Randomized Clinical Trial (RCT)
- Introduced by
- Austin Bradford Hill; MRC Streptomycin Trial team
- Year
- 1948 (first rigorously conducted RCT — MRC streptomycin trial)
- Type
- Interventional experimental study
- Family
- Clinical / epidemiology
What does it do?
A randomized clinical trial (RCT) is an experimental study design in which participants are randomly assigned to an intervention group or a control group, then followed prospectively to compare outcomes.
Random allocation is the defining feature: it distributes known and unknown confounders across groups by chance, making the RCT the strongest individual study design for establishing causal efficacy of a treatment or intervention under controlled conditions.
The intuition behind the method
Think of an RCT as flipping a coin for every participant — heads goes to the treatment group, tails to the control group. Because the coin has no memory of age, sex, smoking history, or any other characteristic, those factors end up roughly balanced between the two groups without the researcher having to measure or control for each one explicitly. After the trial period, any systematic difference in outcomes between the groups can be attributed to the intervention rather than pre-existing differences between participants. This coin-flip logic is what sets RCTs apart from observational designs and why they sit at the top of most evidence hierarchies.
History
The modern RCT is conventionally traced to the 1948 Medical Research Council trial of streptomycin for pulmonary tuberculosis, designed by Austin Bradford Hill and colleagues, which introduced rigorous randomisation and allocation concealment into clinical research. Hill later formalised the statistical principles of clinical trial design in his landmark textbook. The CONSORT (Consolidated Standards of Reporting Trials) statement, first published in 1996 and revised in 2001 and 2010, standardised RCT reporting worldwide. RCTs are now the regulatory gold standard for evaluating new therapies across medicine.
When to use it
Use an RCT when you need to establish whether an intervention causes an improvement in outcomes — for drug approval, device evaluation, surgical technique comparison, or behavioural intervention testing. It is the appropriate design when randomisation is ethically feasible, when there is genuine clinical equipoise, and when the research question is specific enough to be answered by a defined protocol. Do NOT use an RCT when the exposure cannot be randomly assigned (e.g., genetic traits, smoking history), when the condition is too rare to recruit sufficient participants, when long latency between exposure and outcome makes follow-up impractical, or when withholding a proven effective treatment from a control group would be unethical. Observational designs (cohort, case-control) are more appropriate in those scenarios.
Limitations and considerations
- Strict eligibility criteria and controlled settings produce highly selected samples that may not represent routine clinical populations, limiting external validity.
- Expensive, time-consuming, and logistically complex — particularly for long-term outcomes or rare diseases.
- Ethical constraints prevent randomisation for many exposures and situations where treatment withholding is not justifiable.
- Participant adherence and crossover between arms can dilute the estimated treatment effect in ITT analyses.
- Short follow-up windows may miss delayed adverse events or long-term efficacy differences.
Source conditions do not replace independent scientific review or validation by an analysis engine.
Strengths
- Random allocation controls for both measured and unmeasured confounding, providing the strongest evidence for causality of any individual study design.
- Pre-registration and prospective protocols reduce researcher degrees of freedom and selective reporting.
- Blinding reduces performance, placebo, and ascertainment bias when feasible.
- Regulatory bodies (FDA, EMA) accept well-conducted RCTs as the primary evidence basis for drug and device approval.
- Results from multiple RCTs can be pooled in systematic reviews and meta-analyses to increase precision.
Applications
- Evaluating the efficacy and safety of new drugs or vaccines before regulatory approval.
- Comparing surgical procedures or medical devices against standard-of-care alternatives.
- Testing behavioural, psychological, or public-health interventions in community or clinical settings.
- Assessing non-inferiority or equivalence of a new treatment to an established one.
- Generating evidence for clinical practice guideline recommendations rated as Grade A or 1a.
How it works
How it works
The research team specifies a PICO framework: Population (eligibility criteria), Intervention, Comparator (placebo, standard of care, or active control), and primary Outcome(s). Sample size is calculated a priori from the expected effect size, desired power (commonly 80–90%), and significance level (commonly 0.05). A pre-registered analysis plan and data safety monitoring board are established before enrolment begins.
Eligible consenting participants are allocated to study arms using a random mechanism — computer-generated sequences, block randomisation, or stratified randomisation — to achieve balance across known prognostic factors. Allocation concealment (e.g., sealed opaque envelopes or central randomisation) ensures that the person enrolling a participant cannot predict or influence which arm they enter.
