Randomization

Randomization is a random distribution of subjects or study patients to two or more groups in a clinical study. Randomisation is intended to exclude influencing variables on the result, so that it is considered a control technique for selecting a representative sample from the population. All confounders on the result are distributed equally to all groups by randomisation. The aim of randomisation is thus to exclude confounders and to prevent bias, i.e. systematic distortion. Randomised controlled studies are considered the gold standard in clinical research. In accordance with the CONSORT statement, a presentation of the methodology used in randomisation is required in reporting.

Randomization was developed in the 1920s by R.A. Fisher as a basic principle of experimental research and was introduced in medical studies in the 1940s.

A prerequisite for valid randomisation is a sufficiently large number to prevent random distributions. Furthermore, an equally large group allocation should be made (balance) so that interim analyses can be carried out.

In principle, there are two types of group assignments:

• In the case of centralized randomization, independent group assignment is carried out by an independent clinical research organization or software.
• With decentralised randomisation, group allocation is carried out on site, e.g. by drawing envelopes, and is often predictable and can therefore be influenced by the investigator.

The following procedures are distinguished:

• Quasi-randomisation: Here the allocation is carried out by a predictable mechanism (e.g. alternating allocation or allocation according to even and odd birthdays). Disadvantages are the high risk of a selection bias and the numerous disruptive effects.
• Simple, unrestricted randomisation: Here, group allocations are randomly made without any balance.
• Block randomisation with balancing: In contrast to simple procedures, here attention is paid to group allocation in proportion. Each defined block restores balance.
• Adaptive Randomization: Here the group allocation is dynamic, taking into account the previous distributions.

Randomization methods can thus range from very easy and manipulable group assignments to complex, elaborate algorithms. In practice, allocation has been established by means of software into which different designs can be programmed.