Studies
- Epidemiology Studies References
Epidemiology is the study of the distribution and determinants of health-related states and events in specified populations, and the application of this study to the control of health problems.(5) The goal of all epidemiology studies is to uncover relationships between exposure to a specific agent and changes in health status.
Epidemiologic data are important to CFSAN in assessing safety and have been used by the Agency as indicators where avenues of research and further human studies would be most productive. Guidelines for the proper conduct and documentation of epidemiology studies, such as selection of the study population, selection of appropriate controls, exposure assessment, methods used to adjust or control for confounding variables, and statistical analyses will not be discussed here. Appropriate guidelines have been published elsewhere,(1) and should be consulted by the petitioner before submitting epidemiology data for consideration by the Agency.
There are two main categories of epidemiology studies, descriptive and analytic. Descriptive studies are concerned with the existing distribution of variables; they do not test hypotheses or make inferences concerning causality. Analytic studies are designed to examine associations, particularly hypothesized causal relationships, and focus on identifying or measuring the effects of specific risk factors.
1. Descriptive Epidemiology Studies
Descriptive epidemiology studies are relatively inexpensive to conduct and are usually of short duration. However, such studies are limited in their usefulness since no inferences can be made concerning causality. Generally, descriptive epidemiology studies are sentinel devices used to generate hypotheses or to provide evidence that indicates whether there is sufficient cause for conducting a lengthier and costlier analytic study.
a. Correlational Studies
Correlational studies, also called ecological studies, use grouped population data to relate exposure patterns of whole populations to disease incidence or mortality rates for whole populations. Because these studies do not examine the relationship between exposure and disease among individuals, the studies have been traditionally regarded as useful for generating, rather than definitively testing, a scientific hypothesis. Thus, the results of correlational studies would be insufficient to demonstrate a relationship without other types of data to support them.
b. Case Reports
Case reports are a type of descriptive epidemiology study frequently evaluated by CFSAN. Strongly suggestive anecdotal or clinical observations may indicate a possible causal relationship. Analytic epidemiology studies can then be designed to verify and quantify the risks, and to determine the role of confounding factors.
There are two principal avenues through which case reports come to the attention of CFSAN: first, reports published in the peer-reviewed medical literature, and second, reports captured in one or more of CFSAN's ongoing voluntary (also called "passive") adverse event monitoring systems, which include:
The Adverse Reaction Monitoring System (ARMS) - collects spontaneous reports from consumers and health professionals regarding alleged adverse effects from food products.
2. Analytic Epidemiology Studies
Although analytic epidemiology studies are more informative than descriptive studies, they are expensive and time-consuming to conduct. The types of analytic epidemiology studies commonly considered by CFSAN in safety evaluations include cross-sectional, prospective, and retrospective studies. Results from such studies, when available, are used in the overall safety evaluation of regulated products. In addition, analytic epidemiology studies constitute the scientific base for the Agency's regulation of health claims on food and food labeling authorized by the Nutrition Labeling and Education Act of 1990.
a. Cross-Sectional Studies
Cross-sectional studies are those in which individuals are observed at only one point in time; such studies are commonly known as surveys. The presence or absence of disease and the presence or absence of suspected etiologic factors are determined in each member of the study population or in a representative sample at one particular time. The advantages of cross-sectional studies are that they are relatively inexpensive to conduct, and can be completed relatively quickly. However, cross-sectional studies reveal nothing about the temporal sequence of exposure and disease, and necessarily use current exposure as a surrogate for past exposure. Also, cross-sectional studies can only measure disease prevalence rather than incidence.
b. Prospective Studies
In prospective studies, also called cohort or follow-up studies, the investigator selects a study population of exposed and non-exposed individuals and follows both groups to determine the incidence of disease. The group can be characterized by factors thought to influence the development or course of the disease and by the presence or absence of risk factors (e.g., exposure or nonexposure to some agent). Prospective studies generally imply study of a large population, study for a prolonged period of years, or both. This type of study design is effective when there is good evidence of an association of the disease with a certain exposure (from clinical observations or from descriptive epidemiology studies), when exposure is rare, but incidence of disease among the exposed is high, and when the time between the exposure and disease is short. The major advantage of prospective studies is that the incidence rates of the disease under study can be measured directly; therefore, absolute and relative risks also can be measured directly. In addition, it is possible to analyze the association of a particular exposure with several diseases, and a temporal relationship between exposure and disease can be established.
There are a number of disadvantages to prospective studies, including: 1) The difficulty and expense of conducting the studies, since both large study populations and long periods of observation are required for definite results; 2) bias may be introduced if every member of the cohort is not followed; 3) the length of the study may be less than the latency period of the disease; for example, if the study is stopped before old age, many important diseases such as cancer may be missed; and, most importantly, 4) prospective studies are very inefficient for studying rare diseases.
