Within medical research, trial registers provide a perfect example of structured reporting. So, for example, registering a study on clinicaltrials.gov requires completion of a structured template. Likewise, in the section for subsequent reporting of the results, researchers have to provide the study findings in a structured way. It has been shown that some results posted on clinicaltrials.gov, especially harms, are more complete than those in corresponding journal articles reporting the same trials [23,24].
Journal articles that report research findings are currently in a mainly narrative style, usually supported by tables and perhaps also figures. As Riveros and colleagues observed, ‘using templates with mandatory reporting of some elements may facilitate the work of researchers by reminding them what they need to report and by standardising their reporting’ [23]. It is timely to consider whether journal articles would be more informative if they had more structure.
There are two main forms of structure. First, more structure can be applied to the main text by creating many more sections. One of the antecedents of CONSORT was SORT (Standards of Reporting Trials), a reporting guideline that referred to structured reporting in its title [25]. The authors defined structured reporting as providing sufficiently detailed information about the design, conduct and analysis of the trial for the reader to have confidence that the report is an accurate reflection of what occurred during the various stages of the trial. A single published case study of using each item of the SORT checklist as a heading for a short section was widely considered to be a failure, as it made the article longer and less readable [22]. However, some of the criticisms were related more to the content of SORT than the format. Three subsequent generations of the development of CONSORT have addressed those concerns; it is not reasonable to continue to dismiss the principle of structure based on a single case study published 20 years ago using a different checklist. In fact, much the same idea has been implemented elsewhere. The journal PLoS Clinical Trials, published from 2006–2007, created a template based on the then current CONSORT checklist of 2001. Each item in the CONSORT checklist was used a heading in the article. Although this was in essence a repeat of the SORT approach, there were fewer headings and they were structured within the IMRAD format.
Alternatively, authors can identify in the text where each element of the CONSORT checklist is addressed [26]. This approach has no impact on the format of the article, but requires the authors to mark up the text. While that format should help to avoid omissions, some checklist items do not refer to a specific piece of information in a single location, and so this is not in my view a workable solution in general.
A different approach to structure is to adopt a more tabular format akin to completing a tax form, as is required for example on clinical trials registers. In fact, several structural elements are already routinely used in reports of randomised trials. Almost certainly the most successful impact of CONSORT has been the flow diagram, which is now included in the majority of published trial reports. The diagram depicts the flow of participants through the trial, from enrolment through allocation and follow up to analysis. Readers, including peer reviewers, can quickly see the numbers of randomised participants, identify when and why some were lost to follow-up, and the extent to which the numbers analysed reflect the numbers randomised. Other structures within most reports of trial results are a structured abstract and a table showing the baseline characteristics of each intervention group (usually as Table 1). But it is simple, and I believe desirable, to include other elements within explicit displays (tables or boxes), such as eligibility criteria, details of interventions, outcomes, and primary results. Currently, such information is rarely included as a display in published articles. The suggestion is intended to make key information easy to locate rather than to require a rigid format. Thus the same displays can easily accommodate the modified recommendations in several extensions to CONSORT (http://www.consort-statement.org/extensions). A similar idea has been proposed for articles reporting prognostic studies of tumour markers [27].
Readability is often put forward as an argument against structure. This would be a stronger argument if current research reports were highly readable, but many are not. One reason is that much factual information is embedded within the text, which works especially poorly for trial results. Readability is certainly desirable, but far more important are reliability and relevance [28], and crucially important too is reproducibility, which of course requires complete reporting. One reason for suggesting greater use of structural elements is to pull apart factual information from narrative, making it easier both to locate specific information and also to understand the broad sweep of a study without being diverted by all the details, important though they are.
Greater use of structure within research articles (not just for randomised controlled trials) would improve completeness by helping authors ensure they address key issues. It would also greatly aid reviewers and editors in appraising articles, and it would assist future systematic reviewers who currently struggle to find the key information they seek. Indeed, more structure would also aid text-mining. In the future, trial results will perhaps appear only in registries, presumably in a highly structured format, with journals carrying only narrative discussions of their findings [29]. Such a radical change does not seem imminent, however.
In the initial editorial published when Trials was launched, the editors wrote that ‘we believe that there is scope for new and better ways to report the findings of trials. Trials will develop and refine innovative approaches to improving communication about trials’ [30]. As yet there has been very little activity on this theme. We encourage contributions, both suggestions for suitable formats and also specific examples of real trial results presented in alternative formats. This editorial is thus the first contribution in a new Trials series on ‘New ways to publish research findings’.