Randomized controlled trial

Randomized controlled trials (RCTs) are the gold standard for determining whether an intervention actually causes any effects. The use of randomization and control groups helps eliminate the effects of factors other than the treatment being studied.


What are trials?

Suppose you want to know whether a new supplement, Burnz, promotes fat loss. If you know people who take Burnz, you could wait to see if they lose fat; in so doing, you’d be running an informal kind of observational study. Alternatively, you could give Burnz to people who don’t yet take it, then wait to see if they lose fat; you’d be running a trial (a type of study in which researchers intervene instead of just observing).

What are controlled trials?

In either case, however, you won’t know if the people who lost fat wouldn’t have lost fat even without Burnz. You need to create a comparison point. So instead of giving Burnz to everyone, you could ask people whether they’d like to take Burnz. You could then compare, later, the fat loss of the people who took Burnz (the intervention group) with the fat loss of the people who didn’t (the control group).

What are randomized controlled trials?

Yet you still won’t know for sure if any fat loss can be credited to Burnz. Other factors might be at play. The intervention group may be exercising more (or less) than the control group. Or their diet may be better (or worse). Or one group may be more likely to take other fat burners. Any such variable that can affect the result is called a confounder (aka confounding variable or confounding factor).

To eliminate confounders, you could try matching your intervention and control groups. For instance, you could ask your trial’s participants not to take any other fat burner than Gainz.

Unfortunately, it’s easy to overlook confounders, and some are difficult to assess. As a first step toward solving this issue, you could let luck, not your participants, decide who’s going to take Burnz and who isn’t. You’d be randomizing your participants between the intervention group and the control group, thereby running a randomized controlled trial.

What are randomized placebo-controlled trials?

We still haven’t eliminated two important confounders, one of which is the participants subjectivity. The participants who take Burnz may be more (or less) motivated to exercise, thus skewing the results.

To address this issue, you could give the control group an inert substance — a placebo — that looks and tastes like Burnz, and not tell the participants if they are taking Burnz or the placebo (in other words, not tell them if they’re part of the intervention or control group).

What are randomized, double-blind, placebo-controlled trials?

The last major confounder left is the researchers’ subjectivity. The researchers likely want the supplement to work, if only because studies that show positive results are more likely to get published. For that reason, they could consciously or subconsciously encourage the intervention group to exercise more than the control group.

(There are other ways for unethical researchers to ensure their trial yields positive results, mind you. Data dredging, for instance, or p-hacking.)

To solve this issue, you could hide from the researchers themselves what each participant is receiving until the trial is completed.

  • If only the participants didn’t know what they received (Burnz or placebo), you’ve run a single-blind placebo-controlled trial.
  • If the participants and the researchers in contact with the participants didn’t know what the participants received, you’ve run a double-blind placebo-controlled trial.
  • If the participants, the researchers in contact with the participants, and the researchers analyzing the data yielded by the trial didn’t know what the participants received, you’ve run a triple-blind placebo-controlled trial

By running a randomized, double-blind, placebo-controlled trial with enough participants (a large enough sample size), you’ll gather reliable data on the effects of Burnz on the outcomes measured (e.g., weight, fat mass, BMI, waist circumference).