Results of prospective studies have been used at CFSAN in assessing the potential carcinogenic risk of some compounds; for example, occupational cohort studies and studies of human populations accidentally exposed to a carcinogen have been used in safety assessments of benzene, dioxin, and methylene chloride. FDA has also provided financial support for prospective studies on accidental exposure to PBB's in a Michigan cohort, and exposure to methylmercury in fish in a cohort of pregnant women (and their offspring) in the Seychelles Islands.
c. Retrospective Studies
In retrospective studies, also known as case-control studies, the investigator selects cases with a specific disease, and appropriate controls without the disease, and obtains data regarding past exposure to possible etiologic factors in both groups. The rates of exposure of the two groups are then compared. A case-control approach is preferred when studying rare diseases, such as most cancers, because a very large number of individuals would be needed in order to draw conclusions in a prospective study. Although it is possible to detect the association of multiple exposures or factors with a particular disease, retrospective studies are generally used to study diseases that have some unique and specific cause, such as infectious agents, in order to avoid the problem of confounding etiologic factors.
Case-control studies can not determine directly absolute risk or relative risk because the incidence of disease is not known in either the exposed or unexposed population as a whole. However, the relative risk can be estimated in retrospective studies by the odds ratio, which is the ratio of the odds of exposure among cases divided by the odds of exposure among controls. The odds ratio is a good approximation of the relative risk when the subject cases are representative of all cases with regard to exposure, the controls are representative of all controls with regard to exposure, and the disease being studied is rare.
Retrospective studies are much less expensive and less time consuming to conduct than are prospective studies; usually, a relatively small population is needed for the study. Also, since the study selects only cases of the disease of interest, there is no bias incurred in determining the endpoint. However, bias is frequently incurred during detection and selection of cases, and during assessment of exposure. Controls should be identical to the exposed cases except for the factor under investigation, a requirement which is often difficult to achieve in practice. As with prospective studies, problems are frequently encountered in attempting to control for competing risk factors and confounders. The investigators can adjust for known confounders either by matching when selecting controls, statistically by stratification, or by use of regression models.
Results of case-control studies have been frequently used in safety evaluations at FDA, primarily to add further information to the overall assessment of safety. In the past, FDA has supported case-control studies on compounds of interest, such as the National Bladder Cancer Study and the use of artificial sweeteners. In addition, FDA often looks carefully at the results of case-control studies in the setting of outbreaks of food-borne disease to identify the food vehicle that was most likely responsible for transmitting the infectious agent. The results then can be used to help target specific food vehicles for microbiologic testing as a means of recovering the pathogen from the implicated food.
Meta-analysis has been defined as "the statistical analysis of a large collection of analysis results from individual studies for the purpose of integrating the findings".(3) The results of a well-done meta-analysis may be accepted as a way to present the results of disparate studies on a common scale; however, caution should be exercised before attempting to reduce the results to a single value as this may lead to flawed conclusions.(2)
- Identifying criteria for the inclusion and exclusion of studies and avoiding biases in this process;
- Deciding whether the characteristics of study subjects, their interventions, and outcomes in each study are comparable;
- Using well-defined methods for extracting data from the studies;
- Expressing the results of multiple studies in a consistent fashion;
- Using appropriate statistical methods to assess the data.
Where FDA evaluates a meta-analysis, the Agency considers such an analysis primarily as supporting evidence, rather than as primary evidence, that can confirm the validity of data concerning a hypothesis. The Agency must carefully scrutinize each meta-analysis to assess the soundness of its design and the quality of the data from individual studies to determine the significance of the data. Such scrutiny requires review of the original studies used for the meta-analysis.
Anonymous (1981) Reviews and Commentary. Guidelines for documentation of epidemiologic studies. A report by the Epidemiology Work Group of the Interagency Regulatory Liaison Group. Am. J. Epidemiol. 114(5):609-613 (Return)
Bailar, J.C. (1997) The promise and problems of meta-analysis. New Eng. J. Med. 337:559-561 (Return)
Glass, G.V. (1976) Primary, secondary and meta-analysis of research. Educ. Res. 5:3-8 (Return)
Henkel, J. (1998) MedWatch. FDA's 'heads up' on medical product safety. FDA Consumer 32(6):10-12,15 (Return)
Last, J.M. eds. (1983) In, A Dictionary of Epidemiology. (J.H. Abramson, S. Greenland, M. C. Tuhriaus, associate eds.: J. Amsel, assistant ed.) Oxford University Press, New York, N.Y (Return)
Sacks, H.S., Berrier, J., Reitman, D., Ancona-Berk, V.A., Chalmers, T.C. (1987) Meta-analyses of randomized controlled trials. New Engl. J. Med. 316:450-455 (Return)
- Thacker, SB. (1988) Meta-analysis: a quantitative approach to research integration. JAMA 259:1685-1689 (Return)