Case control study is one of the basic types of observational studies. Due to its simplicity, it is a common study design seen in literature. Some basic features characteristics of case control study design are: <list list-type="bullet"><list-item>Presence of two groups - one with an outcome of interest and another without the outcome.</list-item><list-item>Selection of subjects based on outcome status i.e. retrospective study design</list-item><list-item>Absence of any intervention i.e. natural course of the disease is observed</list-item></list>The basic purpose of the case control study is to determine the association between an outcome and an exposure. The exposure may be positively associated or negatively associated with the outcome. In other words, exposure may be either causing (increasing) the outcome or preventing (decreasing) the outcome.In this article we will be discussing the designing of a case control study and analysis of data collected during the study. Various advantages and disadvantages of this study design and situations where this design will be suitable have already been discussed in a previous article ‘Study designs’ of this series.</sec><sec id="sec001"><title>Steps of a case control study

The basic steps of a case control study are

Selection of cases

Selection of controls

Measurement of exposure to variable(s) of interest

Analysis

Selection of cases

This is an important step in the case-control study. One should have a clear definition of a case and the source population from which cases will be selected.

Definition of a case should be appropriate in that it should not miss out any true case with disease of interest and should be specific so that any patient without the disease of the interest is not enrolled i.e. the criteria for case enrolment should be specific and sensitive. However, the criteria cannot be too stringent that the study is not feasible at all. For example, in a study to correlate PROM with sepsis, one should have a clear definition of sepsis. This is because if case definition itself is not appropriate (either poorly sensitive/specific or the test is already superseded by a newer and better test) then the results of the study may not be the truth. Use of culture positive cases may be ideal in this situation. However, if yield of culture is too low in a particular centre, then it may not be possible to define cases by culture. In this situation, it may be appropriate to use septic screen positive babies as cases.

The source population from which cases will be selected should be defined before the enrollment of cases. The source can be a hospital or a particular community or a general population. Though selection of cases from hospital is convenient, it gives rise to a potential for bias as hospital population may not truly represents the general population. This would affect the external validity of that study. Despite these limitations, hospital sampling offers much convenience and is commonly used source for selection of cases.

Selection of controls

This is the most crucial aspect of the case - control study as this stage is prone for introduction of various biases. An ideal control is the one who is similar to case under study except for the disease state. Care should be taken to exclude sub clinical disease in controls during their enrollment. Control can be selected from hospital or from general population. Again, in a study one can opt for more than one control for each case enrolled. Generally, this increases the power of a study but enrollment of more than 4 controls for each case would not be cost effective and will not add much to the power. There are many caveats in a hospital based case control study. The care seeking behavior and the referral pattern would affect the pattern of cases attending the hospital. Selection of controls also gets distorted as controls also have some other disease and the factors affecting that disease may act as confounders. However, easy availability of cases and general willingness of the hospital population makes the hospital based case control study an attractive design.

Deorari A K et al. studied brainstem auditory evoked responses (BAER) in 18 term infants with neonatal hyperbilirubinemia (NHB) (total serum bilirubin > 15 mg/dl) and controls without hyperbilirubinemia. Seven neonates with NHB had abnormal BAER. In another study, Thomas C investigated risk factors of necrotizing enterocolitis (NEC). 23 infants with NEC born within the hospital were compared with 46 weight-matched controls, there were no significant differences in birth weight, gestational age, or feeding patterns. However, NEC cases had a higher frequency of pregnancy-induced hypertension, low mean Apgar scores, polycythemia, hypothermia, and septicemia than controls. These findings suggest that poor gut blood flow may be another important etiologic factor in NEC.

Matching

Let us consider an example. A study to determine incidence of PROM in cases of sepsis is to be conducted. But we know that incidence of sepsis is high in preterm babies and PROM is also common in preterm babies. So if we take controls only based on absence of sepsis, it is likely that more term babies will be enrolled in controls and hence incidence of PROM may be falsely less in controls. This may show a false association. To avoid this fallacy one can conduct study in a predefined gestational age group of babies or one can match controls with respect to gestational age i.e. for each case of sepsis with a given gestation age one control will be selected without sepsis but of same gestation age. This process is called matching. Matching is a process wherein we select controls in such a way that they are similar to cases regarding a predefined variable. This predefined variable should be able to influence the outcome independently and if not matched would affect the outcome of the study i.e. it should be a confounding factor (for details on confounding factor, refer to previous article on ‘bias and confounding factors in a research’). It should be noted that one cannot determine the association between outcome and matched variable.

Matching helps in avoiding a confounding factor in the study. Usually in a study, one can match controls with respect to one or two variables. If too many matching variables are taken, it would become very difficult to enroll any control.

Measurement of exposure variable

After establishing the protocol for the study, cases are selected if they fulfill the diagnostic inclusion criteria. A control is selected from the same population as the case and matched for predefined variable. The presence of one or more of the exposure variable in each of these two groups is recorded. This data can be collected from existing records or by interview or questionnaire. These collected data are then analyzed appropriately for presence of any association between exposure variable and outcome variable.

Analysis of a case control study

Once the data is collected, it needs to be analyzed to derive the result of the study. The relation between exposure and outcome is expressed as odds ratio in a case control study.

Odds ratio: It is important to understand the concept of odds ratio. The odds ratio is the ratio of the probability that a particular event will occur to the probability that it will not occur. It can be any number between zero and infinity.

While probability is a rate, odds are a ratio.

Persons with disease

Odds ratio is a point estimate. 95 percent confidence interval can then be calculated for this ratio. This is calculated as OR ± 1.96 X standard error. When the risk is rare in both cases and controls, then odds ratio approximates well with relative risk.

Calculation of Odds ratio

Analysis of a simple form of case control study wherein exposure to a single variable is discussed here. However, same principle can be extended to multiple exposures. Also, in the present day of computers, various softwares are available which can do all these calculations. However, knowing the basic principles of these calculations is useful to understand the concept of odds ratio.

The following steps explain the method of calculation of odds ratio.

Odds ratio is the ratio of odds of exposure to odds of controls

The collected data can be summarized in a simple 2 X 2 format as below.

	Exposed	Unexposed	Total
Cases	A	B	A+B
Controls	C	D	C+D
Total	A+C	B+D	A+B+C+D