A case–control study compares the characteristics of a group of patients with a particular disease outcome (the cases) to a group of individuals without a disease outcome (the controls), to see whether exposure to any factor occurred more or less frequently in the cases than the controls (Fig. 16.1). Such retrospective studies do not provide information on the prevalence or incidence of disease but may give clues as to which factors elevate or reduce the risk of disease.
Selection of Cases
The eligibility criteria for cases should be precise and unambiguous (e.g. diabetes mellitus [World Health Organization criteria]: single fasting glucose concentration ≥7 mmol/litre or venous plasma glucose measured 2 hours after ingestion of 75 g oral glucose load ≥11 mmol/litre). In particular, it is important to define whether incident cases (patients who are recruited at the time of diagnosis) or prevalent cases (patients who were already diagnosed before entering the study) should be recruited. Prevalent cases may have had time to reflect on their history of exposure to known risk factors, especially if the disease is a well-publicized one such as cancer, and may have altered their behaviour after diagnosis. It is important to identify as many cases as possible so that the results carry more weight and the conclusions can be generalized to future populations. To this end, it may be necessary to access hospital lists and disease registries, and to include cases who died during the time period when cases and controls were recruited, because their exclusion may lead to a biased sample of cases.
Selection of Controls
As with cases, the eligibility criteria for controls should also be precise and unambiguous. Controls should be screened at entry to the study to ensure that they do not have the disease of interest. Where possible, controls should be selected from the same source as cases. Controls are often selected from hospitals. However, as risk factors related to one disease outcome may also be related to other disease outcomes, the selection of hospital-based controls may over-select individuals who have been exposed to the risk factor of interest, and may, therefore, not always be appropriate. It is often acceptable to select controls from the general population, although they may not be as motivated to take part in such a study, and response rates may therefore be poorer in controls than cases. The use of neighbourhood controls may ensure that cases and controls are from similar social backgrounds. Of note, it is important to avoid the temptation to relax the criteria for eligibility of controls part-way through a study simply to speed up the process of recruitment.
Although most case–control studies include only a single control for each case (often referred to as a 1:1 case–control study), it is possible to include multiple controls for each case (a 1 : n case–control study). Increased numbers of controls per case will provide the study with greater power (Chapter 18), although any such gains in power are likely to be fairly small beyond four controls per case1. Where a greater number of individuals are eligible to be selected as controls than is required, it is important to document how the controls should be selected (e.g. by random selection from all eligible individuals).
Identification of Risk Factors
As in any epidemiological study, the potential risk factors should be defined before conducting the study. The definition of these factors of interest should be clear and unambiguous (e.g. in a case–control study for the development of diabetes mellitus, where ‘exercise’ is the factor of interest, there should be a clear explanation of how exercise is to be measured and categorized). A pilot study may help to ensure that the definition will be feasible given the need to rely on retrospectively collected data and/or memory. Other factors which may have an impact on the outcome (i.e. case–control status), either as confounders (Chapter 34) and/or effect modifiers, should also be listed and defined.
Matching
Many case–control studies are matched in order to select cases and controls who are as similar as possible. We may have frequency matching on a group basis (i.e. the average value of each of the relevant potential risk factors of the whole group of cases should be similar to that of the whole group of controls) or we may have pairwise matching on an individual basis (i.e. each case is matched individually to a control who has similar potential risk factors). In general, when performing individual matching, it is useful to sex-match individuals (i.e. if the case is male, the control should also be male), and, sometimes, patients will be age-matched. However, it is important not to match on the basis of the risk factor of interest, or on any factor that falls on the causal pathway of the disease (Chapter 34), as this will remove the ability of the study to assess any relationship between the risk factor and the disease. Furthermore, it is important not to match on too many factors, as this may restrict the availability of suitable controls. Unfortunately, matching does mean that the effect on disease of the variables that have been used for matching cannot be studied.
Analysis of Unmatched or Group-Matched Case–Control Studies
Table 16.1 shows observed frequencies. Because patients are selected on the basis of their disease status, it is not possible to estimate the absolute risk of disease. We can calculate the odds ratio, which is given by:
where, for example, the odds of being a case in the exposed group is equal to
Table 16.1 Observed frequencies (see Fig. 16.1).
The odds of being a case in the exposed and unexposed samples are
and therefore the 
When a disease is rare, the odds ratio is an estimate of the relative risk, and is interpreted in a similar way, i.e. it gives an indication of the increased (or decreased) odds associated with exposure to the factor of interest. An odds ratio of one indicates that the odds is the same in the exposed and unexposed groups; an odds ratio greater than one indicates that the odds of disease is greater in the exposed group than in the unexposed group, etc. Confidence intervals and hypothesis tests can also be generated for the odds ratio.
Analysis of Individually Matched Case–Control Studies
Where possible, the analysis of individually matched case–control studies should allow for the fact that cases and controls are linked to each other as a result of the matching. Further details of methods of analysis for matched studies can be found in Chapter 30 (see Conditional logistic regression) and in Breslow and Day2.
Advantages of Case–Control Studies
- They are generally relatively quick, cheap and easy to perform.
- They are particularly suitable for rare diseases.
- A wide range of risk factors can be investigated in each study.
- There is no loss to follow-up.
Disadvantages of Case–Control Studies
- Recall bias, when cases have a differential ability to remember certain details about their histories, is a potential problem. For example, a lung cancer patient may well remember the occasional period when she or he smoked, whereas a control may not remember a similar period. When preparing the protocol for a case–control study, it is important to describe any attempts that will be made to reduce the possibility of recall bias by ensuring that exposure data are collected in an identical manner from cases and controls.
- If the onset of disease preceded exposure to the risk factor, causation cannot be inferred.
- Case–control studies are not suitable when exposures to the risk factor are rare.
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
