Statistics from Altmetric.com
This paper describes problems that arose over the publication of a study of chlamydia screening, which was completed in 2000. The account has been agreed by the authors of the papers, the editors of the two journals involved and by the Council of the Committee on Publication Ethics (COPE). The aim of revisiting events that occurred several years ago is to describe some of the difficulties that can arise, even when all parties act in good faith, with the hope of preventing similar problems in the future and to make readers aware of the publication history of the three papers. The UK Department of Health (DH) funded a large-scale study of opportunistic screening for chlamydia. One of the two centres involved in the study was Portsmouth. It was agreed from the outset that the Portsmouth data would be published separately, because screening methodology had differed between the two centres. The Portsmouth investigators therefore felt their presentation would be a valid separate analysis, with important messages for a different audience from the paper describing the national findings. A paper1 describing the national results from the study including data from Portsmouth was submitted to the BMJ in April 2002 together with an accompanying paper2 describing the methodology. The two papers were rejected by the BMJ and then submitted to Sexually Transmitted Infections (STI) in June 2002, which accepted both papers in October 2002. Over the same period, the investigators from Portsmouth prepared their own paper,3 which they discussed with the DH, who agreed that it could be submitted but advised that it should not be published before the other papers. The paper from Portsmouth was submitted to STI in March 2002. However, on receiving the paper describing the national findings (in June 2002), the editor of STI (Dr Shahmanesh) considered that the paper from Portsmouth overlapped too much with the national results paper to be published separately in his journal and therefore asked the Portsmouth authors to withdraw their paper from STI, which they did in July 2002. Subsequently, the Portsmouth authors submitted their paper (unchanged) to the Journal of Family Planning and Reproductive Health Care (JFPRHC) together with copies of the national results paper that was under consideration at STI. The editor of JFPRHC agreed with the Portsmouth authors that their paper, although overlapping to some extent with the other paper, was sufficiently different to warrant publication and therefore accepted it in October 2002. Unfortunately, although the authors had been open with the editor of JFPRHC and had supplied a copy of the related paper with their submission, because they did not know the bibliographical details of the STI papers they did not reference them in the JFPRHC paper but cited a DH website, which, at the time of publication, gave details of the study. However, the DH website cited in the JFPRHC paper no longer includes information about the study, so later readers could not determine the relationship between the papers.
The Portsmouth authors also kept the trial sponsor (the DH) fully informed of their publication plans, and the DH approved the publication of the Portsmouth data on the understanding that these would be published after the national findings. The contact at the DH mistakenly believed that the papers describing the national study would appear in STI in December 2002, and this date was not checked by the Portsmouth authors (who were authors on the other publication and therefore could have obtained this information from STI) or by the editor of JFPRHC. The paper from Portsmouth was published in JFPRHC in January 2003, whereas the papers describing the national findings were published in STI in February 2003. The papers published in STI made no reference to the Portsmouth paper, although all six of the Portsmouth authors were also authors of the STI paper; however, the lead author of the STI papers was unaware that the Portsmouth paper had been accepted by JFPRHC. A reader alerted the editors of STI (Drs Miller and Ward, who had taken over from Dr Shahmanesh) to the overlap between the Portsmouth paper in JFPRHC and the paper published in STI. The editors of STI discussed the case with their journal ombudsman (Dr Shahmanesh, the previous editor), the editor of JFPRHC (Dr Reader) and two independent expert reviewers. Dr Shahmanesh reiterated his view that there was too much overlap between the Portsmouth paper and the primary paper to warrant separate publication. Dr Reader, however, maintained that the separate publication was justifiable in the light of the additional data obtained and different methods used in Portsmouth. The two expert reviewers disagreed: one reviewer performed a systematic critique and concluded that the Portsmouth paper constituted redundant publication, whereas the other reviewer, although admitting there was some overlap, felt it was not redundant. Faced by these divided opinions, Drs Miller and Ward brought the case to COPE. It was discussed at a COPE meeting on the basis of a short, anonymised summary written by Drs Miller and Ward (following COPE’s normal practice). This summary stated that the journal (and its ombudsman, Dr Shahmanesh) considered this to be a case of redundant publication but did not represent the view of the authors of the Portsmouth paper, the editor of JFPRHC and one of the independent reviewers that it was an acceptable secondary publication. On the basis of the case as described by Drs Miller and Ward, COPE recommended