ABSTRACT
Random allocation of experimental units to the different treatments under study is the defi ning feature of the randomised controlled trial. The rationale for randomisation in individually randomised trials has been widely discussed. The main reasons for adopting randomisation include the following:
Impartiality: Random allocation provides assurance to the investigators and to the scientifi c community that the treatments have been allocated fairly and impartially. If non-random methods are used, it is diffi cult to avoid the suspicion that the investigators may have been biased, either consciously or unconsciously, in their allocation of units to treatment arms, probably in a direction that would favour the intervention under study. Transparency: Random allocation provides a convenient method of allocating treatments that is transparent, objective and easily described. The method could in principle be replicated by other investigators, and this is an important principle of medical experimentation. Balance: As noted in Chapter 5, random allocation ensures, if the sample size is large, that the treatment arms are similar with respect to all factors other than the treatment under study. In other words, they are balanced on potential confounding factors, both known and unknown. This helps to ensure the credibility or face validity of trial results and may also improve precision. Blinding: Use of a random allocation scheme that is concealed from participants or study personnel may reduce their ability to guess which participants have received which treatment condition and thus help to ensure blinding. Inference: The use of randomisation can provide a formal justifi cation for the use of statistical inference to make probability statements about study hypotheses or confi dence statements about parameter
•
•
•
•
•
values. This applies particularly to the permutation test and related statistical methods. Model-based approaches to analysis do not formally rely upon randomisation, however.
