verifiedEvidence-Based Practice

Randomized Controlled Trial

The random draw that separates the effect of a treatment from what merely happened alongside it — the strongest design for answering whether an intervention works.

Translated from the Portuguese original. Read the original (Português)

boltIn short

The randomized controlled trial (RCT) is the study in which it is randomly decided who receives the tested treatment and who stays in the comparison group. Random allocation is the heart of the method: it distributes by chance everything that could differentiate the groups — including what no one thought to measure. So if a difference shows up between the groups at the end, the most plausible explanation is the treatment itself. That is why the RCT is considered the strongest design for answering "does this work?" — provided it is well conducted and well reported.

Technical definition

The randomized controlled trial (RCT) is an experimental and prospective study in which eligible participants are allocated through a random process to two or more arms — at least one receiving the intervention under investigation and at least one serving as a comparison (placebo, usual care, waiting list, or another active treatment) — and are then followed over time so that outcomes defined before the study began can be compared. Random allocation, together with allocation concealment and, when feasible, blinding, is what authorizes attributing the observed difference between groups to the intervention. The first modern RCT is usually credited to the British Medical Research Council trial of streptomycin for pulmonary tuberculosis, conducted in 1947–1948 and designed by statistician Austin Bradford Hill.

Why randomizing changes everything

Suppose a clinic offers a new program and, months later, shows that those who took part are doing better. It sounds convincing — until we notice how many other explanations are still standing. Maybe the more organized families were the ones who sought out the program. Maybe the milder cases were the ones accepted. Maybe everyone would have gotten better anyway, with time. The challenge of any research on treatment is this: separating what the intervention caused from what merely happened alongside it.

The way out would be to compare each person with themselves in two parallel universes: one in which they receive the treatment and another in which they don't. Since that is impossible, the RCT does the best available substitute — it builds two groups that differ only by chance. And the tool for that is precisely giving up the power to choose: no one decides who goes where; a random procedure decides.

The gain is subtle and enormous. Matching groups "by hand" only balances what we remember to measure — age, severity, education. Random allocation also balances what we don't even know how to name: motivation, family support, sleep, a thousand invisible factors. No other design can do that.

Eligible participants RANDOM DRAW Group A · intervention receives the tested treatment Group B · control placebo or usual care Follow-up over time — the same measures in both groups Compare the outcomes: the difference estimates the effect of the intervention
The path of a randomized trial. Since only chance separated the two groups at the start, the difference found at the end can be credited to what distinguished them in between — the intervention.

The safeguards that protect a trial

Randomizing is necessary, but not enough. A reliable trial combines random allocation with four safeguards, each one guarding a point where bias could creep back in. Tap the tabs to learn about each one:

No one can guess the next draw

Allocation concealment ensures that whoever recruits and enrolls participants cannot predict which arm the next one will go to — through opaque, numbered, sealed envelopes, or through an independent center that discloses the allocation only after enrollment. Without this safeguard, someone could, even without ill intent, hold back a more severe case until the arm they consider better for it comes up — and randomization loses all its value. Note: allocation concealment is not the same thing as blinding; it acts at the moment the person enters the study.

Who knows what, and why it matters

Blinding (or masking) means keeping participants, those delivering the intervention, those assessing outcomes, and those analyzing the data unaware of who received what. It serves to contain the expectations on both sides — which influence both what the person feels and what the assessor records. In psychotherapy, education, and behavioral interventions, blinding the participant and the clinician is usually impossible: everyone knows therapy is taking place. The accepted solution is to blind whoever assesses the outcome and to prefer measures that are less susceptible to the observer's interpretation.

Everyone stays in the group they were assigned to

In an intention-to-treat analysis, each participant is analyzed in the group to which they were randomized — even if they missed sessions, dropped out of the study, or switched treatments. It seems counterintuitive, but it is what preserves randomization: analyzing only those who "did everything right" recreates precisely the differences between groups that randomization had eliminated, since adhering to treatment is not a matter of chance. The result tends to be more conservative — and closer to what happens in real life.

The plan is published before there is a result

Before recruiting the first participant, the protocol and the primary outcome are registered publicly on platforms such as ClinicalTrials.gov or, in Brazil, ReBEC (Registro Brasileiro de Ensaios Clínicos — the Brazilian clinical trials registry, maintained by Fiocruz and integrated into the WHO's international network of registries). This exposes two common maneuvers: switching the outcome after seeing which measures "worked out" and shelving a study whose result was unwelcome. Major scientific journals require this prospective registration as a condition for publication.

info

Randomization is no miracle in a small sampleRandomizing makes groups comparable on average, not identical. In studies with few participants, chance alone can leave one arm with more severe cases, simply by bad luck. That is why every good trial publishes a table comparing the baseline characteristics of the two groups, and many use stratified randomization — randomizing within subgroups (by age, severity, or treatment site) so they end up well distributed.

