Independent discussion sections for improving inferential reproducibility in published research

Open ArchivePublished:January 22, 2019DOI:https://doi.org/10.1016/j.bja.2018.12.010

      Summary

      There is a reproducibility crisis in science. There are many potential contributors to replication failure in research across the translational continuum. In this perspective piece, we focus on the narrow topic of inferential reproducibility. Although replication of methods and results is necessary to demonstrate reproducibility, it is not sufficient. Also fundamental is consistent interpretation in the Discussion section. Current deficiencies in the Discussion sections of manuscripts might limit the inferential reproducibility of scientific research. Lack of contextualisation using systematic reviews, overinterpretation and misinterpretation of results, and insufficient acknowledgement of limitations are common problems in Discussion sections; these deficiencies can harm the translational process. Proposed solutions include eliminating or not reading Discussions, writing accompanying editorials, and post-publication review and comments; however, none of these solutions works very well. A second Discussion written by an independent author with appropriate expertise in research methodology is a new testable solution that could help probe inferential reproducibility, and address some deficiencies in primary Discussion sections.

      Keywords

      There is an increasing recognition of a reproducibility crisis in the biomedical sciences.
      • Collins F.S.
      • Tabak L.A.
      Policy: NIH plans to enhance reproducibility.
      There are many potential contributors to replication failure in research across the translational continuum. These include biological variability, random chance, laboratory measurement errors, sloppy research conduct, small sample sizes, scientific fraud, different population characteristics, imprecise clinical outcomes ascertainment, imperceptible technical differences, and environmental factors. These many considerations are beyond the scope of this perspective piece, which focuses on the narrow topic of inferential reproducibility.
      Although replication of methods and results is necessary to demonstrate reproducibility, these two cardinal pillars of reproducibility are not sufficient. The third such pillar is consistent interpretation within a theoretical framework and is referred to as inferential reproducibility
      • Atmanspacher H.
      • Bezzola Lambert L.
      • Folkers G.
      • Schubiger P.A.
      Relevance relations for the concept of reproducibility.
      (Fig. 1). There are many reasons that scientists might not draw the same conclusions from the same results.
      • Goodman S.N.
      • Fanelli D.
      • Ioannidis J.P.
      What does research reproducibility mean?.
      These include different assessments of prior probability and different choices regarding what to focus on in the analysis and reporting of data.
      • Goodman S.N.
      • Fanelli D.
      • Ioannidis J.P.
      What does research reproducibility mean?.
      Most critiques regarding the quality of research publications focus on the Methods and Results sections, whereas the Introduction and Discussion sections have received less systematic scrutiny.
      • Clarke M.
      • Chalmers I.
      Discussion sections in reports of controlled trials published in general medical journals: islands in search of continents?.
      Some of the ongoing deficiencies of Discussion sections in medical journals include lack of contextualisation of results using systematic review, selective citation of prior studies that agree with the results, overinterpretation or misinterpretation of findings, and incomplete acknowledgement, analysis, and assessment of study limitations.
      • Clarke M.
      • Chalmers I.
      Discussion sections in reports of controlled trials published in general medical journals: islands in search of continents?.
      • Clarke M.
      • Hopewell S.
      • Chalmers I.
      Reports of clinical trials should begin and end with up-to-date systematic reviews of other relevant evidence: a status report.
      • Clarke M.
      • Alderson P.
      • Chalmers I.
      Discussion sections in reports of controlled trials published in general medical journals.
      Considering that many clinicians and policy makers might form an opinion of the importance of a study based on the contents and tone of the Discussion and the Abstract alone, it is important that these sections should be dispassionate, appropriately informative, and demonstrably reproducible. Misleading Discussions can artificially fuel premature conclusions along the translational continuum.
      Fig 1
      Fig 1Three pillars of reproducibility. Rigorous (i) methodological approaches, (ii) analysis and presentation of results, and (iii) interpretation and discussion of findings are all essential pillars to promote reproducibility in scientific research.

      Current deficiencies of discussion sections

      Discussion sections of research articles can be subject to many problems. Interpretation by the investigators of their own results may be biased and researchers often do not appropriately contextualise the findings of studies within a broader scope of scientific evidence.
      • Clarke M.
      • Chalmers I.
      Discussion sections in reports of controlled trials published in general medical journals: islands in search of continents?.
      • Clarke M.
      • Alderson P.
      • Chalmers I.
      Discussion sections in reports of controlled trials published in general medical journals.
      • Goodman S.N.
      Of P-values and Bayes: a modest proposal.
      • Goodman S.N.
      Toward evidence-based medical statistics: 1. The P value fallacy.
      It has been suggested that failure to refer to relevant published literature in the Discussion section is unacceptable research practice, and authors should clarify what strategy they used to identify all relevant publications,
      • Clarke M.
      • Chalmers I.
      Discussion sections in reports of controlled trials published in general medical journals: islands in search of continents?.
      including unpublished (often negative) findings. This practice is not currently required by most journals. A conclusion is often preconceived in scientific manuscripts, and the Discussion misinterprets the results to yield the desirable conclusions.
      • Goodman S.N.
      Of P-values and Bayes: a modest proposal.
      • Goodman S.N.
      Toward evidence-based medical statistics: 1. The P value fallacy.
      Results and their implications for clinical practice are interpreted based on unbalanced assumptions fuelled by confirmation or allegiance bias, sponsor interests, or both. The latter may be particularly influential when research has practical financial implications. Furthermore, exaggerations or misrepresentations in manuscript Discussions can be perpetuated or even further distorted in academic press releases and the resulting science news stories.
      • Sumner P.
      • Vivian-Griffiths S.
      • Boivin J.
      • et al.
      The association between exaggeration in health related science news and academic press releases: retrospective observational study.
      The main factor that has been associated with selective reporting of positive effects of experimental treatments has been the presence of such selective emphasis in an article's conclusion.
      • Yavchitz A.
      • Boutron I.
      • Bafeta A.
      • et al.
      Misrepresentation of randomized controlled trials in press releases and news coverage: a cohort study.

