Randomized controlled trial

A randomized controlled trial estimates the effect of an intervention by comparing groups formed through random allocation. Participants might receive a drug, procedure, behavioral program, usual care, placebo, or another active treatment; the comparator defines what the intervention is being judged against.¹ Randomization balances measured and unmeasured prognostic factors in expectation, so a difference in outcomes can more plausibly be attributed to the assigned intervention rather than to initial health, preferences, or chance imbalances.

Researchers specify eligibility criteria, outcomes, follow-up, and an analysis plan before examining results. Allocation concealment prevents recruiters from knowing the next assignment, whereas blinding conceals assignments after enrollment; the two safeguards address different sources of bias.² Blinding may involve participants, clinicians, outcome assessors, or statisticians, but is not always feasible for surgery or public-health policy.

A trial’s credibility rests on the alignment of randomization, follow-up, and analysis. A computer-generated sequence, centralized service, or sealed method can implement assignment; stratification or minimization may improve balance for especially important characteristics. The primary outcome and time point should be prespecified, and sample size is usually calculated from an anticipated effect, outcome variability, statistical power, and chosen error rate.¹

The intention-to-treat principle analyzes participants according to their assigned groups, preserving the benefit of randomization even when treatment is stopped, crossed over, or not taken. Per-protocol analyses can describe effects among participants who followed the regimen, but may reintroduce confounding. The CONSORT statement asks reports to show participant flow, exclusions, harms, registration, and protocol-relevant details so readers can judge selective reporting and attrition.²

Randomization supports causal inference within the trial, but it does not guarantee a clinically important or universally applicable result. A statistically significant estimate can be too small to matter to patients, while a wide confidence interval may leave important effects unresolved. Missing outcomes, deviations from treatment, early stopping, multiple unplanned analyses, and selective publication can distort the apparent result.¹

External validity depends on who was enrolled, what care they received, and whether the trial setting resembles ordinary practice. Pragmatic trials often prioritize routine patients and clinically meaningful outcomes, whereas explanatory trials test efficacy under tightly controlled conditions. Ethical review, informed consent, independent monitoring, and genuine clinical equipoise are expected; equipoise means that responsible experts remain uncertain about which intervention is preferable at the start of the trial.⁵

RCTs have several designs beyond the familiar parallel-group format. In a cluster-randomized trial, whole clinics, schools, or communities are assigned, requiring analysis that accounts for correlation among people in the same cluster. A crossover trial gives participants more than one intervention in sequence and can be efficient for stable, reversible conditions, but carryover effects and period effects can invalidate comparisons. Factorial trials test combinations of interventions, while noninferiority trials ask whether a new treatment is not unacceptably worse than an established one under a prespecified margin.¹

The 1948 streptomycin trial for pulmonary tuberculosis, associated with the Medical Research Council, helped establish randomized allocation and blinded assessment as practical tools in clinical research.³ Modern trials may also be adaptive, platform-based, or embedded in routine health systems; these approaches can evaluate several questions efficiently, but require prespecified rules and careful control of statistical error.⁴

Allocation concealment

Keeping the upcoming treatment assignment unknown to recruiters until a participant has been enrolled.

Blinding

Withholding knowledge of assigned intervention from specified participants, clinicians, assessors, or analysts.

Intention-to-treat

Analyzing participants in the groups to which they were randomized, regardless of adherence or later treatment changes.

Clinical equipoise

A state of genuine professional uncertainty about which study intervention is superior.