| Sign In to gain access to subscriptions and/or personal tools. |
DOI: 10.1016/J.JHSB.2004.09.015
The Consolidated Standards of Reporting Trials (CONSORT): Better Presentation of Surgical Trials in the Journal of Hand SurgeryFrom the Biochemical and Experimental Division, Medical Faculty, University of Cologne, Germany and the Journal of Hand Surgery Editorial Office, Queens Medical Centre, Nottingham, UK Correspondence: Dr Stefan Sauerland, MD, MPH, Biochemical and Experimental Section, Medical Faculty, University of Cologne, Ostmerheimer Straße 200, D-51109 Köln, Germany. Tel.: +49221 98957 19; fax: +49221 98957 30. E-mail: s.sauerland{at}uni-koeln.de
To assure readers that study results are scientifically valid, the methods of a clinical trial should be described adequately. Since randomization, blinding, and intention-to-treat-analysis are major bias-reducing techniques, these aspects should be reported most accurately. The Consolidated standards of reporting trials (CONSORT) are recommendations to improve the reporting of trials. CONSORT requires that trial authors describe basic methodological aspects that readers need to appraise the strengths of report ed clinical trials. This article presents the CONSORT recommendations and explains some of their main aspects. From now on, the Journal of Hand Surgery will use CONSORT to assist authors of randomized controlled trials in improving the description of their studies. We believe that this decision increases the scientific validity of study reports and helps readers when critically appraising articles.
Key Words: publishing standards • randomized controlled trials • guidelines
Health-care providers often remain limited in their ability to make informed inferences about the validity of the studies upon which they base their clinical practice. One reason for this situation is the scarceness or complete lack of important information about design, conduct, and analysis of a published article. Surgical investigators generally aim to disseminate their findings in the most detailed way to their readership, but sometimes knowledge gaps or misconceptions concerning research terminology lead to an incomplete or misleading description of a clinical study. When considering therapeutic decisions, randomized controlled trials (RCTs) constitute the highest rung on the hierarchy of evidence. Therefore, the proper reporting of RCTs is critically important for guiding clinical patient management. Various studies have demonstrated that the following three factors are especially relevant in assessing the validity of a RCT:
Experienced readers are therefore looking for these features and have a distrust of articles that fail to describe them. Previous research in the field of surgery, however, has highlighted many more deficiencies in descriptions of methods used (Bhandari et al., 2002; Gerritsen et al., 2002) and the analysis and presentation of results (Chung et al., 1998; Gummesson et al., 2004; Martin et al., 2002; Schumm et al., 1999). It is clear that the lack of information about study methods can also make the interpretation of hand surgery trials inconclusive.
You are searching the literature for articles on the usefulness of antibiotics in treating open finger fractures. Using the PubMed website, the search strategy "Finger Injuries" [MeSH] AND antibiotic* yields 54 articles. After limiting the search to RCTs, only one article remains (Stevenson et al., 2003). You read this article and learn from the article's introduction that two more RCTs have been published (Sloan et al., 1987; Suprock et al., 1990), but they were obviously not indexed as RCTs in Medline. You decide to critically appraise all three studies and start with the oldest study by Sloan et al. The study by Sloan et al. describes that 85 patients were randomly allocated to four treatment groups, with one of these group receiving no antibiotics. Since the abstract failed to mention the randomized design, it becomes understandable why this study was difficult to find in Medline. The exact modalities of creating the randomization scheme and hiding the sequence from the study physicians were not described. Therefore, it may well be that some form of pseudo-randomization was employed, e.g. simple alternation or allocation according to date of birth (odd or even) or date of admission. Since trials with unconcealed randomization may show highly exaggerated treatment effects, you are sceptical about the validity of the trial, also because the trial did not have any form of blinding. When further looking at the results of the study by Sloan et al., you discover that infection rates are presented only for 83 of the 85 patients. No information is provided by the authors regarding the reasons for having excluded the missing two patients from analysis. Most likely, they simply withdrew from the study, but other underlying reasons (e.g. wound infection, drug side effects) might also be speculated on. Since group sample sizes are quite small in this trial, any adverse event in one of the treatment arms might easily affect the overall result and conclusion of the study. In summary, the trial by Sloan et al. appears methodologically questionable, but this is not primarily due to the apparent use of faulty methods, it is due to the pure lack of critical information about methods and results. The quality of the trial cannot be reliably assessed. How could this unsatisfactory situation have been prevented?
