Skip to main content
    Path to Science

    Thinking Dictionary

    Randomized controlled trial

    What it is

    A randomized controlled trial, often shortened to RCT, is a way of testing whether a treatment really works. The people taking part are split into groups by chance, as if by tossing a coin.

    • One group gets the treatment being tested.
    • The other group is the control group. It gives the researchers something to compare against. Depending on the trial, it might get no treatment, a placebo (a fake treatment), or the usual treatment that already exists.

    Because chance decides who goes where, the groups start out alike. So if they end up different, the difference is most likely due to the treatment.

    Why random allocation works

    Imagine a new medicine is tested by letting people choose whether to take it. The people who choose it might be healthier, richer, or more careful about their health. If they do better, you can't tell whether it was the medicine or those other differences.

    Random allocation solves this problem. When chance decides, the groups end up alike, on average, in everything: age, health, habits, and even things nobody thought to measure.¹ Such hidden differences are called confounders; see the Confounding variable entry. Statistical tricks can only adjust for confounders that were measured. Randomization takes care of the unmeasured ones too.²

    Notice the words "on average." In a small trial, chance alone can still leave the groups a bit unbalanced. That is one reason trial size matters; see the Sample size entry.

    It also matters that nobody can predict which group the next person will join. Otherwise, a doctor might steer sicker patients away from the new treatment, which would spoil the comparison. A famous 1948 trial of the antibiotic streptomycin for tuberculosis, run by the UK's Medical Research Council, is remembered as a landmark partly because it took careful steps to keep the next allocation secret.³˒⁴

    Trials can overturn what everyone believed

    Because they are such fair tests, trials sometimes contradict what doctors and researchers thought they knew.

    • A trial of knee surgery for worn-out knees found the real operation did no better than a fake one. The Placebo effect entry tells that story.⁵
    • A large trial of hormone replacement therapy found it raised the risk of heart disease, when earlier studies had suggested it protected the heart. The Correlation vs causation entry covers this.⁶

    How to read a trial

    There is an agreed checklist, called CONSORT, of what a good trial report should include.¹ You don't need the full list. These questions cover the basics:

    • Was it really randomized? Who decided who got the treatment, and how?
    • What did the control group get? No treatment, a placebo, or the usual care?
    • Was it blinded? In a blinded trial, people don't know which group they are in. This stops expectations from colouring the results.
    • How big was it? Small trials are more easily swayed by chance.
    • Who took part? The results apply best to people like those in the trial.

    RCTs are powerful, but not perfect

    A randomized trial isn't always possible. You can't ethically assign people to start smoking to see whether it harms them. Trials can also be too small, too short, or carried out on people who are unlike you. A badly run trial can mislead, just like any other study.

    And "random" doesn't mean careless or haphazard. Randomization is a deliberate, carefully recorded procedure.

    In everyday life

    • Medicine adverts. "Clinically proven" sounds impressive. Ask: was there a randomized trial, with a control group?
    • Your own experiments. You can borrow the idea of random allocation for small tests at home, like the one below.
    • News about studies. When a news story says a study "found" something, check whether it was a randomized trial or simply observed what people chose to do.

    The common trap

    The first trap is thinking "randomized means perfect." Size, blinding, and how well the trial was run all matter.

    The second trap is thinking that "control group" means "gets nothing." Often the control group gets a placebo or the current standard treatment.

    Some ways out:

    • Use the five reading questions above.
    • Look for more than one trial pointing the same way.
    • Remember that observational studies, where researchers simply watch what people do, are more likely to be confounded.

    Try it

    Does music help you focus? Make 10 slips of paper: 5 that say "music" and 5 that say "silence." Fold them and put them in a cup. Before each study session, draw a slip without looking. That is your randomization. After each session, rate your focus from 1 to 10.

    When all 10 slips are used, compare the average scores. Then list what could still bias the result. For example, you know which condition you're in, so you can't be blinded. And 10 sessions is a small sample.

    Related concepts

    Placebo effect · Correlation vs causation · Confounding variable · Sample size · Indicators of causation · Reverse causation · Simpson's paradox · Regression to the mean · Blinding · Nocebo effect

    Sources

    1. Schulz, K. F., Altman, D. G., & Moher, D. (2010). CONSORT 2010 statement. BMJ, 340, c332. https://doi.org/10.1136/bmj.c332

    2. Greenland, S., Pearl, J., & Robins, J. M. (1999). Confounding and collapsibility in causal inference. Statistical Science, 14(1), 29–46. https://doi.org/10.1214/ss/1009211805

    3. Medical Research Council. (1948). Streptomycin treatment of pulmonary tuberculosis: A Medical Research Council investigation. BMJ, 2(4582), 769–782. https://doi.org/10.1136/bmj.2.4582.769

    4. James Lind Library. (n.d.). Medical Research Council (1948b). https://www.jameslindlibrary.org/medical-research-council-1948b/

    5. Moseley, J. B., O'Malley, K., Petersen, N. J., Menke, T. J., Brody, B. A., Kuykendall, D. H., Hollingsworth, J. C., Ashton, C. M., & Wray, N. P. (2002). A controlled trial of arthroscopic surgery for osteoarthritis of the knee. New England Journal of Medicine, 347(2), 81–88. https://doi.org/10.1056/NEJMoa013259

    6. Writing Group for the Women's Health Initiative Investigators. (2002). Risks and benefits of estrogen plus progestin in healthy postmenopausal women. JAMA, 288(3), 321–333. https://doi.org/10.1001/jama.288.3.321