html xmlns=”http://www.w3.org/1999/xhtml” xmlns:mml=”http://www.w3.org/1998/Math/MathML” xmlns:epub=”http://www.idpf.org/2007/ops”>
Case-control studies are designed to assess whether there is a link between one or more exposures and a particular outcome. The purpose of the control group is to determine whether this exposure is equally common in individuals without the outcome compared to individuals with the outcome. The study must do this without introducing bias or potential confounding that could either mask or accentuate the relationship. Ideally, individuals would be controls only if they would have been included as cases if they had the outcome of interest. Often it is desirable to have more than one control group, selected in different ways, to provide more convincing evidence for the relationship.
26.1 Identifying Cases and Controls
26.1.1 General Considerations
Both cases and controls must be selected according to inclusion and exclusion criteria. The criteria will be the same between the groups, with the exception of the outcome. There should be a great deal of confidence that the cases have the outcome of interest. This can usually be ascertained through records and physicians’ referrals. Equally important, there should be no evidence that the controls have the outcome of interest. This may be determined by an interview with the potential control, but sometimes controls must be tested to confirm that they do not have the outcome of interest. Moreover, it is critical that there be no other differences between the groups that would bias the results. The potential for deliberate or subconscious bias is particularly strong in case-control studies because the outcome is known. Even if bias is not intended, it may be present when cases and controls are obtained from different populations.
When there are multiple control groups, the criteria that define the different control groups will include a specific criterion for each of the control groups in addition to the general control criteria.
26.1.2 Identifying Cases
Cases are defined by specific inclusion and exclusion criteria (Chapter 12). For some studies, cases may be identified by appealing to the general public for volunteers with the outcome of interest. Methods include mass mailings, advertisements in the media, or posters in selected sites, such as senior centers, day care centers, clinics, religious organizations, and the like. This may be the best way to identify potential participants if the outcome of interest is not one that would be recorded in a clinical or public record. The disadvantage of this method of identifying cases is that you may have to screen many individuals to get even one individual who fits the criteria for being a case in the study.
To study the effect of certain foods on the development of osteoarthritis (OA), investigators placed advertisements for individuals with OA in several local newspapers. It was not clearly specified that the condition had to have been medically diagnosed, so they were flooded by responses from people with some kind of pain in their joints, hoping for a free consultation. Even after the specification of medical diagnosis of OA was added to the advertisement, there were many calls from individuals who did not meet this basic criterion for cases.
When the outcome being studied is a condition that is likely to be underreported, such as drug or alcohol dependency or childhood sexual abuse, then a general call may be the best way to recruit participants. Recruitment through support groups and rehabilitation programs is possible, but this would only identify a subset of the population. Although confidentiality is required in all studies (Section 2.1.2), it is a particularly sensitive issue for these participants, and the text of the advertisement should emphasize that privacy will be respected. Sometimes it is possible to get special authorization to guarantee confidentiality to participants even from mandatory legal disclosure; your local Institutional Review Board (IRB) will be able to give you more information about this, which is termed a “certificate of confidentiality” in the United States.
You may introduce some bias with a general appeal because the people who respond may be different from the general population with the condition, particularly if the media used reaches only a specific population. There may also be a problem obtaining adequate, reliable records for ascertaining some of the study exposures.
The investigators worked in an area with a large immigrant population. Therefore, they placed the advertisements in several different foreign-language newspapers as well as English-language newspapers. Because they required medical records to confirm the diagnosis, the investigators had to exclude many individuals who had recently arrived in the country and did not have any records of their medical history.
Most investigators prefer to identify cases through health care providers. This can include individual physicians who treat patients with the condition, as well as clinics or hospitals. Potential cases identified in this way are more likely to have the condition of interest. Moreover, these individuals are more likely to have records of medical problems and treatment available. If the participant is recruited through a hospital or clinic, normally there will have to be institutional approval of the record screening (either for this specific study or as a general policy that patients have already allowed their records to be used for medical research). The investigators will normally provide material for the health care provider to give to patients, which specifies that the patients are under no obligation to participate and should contact the investigators directly if they wish to participate. IRB approval of the precise material the health care provider will forward to potential cases is required, and for advertisements in general, in addition to IRB approval of the study itself. The investigators will have to obtain informed consent as necessary and must protect the privacy of the individual in accordance with government regulations (Section 2.1.2) for protecting patient privacy. More detail on recruiting participants is given in Chapter 13.
Public records, such as birth or death certificates, registries, and health surveillance files, can provide a large pool of potential participants, but the information may be limited and it may be necessary to find other sources to complete the exposure information. Most of these public records are computerized, and investigators can obtain access under specific conditions, but usually the identification of the participant is not included, which may be a problem if you need more data.
In a case-control study of non-Hodgkin’s lymphoma, cases were identified from the area’s rapid case ascertainment system and from monthly surveillance, epidemiology, and end results program abstracts. This should represent virtually all cases in the area. These systems allowed cases to be contacted through their primary health care provider. After consenting to participation, a subset of cases had their diagnosis verified by an independent review of the pathology.
Another source of cases is laboratory specimens, including autopsies. This would of course not require consent from the participant, but you may have to contact family members if additional information is needed, which may present disclosure issues as well. Another disadvantage is that the specimens needed may only be available for a limited time or cannot be found.
To study the activation of microglia cells in sepsis, the investigators used data from brain autopsies of 170 patients who had died in their hospital during a 4-year period and had given permission for the use of their autopsy material in research at some time prior to death. All individuals who had material for further testing, and did not have evidence of confounding diseases such as CNS disease or some dementias, were selected as potential cases or controls. There were 36 individuals in this pool. Cases were those who were determined to have sepsis, based on clinical and laboratory hospital records and post-mortem cultures of spleen, kidney, or lung, providing a total of 13 cases.
26.1.3 Identifying Controls
Controls may be identified from any of the sources noted above. In many ways the identification of controls is more difficult than the identification of cases. Ideally, the controls should represent the same population as the cases and satisfy the same inclusion and exclusion criteria as the cases, with the exception of the outcome. Although there may be exceptions in special circumstances, in general you would want as potential controls only people who would be identified as potential cases if they had the disease.
In practice, when cases are identified through a health care provider, there are advantages to identifying controls from the same population – that is, people who are being seen at the same facility for a health condition at the same level of seriousness. This will generally make other participant characteristics more comparable.
Example 7C described a study to determine if lack of exercise is associated with the onset of Type II diabetes in adults younger than 40 years. Cases were identified and recruited from the diabetes clinic in a public hospital. Controls would be matched (Chapter 27) to the cases on sex and age within 5 years. Since the investigators wanted the controls to be identifiable as cases if they had the disease, they wanted to select as controls individuals who used the hospital for their own medical care as well. One possible control group was patients in other clinics in the hospital. Because they were using the hospital for one condition, it is likely that they would use it for other conditions as well. Thus, they would have been identified as cases if they were detected with diabetes. Such individuals would likely be similar to the cases in terms of income, education, access to care, and so forth. However, they all have some other condition, and these other conditions might be associated with exercise or lack of exercise as well. Another possible control group is parents of children being treated in the hospital clinics. Such individuals are likely to use the hospital for their own care, and would likely be similar in overall demographic factors, access to care, general knowledge, and health behaviors as patients in hospital clinics, so that they would have been identified as cases if they had diabetes.
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
