Reporting of Randomised Controlled Trials 236
The CONSORT Statement: A Reporting Guideline for Randomised Trials 238
Studies Other Than RCTs 241
Medical research seeks to expand scientific understanding and guide clinical care. However, poor scientific reporting of research, including nonreporting, hinders understanding and impedes the potential benefits to clinicians, researchers, and patients.
Deficient reporting is widespread. It ensues when a key aspect of the methods or results is omitted, incomplete, or vague. Selective reporting results from publication bias or biased reporting within publications. Unfortunately, investigators of many research studies, including randomised controlled trials (RCTs), never publish their findings. Indeed, if a study is not published, it did not happen in the eyes of the research community. Publication bias results from the preferential publication of studies with statistically significant outcomes. Biased reporting within publications appears when outcomes that were statistically significant were more likely to be reported, regardless of their status in the original trial protocol (‘cherry picking’).
We focus on reporting of randomised trials because of their importance to medical research and because most of the early work on improving reporting related to randomised trials. In the early 1990s, a group of researchers and editors started a guideline process that ended with the CONsolidated Standards Of Reporting Trials (CONSORT) Statement. The CONSORT Statement includes a checklist and a flow diagram. CONSORT enables the accurate, complete, and transparent reporting of a trial that facilitates critical appraisal and interpretation.
The CONSORT 2010 Statement provides guidance for reporting all RCTs, regardless of design. It pertains fundamentally to all randomised trials. However, it concentrates on the most common design type, individually randomised, two-group, parallel trials.
Investigators should use a reporting guideline for both planning and reporting all types of studies, not just randomised trials. Usually, as a rule of thumb, a reporting guideline that also has an explanation and elaboration (E&E) paper for support will be more useful. We recommend STARD for diagnostic test accuracy studies, STROBE for observational epidemiological studies, and PRISMA for systematic reviews including meta-analyses of randomised trials ( Panel 22.1 ).
Empirical investigations substantiate that CONSORT leads to improved reporting in medical journals. However, the improvements appear moderate, and substantial room for improvement remains. CONSORT has spawned other useful guidelines for studies besides randomised trials. Authors should utilise appropriate reporting guidelines. Such use aids in study design and manuscript writing, and, in the end, likely leads to a higher probability of publication success.
Medical research seeks to expand scientific understanding and guide clinical care. However, poor scientific reporting, including nonreporting, hinders understanding and impedes the potential benefits to clinicians, researchers, and patients.
Principally, medical journals convey research results. The report in a journal of a research study is almost always the only documentation of the study results. Indeed, if a study is not published, it remains invisible to the research community. Unfortunately, many investigators never publish their findings.
Published articles of research studies offer potential benefits to multiple audiences. For example, other researchers use the methods and results of articles to guide their forthcoming research. Clinicians use journal articles to choose the treatment of patients. Systematic reviewers examine journal articles for inclusion in their reviews. They also need sufficiently detailed information on methods and results for potential integration into a meta-analysis. Many patients access the medical literature to guide their treatment decisions.
Publication is necessary but not sufficient. A journal article must contain full descriptions of the methods used and the results found. Readers need clear, accurate, complete, and transparent reporting. Only then can they correctly interpret the study findings.
These general reporting principles pertain basically to all types of healthcare research. However, in this chapter we focus on reporting of RCTs for three reasons. First, one of us (K.F.S.) has been involved in setting the standards for reporting trials. Second, throughout this book we intentionally devote more attention to trials. Third, the reports of randomised trials have the highest likelihood of all published research of having an immediate effect on patient care. Indeed, as noted by a famous editor, ‘The whole of medicine depends on the transparent reporting of clinical trials’. Nevertheless, in a later section, we will recommend reporting guidelines for studies other than RCTs.
Reporting of Randomised Controlled Trials
RCTs, when appropriately designed, conducted, and reported, embody the gold standard in evaluating healthcare interventions. That distinguished position in the medical research taxonomy does not mean, however, that readers should uncritically accept the results of all RCTs. Indeed, randomised trials can yield biased results if they lack methodological rigour. Although readers of journal articles need complete, clear, and transparent information on their methodology and findings, many authors fail to report critical information.
