Block Randomization

A randomisation technique assigning participants to groups in balanced blocks of a set size, keeping numbers roughly equal throughout enrolment.

Source: Friedman LM, Furberg CD, DeMets DL, Reboussin DM, Granger CB. Fundamentals of Clinical Trials. 5th ed. Springer; 2015. doi:10.1007/978-3-319-18539-2.

Simple randomisation, like tossing a coin for each patient, can by chance leave the groups markedly unequal in size, especially in a small trial or at any interim point. Block randomisation solves this by allocating patients in balanced blocks, so that after each block the groups are even, keeping their sizes close throughout enrolment. This guards against the imbalance that simple randomisation can produce and ensures adequate numbers in each arm at all stages. It trades a little unpredictability for balance. Schulz and Grimes (2002) describe it.

Source: Schulz & Grimes 2002

Block randomisation works by dividing enrolment into blocks of a chosen size and, within each block, randomly ordering a fixed set of assignments that balances the treatments, so that after each completed block the groups are equal in size. As participants are enrolled, they are assigned according to the random order within the current block, and a new block begins when one is filled. This guarantees balance at the end of every block, keeping the group sizes close throughout the trial rather than only in expectation as with simple randomisation.

Source: Friedman, Furberg & DeMets 2015

Block randomisation is used to keep the treatment groups balanced in size throughout enrolment, avoiding the sizeable chance imbalances that simple randomisation can produce, especially in smaller trials or when analyses occur before enrolment is complete. Balanced group sizes improve statistical efficiency and ensure that if the trial stops or is analysed partway, the groups are comparable in number. Block randomisation thus provides the benefits of randomisation while controlling group sizes, making it a common choice when maintaining balance during the trial is important.

Source: Friedman, Furberg & DeMets 2015

The risk of predictability in block randomisation is that, if the block size is known and fixed, the assignments toward the end of a block can become predictable, since the need to balance the block determines the remaining assignments, which could allow those enrolling participants to guess upcoming allocations and undermine allocation concealment. This risk is reduced by using randomly varying block sizes and by keeping the block size unknown to those enrolling. Managing this predictability is important so that block randomisation does not compromise concealment and introduce selection bias.

Source: Friedman, Furberg & DeMets 2015

Block randomisation differs from simple randomisation in that it assigns participants in balanced blocks to keep group sizes equal throughout enrolment, whereas simple randomisation assigns each participant independently, balancing groups only in expectation and allowing chance imbalances, especially in small trials. Block randomisation guarantees balance at the end of each block but can risk predictability if blocks are fixed, while simple randomisation avoids predictability but permits imbalance. So block randomisation trades a small predictability risk for guaranteed balance, making it preferable when controlling group sizes during the trial matters.

Source: Friedman, Furberg & DeMets 2015