When we hear about exciting breakthroughs in the news or social media, we want to be able to critically understand and evaluate the applicability of the research discoveries to our own lives. However, many of us have not taken statistics or research methods classes (or may not remember them if we did). So, we are going to spend a little time in this blog series discussing some key considerations in deciphering research jargon so that you can better interpret study findings presented to you. We hope that this can guide you in fending off misinformation and help you live a more evidence-based life, especially when it comes to making decisions regarding your health.
Always remember the old adage “correlation does not imply causation”
Correlation and causation are not interchangeable. Correlation is far easier to demonstrate than causation in research and may not give us any valuable information at all. We can think about this in our own lives. The news is often filled with headlines like “Higher caffeine intake associated with higher income”. The article may go on to imply that we should all start drinking more coffee so that our incomes will magically start rising. Sound advice, right? Here’s where the news led you wrong. Just because something is associated or correlated statistically, does not mean that there is a causative link between these two things or any meaningful connection at all.
So, before you all begin doubling your morning serving of coffee, let’s discuss causation. The first thing you need to consider is the study design. Our strongest evidence for determining causation is through randomized, double-blind, placebo-controlled clinical trials. These are considered the gold standard of research. If well-designed, these types of studies have the opportunity to provide the highest level of evidence and certainty as to whether any intervention can produce a desired effect. Most of our concepts of clinical trials arise from pharmaceutical (drug) studies where clinical trials are not only preferred but required by regulatory authorities as evidence that a drug is both safe and effective.
What exactly is a clinical trial?
In simple terms, “clinical trials are research studies performed in people that are aimed at evaluating a medical, surgical, or behavioral intervention.” - National Institutes of Health
So, we could have several patients who have a health condition and give them some intervention (whether that be a drug or exercise regimen, etc). Then, we see if their condition improves according to some measure.
There are a few other important factors in determining whether you should start applying the results of a study to your own life.
First, is the clinical trial controlled, meaning did they compare the patients receiving this treatment with another group of patients who did not receive this treatment (controls)? There are two main ways this is done, either with placebos or with active comparators. Placebos are inactive substances that have no known mechanism in affecting the outcome. Usually, the placebo is a sugar pill (for drug studies at least). This in many ways would be ideal, because it allows us to determine the difference between the group of patients taking an inactive substance and the group receiving the treatment. However, there are some instances in which this would be considered unethical or impractical. For example, if you already have a drug on the market (i.e. shown to be both safe and effective) for a particular disease, then it would be unethical to withhold that treatment for a control group of patients participating in a clinical trial. For that reason, many drug trials use an active comparator, which is an intervention that is currently used in medical practice as part of the standard treatment for that condition. In those cases, a new drug could be approved by regulatory authorities if it is better or at least not worse than another treatment already on the market.
Now, there is another reason having a placebo or active comparator is important and that is the placebo effect. This, in truth, is not fully understood. However, for some reason, when people are given anything that they believe will make them better, a certain amount of patients actually will get better. This happens even if people are only given a sugar pill. It is attributed to the power of your brain to affect your health not only in perception but also in objective ways. The placebo effect is real, though it will vary with the condition being studied and the measures used to determine the outcome. Regardless, it is important that a study takes the placebo effect into consideration. Otherwise, the study may falsely attribute the benefits of the placebo effect to the intervention.
So, now we have a controlled clinical trial, but was the study randomized? Randomization means that every patient who enrolled to participate in the study was randomly assigned to study groups, which in most cases include a treatment and control group. In most large, multi-centered clinical trials, a software algorithm will be used to randomly assign patients upon enrolling in the trial.
Randomization is done in order to reduce bias and confounding.
Let’s consider a specific type of bias called selection bias as an example. If an investigator truly believed that an intervention was going to work, he or she may inadvertently (or even knowingly) direct healthier patients to the treatment group and sicker patients to the control group. Healthier patients (because they are healthier to start) will have better results, so this may look like the treatment is more effective when in actuality it is not. Randomization removes the influence of the investigator and study team on the patient selection, thereby minimizing this bias.
There is often an array of factors which contribute to a patient’s health. Known factors that are within our control are called modifiable risk factors and an example of this would be smoking cigarettes. Non-modifiable risk factors on the other hand are not within our control, such as age. Confounding is the effect of outside factors, other than the one we are studying, on the outcome. Randomization reduces the impact of both confounders we know as well as the ones we don't know. Through randomization, we expect these differences to distribute themselves into the different study groups relatively equally by chance.
The final consideration in evaluating a clinical trial is the use of blinding. Blinding means that individuals in the study are not told which group they have been assigned. If the treatment assignment is only hidden from patients, this is termed single blinding. If assignments are hidden from both the patients and the study team, this is termed double blinding. In a final, highest level called triple blinding, study assignment is concealed even from those who organize and analyze the data. Blinding, in combination with controls and randomization, helps to reduce bias, confounding, and account for the placebo effect.
Now, does this mean that we should not trust in the results of any research that does not come from randomized, double-blind controlled clinical trials?
Of course not!
There are tons of research including epidemiological studies that are credible especially when you evaluate them by certain criteria and take into account statistical considerations. For example, the studies initially demonstrating that smoking causes lung cancer was not a randomized, controlled clinical trial, though now the causative effect of smoking on lung cancer is considered a fact.
So, how do we interpret findings of studies that are not clinical trials, and how do we know if any study results are clinically meaningful? We will leave that for later parts of this blog series.
If you would like to sponsor research that matters to you, visit our main page at www.researchlead.com.