      Accounting for the totality of evidence

      Clinical research guidelines, such as the CONSORT statement, suggest that the results of a study should be interpreted in the context of the totality of available information.
      • Begg C.
      • Cho M.
      • Eastwood S.
      • et al.
      Improving the quality of reporting of randomized controlled trials. The CONSORT statement.
      Contextualising the results of studies using systematic review of relevant literature can help to ensure that new research builds on lessons from earlier research, and can assess whether findings from previous studies are replicated or contradicted. In three studies conducted over a decade, Clarke and Chalmers
      • Clarke M.
      • Chalmers I.
      Discussion sections in reports of controlled trials published in general medical journals: islands in search of continents?.
      and Clarke and colleagues
      • Clarke M.
      • Hopewell S.
      • Chalmers I.
      Reports of clinical trials should begin and end with up-to-date systematic reviews of other relevant evidence: a status report.
      • Clarke M.
      • Alderson P.
      • Chalmers I.
      Discussion sections in reports of controlled trials published in general medical journals.
      showed that manuscripts in major medical journals have not embraced this recommendation, and most Discussion sections do not include systematic review in interpreting findings. It is clear, however, that the practice of systematic review or meta-analysis is not uniform or required in all major medical journals. It is unlikely that the situation is better in other journals. Absent systematic review, there is a tendency to preferentially cite studies with positive findings or findings congruent with authors' perspectives.
      • Perino A.C.
      • Hoang D.D.
      • Holmes T.H.
      • et al.
      Association between success rate and citation count of studies of radiofrequency catheter ablation for atrial fibrillation: possible evidence of citation bias.
      • Kivimaki M.
      • Batty G.D.
      • Kawachi I.
      • Virtanen M.
      • Singh-Manoux A.
      • Brunner E.J.
      Don’t let the truth get in the way of a good story: an illustration of citation bias in epidemiologic research.
      • Jannot A.S.
      • Agoritsas T.
      • Gayet-Ageron A.
      • Perneger T.V.
      Citation bias favoring statistically significant studies was present in medical research.
      Selective citation in Discussion sections has the potential to distort scientific knowledge.
      • Kivimaki M.
      • Batty G.D.
      • Kawachi I.
      • Virtanen M.
      • Singh-Manoux A.
      • Brunner E.J.
      Don’t let the truth get in the way of a good story: an illustration of citation bias in epidemiologic research.
      Such distortion can lead to the inappropriate translation of findings, such as the premature or indiscriminate clinical implementation of a protocol from a clinical trial.

      Overinterpretation and misinterpretation of results

      There are several motivations for overinterpreting and misinterpreting results, including confirmation bias, allegiance bias, and the desire to impact practice with positive results. For example, it is common to overinterpret the clinical applicability of diagnostic tests.
      • Lumbreras B.
      • Parker L.A.
      • Porta M.
      • Pollan M.
      • Ioannidis J.P.
      • Hernandez-Aguado I.
      Overinterpretation of clinical applicability in molecular diagnostic research.
      Indeed, the problem of overinterpretation and misreporting of results in Abstracts and Discussions in diagnostic accuracy studies has been found to be common even in journals with high journal impact factors.
      • Ochodo E.A.
      • de Haan M.C.
      • Reitsma J.B.
      • Hooft L.
      • Bossuyt P.M.
      • Leeflang M.M.
      Overinterpretation and misreporting of diagnostic accuracy studies: evidence of “spin”.
      An analysis of 76 trials of analgesic interventions revealed that treatment benefit is often portrayed despite non-‘significant’ primary findings, and even without ‘significant’ findings of effectiveness, investigators often recommend clinical adoption of analgesic treatments.
      • Gewandter J.S.
      • McKeown A.
      • McDermott M.P.
      • et al.
      Data interpretation in analgesic clinical trials with statistically nonsignificant primary analyses: an ACTTION systematic review.
      A concerning investigation of 72 RCTs with non-‘significant’ results revealed that 43% of the Discussion sections contained ‘spin’ or interpretations of findings that were inconsistent with the non-‘significant’ results.
      • Boutron I.
      • Dutton S.
      • Ravaud P.
      • Altman D.G.
      Reporting and interpretation of randomized controlled trials with statistically nonsignificant results for primary outcomes.
      Such ‘spin’ has also been reported in the titles and abstracts of papers in the leading anaesthesia journals.
      • Kinder N.
      • Weaver M.D.
      • Wayant C.
      • Vassar M.
      Presence of spin in the abstracts and titles of anaesthesiology randomized controlled trials.
      Industry funding can produce allegiance bias, and numerous studies by Bero and colleagues
      • Bero L.
      • Oostvogel F.
      • Bacchetti P.
      • Lee K.
      Factors associated with findings of published trials of drug–drug comparisons: why some statins appear more efficacious than others.
      and Lundh and colleagues
      • Lundh A.
      • Sismondo S.
      • Lexchin J.
      • Busuioc O.A.
      • Bero L.
      Industry sponsorship and research outcome.
      have shown that industry funding is likely to influence the interpretation of clinical trials (more so than their results). The overinterpretation of ‘significant’ results also has the potential to create inaccurate scientific narratives that can persist in research or clinical practice. There is a false belief that the probability of a postulate being in error or that probable causality can be inferred from the data in a single experiment.
      • Goodman S.
      A dirty dozen: twelve p-value misconceptions.
      It is a common misconception that the P-value can be used for null hypothesis testing, and to suggest that it reflects the probability that the null hypothesis is true.
      • Goodman S.
      A dirty dozen: twelve p-value misconceptions.

      Lack of appropriate discussion of limitations

      One of the major problems with Discussion sections is that limitations of studies are not thoroughly addressed.
      • Ioannidis J.P.
      Limitations are not properly acknowledged in the scientific literature.
      Although relatively common in clinical research, deficiency in addressing limitations is probably even more of a problem in basic (laboratory) research.
      • Yazici H.
      • Gogus F.
      • Esen F.
      • Yazici Y.
      There was less self-critique among basic than in clinical science articles in three rheumatology journals.
      Insufficient acknowledgement of limitations in Discussion sections has previously been found to be among the most important deficiencies in scientific articles.
      • Goodman S.N.
      • Berlin J.
      • Fletcher S.W.
      • Fletcher R.H.
      Manuscript quality before and after peer review and editing at Annals of Internal Medicine.
      An examination of 400 articles published in 2005 in eight leading journals showed that only 17% used at least one word denoting limitations in the context of the presented work in the manuscript, and only five articles had a separate section on limitations.
      • Ioannidis J.P.
      Limitations are not properly acknowledged in the scientific literature.
      These findings were reinforced in an analysis of 300 biomedical publications using linguistic software to detect expressions of uncertainty.
      • Ter Riet G.
      • Chesley P.
      • Gross A.G.
      • et al.
      All that glitters isn't gold: a survey on acknowledgment of limitations in biomedical studies.
      Researchers found that more than a quarter of studies did not acknowledge any limitations expressed in any way with some ‘hedge’ wording, and this was more likely in industry-funded studies.
      • Ter Riet G.
      • Chesley P.
      • Gross A.G.
      • et al.
      All that glitters isn't gold: a survey on acknowledgment of limitations in biomedical studies.
      A compelling argument has been made that articles that fully attempt to address limitations are likely to be widely cited and to shape the future research agenda.
      • Puhan M.A.
      • Akl E.A.
      • Bryant D.
      • Xie F.
      • Apolone G.
      • ter Riet G.
      Discussing study limitations in reports of biomedical studies—the need for more transparency.
      However, it might remain difficult for investigators to address honestly the limitations of their research.

