A study that randomly assigns people to treatments to test cause and effect.
A randomised controlled trial (RCT) is a study design where participants are allocated to an intervention group or a comparison group purely by chance, typically using a computer-generated sequence. Because allocation is random rather than chosen by clinician or patient, the groups should be similar at baseline in both known and unknown factors that could affect outcomes. This means any difference seen afterwards can more confidently be attributed to the intervention itself rather than to who ended up in which group.
Randomisation guards specifically against selection bias and confounding, the problems that arise when sicker, more motivated, or otherwise different patients end up disproportionately in one group. Without it, an apparent treatment effect could simply reflect pre-existing differences between groups rather than the intervention. Blinding, where used alongside randomisation, further protects against performance and detection bias by stopping participants or assessors from behaving or scoring differently based on knowledge of group assignment.
In a paper, look for a clear description of the randomisation method, allocation concealment, and who was blinded (participants, treating clinicians, outcome assessors), often summarised against a checklist such as CONSORT. Risk-of-bias tools like the Cochrane RoB 2 assess exactly these domains, alongside deviations from intended intervention, missing outcome data, and selective reporting. A baseline characteristics table is usually presented so readers can judge whether randomisation produced comparable groups in practice.
Randomisation controls for confounding but does not guarantee validity: a trial can still be undermined by an inadequate sample size, poor adherence to the allocated treatment, high or differential dropout, or a study population that does not resemble the patients seen in everyday clinical practice. Blinding is often difficult or impossible in physiotherapy trials involving manual therapy or exercise, which leaves performance and detection bias as a persisting concern even in an otherwise well-randomised trial. A single RCT, however well conducted, should also be weighed alongside the wider body of evidence rather than treated as definitive on its own.
This guide was auto-drafted and is pending editorial review.