Randomized Controlled Trials
We finally arrive at the randomized, controlled trial (RCT). It goes by many names (randomized trial, clinical trial, intervention study, experimental study, trial).
- Start with eligibility criteria to determine who is permitted to enter this study population
- Then the study population is randomized into the experimental group (or exposed group or intervention group) and the control group (or unexposed group).
- The experimental group is exposed to an intervention, such a particular drug or therapy.
- Then you see who develops the outcome of interest (such as who develops a disease or gets cured).
The main difference between this and a cohort study is the randomization and the fact that there is an intervention that is introduced by the researcher. In the observational study, the researcher just watches what happens naturally – they don’t intervene in any way.
The randomization is meant to eliminate confounding variables. If there is a particular factor that may bias outcome one way, the randomization would hopefully distribute that factor evenly between both the control and experimental groups.
Randomization can also prevent selection bias. Perhaps you have a particular therapy you really believe in, so much so you want to give it to the sickest patients. In a study, your patients in the experimental arm would be sicker than those who didn’t get the drug (the control group). This would not be a fair comparison for the drug. Randomization takes the assignment of who gets the drug and who doesn’t away from the researcher and puts it in the hands of chance.
Given your preference for your favorite drug, you may give extra special attention to the patients who received your pet drug. This extra special attention may be the reason why they do better. Blinding helps prevent this. No one knows who got the drug and who didn’t, they are blind to this information. There are several groups who can be blinded:
- Clinicians – the treating doctors don’t know who got the drug and who didn’t, so they will likely treat both the same
- Patients – the patients don’t know, the positive energies come from the hope arising from getting the drug will be neutralized. Don’t discount this placebo effect. It is real. And it can be fairly impressive.
- Adjudicators – the people who are looking at how the patients did, that is which patients got better and which didn’t, don’t know who got the drug. So any prejudice to swing ambiguous cases in favor of the drug is gone.
- Data analyzers – the people who calculate the statistics after the study can also be biased, by not letting them know who got the drug and who didn’t, they can’t introduce their own preferences into the study.
Intention To Treat
Once randomized, always analyzed.
Patients may decide that they don’t want to cooperate with the study. Or they may leave the study altogether. How do you count those people? If they were randomized to receive the treatment but they left the study and technically didn’t get the treatment, where do you count them? In the treatment group? In the non-treatment group?
It may make sense to put them in the non-treatment group because they didn’t get the treatment, however that is not what we do. Adherence to treatment is a big part of the treatment. If a pill has a 100% cure rate but 0% of the people take it, it is a failure. So you need to count the people who were randomized to the treatment group but didn’t follow-through in the treatment group — within the group they were initially randomized.
There may be some characteristic of the people who are able to comply with the treatment versus those who can’t. Perhaps the patients with no taste buds can swallow the bitter pill and those with taste buds leave the study. Or its some other factor that we don’t know about. Or maybe some couldn’t afford treatment. Or they had worse side effects. These are confounding factors!! This is why we randomized in the first place. To exclude these people would artificially inflate the benefits of your drug.
patwari 