      Potential solutions proposed to date

      Several solutions have been proposed in the past to address the problem of misleading Discussions, but none seems to be fully satisfactory. One proposed solution is for readers to read only Methods and Results sections, and to interpret the manuscript by themselves without reference to the paper's Discussion.
      • Montori V.M.
      • Jaeschke R.
      • Schunemann H.J.
      • et al.
      Users' guide to detecting misleading claims in clinical research reports.
      It has also been argued that Discussion sections can be entirely eliminated. However, this solution seems suboptimal because an impartial and informative Discussion can and should provide readers with critical insights and context. Even an educated reader certainly cannot be expected to perform a systematic review of other available evidence. Furthermore, many consumers of scientific literature lack the analytical tools for evaluation and critical interpretation, and as such often read only the Abstract, Introduction, and Discussion. Scientific manuscripts can also be placed in context by editorials. However, an editorial typically focuses on limitations, contextualisation, and clinical implications; it can seldom explore the methods, results and interpretations of specific papers both comprehensively and in sufficient depth. Moreover, editorials legitimately can, by their very definition, be biased and are often intended to reflect one opinion. Post-publication review could also, in theory, achieve similar goals by providing critical review and alternative viewpoints as a paper evolves. Unfortunately, post-publication peer review has not been widely embraced in mainstream journals, and it might be perceived as powerless, as it often does not mandate response from the authors.
      • Schriger D.L.
      • Altman D.G.
      Inadequate post-publication review of medical research.
      Initiatives such as PubMed Commons (now dismantled) and F1000 (https://f1000.com) have also aimed at enhancing post-publication peer review and discussion of the scientific literature.

      An alternative approach: an independent second discussant

      We propose that there should be two Discussion sections for scientific manuscripts, one of which would be written by the investigators, as is current practice, and the other would be authored independently of the research team. The investigators would make the protocol, Introduction, Methods, Results, and raw data (but not the Discussion section) available to an appropriate expert (or experts), who would be expected to write a second Discussion section for the manuscript. The primary author may propose potential independent second discussant(s), provided that transparent safeguards are in place about the impartiality of the selected scientist(s). Final choice would be up to the journal editors. As part of the second Discussion, the discussant could also specify whether and how he/she examined, probed, or scrutinised any of the data, analyses, and findings. With an increasing number of journals now requiring investigators to make available raw data and all versions of research protocols,
      • Iqbal S.A.
      • Wallach J.D.
      • Khoury M.J.
      • Schully S.D.
      • Ioannidis J.P.
      Reproducible research practices and transparency across the biomedical literature.
      • Taichman D.B.
      • Backus J.
      • Baethge C.
      • et al.
      Sharing clinical trial data: a proposal from the International Committee of Medical Journal Editors.
      • Hrynaszkiewicz I.
      • Norton M.L.
      • Vickers A.J.
      • Altman D.G.
      Preparing raw clinical data for publication: guidance for journal editors, authors, and peer reviewers.
      • Ioannidis J.P.
      • Greenland S.
      • Hlatky M.A.
      • et al.
      Increasing value and reducing waste in research design, conduct, and analysis.
      • Laine C.
      • Goodman S.N.
      • Griswold M.E.
      • Sox H.C.
      Reproducible research: moving toward research the public can really trust.
      including for observational studies,
      • Eisenach J.C.
      • Kheterpal S.
      • Houle T.T.
      Reporting of observational research in anesthesiology: the importance of the analysis plan.
      it is becoming feasible for a second discussant to judge the potential for selective reporting of outcomes, changes in analytic plans, and deviations from the protocol. The process will also make the original authors more likely to be transparent about such discrepancies with the original protocol and hopefully even prevent deviations from the original protocol.
      The second discussant may communicate with the primary team to obtain clarifications on important but unclear matters. Primary investigators should also convey to the discussant any additional information that pertains to possible weaknesses of their study, especially those that may not be readily discernible to an outsider. Eventually, the second discussant should be able to make a balanced appraisal of the study, place it in the context of other literature and discuss its implications in a thorough and impartial way. We suggest that this novel approach to writing manuscripts could complement the tools and guidelines that are available for reporting of studies.
      • Simera I.
      • Moher D.
      • Hoey J.
      • Schulz K.F.
      • Altman D.G.
      A catalogue of reporting guidelines for health research.
      Importantly, such an approach would introduce another layer of checks and balances, which could help to prevent scientific misconduct through an impartial third party having direct access to the data and other study materials. As inferential reproducibility is the explicit goal of this approach, it is likely to influence the primary discussants to adopt a more objective and less sensationalist tone, and it would probably promote more stringent thresholds for claiming ‘success.’
      • Ioannidis J.P.
      How to make more published research true.
      If the Discussion were demonstrably reproducible with this parallel approach, the credibility of the implications for medical practice would be enhanced, and suggestions regarding whether or not the findings of a study warranted a change in clinical care.
      • Lynn J.
      • Owens A.P.
      • Bartunek J.M.
      Clarity and strength of implications for practice in medical journal articles: an exploratory analysis.
      Conversely, a divergent second Discussion would represent a challenge to inferential reproducibility that might prompt a cautious re-evaluation of the original interpretation.

      Ideal attributes of an independent discussant

      It is important to consider what the essential attributes would be for a second discussant. Our perspective is that the chosen individual should certainly have expertise in scientific methods and interpretation, but would not necessarily need deep content (subject matter) expertise. The discussant could, for example, clarify the limitations of the P-values in a scientific manuscript, and inject a cautionary tone when interpreting results.
      • Goodman S.N.
      Toward evidence-based medical statistics: 1. The P value fallacy.
      • Halsey L.G.
      • Curran-Everett D.
      • Vowler S.L.
      • Drummond G.B.
      The fickle P value generates irreproducible results.
      Some content expertise would be useful however, because otherwise it may be difficult to contextualise the findings of a study using systematic review, as has been recommended.
      • Clarke M.
      • Hopewell S.
      • Chalmers I.
      Clinical trials should begin and end with systematic reviews of relevant evidence: 12 years and waiting.
      Importantly, even without vested interests in a particular manuscript, scientists may still have a priori beliefs and could unwittingly inject their partiality in a Discussion they intended to be impartial. Therefore, thoughtful selection of the discussants would be important and peer review of the independent Discussion would also be necessary, as with any scientific writing.

