A randomised controlled trial, commonly abbreviated RCT, is an experimental research study in which participants are allocated to different intervention groups using a random process and the outcomes are compared. Typically, one group receives the intervention being evaluated while another receives a control condition, which may be usual care, placebo, another treatment or sometimes no intervention.
Randomisation is the defining feature because it aims to distribute known and unknown confounding factors between groups by chance, making systematic baseline differences less likely. RCTs are particularly important for determining whether an intervention causes an outcome rather than merely being associated with it. They are widely used for medicines, procedures, behavioural interventions, screening strategies and healthcare policies, although their suitability depends on the research question and ethical considerations.
Randomisation differs from simply assigning people arbitrarily or alternately. A properly generated random sequence determines allocation, ideally with allocation concealment so that researchers enrolling participants cannot predict the next assignment. Concealment helps prevent selection bias before participants enter the intervention groups. After allocation, researchers compare outcomes according to the prespecified analysis plan. Intention-to-treat analysis commonly preserves participants in the groups to which they were originally randomised, regardless of whether they fully adhered to the assigned intervention, helping preserve the benefits of randomisation. Per-protocol analyses can provide additional information about effects among people who followed the protocol, but they may reintroduce bias because adherence is not randomly determined.
A control group provides the counterfactual that researchers need: what would have happened to comparable participants had they not received the experimental intervention? Because the same study population is followed under different assigned conditions, differences in outcomes can be attributed more confidently to the intervention than in many observational designs. However, the control condition matters greatly. A placebo-controlled trial asks a different question from a trial comparing a new medicine with an established standard treatment. A usual-care control may better reflect real-world practice, whereas a placebo can help isolate pharmacological effects. The ethical acceptability of withholding an effective treatment depends on the clinical context and existing evidence.
Blinding is another important feature that can reduce bias. In a single-blind design, participants or investigators may be unaware of allocation; in a double-blind design, both participants and relevant study personnel may be unaware; terminology varies between research traditions, so the exact procedures should be examined rather than relying solely on labels. Blinding can reduce differences in expectations, treatment delivery, outcome assessment and reporting. It is not always possible, particularly for surgery, physiotherapy or behavioural interventions. When blinding cannot be achieved, objective outcomes, blinded outcome assessment and rigorous protocols can still reduce bias. Lack of blinding does not automatically invalidate an RCT, but it can affect confidence in subjective outcomes.
RCTs have limitations as well as strengths. They can be expensive, time-consuming and difficult to conduct, and strict eligibility criteria may make participants less representative of the wider population. Follow-up may be too short to detect rare or long-term harms. Some interventions cannot ethically be randomised, such as deliberate exposure to harmful substances, and some questions concern prognosis or risk factors rather than treatment efficacy. Adherence, loss to follow-up, crossover between groups and deviations from protocol can complicate interpretation. Even a well-designed RCT therefore answers a particular question in a particular population under particular conditions; it does not automatically establish that the same effect will occur for every patient or in every healthcare setting.
RCT results are interpreted using effect estimates and measures of uncertainty rather than simply asking whether a p-value is below a threshold. Outcomes can be reported as absolute risk differences, relative risks, odds ratios, hazard ratios, mean differences or other appropriate measures. Absolute effects are particularly useful for clinical decisions because they show how many additional or fewer events occur.
Number needed to treat can sometimes summarise how many people would need an intervention for one additional beneficial outcome, although its interpretation depends on the time period and baseline risk. Confidence intervals indicate the range of values reasonably compatible with the observed data under the statistical model. Clinical importance should therefore be considered alongside statistical significance.
RCTs are often regarded as particularly strong evidence for intervention efficacy because randomisation reduces confounding, but trial quality varies substantially. Important methodological considerations include adequate random sequence generation, allocation concealment, appropriate control groups, blinding where feasible, complete outcome reporting, prespecified outcomes, adequate sample size and appropriate statistical analysis. Selective reporting can make an intervention appear more effective if favourable outcomes are emphasised while unfavourable outcomes are omitted. Trial registration and publication of protocols help improve transparency. Replication across independent trials and populations can also strengthen confidence that an observed effect is robust rather than a consequence of chance, bias or unusually favourable study conditions.
RCTs are a central component of evidence-based medicine because they can provide direct evidence about causal effects, but they form only one part of the evidence ecosystem. Systematic reviews and meta-analyses can synthesise multiple trials, observational studies can provide information about rare harms and long-term effects, and qualitative research can explain patient experiences and implementation barriers. Medical guidelines integrate different forms of evidence according to the question being addressed.
NICE, for example, describes guideline development as involving systematic identification and appraisal of evidence, interpretation by multidisciplinary committees and consideration of factors such as benefits, harms and applicability. Thus, an RCT is not synonymous with perfect evidence; its value depends on its design, conduct, relevance and the consistency of its findings with the broader evidence base.