In 1996, journal editors and researchers developed the consolidated standards of reporting trials (CONSORT; Begg et al., 1996). The CONSORT check list aims at total transparency regarding design, conduct, analysis, and interpretation of a trial. Subsequently, several major health-care journals have adopted the CONSORT recommendations to make it easier for readers to assess the quality of controlled trials. In 2001, the CONSORT recommendations were revised to incorporate some new information (Moher et al., 2001b). CONSORT now consists of a 22-item checklist (Table 1) and a trial flow diagramme (Fig 1).
While some of the CONSORT items are easily understandable and acceptable, other require comment. Under the heading "outcomes" (item 6) trialists are required to define one single primary outcome. This meant to prevent any form of "data-dredging" by performing comparisons for multiple outcome variables and then selecting the most significant difference to represent the primary outcome. It was recently demonstrated that such practice is not as seldom as one would expect (Chan et al., 2004). Items 8 to 10 relate to the randomization procedure. It is essential to understand the two different steps of randomization. First, the randomization scheme has to be generated, e.g. by drawing cards or using random tables. Secondly (and more importantly), the sequence of randomization must be concealed. Otherwise, the study physician foresees the next treatment assignment, and his decision to include a patient for the study will be influenced by this knowledge. The higher likelihood to accept sicker patient for the study, each time a control group assignment is coming up next on the random list, will bias treatment results in favour of the experimental therapy. Since all methods of pseudo-randomization, such as assignment according to date of birth, are unconcealed, they should be avoided wherever possible. The issue of blinding is addressed by item 11. Of note, the use of the descriptors "double or single blind" is no longer advisable, because blinding of study physicians has two separate aspects. Of those who administer the treatment and those who evaluate its effects, each single, both, or neither of the two groups can be blinded. Although in studies on surgical procedures the first group can of course not be blinded, it still is possible to blind patient and treatment evaluator, thus theoretically allowing double–blinded surgical trials. In item 13, a diagram is proposed to describe the flow of study participants (Fig 1). If a study has a straightforward design without major protocol violations, the diagram is unnecessary, but larger trials with prolonged follow-up examinations usually benefit from a clearly structured presentation of patient flow (Egger et al., 2001). The diagram is also useful to detect selection bias. This would be the case, if only a small percentage of eligible patients were recruited for the trial. Until today, many trials fail to keep a log of uneligible so-called "escape" patients, although this is crucial to ensure generalizability of trial results. In item 16, it is suggested that results are presented as absolute numbers, not as percentages. One example may explain the rationale behind this. In a series of 100 operations, it would be deceptive to state the clinical success rate to be 50%, when in reality this percentage has been calculated only on the basis of those 20 patients, who could be followed-up (with 10/20=50%). In hand surgery, it sometimes will be necessary here to differentiate between patients and hands as the denominator used.
From now on, authors of randomized trials should check before submission of a manuscript to the Journal whether their manuscript fulfills the 22 CONSORT items. Reviewers and editors will additionally inspect study reports for compliance with CONSORT. We believe that this will foster the scientific quality of articles in the Journal. Future studies should examine whether this aim is really being achieved. Several studies have already found that the quality of reporting of clinical trials increases after a journal's decision to adopt CONSORT (Devereaux et al., 2002; Moher et al., 2001a). CONSORT also makes it easier for meta-analysts to assess clinical trial quality and to use outcome data for statistical analysis (Gerritsen et al., 2002; Gotzsche, 2001). In most cases, CONSORT does not reveal hidden faults in study design or execution, rather it reassures readers about the good quality of a study (Soares et al., 2004). Additionally, compliance with CONSORT will make it easier to identify high quality evidence in medical database searching. Of course, a trial report that complies with CONSORT will cover somewhat more space in a journal than a non–complying report: The publication by Sloan et al. covered two pages, whereas the article by Stevenson et al. extended over seven pages. This may correspond to insignificantly more work for authors, reviewers, and readers, but the increased reliability of study reporting will certainly guarantee the many advantages of embracing and using CONSORT in the Journal.
  CiteULike  Connotea  Del.icio.us  Digg  Reddit  Technorati What's this?
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
|
|
 |
 |
Sign In to gain access to subscriptions and/or personal tools.
TOP
Abstract
INTRODUCTION