      Barriers to implementation

      The proposal of having a second independent discussant faces important barriers, some of which should be emphasised. If the primary authors do not include weaknesses in their own Discussion, they might be reluctant to reveal them to the independent discussant. This is a problem in particular for weaknesses that require insider knowledge of the study and its implementation. Hopefully, with a move towards more transparency and pre-registration for clinical trials and other types of research and availability of full protocols and of their revisions, this will be less of an issue. The approach of adding a discussant separate from the investigative team would create additional complexity and could result in delays to publication. However, having independent second discussants for an article is not without precedent, as some journals already publish additional Discussions immediately after the one prepared by the authors or they have influential papers accompanied by editorials. Our recommendation is that an impartial second Discussion complementing the current standard could build on this practice by providing the discussant with the protocol and raw data, rather than merely the narrative of a final article that already includes a Discussion. In order to be efficient, the process would need to be conducted expeditiously and with early engagement of the second discussant, perhaps during the revision phase of the peer review process as is common with editorial writers. Much like an editorialist, an independent discussant would fulfill the requirements for authorship as stipulated by the International Committee of Medical Journal Editors, so there should be also an academic incentive for becoming a good discussant and writing balanced, carefully thought Discussions.
      Finally, an open question is whether the independent discussant may be asked also to perform the systematic review that, as was noted, is indispensable for placing results in context. Such a review may take considerable time and effort, which could delay the publication of the work. In fact, we think that the primary authors would be better suited to perform this systematic review, as ideally a systematic review should be done before the study design phase. The independent discussant may then use this review and may also judge its completeness and impartiality.

      Testing whether discussions are improved

      Both the investigators and an independent discussant would bring strengths to their Discussion sections. The investigators would have content and context expertise, and would be familiar with the intricacies of the study. An independent second discussant would hopefully have expertise in research methodology and would regard a study through unbiased lenses. Although much of the material included in Discussions is subjective and qualitative, there are methods by which to evaluate independent Discussions compared with the current status quo in order to assess whether the proposed methods offer any tangible improvements. For example, natural language processing (NLP), a form of computational linguistics, is one strategy to measure the tone of a Discussion with relative objectivity. NLP and machine learning approaches to text analysis have been used (among numerous uses in clinical medicine and research) to identify emotional tone in dream reports,
      • Amini R.
      • Sabourin C.
      • De Koninck J.
      Word associations contribute to machine learning in automatic scoring of degree of emotional tones in dream reports.
      evaluate psychological states,
      • Kats Y.
      Towards NLP-based psychological evaluation.
      classify emotional expression in suicide notes,
      • Sohn S.
      • Torii M.
      • Li D.
      • Wagholikar K.
      • Wu S.
      • Liu H.
      A hybrid approach to sentiment sentence classification in suicide notes.
      and identify positive and negative emotional polarity.
      • Wang B.
      • Huang Y.
      • Wu X.
      • Li X.
      A Fuzzy computing model for identifying polarity of Chinese sentiment words.
      Qualitative analysis of text using NLP algorithms could allow comparisons of Discussions by study authors vs independent authors. Furthermore, standardised NLP programs could ultimately be used by authors, reviewers, or editors to screen for inappropriate levels of positive tone in article Discussions in the same way that investigators and journals currently screen for inappropriate levels of text reproduction (i.e. plagiarism). Furthermore, empirical studies by meta-researchers (e.g. observational assessments or even randomised trials comparing Discussion strategies) may continue to assess whether systematic reviews of the evidence are properly done and incorporated in Discussions and whether the conclusions and recommendations for practice are commensurate with the evidence.

      An example of an independent discussion

      In order to demonstrate how this proposal can be operationalised, we provide an example of an independent Discussion of a recent scientific manuscript, published in the journal JAMA.
      • van den Boogaard M.
      • Slooter A.J.C.
      • Bruggemann R.J.M.
      • et al.
      Effect of haloperidol on survival among critically ill adults with a high risk of delirium: the REDUCE randomized clinical trial.
      Following an established structure for a Discussion section (Table 1), one of the authors (M.S.A.) wrote the independent Discussion based on the Introduction, Methods, and Results sections of the manuscript, without reading the Discussion section. Another author (G.A.M.) then extracted the relevant sections from the manuscript's Discussion (Original Discussion), and assessed commonalities and differences between the separately written Discussions (Commentary). The third author (J.P.A.I.) reviewed and edited these Commentary assessments. The example of an independent Discussion with commentary is presented in Table 2.
      Table 1Relevant extracts from the British Journal of Anaesthesia Instructions to Authors, detailing the desirable structure and contents of a Discussion section of a scientific manuscript
      State main findings—This does not mean a repetition of all the results with their statistics. It should provide a concise overview of the study. For example, ‘Drug X produced a greater haemodynamic change on induction of anaesthesia than occurred with drug Y, resulting in a greater fall in arterial pressure and a higher incidence of tachycardia’. Do not repeat data already presented in the Results, figures, and tables.
      Relation to previous studies—This section should relate directly to the statements made in the Introduction and qualify your findings in relation to previous studies of the subject. Mention any important uncertainties in the methods of measurement. In laboratory studies, try to relate the concentrations used to those encountered clinically.
      Why are they different/same—Avoid excessive speculation or speculation beyond your results. It is quite reasonable to suggest possible explanations for your findings and any differences from previous studies, but the 'missing parts' of such reasoning must be acknowledged.
      Additions to knowledge of the subject—Summarise previous sections by pulling together the implications of your main findings, studies by other workers and their combined contribution to our knowledge of the subject. This should not be just another repetition of the results and preceding discussion but more of an expanded conclusion. A meta-analysis incorporating the results of the new study and previous relevant studies could be included in this paragraph.
      Weaknesses and strengths in study—This is the most important part of the Discussion. Acknowledge any limitations of your study at this point. Examples here could include the generalisability of the patient population, patients lost to follow-up (or missing data in general), and limitations of analytical test. Authors are advised to be honest but succinct in this section. It is also reasonable to comment briefly on the particular strengths of the study, especially in comparison with comparable studies.
      Future studies—Identify future studies that would address some of the potential explanations and limitations discussed earlier. This section should be very brief.
      Conclusions—The original contribution to knowledge from the present study should be stated. A common fault here is to overstate the findings from a study. It may be appropriate to give the implications of the conclusions for (anaesthetic) practice and the indications for further enquiry in this area of interest.
      Table 2Original and Independent Discussions, and interpretive Commentary
      Title of Manuscript: Effect of Haloperidol on Survival Among Critically Ill Adults With a High Risk of Delirium: The REDUCE Randomized Clinical Trial
      • van den Boogaard M.
      • Slooter A.J.C.
      • Bruggemann R.J.M.
      • et al.
      Effect of haloperidol on survival among critically ill adults with a high risk of delirium: the REDUCE randomized clinical trial.
      Main Finding

