How to spot bias and other potential problems in randomised controlled trials
- 1Neurosciences Trials Unit, Division of Clinical Neurosciences, Western General Hospital, Edinburgh, UK
- 2Division of Clinical Neurosciences, University of Edinburgh
- Correspondence to: Dr Steff C Lewis Neurosciences Trials Unit, Division of Clinical Neurosciences, Western General Hospital, Crewe Road, Edinburgh EH4 2XU, UK; steff.lewised.ac.uk
- Received 19 August 2003
- Accepted 24 November 2003
- Revised 24 November 2003
These days, all clinical trials should be reported using the CONSORT guidelines1 (table 1); indeed JNNP recommends this in its instructions for authors. However, not all trials are reported in this way, and many journals do not insist on it. Thus some trials may have been carried out adequately but reported inadequately, while others have been carried out inadequately. Our aim in this article is to guide clinicians in what to look for in a report of a randomised controlled trial (RCT), so they can assess whether the trial was done adequately; we do not intend it to be a guide on how to do an RCT, as there are many such guides available.2
- In this window
- In a new window
Checklist of items to include when reporting a randomised trial (from the CONSORT statement)
The two crucial principles in clinical research are to minimise bias and to increase precision. If a study is not designed with these two principles in mind, no amount of analysis will sort them out. We will discuss some of the major biases to look out for, issues related to precision, and some other aspects of statistical analysis.
BIAS
Bias is any departure of results from the truth. An RCT is less susceptible to bias than other study designs for assessing therapeutic interventions. However, just because a study is randomised does not mean it is unbiased. There are at least seven important potential sources of bias in RCTs, which are discussed below. When assessing bias, it is important to consider its magnitude as well as its direction. Trials that have shown large treatment effects may still be positive after a small bias has been accounted for.
Poor allocation concealment
In a good trial, the treatment allocation is concealed during the randomisation procedure. In other words, at the time a clinician randomises a patient they …
