*Geek Box: Randomisation Methods
Randomisation provides a means of minimizing the risk of bias, and is most effective particularly where there are unknown factors which could influence the results: randomisation then balances the distribution of these unknown factors between groups. There are a number of type of randomisation: simple, block, and stratified, being common randomisation methods.
Simple randomisation involves dividing subjects at random to allocation of either treatment group or control, either using computer generated or numbered tables. The essential characteristic of simple randomisation is that all subjects have an equal chance of allocation to either group. A potential disadvantage may be observed in trials with small sample sizes, where simple randomisation may result in unequal numbers between groups.
To address this, block randomisation divides subjects into multiple, smaller groups of equal subjects based on a predetermined group size, allowing for control of balance across similar-sized groups. A potential disadvantage to block randomisation is that the process of randomisation may lead to different covariates between block groups. To address this potential, stratified randomisation is used to achieve balance of characteristics between block groups.
Stratification allows for covariate baseline characteristics, for example age, cholesterol, hypertension, and/or sex, to be balanced by grouping together particular characteristics which could influence the outcome. Subjects would then be randomised to blocks, within that particular stratum identified by the covariate characteristic, i.e., similar numbers of subjects with the same level of blood cholesterol. Disadvantages to stratified randomisation arise where there are multiple covariates in small sample sizes, or where subjects are enrolled on an ongoing basis, and baseline characteristics of all subjects are not available prior to randomisation.