Deficient reporting ensues when a key aspect of the methods or results is vague, incomplete, or omitted. Certainly, some overlap occurs among all these terms. For example, a description of a trial only as ‘randomised’ is vague, incomplete, and omits critical information. Clearly, critical information on the details of the randomisation methods is missing.
Deficient reporting is common. For example, one group of investigators found 177 reviews assessing the quality or reporting of RCTs published just between 1987 and 2007. The consistent theme from this enormous body of research is that the reporting of RCTs in the literature is greatly deficient.
To exemplify the level of poor reporting, two of the more recent studies provide the best general estimates. These two studies examine trial reports indexed in PubMed. As such, they provide information on all journals indexed in PubMed, including the most elite weekly general medical journals as well as the less recognised, lower-impact-level specialty journals. Overall, the level of reporting was dismal with only some small improvement over time.
The glaring deficiency was that authors omitted critical information. For example, information on random sequence generation and allocation concealment, arguably the most important ingredients in a randomised trial, was not described in about two-thirds and three-quarters, respectively, of trial reports.
Unfortunately, the true state of conduct was even more dire. One would reasonably expect that when allocation concealment was reported in about one-fourth of the reports, it was an acceptable approach. That represents a rational expectation. Not so. Another study found the same level of good reporting but then further examined the quality of the method described. More than half the authors described an inadequate approach to allocation concealment. In other words, when the reporting was good, the actual method revealed was frequently bad. That sad fact further illustrates the need for good reporting for two reasons. First, with good reporting, the reader at least can identify a trial with poor methods and thereby properly interpret the results. Second, trials that report well but reveal poor methods provide an indication of the poor methods likely used in all those trials that report no information on allocation concealment.
Poor reporting extends beyond trial design and statistical methods. Medical research publications guide clinical practice. Published articles on randomised trials should provide adequate details of treatments studied such that they could be used by clinicians upon reading the article. That seems a minimum expectation. However, when investigators assessed descriptions of treatments in 80 published articles, critical elements of the treatments were not described in 41 of those studies. Readers would not be able to replicate the treatments.
The most recognised form of selective reporting is termed ‘publication bias’. This bias results from authors of completed trials preferentially not publishing the trials with statistically nonsignificant findings. Extensive investigations support the existence of publication bias. A large review that assessed study publication bias found persuasive support that trials with statistically significant results were more likely to be published. Of note, some maintain that the term ‘publication bias’ is confusing and prefer the term ‘nonpublication bias’.
Another form of selective reporting, biased reporting within publications, is gaining recognition. It might be the most insidious form of selective reporting. The culprit appears to be that outcomes that were statistically significant were more likely to be reported, regardless of their status in the original trial protocol (‘cherry picking’). Moreover, other investigations have demonstrated that the primary outcome described in the trial protocol frequently is not necessarily the primary outcome reported in the trial publication. Indeed, 40%–62% of studies had at least one primary outcome changed, newly introduced, or excluded.
The CONSORT Statement: A Reporting Guideline for Randomised Trials
As the burgeoning evidence pointed to the poor reporting of RCTs, investigators considered solutions. Poor reporting may simply signify ignorance on the vital information that must be included in the report of an RCT. Naivety may be more responsible than malevolence. A reporting guideline could provide insight to that vital information, and, thus, theoretically help improve reporting.
However, not all reporting guidelines are created equal. Early guidelines for RCT publications seemed to have minimal effect. In the early 1990s, a group of researchers and editors started a guideline process that resulted in the CONsolidated Standards Of Reporting Trials (CONSORT) Statement. The CONSORT Statement includes a checklist and a flow diagram. CONSORT enables the accurate, complete, and transparent reporting of a trial. It was designed to help authors, peer reviewers, and editors, and facilitates critical appraisal and interpretation.