that STI should publish a notice of redundant publication4 and contact the authors’ institution to alert them to this. The authors of both papers were offered the opportunity to respond to the notice of redundant publication, which was published as a rejoinder alongside the notice in STI.5 The anonymised case was published on the COPE website (again, following COPE’s normal practice). When it became aware that its recommendations had been challenged by the authors, and were not accepted by Dr Reader (the editor of JFPRHC), COPE invited Dr Reader to attend a COPE meeting to present her (and the authors’) side of the story. On learning about the different opinions on whether the papers constituted redundant publication and the authors’ concerns that, even after anonymisation, the case might be recognisable, COPE removed the case from its website. Following the meeting, Dr Reader wrote to COPE expressing her concerns about the way in which the case had been handled. Sometime later, her letter was passed to the newly appointed COPE ombudsman. A considerable delay ensued as a result of the fact that the COPE ombudsman was unable to consider cases until COPE’s constitution and legal status provided appropriate indemnity. As soon as COPE was able to provide the necessary legal indemnity, the ombudsman tried to arrange a meeting between the authors and the journal editors to understand their complaints about COPE’s handling of the case and considerations of possible redundant publication. It took some time to find a mutually convenient date, but this meeting eventually took place in May 2008, at which it was agreed that this account should be written.
COPE accepts the criticism that it should try to ensure that cases are properly represented and not oversimplified. Believing that this was a clear case of redundant publication, COPE recommended that STI should publish a notice of redundant publication and inform the authors’ institutions. Although this is consistent with COPE’s guidelines, COPE accepts that its recommendation was perhaps inappropriate, given the complexity of the case and the lack of agreement about whether the papers really were redundant. However, the COPE ombudsman considers that the papers do constitute redundant publication and therefore that the actions of the STI editors in issuing a notice of redundant publication (on COPE’s advice) were therefore appropriate.
COPE posts all cases it receives, together with a summary of its discussion and advice, onto its website to form a resource for editors and researchers. As both journals were COPE members, COPE accepts that it should have notified the editor of JFPRHC before posting this case on its website. Every effort is made to ensure that the cases are anonymised before posting and details of the submitting journals’ names are never included. However, although anonymisation usually prevents the great majority of readers from identifying an individual case, it is almost impossible to prevent people who have been involved in a case, or who work with the individuals concerned, from recognising it. In this case, the posting (which was withdrawn from the COPE website) referred to “a pilot study on two sites to examine the feasibility and acceptability of screening for infection X” set up by “a country’s national health ministry”. The Portsmouth authors consider that, given this information, the case could be identified as the UK Department of Health Chlamydia Screening Study. COPE accepts that people familiar with the study or the publications might guess its identity but believes its practice of posting anonymised cases is appropriate, given the usefulness of the resulting database. However, COPE accepts that it must continue to be vigilant to ensure that cases are anonymised to the fullest extent possible while still retaining useful information.
No objective definition of redundant publication exists. The World Association of Medical Editors guidelines state “Redundant publication occurs when multiple papers, without full cross reference in the text, share the same data, or results”.6 The International Committee of Medical Journal Editors uniform requirements state that “Redundant (or duplicate) publication is publication of a paper that overlaps substantially with one already published in print or electronic media”.7 The uniform requirements go on to state that “Duplicate publication of original research is particularly problematic, since it can result in inadvertent double counting or inappropriate weighting of the results of a single study, which distorts the available evidence”. The COPE flowcharts define “major overlap/redundancy” as attempting to publish identical findings especially when there is “evidence that authors have sought to hide redundancy (eg, by changing author order and title)”. The flowcharts define “minor overlap” as “salami publishing with some element of redundancy or legitimate re-analysis (eg, subgroup/extended follow-up)”. The flowchart recommends that editors should consider issuing a notice of redundancy only in the case of major overlap, whereas for cases of minor overlap they should simply “discuss publishing [a] correction giving [the] reference to [the] original paper”. All the available guidelines therefore indicate that some forms of secondary publication are legitimate and will, necessarily, contain some degree of overlap with the primary publication. However, it remains at the discretion of editors to decide exactly how much overlap is permissible and how much constitutes a redundant publication.
