Owen Yang
Some clinicians, when doing research, is underinvesting their resource to obtain good quality study controls. And it is a shame.
Research 101: experiment group vs control group
Clinicians tend to focus on patients with diseases, or those who are on the new treatment. We tend to feel that a good study relies on how we characterise these patients.
When it comes to science, we need to go back to the basic and compare the experiment group and the control group. A good control group is fundamental to demonstrate a real effect of the experiment. In clinical medicine it is also called treatment group vs comparison group, or exposed group versus non-exposed group. It is important to make sure the two groups are only different by one factor (i.e. the treatment or the exposure), so that the difference of the two groups can be attributed safely to that one factor.
Most of us know this when we were asked in an exam, but when it comes to real life, for some reason many of us feel it is okay to overlook this basic principle and thinks it is okay to look away.
A pre- post- study is almost always unacceptable. A pre- post- study is where you measure something before giving the treatment, say blood pressure, and then take another measure after giving the treatment, and conclude any effect based on the different before and after the treatment. In this case you have a blood pressure of 140 before taking a tablet, and a blood pressure of 120 after taking a tablet. There is no control group here, and it will generally not be acceptable to attribute the difference from 140 to 120 to the tablet because of this.
Measure the right factors
We also tend to measure the factors that we think that is related to this treatment. I would say this is as important as measuring the factors that we think that is related to the outcome. It is common to see someone conduct study about a diabetes drug on a fancy pathway, and only measure that pathway in their study. What is equally required is all the other pathways that may lead to your outcome. If the purpose of the study is to demonstrate blood sugar control, the outcome is blood sugar control, and the factors related to blood sugar control will have to be recorded or measured. If the outcome is heart attack, then the factors related to heart attack will have to be recorded or measured.
What is described above is basic understanding of confounding.
Do not kill the mediator
This is itself a little bit of an advanced topic, but I guess there is no harm to mention here. If you are a well-behave, high integrity researcher, you tend to be doing your best to do a nice exposed versus non-exposed comparison. One thing that can come into confusion is ‘the mediator.’ If you are testing whether regular exercise can reduce the risk of diabetes, how do you take into account the body weight and body weight changes during the time? Could it be that exercise help to reduce body weight, and reduce the risk of diabetes? In this case, who is the perfect control or comparison group?
Let me know what you think. (hint: do not kill the mediator)
