Randomisation is the process of assigning clinical trial participants to treatment groups such that each participation has a known (usually equal) chance of being assigned to any of the groups
- successful randomisation requires that group assignment cannot be predicted in advance
- knowledge of group allocation should be kept secure until after patient is enrolled; i.e. allocation concealment
- lack of allocation concealment may affect the decision to recruit a patient and so distort the accuracy of the study’s findings (loss of internal validity due to bias from systematic variation)
- minimises selection bias (study sampling involved selection of participants that are not representative of the study population)
- balances known and unknown prognostic factors
- allows probability theory to express the likelihood that any difference in outcome between groups (other than intervention) reflects chance
- facilitates blinding of investigators, participants and evaluators
- trials that lack randomisation introduce bias (systematic variation)
TYPES OF RANDOMISATION
There are 4 main types of randomisation, and all types of randomisation other than ‘simple’ are termed ‘restricted’
- commonest method
- each patient has an equal chance of being allocated to each group
- random numbers or computer-generated list
- may result in unequal numbers allocated to each group or unequal distribution of potential confounding factors (especially in smaller trials)
- aims to keep numbers of patient in each group approximately the same
- e.g. for every block of 8 subjects, 4 must be allocated to each treatment group
- blocks can vary in size
- risks subversion if someone discovers the block size
- pre-identified confounding factors act as criteria for separate randomisation schedules, ensuring they are equalised between the groups
- e.g. block randomisation based on center in multi-center trials; also: age, stage of disease or other potential confounders
- aims to minimise confounding and achieve baseline balance between study groups
- useful for small studies to ensure different strata are represented in the study sample
- not true randomisation, a method of equalising baseline characteristics
- only the first patient is truly randomly allocated, then allocation of subsequent patients is weighted (e.g. 0.8) to minimise imbalance of pre-selected factors at that time
- when wanting to equalise for several confounding variables
- if randomisation is unacceptable due to ethical concerns
- use of a random component is preferred
- unlike true randomisation does not have the theoretical basis for eliminating bias based on all known and unknown factors
- generally considered methodologically equivalent to true randomisation
Other rarely used types of restricted randomisation include the replacement, biased coin, and urn randomisation methods.
REPORTING OF RANDOMISATION IN PUBLICATIONS
- All trials that are not truly randomised are “non-randomised” (terms like “quasi-randomised” are unacceptable)
- method of sequence generation must be specified (e.g. random-number table or a computerised random number generator)
- If study participants are intentionally allocated to groups in different numbers:
- randomisation ratio should be reported
- For block randomisation provide details on:
- how the blocks were generated
- the block size or sizes
- whether the block size was fixed or randomly varied
- trialists becoming aware of the block size(s) (due to risk of code breaking)
- whether stratification was used, and if so:
- which factors were involved
- the categorisation cut-off values within strata
- method used for restriction
- minimisation, If used:
- variables incorporated into the scheme
- if a random element was used
REFERENCES AND LINKS
- Armitage P. The role of randomization in clinical trials. Stat Med 1982;1:345-52.
- Enas GG, Enas NH, Spradlin CT, Wilson MG, Wiltse CG. Baseline comparability in clinical trials: prevention of poststudy anxiety. Drug Information Journal 1990;24:541-8.
- Kleijnen J, Gøtzsche PC, Kunz R, Oxman AD, Chalmers I. So what’s so special about randomisation. In: Maynard A, Chalmers I, eds. Non-random reflections on health services research. BMJ Books, 1997:93-106.
- Lachin JM. Properties of simple randomization in clinical trials.Control Clin Trials 1988;9:312-26.
- Greenland S. Randomization, statistics, and causal inference.Epidemiology 1990;1:421-9.
- Schulz KF. Randomized controlled trials. Clin Obstet Gynecol 1998;41:245-56.
- Schulz KF. Subverting randomization in controlled trials. JAMA. 1995;274:1456-8.
- Treasure T, MacRae KD. Minimisation: the platinum standard for trials?. Randomisation doesn’t guarantee similarity of groups; minimisation does. BMJ. 1998;317:362-3.
FOAM and web resources
- CONSORT Statement — Randomisation and minimisation