For more information on how to effectively use tables and figures in scientific papers:61–63
When new healthcare interventions are studied, researchers tend to focus more on efficacy than safety. There is poor reporting of harms in trial reports across many clinical areas,64 which makes it difficult to obtain a true estimate of the benefit-harms ratio. The CONSORT extension for harms65 was developed to improve reporting of harms-related data in trials. Because the main focus of the CONSORT checklist is efficacy reporting, we suggest you supplement your trial reporting with the CONSORT extension for harms65 to improve reporting of harms-related data.
For more information on reporting of harms-related data: 65
The CONSORT checklist46 holds the overall framework for the discussion and items you should address, but scientific journals may have additional requirements. We suggest you use the CONSORT checklist to structure the discussion, and supplement with requirements from your target journal, if needed. We would like to highlight two important items: clinical relevance and confirmation bias.
We suggest you focus on the primary analysis and outcome. Your trial was designed first and foremost to provide a reliable answer in terms of the hypothesis for this analysis and outcome. The test statistics will determine if the difference between groups is statistically significant. Judging and discussing whether a statistically significant difference between groups is also clinically relevant should be easy at this point. You will already have argued in your trial protocol and sample size paragraph what minimum theoretical difference between groups you consider clinically relevant and why. Now that you have the observed difference between groups, the main issue is to compare the two and discuss the size of the observed effect. An important aspect of this discussion is the precision of the observed effect. In general, the larger the sample size of your trial, the greater the precision of the observed effect. The precision is reflected in the 95% CI of the observed effect. The greater the precision, the smaller the 95% CI and vice versa. We suggest a balanced discussion of the clinical relevance of the observed effect to include both its size (in relation to the predefined minimal clinically importance difference) as well its precision. It will help you avoid unintentional confirmation bias (please see below).
Biases come in many forms and can affect healthcare in many ways. There may be biases that you want to acknowledge specifically under ‘Limitations’ in the discussion because you think they may have influenced trial procedures or outcomes. We suggest you consider your own ‘confirmation bias’ when writing the discussion—or the whole trial report for that matter. As stated by the Catalogue of Bias Collaboration37: ‘Confirmation bias occurs when an individual looks for and uses the information to support their own ideas or beliefs. It also means that information not supporting their ideas or beliefs is disregarded.’ Being researchers, we argue that most of us unintentionally wish for our intervention to be superior to the comparator for several reasons: (1) we want to help patients by advancing the field, or (2) we think it will bring promotion or other academic reward. By being intentionally aware of our own confirmation bias, we can better stay clear of issues such as unintentional secondary analysis emphasis (spin) and selective referencing of work that support our own findings.
For more information on statistical significance, clinical relevance, spin and confirmation bias: 17 37 66
When you write the trial report conclusion, we encourage you to think ‘aim’, ‘hypothesis’ and ‘trial design’. What was your trial designed to test primarily and how was this formulated in the aim? Was it to assess if the intervention of interest was better than (superiority trial), no worse than (non-inferiority trial), or whether it was as equally effective as (equivalence trial) the comparator? Using this line of thinking will help create a strong connection between aim, hypothesis and conclusion. It will also help you conclude only what the trial data support. If the aim of a superiority trial was ‘To investigate if I (Intervention of interest) is superior to C (comparator) in improving O (primary outcome) at T (timepoint) in P (population) and there was no difference in response between groups, the conclusion could start with: ‘Compared with C (comparator), I (Intervention of interest) was not superior in reducing O (primary outcome) at T (time point) in P (population). A very common mistake is to interpret the absence of evidence of superiority as evidence of equivalence or non-inferiority and conclude that the intervention of interest and comparator were equally effective or no different (for more details, please see refs 1 67).
Having addressed the main hypothesis, analysis and outcome the trial was designed to assess, we encourage you to proceed with interesting secondary analyses and—at the same time—inform the reader about the increase in risk of bias for these analyses: ‘For the secondary outcomes, X, Y and Z, we found that (………).’. When you conclude first on the primary analysis, you minimise the risk of unintentional reporting31 or spin32 biases. If your trial was more exploratory than confirmatory1—or had a flat outcome hierarchy with no single primary outcome—you may want to consider finishing the conclusion by acknowledging this. For example, ‘This finding needs replication in future trials’. Readers will often be interested in your thoughts on the implications of your trial findings. Some journals allow implication statements and others do not. If you do write about implications, we suggest you make it clear that this part of the conclusion is you speculating and conveying your expert opinion with phrasing like: ‘These findings may have implications for (……) insofar as (……).’. When you have finished writing the conclusion, check that it matches the trial aim and conclusion in the abstract.
