.png){kind=icon}
In medical and veterinary research testing the efficacy of a new treatment, there are ethical issues associated with using a negative control, since this will prolong suffering compared to a positive control consisting of an existing treatment. Assuming there is such an existing treatment that is considered generally effective, a positive control seems the more attractive option ethically. However, there may be times when a negative control could be justified. Say the existing treatment was problematic in some way (e.g. because of highly unpleasant side effects), then the new treatment might still be attractive even if it was less effective that this existing treatment, providing it avoided the side effects and was still better than no treatment at all. To investigate this last point a negative control seems advantageous. In such a case there are steps that can be taken to minimize ethical concerns: the negative control group should be no bigger than absolutely essential; you should make sure that the trial is of the minimum required duration and those in the negative control group are switched to an effective treatment immediately afterwards or if they show predefined symptoms of deterioration during the trial.
In epidemiological studies in human health, case-control studies are often used. In this case the purpose of controlling is similar but the organization is different. Say we are interested in exploring the factors that might pre-dispose someone to a stroke. We could recruit individuals being treated in a hospital unit for stroke victims into the trial. For each of these case individuals we might use a mixture of medical records and interview to explore whether factors such as previous medical history or occupation might be risk factors that predict likelihood of a stroke. However, to make this comparison we also need to explore how common a given factor is in control individuals who have not had a stroke to that point. Thus, for each case individual we identify a matched control individual and we will compare the factors of interest to us between these two. We might choose to match on factors that we think might affect strokes but which we are not especially interested in (like age and sex). So for each case individual recruited from the specialist stroke unit we also recruit a control individual who is the same sex, has never had a stroke, and has a birthdate within one year of the case individual. Over a number of such matched pairs we can look to see if factors that were of interest to us (e.g. smoking, having a sedentary occupation) are more or less common in case individuals than controls. Care needs taken in how the controls are recruited to avoid introducing any biases. Case-control methodology could be used prospectively also. You could imagine exploring how a stroke affects subsequent medical conditions using exactly the same approach and following individuals from the point of recruitment until sufficient time has elapsed to get a measure of their health; this can then be compared across case-control matched pairs.
One reason for having control groups in some experiments may be to test that any assays you do are working properly; if we were testing amount of enzyme activity in samples from two different populations that we are interested in comparing, we might also include a control group of material that has been pre-treated to remove any enzyme activity, and another control group of material that has been pre-treated to enhance enzyme activity; comparison of the results of these control groups as well as the two groups we are interested in helps increase confidence that our assay of enzyme activity at least gives sensible readings in the two extreme cases of the controls, so our confidence in their reliability with respect to the groups of interest should increase.
There is a strong need for placebos in research in human medicine because of the placebo effect. The placebo effect is an umbrella term for a great range of studies that show that if someone receives a treatment that may not have many active ingredients but which the patient believes will be effective then they can experience genuine improvement in their condition. Thus, when testing a new drug given to patients in capsule form it is useful to make capsules identical in as many ways as possible but without the active ingredient, and have them administered to control patients in exactly the same way as the ‘real’ capsules. Indeed, it would be best if neither the practitioners nor the patients knew whether an individual patient was given the new drug or the placebo (a blind procedure - see Section 11.6.3).
You might expect a placebo effect in talking therapies like hypnotherapy, psychotherapy, and counselling but here design of placebos seems challenging. You might consider whether meeting with someone with no specific training simply to chat would be a suitable placebo, or whether an actor playing the part of a therapist and thus duping the subject would be a better but still ethical control.
Given the placebo effect you might want also to mull on whether a placebo pill is a negative or positive control. We are not sure you should devote hours of your life to this pondering however!