Randomized Controlled Trial (RCT)

What makes a trial “randomized”

Randomization means the sequence that decides which arm each participant is allocated to is generated purely by chance — a computer-generated random number sequence, a random-permuted-block schedule, or (for multi-site or unbalanced-covariate designs) a minimization or stratified-randomization algorithm. Two things have to hold for the randomization to be sound: the sequence itself must be genuinely unpredictable, and its implementation must be concealed from whoever is enrolling participants until the moment of assignment (allocation concealment). Concealment is what stops a coordinator from, even unconsciously, steering healthier or sicker participants toward one arm. In multi-site trials, randomization is typically centralized through an Interactive Response Technology (IRT/IWRS) system rather than left to site-level sealed envelopes, which also ties directly into blinded drug supply and kit dispensing — see Clinical Trial Supply Management and Clinical Trial Management System (CTMS).

What makes a trial “controlled”

A trial is “controlled” when it includes a comparison group — the control arm — that does not receive the intervention under investigation, so that the outcome in the intervention arm can be measured against a genuine comparator rather than against history, expectation, or the natural course of the condition. A trial can be randomized without being controlled only in the narrow sense of comparing two active doses of the same drug against each other (which is itself usually classified as a form of active control); a single-arm study with no concurrent comparison group, however doses or sequences are assigned within it, is not an RCT.

Types of control arms

ICH E10 (“Choice of Control Group and Related Issues in Clinical Trials,” the harmonized guideline jointly adopted by FDA, EMA, and Japan’s PMDA) sets out the standard categories:

• Placebo concurrent control — an inert or sham treatment, indistinguishable in appearance/administration from the study intervention, used where withholding active treatment is ethically acceptable (e.g., no proven effective therapy exists, or the added risk of a short placebo period is low).
• Active (positive) comparator control — an existing, approved treatment for the condition, used either to show superiority over current standard care or, in a non-inferiority design, that the new intervention is not unacceptably worse.
• Standard-of-care control — participants continue to receive usual clinical management (which may itself include an active drug) rather than a specifically constructed comparator arm; common in pragmatic and behavioral-intervention trials where a placebo is impractical or unethical.
• Dose-response concurrent control — two or more doses (including possibly a zero dose) of the study drug are compared against each other.
• No-treatment concurrent control — participants receive no intervention at all, used only where blinding is not feasible or necessary and an unblinded no-treatment arm doesn’t itself bias reporting.
• External (including historical) control — outcomes are compared against a separate, previously studied population rather than a concurrently randomized group; ICH E10 treats this as the weakest form of control because it is not protected by randomization at all, and it is generally reserved for settings where a concurrent control is infeasible (e.g., some rare-disease trials).

Why randomization matters

Random allocation is the mechanism that protects a trial against selection bias — the systematic error introduced when the characteristics of participants assigned to one arm differ, on average, from those assigned to the other in ways that also affect the outcome. Because assignment is left to chance rather than to clinician judgment, patient preference, or disease severity, randomization tends to balance both known and unknown confounding variables across arms in expectation, which is precisely what lets investigators attribute a measured difference in outcomes to the intervention rather than to pre-existing differences between groups. This is the core reason RCTs sit above observational designs (cohort studies, case-control studies) in most evidence hierarchies for causal claims about an intervention’s effect, and why CONSORT’s reporting checklist requires authors to describe the randomization method, allocation-concealment mechanism, and any deviations from the intended sequence — see CONSORT 2010.

Blinding is related, but a separate design choice

Blinding (also called masking) concerns who knows which arm a participant was allocated to after randomization has already happened — participants, treating clinicians, outcome assessors, and/or data analysts. It is a distinct axis from randomization and control: an RCT can be single-blind (participants unaware), double-blind (participants and assessors/clinicians unaware), or open-label (no one blinded, as is common in surgical, behavioral, or pragmatic-care trials where blinding is not feasible) and still be a fully valid randomized controlled trial. Blinding reduces a different source of bias — performance bias and detection/ascertainment bias arising from expectation effects during the trial — rather than the selection bias that randomization addresses at the point of enrollment. A trial that is randomized and controlled but not blinded is still an RCT; it simply carries a different, explicitly reportable risk-of-bias profile.

How RCT relates to other clinical-research concepts on this site

RCT describes a study’s design (how participants are assigned and compared), which is a different axis from a trial’s regulatory phase (how far along the intervention is in development) — see Clinical Trial Phases. A trial can be, and often is, both an RCT and a Phase 2 or Phase 3 study at the same time; the two labels answer different questions. For the broader regulatory definition of what counts as a “clinical trial” at all (a question NIH and FDAAA answer slightly differently), see What Is a Clinical Trial? The NIH Definition Explained. Once an RCT is underway, its registration and results-reporting obligations are covered separately — see Clinical Trial Registration and Reporting Compliance. For the full clinical-research vocabulary and standards landscape, see the Clinical Research pillar.

References

• ICH E10, “Choice of Control Group and Related Issues in Clinical Trials” (step 5, adopted by FDA/EMA/PMDA) — EMA guideline text.
• Schulz, Altman, Moher, “CONSORT 2010 Statement: updated guidelines for reporting parallel group randomised trials,” BMJ (2010) — consort-statement.org.
• Cochrane Handbook for Systematic Reviews of Interventions, chapters on randomization sequence generation, allocation concealment, and risk-of-bias assessment.