Randomization in clinical research

Sir:

I read the article by Dr Johanna van Rhijn and colleagues [1] with interest, especially because the methods section in the abstract stated “Twenty-one patients were randomized to … ” [my italics]. I am interested in all randomized studies. However, on continuing the sentence I read, “ … to three groups according to impairment severity … ”, which greatly concerned me. The detailed description in the paper confirmed my concerns—this was not a study using randomization. The abstract is in error.

As a reader of many research articles, a reviewer of many articles and grant applications, and as the editor of a journal, I am repeatedly struck by how unaware many people are both of the meaning of the word random, and of the reason for using randomization.

Random means “by chance”. It implies unpredictability. Randomization in the context of scientific research means that chance (in the true sense of the word) determines the outcome of a decision. It means that a stochastic process must be involved.

Random does not equate to:

• arbitrary (i.e. not due to any obvious system);

• alternate, or based on order of recruitment in some way;

• being based on some variable or fact thought to be un-associated with outcome, such as time of day, birth date, availability of a bed, etc.;

• being based on some patient variable (such as in this study [1]).

Randomization is used as one method to reduce bias, a systematic error, or variance in the collection or interpretation of data. Classically, it is used to determine which of two or more treatments someone has, but it can be used wherever a choice or decision can be made. One might randomize:

• different treatments/interventions;

• timing of starting a treatment (in a single case-study);

• order of treatments/interventions/observations/tests;

• the observer making observations.

In this study the patients were allocated to groups according to the nature and severity of their impairments, and also the nature of the goals of treatment. This is the opposite of random allocation; the group of each patient was defined precisely by the rules.

I hope that researchers and authors will restrict their use of the words random and randomization to decisions that are indeed unpredictable and governed entirely by a stochastic process.

Derick T Wade

Consultant and Professor in Neurological Rehabilitation Neurological Rehabilitation Service, Oxford Centre for Enablement

Windmill Road, Oxford OX3 7LD, UK