The new CONSORT statementBMJ 2010; 340 doi: http://dx.doi.org/10.1136/bmj.c1432 (Published 24 March 2010) Cite this as: BMJ 2010;340:c1432
- Gerd Antes, scientist and director of German Cochrane Centre
Full and transparent reporting of results of clinical trials is essential for assessing the quality of healthcare interventions. Inadequate reporting of trials is common, and it impedes the use of trial results in healthcare research and practice.1 Consequently, during the past 15 years, a series of reporting guidelines have been developed and recently collected by the Equator Initiative (www.equator-network.org).2 The pioneering first step of this framework was the CONSORT statement in 1996,3 which provided recommendations for the publication of randomised controlled clinical trials, the gold standard to assess healthcare interventions. The 2010 update of the statement (doi:10.1136/bmj.c332)4 is accompanied by a comparative study by Hopewell and colleagues (doi:10.1136/bmj.c723),5 which assesses whether the quality of reporting of randomised trials has improved since publication of the CONSORT statement in 2001.
Concerns have been raised that such publication guidelines may be too prescriptive, impede the creativity of research, and lack authorisation of the self selected author groups who write them.6 However, these concerns are clearly outweighed by the harm caused by poor reporting. Thus, broad acceptance of CONSORT and other statements has been a natural step towards achieving better quality of trial reporting. That is the theory, but what about the practice?
Hopewell and colleagues compare the reporting characteristics and methodological details of randomised trials indexed in PubMed in 2000 and in 2006.5 By looking at 616 trials indexed in December 2006, and a similar sample from 2000, they assessed whether the quality of reporting has improved since publication of the revised CONSORT statement in 2001.3 Two key messages emerged: firstly, improvements occurred in the reporting of several items that are crucial for the assessment of trial quality; and secondly, these improvements generally started from an embarrassingly low level. Improving the quality of reporting from very poor to poor could be interpreted as a positive trend, but essential items like sample size estimation (27% v 45%), description of the randomisation procedure (21% v 34%), or description of the concealment of treatment allocation (18% v 25%) are described in an unacceptably low number of reports. An optimistic perspective would be to view the glass as half full; a realistic one would have to admit that the glass is more than half empty.
A look at the proportion of journals that support the CONSORT statement may help to gain more insight into these disappointing figures. Hopewell and colleagues showed that even among high impact journals, fewer than 50% recommend that authors comply with the CONSORT statement.7 8 Of those, only a minority have procedures that support adherence to the requests from the CONSORT statement by authors and reviewers.7
So what has changed in the new update of the CONSORT statement? In addition to improvements of wording and consistency across items, several items were extended to include sub-items that concern, for example, allocation concealment, blinding, or other methodological matters. A few new items were introduced: registration is now required before inception, and researchers must state where the protocol can be accessed (if this is possible) and where the funding comes from.
Why is better endorsement and adherence to the CONSORT statement not achieved although the evidence of benefit is impressive?9 The answer is unclear and therefore speculative. Even after almost 15 years, the main reason seems to be the lack of awareness and a considerable amount of ignorance about recommendations like CONSORT. Journal editors have an essential role, and their reluctance may be based on the expectation of a serious increase in workload. However, much of the workload could be shifted to authors. Incorporating the CONSORT checklist into the peer review process might increase the workload of authors initially, but it would help to simplify and structure the review process by enabling reviewers to check trial reports against the CONSORT items. This would establish a solid communication structure between authors, reviewers, and editors, but this approach has not been implemented on a large scale so far. The simplest explanation for this may be mere lack of appreciation of the importance of publication ethics. This was confirmed by a survey of editors, which concluded that most editors of science journals are not very concerned about publication ethics and believe that misconduct occurs only rarely in their journals.10
Finally, a problem with research in this field is the almost complete restriction to articles in the English language. Several mechanisms, especially the pursuit of academic incentives and maximum promotion of trial results, has led to reduced publication in non-English journals.11 The migration of original research publications into English may solve this problem eventually. However, this risks the danger of moving towards double standards of quality requirements. Although global knowledge is mostly exchanged in the English language, a large number of local non-English language journals remain beyond systematic empirical investigations of their quality and beyond the reach of publication guidelines.12 CONSORT has been translated into 10 other languages (www.consort-statement.org/consort-statement/translations), but not much is known about endorsement and adherence in those areas.13
Cite this as: BMJ 2010;340:c1432