      Original Discussion: ‘In this large multicenter double-blind randomized clinical trial involving critically ill adults at high risk of delirium, there was no significant difference in the number of days survived at 28 days following inclusion between patients who received prophylactic haloperidol therapy and patients who received placebo. The pathophysiological mechanism of delirium is poorly understood. Delirium is considered a multifactorial disorder and many different pathways to its occurrence have been postulated resulting in many hypotheses. The fact that the average of 11 risk factors are present at the same time in ICU patients with delirium suggests the involvement of multiple pathways in its development. Therefore, it seems plausible that a mediator that alters causal pathways of delirium may be helpful. Haloperidol is an antipsychotic agent with antidopaminergic, antiadrenergic, limited anticholinergic properties and possibly has anti-inflammatory effects that potentially antagonizes multiple pathways of delirium.’
      • van den Boogaard M.
      • Slooter A.J.C.
      • Bruggemann R.J.M.
      • et al.
      Effect of haloperidol on survival among critically ill adults with a high risk of delirium: the REDUCE randomized clinical trial.


      Independent Discussion: Delirium is associated with death. Haloperidol might prevent delirium in critically ill patients. Therefore, if haloperidol prevents delirium, it might also prevent death, assuming that delirium causes death. There was no evidence in this study that haloperidol compared with placebo increases or decreases 28 day mortality. Based on a precision estimate (95% confidence interval for the hazard ratio of 0.78–1.30), the results of this study remain consistent with a clinically meaningful decrease or increase in 28 day mortality attributable to haloperidol.

      Commentary: The original investigators provided rationale for the study in relation to delirium, but did not provide rationale in relation to the primary outcome, which was mortality. The summary of the main finding was the same (i.e. that there is no difference in mortality), although the independent discussant acknowledged also that clinically meaningful changes in mortality cannot be excluded.
      Relationship of Main Finding to Previous Studies

      Independent Discussion: Based on systematic review of the scientific literature, there is no previous evidence on the effectiveness of haloperidol prophylaxis in preventing mortality in critically ill patients. In other settings (e.g. those with myocardial infarction, dementia, or in nursing homes), there is some evidence that prophylactic haloperidol is associated with increased mortality. Overall, the evidence for a causal link between haloperidol and either increased or decreased mortality is weak.
      • Park Y.
      • Bateman B.T.
      • Kim D.H.
      • et al.
      Use of haloperidol versus atypical antipsychotics and risk of in-hospital death in patients with acute myocardial infarction: cohort study.
      • Maust D.T.
      • Kim H.M.
      • Seyfried L.S.
      • et al.
      Antipsychotics, other psychotropics, and the risk of death in patients with dementia: number needed to harm.
      • Gerhard T.
      • Huybrechts K.
      • Olfson M.
      • et al.
      Comparative mortality risks of antipsychotic medications in community-dwelling older adults.
      • Hulshof T.A.
      • Zuidema S.U.
      • Ostelo R.W.
      • Luijendijk H.J.
      The mortality risk of conventional antipsychotics in elderly patients: a systematic review and meta-analysis of randomized placebo-controlled trials.


      Commentary: In the original manuscript the Discussion did not contextualise the main finding in relation to previous studies, perhaps because this specific question has not previously been addressed.
      Additional (Secondary) Findings

      Original Discussion: ‘Also, no significant differences were found in the number of days patients survived at 90 days between the haloperidol group and the placebo group. No differences were found for any other reported secondary end points. Furthermore, across predefined subgroups, the lack of a prophylactic effect was very consistent. Prophylactic haloperidol therapy was not associated with haloperidol-induced adverse effects.’
      • van den Boogaard M.
      • Slooter A.J.C.
      • Bruggemann R.J.M.
      • et al.
      Effect of haloperidol on survival among critically ill adults with a high risk of delirium: the REDUCE randomized clinical trial.


      Independent Discussion: There were no differences between groups in 90 day mortality, safety outcomes, delirium incidence, or any delirium related outcome. Higher APACHE-II (severity of acute illness) and PRE-DERELIC (delirium prediction) scores were independently associated with increased delirium risk and decreased survival. Delirium incidence was high (33%), as expected, in these critically ill patients. Surprisingly, the side-effects anticipated with haloperidol (sedation and prolonged QTc) were not seen in this study.

      Commentary: The summary of additional findings was similar in both Discussions. In both Discussions, surprise was expressed that there were no apparent side-effects or safety concerns with prophylactic (low dose) haloperidol. The independent Discussion addressed some predictive analyses that were not discussed in the original.
      Relationship of Additional (Secondary) Findings to Previous Studies

      Original Discussion: ‘Haloperidol has been the first-line drug of choice to treat delirium for decades despite the lack of evidence that haloperidol is effective. For this reason, the Society of Critical Care Medicine in its last guideline on pain, agitation, and delirium did not recommend the use of haloperidol for treatment or for delirium prevention for critically ill adults. However, several ICU studies have evaluated possible prophylactic effects of haloperidol but demonstrate contradictory effects. In one randomized clinical trial, postoperative ICU patients received a maximum of 1.2 mg of haloperidol and showed a reduced delirium incidence and more delirium-free days. Another randomized clinical trial involving severely ill medical ICU patients receiving 2.5 mg 3 times daily, showed no beneficial effect, and no effect was found in reducing subsyndromal delirium with prophylactic haloperidol. Although a previous before–after study showed clinical relevant and favorable effects in a similar group of high-risk and critically ill adults, these beneficial effects could not be replicated in the current randomized clinical trial. However, the findings of this study corroborate the findings of other randomized clinical trials involving critically ill adults.’
      • van den Boogaard M.
      • Slooter A.J.C.
      • Bruggemann R.J.M.
      • et al.
      Effect of haloperidol on survival among critically ill adults with a high risk of delirium: the REDUCE randomized clinical trial.


