randomised controlled trial

When a treatment is compared with something else, the two groups have to be alike in everything except the treatment, or a difference in outcome could just as easily be explained by who ended up in which group. A randomised controlled trial does that by dividing participants between the groups by chance rather than by choice. In the National Cancer Institute's words, using chance to divide people into groups means that the groups will be similar and that the effects of the treatments they receive can be compared more fairly.

The comparison group, or control, receives a dummy treatment where no proven option exists, or the existing standard treatment where one does. Allocating by chance also removes the influence of what a doctor or a researcher expects of a particular patient. Many trials go further and conceal the assignment from the participants and the staff, which is called blinding; a trial where everyone knows is an open-label trial. At the point a trial begins it is not known which treatment is better, and that uncertainty is what makes randomising people ethical.

Randomisation is why this design carries the most weight in assessing a treatment, and why studies without it are read more carefully. Work that compares people who happened to receive different treatments, or that compares routine records before and after something changed, can be large and informative but cannot rule out that the groups or the periods differed in some other way. Honest authors of such studies say so, which is why their papers name the trends and reporting shifts they could not exclude.