Blinding (masking) reduces performance and ascertainment bias. In a double-blind trial neither the participant nor the outcome assessor knows the allocation; in a single-blind trial only one of these parties is masked. Open-label trials (no blinding) are used when masking is impractical (e.g., surgical vs. medical treatment), but they require extra vigilance about objective outcome assessment.
Participants in the intervention arm receive the experimental treatment according to the protocol; the control arm receives the comparator. Protocol adherence is monitored through pill counts, drug level assays, or attendance records. Deviations and unscheduled withdrawals are recorded to support later intention-to-treat (ITT) and per-protocol analyses.
Outcomes are assessed at pre-specified time points by blinded or independent assessors. The primary endpoint — defined before the trial — drives the confirmatory analysis. Secondary and exploratory endpoints are pre-specified to limit multiplicity. Adverse event monitoring runs continuously; serious adverse events trigger safety reviews.
The primary analysis compares the primary outcome between arms, typically using a two-sample t-test, chi-squared test, log-rank test, or Cox regression, depending on outcome type. The ITT principle (analysing all randomised participants in their assigned arm regardless of adherence) preserves the prognostic balance created by randomisation and is the regulatory standard. Effect sizes are reported with confidence intervals and p-values; absolute risk differences are preferred alongside relative measures.
Common pitfalls
- Inadequate allocation concealment — if the enrolling clinician can predict group assignment, they may selectively enrol patients, reintroducing selection bias and negating the randomisation benefit.
- Underpowering — a sample size not calculated from a realistic effect size often produces inconclusive results; post-hoc power calculations do not rescue an underpowered study.
- Multiple primary endpoints without multiplicity adjustment inflate the Type I error rate and make it impossible to interpret significance claims.
- Reporting per-protocol results as the primary analysis — this excludes non-adherers in a non-random way and may overestimate efficacy; ITT analysis should always be reported as the primary result.
- CONSORT non-compliance — omitting key flow diagram information (screened, excluded, randomised, lost to follow-up) makes quality assessment impossible and reduces trust in findings.
Frequently asked questions
What is the difference between a randomized trial and a quasi-experimental study?
A true RCT uses random allocation — every eligible participant has a known, chance-based probability of entering each arm. A quasi-experimental study uses non-random assignment rules (alternating allocation, cut-off scores, historical controls) that can reintroduce systematic differences between groups. Quasi-experimental designs are useful when randomisation is infeasible but they do not provide the same causal guarantee.
What does intention-to-treat mean and why does it matter?
Intention-to-treat (ITT) means analysing all randomised participants in the group they were assigned to, regardless of whether they completed the treatment or switched arms. This preserves the balance achieved by randomisation. Excluding non-adherent participants in a per-protocol analysis introduces selection bias because adherence is often related to prognosis, and it typically overestimates efficacy.
How many participants do I need?
Sample size depends on the expected effect size (from pilot data or the literature), the desired statistical power (usually 80–90%), the significance threshold (usually 0.05), and the anticipated dropout rate. There is no universal minimum. A properly powered superiority trial detecting a clinically meaningful difference may need hundreds or even thousands of participants per arm; an equivalence or non-inferiority trial typically requires a larger sample than a superiority trial.
Is blinding always required in an RCT?
Blinding is strongly preferred because it eliminates performance bias (participants changing behaviour based on group knowledge) and detection bias (assessors rating outcomes differently). However, some interventions cannot be masked (e.g., surgery vs. physiotherapy). In open-label trials, using objective, independently adjudicated primary endpoints partially compensates for the absence of blinding.
Should I pre-register my RCT?
Yes. Pre-registration in a public registry (ClinicalTrials.gov, ISRCTN, WHO ICTRP) before enrolment begins is an ethical requirement in most jurisdictions and is expected by journals following ICMJE guidelines. It locks the primary endpoint and analysis plan, preventing post-hoc outcome switching, and allows readers and reviewers to compare what was planned versus what was reported.
Related methods
Sources
Friedman, L. M., Furberg, C. D., DeMets, D. L., Reboussin, D. M., & Granger, C. B. (2015). Fundamentals of Clinical Trials (5th ed.). Springer.
ISBN: 978-3319185385Copied.Schulz, K. F., Altman, D. G., & Moher, D. (2010). CONSORT 2010 Statement: Updated guidelines for reporting parallel group randomised trials. BMJ, 340, c332.
Open DOI referenceCopied.