      Independent Discussion: Studies have generally not found that haloperidol is effective in preventing or treating delirium.
      • Santos E.
      • Cardoso D.
      • Neves H.
      • Cunha M.
      • Rodrigues M.
      • Apostolo J.
      Effectiveness of haloperidol prophylaxis in critically ill patients with a high risk of delirium: a systematic review.
      Specifically, the evidence for haloperidol being effective at preventing delirium is sparse. Thus, the results of this trial are consistent with previous randomised trials.

      Commentary: There is agreement between the Discussions that the findings in relation to delirium prevention are largely consistent with the findings of previous randomised trials.
      Limitations:

      Original Discussion: ‘This study has several limitations. First, the 1 mg haloperidol group was terminated early, as a predefined consequence of the adaptive design, which is considered a strength of our study. This discontinuation did not affect the presented study findings. Second, the duration of prophylactic therapy (median, 2 days) could be too short to prevent delirium and its deleterious outcome. It cannot be excluded that longer exposure to haloperidol may be needed to influence patient outcome. However, subgroup analysis in patients treated for more than 2 days also did not show any beneficial effect. Third, the dose may have been too low. In a before–after study a 1 mg dose every 8 hours demonstrated beneficial effects without relevant adverse effects. For this reason, using a higher dose was included as part of the current study, similar to the study of Page et al. In both haloperidol dosage groups no beneficial effects were found. Fourth, it was not feasible to collect data for all secondary outcome measures in some centers due to research staff limitations. However, the median number of delirium- and coma-free days between both groups did not differ; therefore, collecting these data in a somewhat smaller group did not affect the results. Fifth, the study population included severely ill ICU adults, whose brains may have been too seriously affected for haloperidol to exert a prophylactic effect, since in non-ICU adults, prophylactic haloperidol may have beneficial effects. But the subgroup of patients with a low severity of illness score also demonstrated no beneficial effects. Nevertheless, it cannot be ruled out that delirium may be more easily and favorably affected in non-ICU adults than critically ill ICU adults. Sixth, the results of the long-term quality of life are not included herein, and therefore the effects of haloperidol on quality of life remain to be determined.’
      • van den Boogaard M.
      • Slooter A.J.C.
      • Bruggemann R.J.M.
      • et al.
      Effect of haloperidol on survival among critically ill adults with a high risk of delirium: the REDUCE randomized clinical trial.


      Independent Discussion: A single study seldom provides definitive evidence,
      • Murad M.H.
      • Montori V.M.
      Synthesizing evidence: shifting the focus from individual studies to the body of evidence.
      • Guyatt G.H.
      • Mills E.J.
      • Elbourne D.
      In the era of systematic reviews, does the size of an individual trial still matter.
      and although this was a ‘negative’ study (there were no estimated differences between the groups in the primary or secondary outcomes), the 95% confidence intervals for differences in both 28 day mortality (–3.4%–4.6%) and delirium (–4.6%–5.4%) included clinically important both benefit and harm. Because haloperidol was administered open label for treatment of delirium in both groups, its potential effectiveness for treatment could not be determined in this study. Delirium was not assessed in all the participating sites resulting in missing assessments; thus, the effective sample size for the delirium outcomes was reduced. It is possible that the dose of haloperidol was insufficient or that the time required for effective prophylaxis was insufficient. The hypothesis underpinning this study appears somewhat implausible. In order for haloperidol to prevent death, it would have to be effective first in preventing delirium, and second delirium would have to be causally implicated in (rather than merely being associated with) mortality. Delirium is a complex syndrome that is likely to have multiple pathophysiological pathways
      • Inouye S.K.
      • Westendorp R.G.
      • Saczynski J.S.
      Delirium in elderly people.
      ; it is therefore improbable that any single intervention will have a marked impact on delirium incidence. Thus, the pre–study likelihood of a positive primary outcome with the intervention was low.
      • Nuzzo R.
      Scientific method: statistical errors.


      Commentary: The main difference in emphasis was that the independent discussant noted the imprecision around the estimates for death and delirium. The independent discussant therefore inferred that this study should not be regarded as providing definitive evidence in relation to these questions. The independent discussant also felt that the initial premise behind this study was not compelling; there was no strong biological plausibility for the hypothesis. Both the independent Discussion and the original listed several limitations, with some overlap, but also differences, and these limitations can form a complementary list.
      Strengths:

      Original Discussion: ‘The large sample size of the current study allowed us to perform several sensitivity analyses, confirming the lack of effect across the different subgroups.’
      • van den Boogaard M.
      • Slooter A.J.C.
      • Bruggemann R.J.M.
      • et al.
      Effect of haloperidol on survival among critically ill adults with a high risk of delirium: the REDUCE randomized clinical trial.


      Independent Discussion: This was a rigorously conducted, large multi-centre pragmatic trial. The protocol for the study was thorough and comprehensible. The study was pre-registered, and the protocol was published ahead of time. Overall there were few deviations from the protocol and outcomes were unchanged with per-protocol analysis. The adaptive design allowed for a more efficient study.

      All these factors increase confidence in relation to the robustness of the findings.

      Commentary: The independent discussant highlighted more strengths of the study than the investigators themselves did. The lists of strengths in the two Discussions are complementary.
      Future Directions:

      Independent Discussion: It will be important to clarify the pathophysiology of delirium so that mechanistically targeted treatments can be developed and tested
      Conclusion:

      Original Discussion: ‘Among critically ill adults with high-risk of delirium, the use of prophylactic haloperidol compared with placebo did not improve 28-day survival. These findings do not support the use of prophylactic haloperidol in critically ill adults.’
      • van den Boogaard M.
      • Slooter A.J.C.
      • Bruggemann R.J.M.
      • et al.
      Effect of haloperidol on survival among critically ill adults with a high risk of delirium: the REDUCE randomized clinical trial.


      Independent Discussion: It is unlikely that prophylactic haloperidol for critically ill patients on intensive care units is effective for preventing either death or delirium.

      Commentary: The take-home message was the same.
      Inferential Reproducibility:

      Commentary: Overall, there is considerable concordance between the two Discussions, and inferential reproducibility is generally endorsed. There is some discordance regarding the extent to which the ‘negative’ findings are interpreted as being ‘definitive’, and also in relation to the biological plausibility of the hypothesis on which this study was predicated.

      Conclusion

      More objective, balanced Discussions are needed in research reporting. Using second independent discussants may improve the quality of research articles and contribute to the evolving culture or research reproducibility. We acknowledge that there are still important questions about the practicality of the proposed approach, and the exact mode in which it would be best implemented. The best way to resolve these questions is by piloting the concept in a number of studies and drawing some endorsement by journals, research organisations, and funding agencies. Notwithstanding the known and unknown constraints, our opinion remains that second independent and dispassionate Discussions are feasible, and would help to combat bias, misinformation, and lack of reproducibility in science. Enhancing inferential reproducibility is necessary to the appropriate determination of whether new knowledge merits advance along the translational continuum.

