Better reporting of randomised controlled trials: the CONSORT statementBMJ 1996; 313 doi: http://dx.doi.org/10.1136/bmj.313.7057.570 (Published 07 September 1996) Cite this as: BMJ 1996;313:570
- Douglas G Altman
- Head ICRF Medical Statistics Group, Centre for Statistics in Medicine, Institute of Health Sciences, Oxford OX3 7LF
Authors must provide enough information for readers to know how the trial was performed
Randomised controlled trials are the best way to compare the effectiveness of different interventions. Only randomised trials allow valid inferences of cause and effect. Only randomised trials have the potential directly to affect patient care—occasionally as single trials but more often as the body of evidence from several trials, whether or not combined formally by meta-analysis. It is thus entirely reasonable to require higher standards for papers reporting randomised trials than those describing other types of study.
Like all studies, randomised trials are open to bias if done badly.1 It is thus essential that randomised trials are done well and reported adequately. Readers should not have to infer what was probably done, they should be told explicitly. Proper methodology should be used and be seen to have been used. Yet reviews of published trials have consistently found major deficiencies in reporting,2 3 4 making the task for those carrying out systematic reviews much harder. Almost 50 years after the first publication of a randomised trial,5 the guarantee of adequate reporting of these important studies is surely long overdue.
In 1994 two groups independently published proposals for requirements for the reporting of randomised trials.6 7 In an editorial in JAMA Drummond Rennie suggested that the two groups should combine to produce a unified statement,8 and the outcome of this process was published last week.9 The new CONSORT statement lists 21 items that should be included in a report (see table 1) as well as a flow chart describing patient progress through the trial (fig 1). In addition, a few specific subheadings are suggested within the methods and results sections of the paper. In the spirit of the times, the recommendations are evidence based where possible, with common sense dictating the remainder.
In essence the requirement is that authors should provide enough information for readers to know how the trial was performed so that they can judge whether the findings are likely to be reliable. The CONSORT statement means that authors will no longer be able to hide inadequacies in their study by omitting important information. For example, at present authors can, and often do, hide their procedures behind the single word “randomised.” Authors will now be required to give details of the randomisation procedure. If authors have used an inferior approach, such as alternate allocation, they will have to say so. The BMJ has in fact refused to publish trials that were not truly randomised since 1991,10 a position justified by subsequent empirical findings.1
As the authors of the CONSORT statement note,9 the checklist applies to the most common design of randomised trial—trials with two parallel groups. Some modification is needed for special types of trial such as crossover trials and those with more than two treatment groups. Also, the list should be taken in conjunction with existing general requirements—for example, the requirement to specify all statistical methods used in the analysis. This and other items appear on the checklist for controlled trials that has been used by the BMJ's statistical referees for over 10 years.11
Some of the items on the checklist would benefit from greater explanation than is possible in the CONSORT statement. In time a fuller accompanying explanatory paper could be valuable. For example, while the advantages of randomisation have been apparent for several decades, understanding the rationale for it remains poor and so its importance is not fully appreciated by researchers.12
The BMJ supports the CONSORT statement and is adopting its recommendations. So too are JAMA, Lancet, and some other journals. Trialists are encouraged to follow the statement right away, but from 1 January 1997 they will be required to do so. Authors should submit with their papers a copy of the completed checklist indicating on which page of the manuscript each item is addressed. The checklist will be used by the editors and supplied to referees. In the published papers the BMJ will use the additional subheadings suggested by CONSORT.
It seems reasonable to hope that, in addition to improved reporting, the wide adoption of this new publication standard will improve the conduct of future research by increasing awareness of the requirements for a good trial. Such success might lead to similar initiatives for other types of research.