randomised controlled trial

A randomised controlled trial allocates each participant to a treatment group or a comparison group using a random mechanism, usually a computer generated sequence. The comparison group may receive nothing, usual care, a dummy treatment, or a rival treatment.

Randomisation is what the design is for. Because chance decides who goes where, the groups tend to arrive balanced on everything that might otherwise explain a difference in outcome, including factors nobody thought to measure. The Cochrane Handbook puts it plainly: randomisation is the only way to prevent systematic differences between the baseline characteristics of the groups, in both known and unknown confounders. That is why a randomised trial supports a claim about cause and effect more strongly than a study that observes people who chose their own treatment. Research comparing the two finds that non randomised studies tend to report more extreme benefits, and that the size and even the direction of that distortion is hard to predict.

Two further safeguards usually travel with randomisation. Allocation concealment keeps the next assignment hidden from whoever enrols participants, so the sequence cannot be steered. Blinding keeps participants, clinicians or outcome assessors unaware of which group a person is in. A trial can be properly randomised and still be poorly conducted, which is what a risk of bias assessment is for.