Blocked randomization in unblinded trials

Some combinations of methods for sequence generation, allocation concealment and blinding act together to create a risk of selection bias in the allocation of interventions. One particular combination is the use of blocked randomization in an unblinded trial, or in a blinded trial where the blinding is broken, for example because of characteristic side effects. When blocked randomization is used, and when the assignments are revealed subsequent to the person recruiting into the trial, then it is sometimes possible to predict future assignments. This is particularly the case when blocks are of a fixed size and are not divided across multiple recruitment centres. This ability to predict future assignments can happen even when allocation concealment is adequate according to the criteria suggested in Table 8.5.d (Berger 2005).