Both journals involved in this case provide guidance to potential authors about redundant publication. The notes for contributors for JFPRHC state that “Papers submitted for publication to the journal are assumed to be submitted exclusively unless it is clearly stated to the contrary. Papers must not have been, or be about to be, published elsewhere, either wholly or in part”. Similarly, STI states “Papers are considered by STI on the basis that they are submitted solely to this journal and do not duplicate material already published elsewhere. In cases of doubt, where part of the material has been published elsewhere, the published material should be submitted together with the manuscript (as a supplementary file) and this fact mentioned in the submission cover letter.”
When several papers are published from the same study, they must be properly cross-referenced. Although the Portsmouth authors were open with the editor of JFPRHC about the existence of the other papers, and included a reference to the DH website, it would have been preferable to include a citation to the national paper (eg, as STI “in press”) and an explicit description of the relationship between the papers. The Portsmouth paper should also have been cited in the national paper. JFPRHC should publish a note directing readers of the Portsmouth paper to the STI paper so they understand that some of the data have been published previously. Undetected redundant publication has been shown to bias systematic reviews if patients are double counted and is therefore potentially harmful to the evidence base.8 The practice of including trial registration numbers in all publications (which has become more common since the chlamydia papers were published)9 10 should reduce the risks of data from the same study being included more than once in data syntheses. It will also help journal editors, reviewers and readers to identify papers arising from a single study.
Editors have a duty to correct the published record and ensure that readers are not misled. Judging the paper to be redundant, the editors of STI were correct to follow COPE’s suggestion that a notice of redundant publication should be published.
Journal editors also have a responsibility to inform authors’ employers or institutions when they suspect that research or publication misconduct has occurred.11 However, they should be mindful that such action can have serious consequences for those accused of misconduct. The COPE flowcharts11 (produced after the initial advice was provided) therefore state only that editors should “consider contacting” the institution. Whereas COPE aims to reduce publication misconduct and to reprimand offenders appropriately, it recognises that inappropriate behaviour can arise through naivity and lack of training (especially when it involves inexperienced researchers). An action that might warrant disciplinary action for a senior author, or a repeat offender, might therefore warrant only a gentle rebuke for a junior researcher.
In this case, publishing the notice of duplicate publication had unusually unpleasant consequences for three of the authors of the Portsmouth paper. A local newspaper was notified anonymously and published a critical article including a photograph of one of the authors. The authors’ employer (the Portsmouth Hospital Trust) received e-mails purporting to come from a senior American physician on behalf of “an international organisation concerned in limiting research fraud” emphasising the seriousness of the authors’ misconduct and threatening to refer them to the General Medical Council (GMC). The Trust was also contacted by somebody who claimed to be from the GMC. These communications were found to be fraudulent (ie, they did not, in fact, come from the named physician nor from the GMC) and were therefore ignored by the Trust, but the affected authors feel they were fortunate to have received support from an understanding employer and that under different circumstances they might have been suspended or undergone disciplinary procedures. Although these consequences are most unusual (especially the fraudulent e-mails apparently sent by somebody who wished to discredit the authors, who may also have notified the local newspaper about the case) they do serve to remind authors and editors of the possible unintended effects of such cases. It is clear, in this case, that the degree of overlap (or redundancy) between the papers falls into a grey area, which some consider acceptable while others do not. No clear and objective definition of redundant publication exists.
Software is now available that can measure the amount of similarity between two papers; however, it assesses similar strings of text rather than common data. This may become a useful tool for alerting editors to overlapping (or plagiarised) papers but it will still depend on the editor’s judgement to determine what degree of similarity is acceptable. We hope that, by explaining this case in detail, we have provided a clear record for readers to understand the precise relationship between the chlamydia papers.1–3 We have also highlighted some events that, with the benefit of hindsight, might have been better handled. We hope that these may provide learning points for journal editors, authors and COPE. We also hope that, by explaining the consequences of the events, we have provided food for thought for potential authors and editors of other journals who might examine their practices and processes to reduce the likelihood of similar problems occurring in the future.
Competing interests: None.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.