A scientific study in which two or more groups of patients receive a different drug, the same drug at a different dose, or a placebo, with neither the investigators assessing the outcome nor the people being treated knowing which of these they are receiving. The aim is to remove any hint of bias due to the investigators’ or patients’ preferences or preconceptions. The results are analysed after all the data have been collected and the code has been broken. Trials should have a separate supervising committee, the members of which know the code but do not take part in the study. Their job is to check the results at intervals so they can stop the trial if one arm of treatment is clearly better than another or appears to be harmful.