      Authors' contributions

      All authors wrote the first draft of the manuscript, contributed to the writing of the manuscript, agreed with the manuscript's results and conclusions, and approved the final version of this article. All authors have read, and confirm that they meet, the ICMJE criteria for authorship.

      Declarations of interest

      MSA is an editor of the British Journal of Anaesthesia. The other authors have no relevant conflicts to declare.

      References

        • Collins F.S.
        • Tabak L.A.
        Policy: NIH plans to enhance reproducibility.
        Nature. 2014; 505: 612-613
        • Atmanspacher H.
        • Bezzola Lambert L.
        • Folkers G.
        • Schubiger P.A.
        Relevance relations for the concept of reproducibility.
        J R Soc Interface. 2014; 11: 20131030
        • Goodman S.N.
        • Fanelli D.
        • Ioannidis J.P.
        What does research reproducibility mean?.
        Sci Transl Med. 2016; 8: 341ps12
        • Clarke M.
        • Chalmers I.
        Discussion sections in reports of controlled trials published in general medical journals: islands in search of continents?.
        JAMA. 1998; 280: 280-282
        • Clarke M.
        • Hopewell S.
        • Chalmers I.
        Reports of clinical trials should begin and end with up-to-date systematic reviews of other relevant evidence: a status report.
        J R Soc Med. 2007; 100: 187-190
        • Clarke M.
        • Alderson P.
        • Chalmers I.
        Discussion sections in reports of controlled trials published in general medical journals.
        JAMA. 2002; 287: 2799-2801
        • Goodman S.N.
        Of P-values and Bayes: a modest proposal.
        Epidemiology. 2001; 12: 295-297
        • Goodman S.N.
        Toward evidence-based medical statistics: 1. The P value fallacy.
        Ann Intern Med. 1999; 130: 995-1004
        • Sumner P.
        • Vivian-Griffiths S.
        • Boivin J.
        • et al.
        The association between exaggeration in health related science news and academic press releases: retrospective observational study.
        BMJ. 2014; 349: g7015
        • Yavchitz A.
        • Boutron I.
        • Bafeta A.
        • et al.
        Misrepresentation of randomized controlled trials in press releases and news coverage: a cohort study.
        PLoS Med. 2012; 9 (e1001308)
        • Begg C.
        • Cho M.
        • Eastwood S.
        • et al.
        Improving the quality of reporting of randomized controlled trials. The CONSORT statement.
        JAMA. 1996; 276: 637-639
        • Perino A.C.
        • Hoang D.D.
        • Holmes T.H.
        • et al.
        Association between success rate and citation count of studies of radiofrequency catheter ablation for atrial fibrillation: possible evidence of citation bias.
        Circ Cardiovasc Qual Outcome. 2014; 7: 687-692
        • Kivimaki M.
        • Batty G.D.
        • Kawachi I.
        • Virtanen M.
        • Singh-Manoux A.
        • Brunner E.J.
        Don’t let the truth get in the way of a good story: an illustration of citation bias in epidemiologic research.
        Am J Epidemiol. 2014; 180: 446-448
        • Jannot A.S.
        • Agoritsas T.
        • Gayet-Ageron A.
        • Perneger T.V.
        Citation bias favoring statistically significant studies was present in medical research.
        J Clin Epidemiol. 2013; 66: 296-301
        • Lumbreras B.
        • Parker L.A.
        • Porta M.
        • Pollan M.
        • Ioannidis J.P.
        • Hernandez-Aguado I.
        Overinterpretation of clinical applicability in molecular diagnostic research.
        Clin Chem. 2009; 55: 786-794
        • Ochodo E.A.
        • de Haan M.C.
        • Reitsma J.B.
        • Hooft L.
        • Bossuyt P.M.
        • Leeflang M.M.
        Overinterpretation and misreporting of diagnostic accuracy studies: evidence of “spin”.
        Radiology. 2013; 267: 581-588
        • Gewandter J.S.
        • McKeown A.
        • McDermott M.P.
        • et al.
        Data interpretation in analgesic clinical trials with statistically nonsignificant primary analyses: an ACTTION systematic review.
        J Pain. 2015; 16: 3-10
        • Boutron I.
        • Dutton S.
        • Ravaud P.
        • Altman D.G.
        Reporting and interpretation of randomized controlled trials with statistically nonsignificant results for primary outcomes.
        JAMA. 2010; 303: 2058-2064
        • Kinder N.
        • Weaver M.D.
        • Wayant C.
        • Vassar M.
        Presence of spin in the abstracts and titles of anaesthesiology randomized controlled trials.
        Br J Anaesth. 2019; 122: e13-e14
        • Bero L.
        • Oostvogel F.
        • Bacchetti P.
        • Lee K.
        Factors associated with findings of published trials of drug–drug comparisons: why some statins appear more efficacious than others.
        PLoS Med. 2007; 4: e184
        • Lundh A.
        • Sismondo S.
        • Lexchin J.
        • Busuioc O.A.
        • Bero L.
        Industry sponsorship and research outcome.
        Cochrane Database Syst Rev. 2012; 12: MR000033
        • Goodman S.
        A dirty dozen: twelve p-value misconceptions.
        Semin Hematol. 2008; 45: 135-140
        • Ioannidis J.P.
        Limitations are not properly acknowledged in the scientific literature.
        J Clin Epidemiol. 2007; 60: 324-329
        • Yazici H.
        • Gogus F.
        • Esen F.
        • Yazici Y.
        There was less self-critique among basic than in clinical science articles in three rheumatology journals.
        J Clin Epidemiol. 2014; 67: 654-657
        • Goodman S.N.
        • Berlin J.
        • Fletcher S.W.
        • Fletcher R.H.
        Manuscript quality before and after peer review and editing at Annals of Internal Medicine.
        Ann Intern Med. 1994; 121: 11-21
        • Ter Riet G.
        • Chesley P.
        • Gross A.G.
        • et al.
        All that glitters isn't gold: a survey on acknowledgment of limitations in biomedical studies.
        PLoS One. 2013; 8e73623
        • Puhan M.A.
        • Akl E.A.
        • Bryant D.
        • Xie F.
        • Apolone G.
        • ter Riet G.
        Discussing study limitations in reports of biomedical studies—the need for more transparency.
        Health Qual Life Outcome. 2012; 10: 23
        • Montori V.M.
