Print This PostLearning how to design experiments or studies is probably the single most important thing when training to become a scientist. Regardless of the methodology being used, or the specific research question being asked, it’s essential to know how to design an experiment in such a way to test a specific hypothesis of interest. Key to that is knowing how to include appropriate controls. However, having worked in laboratory- and clinical-based research, I’m aware that controlling an experiment or study can mean different things depending on the speciality you’re working in. In this blog, I discuss how to use controls in your scientific studies using examples from my own research.
In scientific research, controls are used to help determine a cause-and-effect relationship between two or more things known as variables. Using controls help to isolate the effect of a particular variable and ensure that the results from an experiment or study aren’t random. Controls are an important part of experimental design but are applied differently across research areas and methodologies. In clinical research, the first clear use of a control group was in a study by Lane-Claypon in 1926 when investigating cases of breast cancer. The study compared women with a history of breast cancer to those without, but who were comparable on a range of other measures. The women without a history of breast cancer made up the control group, differing from the test group who were comprised of those with a history of breast cancer. In this type of study design, known as a case-control study, the control group should be as similar as possible to the test group, except on the condition or disease being investigated.
In addition to understanding potential disease causes, this type of simple, yet elegant, study design can answer research questions relating to whether a particular treatment/intervention is better than what’s currently being used. Here, the control group acts as a means to determine to what degree certain outcome measures can be attributed to the treatment/intervention, rather than other variables. As an example, a clinical study might test whether or not a new treatment/intervention has any benefit for people living with dementia. In contrast to the breast cancer study example, the control group isn’t a group without the disease/condition, as all participants included in the study have a dementia diagnosis. Instead, those allocated to the test group will be given the new treatment/intervention, and those allocated to the control group will either be given the currently used treatment/intervention (to determine if it’s better than what we’ve currently got), a placebo, or no treatment/intervention (depending on the specific research question being asked). This type of study design means that if all potential variables between groups are kept constant, and the only key difference is whether they received the treatment/intervention in question, we can be confident that any difference in the results is due to the treatment/intervention.
When I worked in clinical research I was involved in projects where participants were randomly allocated to a particular group, known as a randomised controlled trial. When comparing a drug with either a current medication or placebo, this was done ‘double-blind’ whereby neither the researchers nor the participants knew whether they were allocated to the test or control group. This gold standard method of clinical study design helps to eliminate potential for bias because research has shown that knowing how someone is allocated to a group can influence the outcome measures. This has also become a widely used study design for preclinical testing of new drugs using animal models.
In my research using human post-mortem brain tissue, I’ve carried out retrospective cohort studies with a matched design, whereby existing cases are allocated to different groups, such as whether or not they have a history of traumatic brain injury, for example. In these types of studies, there are several things which could potentially influence the outcome measures, many which can’t be controlled for because it’s retrospective and you are limited in terms of available resources. However, two of the biggest variables that can influence results, and are reasonably easy to control, are age and sex. Using a matched design involves matching individuals on these variables across the groups which are being compared, to help eliminate their effect on the results. We can therefore be relatively confident that the results are not due to a sex effect, or effect of one group being significantly older or younger. It helps to maintain internal validity and reduce variation in the results but can require large case numbers. In my neuropathology studies, there are additional mechanisms to introduce controls in the experimental design when using immunohistochemistry. This involves using particular antibodies to bind to specific antigens (usually proteins) to visualise cells or the proteins expressed by them. This is done using thinly cut samples of brain tissue placed on microscope slides. To be confident that the correct target is being stained, we use positive and negative controls.
A positive control is a tissue sample that we know should show specific staining for our target of interest. It’s stained using the same protocol as the rest of the samples and by the end, if the positive control doesn’t show specific staining, something has gone wrong, and we can’t trust the results from that batch.
A negative control is a tissue sample which follows the staining protocol the same as the other samples, except the primary antibody step is omitted. This leaves nothing for the secondary antibody to bind to, and ultimately no colour reaction can occur. If we see any positive staining in this sample, again, something has gone wrong.
The final example from my experience using controls in scientific studies comes from my animal research. There are inherently more opportunities to control aspects of the experiment compared to human studies, as you can determine the genetic makeup, species, age, sex, housing, husbandry etc. I’ve carried out studies using mice that are genetically modified to overexpress amyloid precursor protein (APP), allowing us to study a characteristic pathology associated with Alzheimer’s disease. When breeding these mice, some in a litter had the genetic modification and some were what we call wild-type. This meant that in our experiments not only could we match our experimental (APP) and control (wild-type) groups for age, we could also control for many other variables because mice in both groups were littermates.
From these few examples I’ve discussed, you can see the different ways in which controls are incorporated into study designs. Each time, careful consideration is made to determine what variables should and can be controlled to minimise bias and variability, based on the limitation of resources and what’s practicable.
Dr Kamar Ameen-Ali
Author
Dr Kamar Ameen-Ali is a Lecturer in Biomedical Science at Teesside University & Affiliate Researcher at Glasgow University. In addition to teaching, Kamar is exploring how neuroinflammation following traumatic brain injury contributes to the progression of neurodegenerative diseases that lead to dementia. Having first pursued a career as an NHS Psychologist, Kamar went back to University in Durham to look at rodent behavioural tasks to completed her PhD, and then worked as a regional Programme Manager for NC3Rs.
Dementia Researcher
NIHR