Epi corner - October 2016

Case-co​​ntrol studies

Case control studies have traditionally been frequently used to determine exposures in outbreaks. Their benefits include that they can be carried out rapidly, usually do not require as large a sample size as cohort studies, and are generally less expensive. They allow for multiple exposures to be studied, and for a rare event, they are often the only practical way to test a hypothesis.

However case-control studies have their limitations; one of the major limitations is their potential for bias, particularly recall and selection bias. In cases of an acute illness like gastroenteritis, patients may search for a cause for their illness, and so be more likely to report an exposure (for example, eating takeaway chicken) than controls; this is a form of recall bias.

It can also be difficult to define an appropriate control group, and methods to select controls also have the ability to introduce selection bias. Traditional sources of controls in case-control studies include:

• population controls – these are people randomly selected from the general population, for example by random-digit dialling by telephone, or knocking on every third door in a neighbourhood. These control selection methods can be very time-consuming and expensive
• friend controls – the case is asked to nominate a friend to serve as their control
• physician nominated controls – this involves asking the patients’ GP to select a similar patient from their records to act as a control

In traditional case-control studies, a significant bias can arise because controls are selected randomly, whereas cases that are notified to public health authorities have undergone a process of self-selection. This arises because not all people who suffer from a disease will present to their GP for testing, and not all tests recommended by GPs will be carried out. For every case of salmonella reported in surveillance systems in Australia, it is estimated that there could be between 3 and 12 cases occurring in the community who do not present to GPs and/or undertake faecal testing. This process can be affected by factors such as occupation, social status, health seeking behaviours; and these factors can also be related to differences in possibly exposures, for example diet.

Therefore in this population, the notified cases do not represent a truly random sample of cases.

Case-case studies and their advantages

As we have discussed, the difference in exposure histories seen in a traditional case-control studies between cases and controls may not be due to true differences in exposure, but to biases involved in the how cases are selected (via surveillance system notifications) vs. controls (which are randomly selected).

An alternative is the case-case study. Case-case studies draw on pre-existing surveillance systems and use notifications of different diseases, or of different serotypes of the same disease, as proxy controls. An example of the use of case-case studies is in the setting of salmonellosis. In Australia, Salmonella strains are subtyped in the reference laboratory by serotyping, allowing further characterisation of salmonellosis as caused by Salmonella Typhimurium, Salmonella Enteriditis, Salmonella Virchow, etc. This subtyping allows for the identification of outbreaks of less common Salmonella serovars.

The data collected on cases via routine surveillance mechanisms can be used in the place of controls to compare exposure histories between different groups of cases. For example, in an ongoing outbreak of one particular Salmonella strain – e.g. Salmonella Hvittingfoss – information on exposures for these cases, as collected using standard questionnaires, can be compared against exposure data from other recent interviewed cases of Salmonella Typhimurium as controls.

This method is time and cost efficient, as it removes the difficulties of identifying community controls for the outbreak and draws on pre-existing surveillance systems. It also helps eliminate the recall and selection biases associated with traditional case-control studies

Benefits of case-case studies

• All cases came through the same selection process; therefore biases that may be introduced by differences in health-seeking behaviours etc. between cases and randomly-selected controls is minimised.
• It reduces differences in recall between cases and controls, as both cases and “controls” in this methods have suffered from a recent gastrointestinal disease, and will have had the same consideration given to potential exposures.

Disadvantages of case-case studies

• Problems can arise with validity – in the case-case study, each “control” will have one exposure which differs systematically from that of the cases – the exposure which led to their own infection. For example, if using Salmonella Typhimurium cases as controls, there may be higher rates of consumption of chicken in the control group compared to the case group. The bias whereby all of the cases form the previous outbreak may differ from the general population in some important aspect could be circumvented by mixing cases from several different serotypes in the pool of controls.
• In small populations it can be difficult to source a large enough group of controls.
• If using historical cases as controls, there is the problem that dietary habits can change over time – therefore “controls” should not be more than a few years old at most.
• There can be seasonal variation in eating habits, particularly around fresh produce and salads, which may lead to systematic differences between cases and control groups if the control group cases occurred at a different time of year.

Question

We have discussed recall bias and selection bias.

What are some other types of biases that can arise in epidemiological studies?