Depending on national legislation, you may or may not be able to share the raw trial data. Data sharing is one way of increasing transparency and maximising the trial participants’ research contribution by making the data they provided broadly available for secondary research purposes.68 69 Data sharing is also expected by some non-private funding bodies.70 If you can share your trial data, there are some things that you may want to consider. They include practical steps to data management, anonymisation and storage.
For more information on data sharing: 71–74
When you are ready to submit your trial report to a scientific journal, consider publishing a preprint. A preprint is scientific work that has not undergone peer review and is not published in a scientific journal.75 It is typically a manuscript draft that is ready to be submitted to a journal for peer review. A preprint can also be an earlier manuscript version that you want to make public. One advantage of publishing a preprint is that it is assigned a DOI,35 which makes it searchable. Most publishers allow preprints,76 but we suggest you check the preprint policy of the scientific journal that you aim to submit your trial report to. Elsevier states: ‘Preprint: Authors can share their preprint anywhere at any time. If accepted for publication, we encourage authors to link from the preprint to their formal publication via its Digital Object Identifier (DOI). Millions of researchers have access to the formal publications on ScienceDirect, and so links will help your users to find, access, cite, and use the best available version. Authors can update their preprints on arXiv or RePEc with their accepted manuscript. Please note: Some society-owned titles and journals that operate double-blind peer review have different preprint policies.’.77
Submission to preprint servers is typically free and it creates an open access option, even if you end up publishing your trial report behind a paywall.78 It allows you to have crowdsourced feedback and to promote your open access research early (eg, during the period of peer review). Based on feedback, you can update your preprint version when you revise your manuscript. Some (but not all) publishers even allow you to update your preprint to the accepted (non-type set) manuscript version with proper reference to the journal publication. Please check the publisher’s preprint policy for guidance. If you look at the bottom of the abstract of this guide, you will see a link to an open access preprint. Had this guide not been published open access, an interested reader could see from the PubMed abstract where to find an open access full text (preprint).
For more information on preprints: 75
Researchers are familiar with social media platforms like Twitter for sharing new scientific content. When posting to social media, make space to include (1) the DOI35 and (2) an image—two simple steps to help make your post visible to attention metrics-aggregators like Altmetric and capture the viewer who might otherwise scroll past your post. Across research areas, the Altmetric score has been associated with number of citations, journal impact factor, press releases and open access status.79 80
Have you considered other forms of media? Researchers who embrace the rich ecosystem of digital media might find themselves partnering with clever infographics designers or using free (or freemium) websites to design their own. Consider writing for trusted outlets like The Conversation81—a news organisation that is dedicated to sharing information from the academic and research community, direct to the public, with ‘academic rigour and journalistic flair’. Sports medicine and sports physiotherapy journals including British Journal of Sports Medicine and Journal of Orthopaedic & Sports Physical Therapy have blogs dedicated to reaching a non-academic audience of clinicians, patients, athletes and coaches.
Consider approaching your academic institution’s media and communications department or press office. The staff are typically pleased to work with you to shape a press release, distribute the press release to mainstream media services, and connect with media contacts. Media and communications departments also share helpful tips for making your research visible to the media.82
After your trial report is published, consider (1) Is the ‘Trial status’ up to date in the trial registry? (2) Do I need to update the trial registry with a link to the published trial report and/or raw data if shared? (3) Do I need to report to funding bodies on the accomplished milestone (publication)? (4) Do I have a plan for disseminating the trial results other than the primary trial report? (5) Do I have a plan for storing and filing essential trial documents and data that adheres to national guidelines?
We hope the REPORT guide is helpful and a valuable supplement to ‘first choice’ trial reporting tools, such as CONSORT. We aimed to incorporate tacit knowledge about reporting, and flag issues we have struggled with. Quality decisions in healthcare depend on reliable evidence of treatment effects. Good research reporting practice does not cure ‘diseases’ that arise from poor research methodology—it helps the reader see the illness and appraise the research quality. No research is perfect. We do not profess to produce and report perfect research that is free from ‘disease’ 100% of the time. We implore all researchers to commit to conducting (and reporting) clear and transparent research.