Not every randomized trial has the same shape

Random allocation is always the same principle, but the design adapts to the question and the context. Tap the cards to see the most commonly used formats:

How the quality of a trial is judged

Reading a trial with a critical eye does not require being a statistician: there are public tools that organize what to look for. Two are indispensable.

CONSORT is the roadmap for how to report a trial — a checklist of essential items plus a diagram showing the path of participants, from recruitment to analysis. First published in 1996, it was updated in 2001, 2010, and 2025, the version in which it grew to 30 items. When an article fails to report something CONSORT asks for — such as the method used to randomize or how many people dropped out along the way — that omission is, by itself, already a warning sign.

The RoB 2, from Cochrane, is the recommended tool for assessing the risk of bias of a randomized trial result. It examines five domains:

  • the randomization process;
  • the deviations from intended interventions — what actually happened after randomization;
  • the missing outcome data — who was lost from the study, and whether that may have distorted the result;
  • the measurement of the outcome;
  • the selection of the reported result — whether what was published is what had been planned.

Each domain receives a judgment — low risk, some concerns, or high risk — and it is the combination of these that indicates how much that specific result deserves to be trusted.

What the randomized trial does not solve

Standing at the top of the hierarchy of evidence does not make the RCT an answer for everything. It has real limits, and knowing them is part of reading it well:

  • Efficacy is not effectiveness. Trials tend to select participants under strict criteria and offer the intervention under ideal conditions, with trained staff and supervision. The result answers "does it work under these conditions?", not necessarily "does it work at the neighborhood clinic, with the people who actually show up there?"
  • Rare events and long-term effects slip through. A study with a few hundred participants followed for a few months is unlikely to detect an uncommon adverse event or what happens ten years later.
  • Not everything can be randomized. It would be unethical to randomize people to a harmful exposure. For many questions, only observational designs remain — and that is where the answer will have to be sought.
  • It is costly and slow. Large trials require funding, staff, and years of work, which helps explain why less-funded fields have fewer RCTs available.
  • The answer is about the group average. Knowing that a treatment works "on average" does not say whether it will work for a specific person. That other question is the territory of single-subject designs, widely used in behavior analysis, which follow the same person, comparing phases with and without the intervention. The two types of study are complementary, not rivals.
warning

Two common mistakes"It's a randomized trial, so it's proven." The label indicates the design's potential, not the care with which it was carried out: an RCT without allocation concealment, with heavy loss to follow-up, or with the outcome switched partway through can mislead just as much as a weak study. And the opposite mistake: "randomized trials are a drug thing." Psychotherapies, educational programs, and behavioral interventions have been tested in RCTs for decades — what changes is how blinding is done and the choice of comparator, not the logic of randomization.

randomization allocation concealment blinding (masking) intention to treat control group CONSORT

Test your understanding

quiz

Quick quiz

Three questions to reinforce the concept. You'll see the explanation right away.

Further reading

  1. Medical Research Council. "Streptomycin treatment of pulmonary tuberculosis: a Medical Research Council investigation". British Medical Journal, 1948.
  2. Hopewell, S., Chan, A.-W., Collins, G. S. et al. "CONSORT 2025 statement: updated guideline for reporting randomised trials". BMJ / The Lancet / JAMA, 2025.
  3. Sterne, J. A. C. et al. "RoB 2: a revised tool for assessing risk of bias in randomised trials". BMJ, 2019.
  4. Schulz, K. F., Grimes, D. A. "Allocation concealment in randomised trials: defending against deciphering". The Lancet, 2002.
  5. Higgins, J. P. T., Thomas, J. et al. (eds.). Cochrane Handbook for Systematic Reviews of Interventions. Cochrane.
  6. Registro Brasileiro de Ensaios Clínicos (ReBEC) — Fiocruz. Available at ensaiosclinicos.gov.br.
info

This entry is educational and does not replace assessment, diagnosis or treatment by a qualified health professional. Every person is unique — for a specific case, seek individual guidance.

Related entries

Looking for professional guidance?

Instituto Walden4 is a psychology and education center in Brasília, Brazil, offering evidence-based assessment and care for autism, psychotherapy and development — in person and online. Our team serves in Portuguese; get in touch and we will find the best way to help.

chat Message us on WhatsApp
chat

Conta Walden4

Entrar

Use o e-mail e a senha da sua conta.