Randomization

Aim

To formulate quality requirements for randomization.

 

Requirements

  • Randomization sequence must be generated adequately
  • Allocation to treatment must be concealed
  • Randomization and allocation performed by an independent researcher.
  • Clear documentation of the randomization procedure.
  • Proof that the randomization procedure has been checked on success.
  • Corrected analyses for confounding when randomization was not succeeded.

 

Documentation

Clear description of the randomization procedure, including randomization check and (possible existing) confounders in research protocol or logbook.

 

Responsibilities

Executing researcher: Follow the steps to ensure random allocation to treatment groups for subjects:
  • Define randomization method;
  • Specify who, when and how the randomization procedure will do/take place;
  • Ask supervisor for checking the randomization procedure;
  • Ask independent individual to perform randomization;
  • Never change sequence allocation of subjects;
  • Save allocation;
  • Consider whether blindness can be realized.
Project leaders:
  • Check the randomization procedure;
  • Determine whether blindness can be realized.
Research assistant: N.a.

 

How To

Randomization is frequently applied in randomized clinical trials (RCTs). Allocation to treatment groups has to occur truly at random. This can be achieved by using a software randomization program. You can download one for free here.
When considering randomization you have to make a decision about (1) unit of randomization (individual level or cluster randomisation), (2) allocation ratio (1:1 or different), (3) the use of blocks (fixed number or random number), and (4) possible stratification.
Block randomization ensures that the number of subjects is (almost) always the same for each group. For instance, in a block of four (or eight, sixteen) there will be two (or four, eight) subjects allocated to the control group and two (or four, eight) to the treatment group. However, the disadvantage is that it is easier to guess what the next randomization outcome will be (in a block of 4 you always know what the 4ht outcome is). To solve this you might use random block sizes.
You use stratification in case you know there is a strong confounder (e.g. effect will be very different for men than for women) and you want to make sure that that confounder is distributed equally. You then randomize within strate (e.g. males and females).
Allocation to treatment must be concealed. This means that the person who includes a patient in a trial is not aware what the next randomisation outcome will be. There are several methods to achieve this e.g. by using sealed envelopes, by asking a colleague to perform the actual allocation, or by using a trial service (e.g. https://www.sealedenvelope.com).
Randomisation is intended to minimalize the effect of confounders (Pocock) since you expect those possible confounders to be distributed equally between the groups. However, by accident the distribution might turn out not to be equal. You can do 2 things to solve this. First, while randomizing you might use stratification. This means that you randomize within certain groups (e.g. males and females) and thus ensure that the variable (gender) is distributed equally. Stratifying becomes more important when there are strong confounders or the study group is small. Second, you might check after randomization if the ‘baseline statistics’ (age, sex, etc), are distributed equally across study groups. When this is not the case, it is necessary to correct in the analysis for the confounders by including them as variables in the data set. To be able to correct for the appropriate confounders, these need to be recorded during the study.
Following next steps will help you to ensure random allocation to treatment groups:
  • Use the CONSORT statement to decide which randomization method fits best within your research;
  • Specify in your protocol who, when and how your randomization and allocation procedure will do/take place;
  • Ask you supervisor to check your randomization procedure;
  • Ask an individual who is not involved in the project, and is blind for characteristics of the subjects, to perform the randomization (and allocation);
  • Never change the sequence of allocated subjects;
  • Save the allocation and use your logbook to note the date of randomization;
  • Think about whether blindness of researchers and subjects for conditions can be realized.

 

Appendices/references/links

  • Pocock SJ. Clinical trials: A statistician’s perspective. In: Adèr HJ, Mellenbergh GJ, eds., Research Methodology in the Social and Behavioural & Life Sciences. London Thousand Oaks New Delhi: SAGE Publ., 1999; 96-109.

 

 Audit questions

  1. Has the randomisaton sequence been generated adequately?
  2. Is the allocation concealed?
  3. Has the procedure followed for randomization been described in detail, and has the researcher maintained a logbook including any potential changes in the protocols?
  4. Have the most important confounders been included as variables in the data set?
  5. Did randomization work? If something went wrong with randomization, how was this corrected for the data analysis?

 

V3.0: 12 Aug 2016: Text updated
V2.0: 12 May 2015: Revision format
V1.4: 31 Oct 2013: Addition of practical steps
V1.3: 1 Jan 2010: Translation into English
V1.2: 11 Oct 2007: Text abbreviated, concept of pre-stratification described
V1.1: 23 Mar 2006: Minor textual modifications and references to randomization software