All the information required to evaluate a trial guided the reasoning for including items in the checklist. The checklist ensures provision of the basic information that readers need to appraise the validity of the findings. Inclusion of items was based on relevant empirical evidence, whenever possible. The flow diagram ( Fig. 22.1 ) depicts the movement of trial participants from recruitment to final analysis. The checklist ( Table 22.1 ) and flow diagram form the CONSORT Statement.
|Section/Topic||Item No||Checklist Item||Reported on Page Number|
|Title and Abstract|
|1a||Identification as a randomised trial in the title||__________|
|1b||Structured summary of trial design, methods, results, and conclusions (for specific guidance see CONSORT for abstracts)||__________|
|Background and objectives||2a||Scientific background and explanation of rationale||__________|
|2b||Specific objectives or hypotheses||__________|
|Trial design||3a||Description of trial design (such as parallel, factorial) including allocation ratio||__________|
|3b||Important changes to methods after trial commencement (such as eligibility criteria), with reasons||__________|
|Participants||4a||Eligibility criteria for participants||__________|
|4b||Settings and locations where the data were collected||__________|
|Interventions||5||The interventions for each group with sufficient details to allow replication, including how and when they were actually administered||__________|
|Outcomes||6a||Completely defined prespecified primary and secondary outcome measures, including how and when they were assessed||__________|
|6b||Any changes to trial outcomes after the trial commenced, with reasons||__________|
|Sample size||7a||How sample size was determined||__________|
|7b||When applicable, explanation of any interim analyses and stopping guidelines||__________|
|Sequence generation||8a||Method used to generate the random allocation sequence||__________|
|8b||Type of randomisation; details of any restriction (such as blocking and block size)||__________|
|Allocation concealment mechanism||9||Mechanism used to implement the random allocation sequence (such as sequentially numbered containers), describing any steps taken to conceal the sequence until interventions were assigned||__________|
|Implementation||10||Who generated the random allocation sequence, who enrolled participants, and who assigned participants to interventions||__________|
|Blinding||11a||If done, who was blinded after assignment to interventions (for example, participants, care providers, those assessing outcomes) and how||__________|
|11b||If relevant, description of the similarity of interventions||__________|
|Statistical methods||12a||Statistical methods used to compare groups for primary and secondary outcomes||__________|
|12b||Methods for additional analyses, such as subgroup analyses and adjusted analyses||__________|
|Participant flow (a diagram is strongly recommended)||13a||For each group, the numbers of participants who were randomly assigned, received intended treatment, and were analysed for the primary outcome||__________|
|13b||For each group, losses and exclusions after randomisation, together with reasons||__________|
|Recruitment||14a||Dates defining the periods of recruitment and follow-up||__________|
|14b||Why the trial ended or was stopped||__________|
|Baseline data||15||A table showing baseline demographic and clinical characteristics for each group||__________|
|Numbers analysed||16||For each group, number of participants (denominator) included in each analysis and whether the analysis was by original assigned groups||__________|
|Outcomes and estimation||17a||For each primary and secondary outcome, results for each group, and the estimated effect size and its precision (such as 95% confidence interval)||__________|
|17b||For binary outcomes, presentation of both absolute and relative effect sizes is recommended||__________|
|Ancillary analyses||18||Results of any other analyses performed, including subgroup analyses and adjusted analyses, distinguishing prespecified from exploratory||__________|
|Harms||19||All important harms or unintended effects in each group (for specific guidance see CONSORT for harms)||__________|
|Limitations||20||Trial limitations, addressing sources of potential bias, imprecision, and, if relevant, multiplicity of analyses||__________|
|Generalisability||21||Generalisability (external validity, applicability) of the trial findings||__________|
|Interpretation||22||Interpretation consistent with results, balancing benefits and harms, and considering other relevant evidence||__________|
|Registration||23||Registration number and name of trial registry||__________|
|Protocol||24||Where the full trial protocol can be accessed, if available||__________|
|Funding||25||Sources of funding and other support (such as supply of drugs), role of funders||__________|