        • Jaeschke R.
        • Schunemann H.J.
        • et al.
        Users' guide to detecting misleading claims in clinical research reports.
        BMJ. 2004; 329: 1093-1096
        • Schriger D.L.
        • Altman D.G.
        Inadequate post-publication review of medical research.
        BMJ. 2010; 341: c3803
        • Iqbal S.A.
        • Wallach J.D.
        • Khoury M.J.
        • Schully S.D.
        • Ioannidis J.P.
        Reproducible research practices and transparency across the biomedical literature.
        PLoS Biol. 2016; 14 (e1002333)
        • Taichman D.B.
        • Backus J.
        • Baethge C.
        • et al.
        Sharing clinical trial data: a proposal from the International Committee of Medical Journal Editors.
        PLoS Med. 2016; 13 (e1001950)
        • Hrynaszkiewicz I.
        • Norton M.L.
        • Vickers A.J.
        • Altman D.G.
        Preparing raw clinical data for publication: guidance for journal editors, authors, and peer reviewers.
        Trials. 2010; 11: 9
        • Ioannidis J.P.
        • Greenland S.
        • Hlatky M.A.
        • et al.
        Increasing value and reducing waste in research design, conduct, and analysis.
        Lancet. 2014; 383: 166-175
        • Laine C.
        • Goodman S.N.
        • Griswold M.E.
        • Sox H.C.
        Reproducible research: moving toward research the public can really trust.
        Ann Intern Med. 2007; 146: 450-453
        • Eisenach J.C.
        • Kheterpal S.
        • Houle T.T.
        Reporting of observational research in anesthesiology: the importance of the analysis plan.
        Anesthesiology. 2016; 124: 998-1000
        • Simera I.
        • Moher D.
        • Hoey J.
        • Schulz K.F.
        • Altman D.G.
        A catalogue of reporting guidelines for health research.
        Eur J Clin Invest. 2010; 40: 35-53
        • Ioannidis J.P.
        How to make more published research true.
        PLoS Med. 2014; 11 (e1001747)
        • Lynn J.
        • Owens A.P.
        • Bartunek J.M.
        Clarity and strength of implications for practice in medical journal articles: an exploratory analysis.
        BMJ Qual Saf. 2011; 20 (i52–7)
        • Halsey L.G.
        • Curran-Everett D.
        • Vowler S.L.
        • Drummond G.B.
        The fickle P value generates irreproducible results.
        Nat Methods. 2015; 12: 179-185
        • Clarke M.
        • Hopewell S.
        • Chalmers I.
        Clinical trials should begin and end with systematic reviews of relevant evidence: 12 years and waiting.
        Lancet. 2010; 376: 20-21
        • Amini R.
        • Sabourin C.
        • De Koninck J.
        Word associations contribute to machine learning in automatic scoring of degree of emotional tones in dream reports.
        Conscious Cogn. 2011; 20: 1570-1576
        • Kats Y.
        Towards NLP-based psychological evaluation.
        AMIA Annu Symp Proc. 2007; : 1005
        • Sohn S.
        • Torii M.
        • Li D.
        • Wagholikar K.
        • Wu S.
        • Liu H.
        A hybrid approach to sentiment sentence classification in suicide notes.
        Biomed Inform Insights. 2012; 5: 43-50
        • Wang B.
        • Huang Y.
        • Wu X.
        • Li X.
        A Fuzzy computing model for identifying polarity of Chinese sentiment words.
        Comput Intell Neurosci. 2015; 2015: 525437
        • van den Boogaard M.
        • Slooter A.J.C.
        • Bruggemann R.J.M.
        • et al.
        Effect of haloperidol on survival among critically ill adults with a high risk of delirium: the REDUCE randomized clinical trial.
        JAMA. 2018; 319: 680-690
        • Park Y.
        • Bateman B.T.
        • Kim D.H.
        • et al.
        Use of haloperidol versus atypical antipsychotics and risk of in-hospital death in patients with acute myocardial infarction: cohort study.
        BMJ. 2018; 360: k1218
        • Maust D.T.
        • Kim H.M.
        • Seyfried L.S.
        • et al.
        Antipsychotics, other psychotropics, and the risk of death in patients with dementia: number needed to harm.
        JAMA Psychiatry. 2015; 72: 438-445
        • Gerhard T.
        • Huybrechts K.
        • Olfson M.
        • et al.
        Comparative mortality risks of antipsychotic medications in community-dwelling older adults.
        Br J Psychiatry. 2014; 205: 44-51
        • Hulshof T.A.
        • Zuidema S.U.
        • Ostelo R.W.
        • Luijendijk H.J.
        The mortality risk of conventional antipsychotics in elderly patients: a systematic review and meta-analysis of randomized placebo-controlled trials.
        J Am Med Dir Assoc. 2015; 16: 817-824
        • Santos E.
        • Cardoso D.
        • Neves H.
        • Cunha M.
        • Rodrigues M.
        • Apostolo J.
        Effectiveness of haloperidol prophylaxis in critically ill patients with a high risk of delirium: a systematic review.
        JBI Database Syst Rev Implement Rep. 2017; 15: 1440-1472
        • Murad M.H.
        • Montori V.M.
        Synthesizing evidence: shifting the focus from individual studies to the body of evidence.
        JAMA. 2013; 309: 2217-2218
        • Guyatt G.H.
        • Mills E.J.
        • Elbourne D.
        In the era of systematic reviews, does the size of an individual trial still matter.
        PLoS Med. 2008; 5: e4
        • Inouye S.K.
        • Westendorp R.G.
        • Saczynski J.S.
        Delirium in elderly people.
        Lancet. 2014; 383: 911-922
        • Nuzzo R.
        Scientific method: statistical errors.
        Nature. 2014; 506: 150-152

      Linked Article

      • Nephrogenic diabetes insipidus associated with prolonged sedation with sevoflurane in the intensive care unit
        British Journal of AnaesthesiaVol. 122Issue 5
        • In Brief
          Practice has changed in ICUs with the goal to minimise sedation in order to reduce length of stay (LOS) and delirium.1 This should result in decreased benzodiazepine use,2 and volatile anaesthetic agents could provide a safe alternative. Sevoflurane appears to be effective in reducing wake-up and extubation times compared with propofol or midazolam,3,4 and was recently described for sedation in acute respiratory distress syndrome (ARDS).5 Studies suggest renal safety,6 but data are lacking on long-term sedation.
        • Full-Text
        • PDF
        Open Archive