Here’s a CLUE …

Journals have a responsibility not to mislead their readers. That seems pretty straightforward and uncontroversial, but achieving this lofty aim can be tricky.

In order to do that, journals need to know if a research report is trustworthy. Peer review is the first step but we know it’s not perfect and, in fact, although it may help to improve reasonable manuscripts and pick up minor glitches, it has a poor track record for detecting serious problems due to either fraud or honest error. One reason is that reviewers usually only see analysed data and summaries of statistics. Even if data are available, I’m not convinced that peer review will develop super powers because most reviewers don’t have the time or the statistical knowledge to review the data. So, journals will probably continue to face situations in which doubt is cast on something they have published.

The traditional wisdom, set out in guidelines such as those from COPE (the Committee on Publication Ethics—some of which I helped to write) is that journals should always refer concerns to the authors’ institution and request an investigation. This sometimes works well, as in the case of Dr Fujii from Japan, whose institutions responded promptly and helpfully to a request from the editors of a large number of anaesthesia journals. But sometimes, the response is much less constructive.

Richard Smith battled for 15 years to get a Canadian institution to examine problems with articles by RK Chandra.

Even when institutions do investigate, the process can take several years. A US university, first notified in 2008 about concerns in an article by one of their researchers, took 4 years to investigate all the 101 papers produced by his group.

Institutions that won’t investigate (or don’t investigate properly), and investigations (and appeals) that take many years are a headache for journals. Editors may decide to publish an Expression of Concern, but may worry that this implies guilt before the trial is over. Institutions may be unable to inform journals that an investigation is taking place because employment laws require them to maintain strict confidentiality.

Last summer (July 2016), I helped bring together a group of people to work on solutions to such problems. The group included a university dean, a vice-chancellor, research integrity officers, a lawyer, publishers, and journal editors. We’ve recently posted our recommendations and hope they’ll be discussed. We’re calling them CLUE: Cooperation & Liaison between Universities & Editors. One of our more radical suggestions is that institutions need to develop processes for assessing the reliability of reported research which is separate from the time-consuming, expensive (and, frankly cumbersome) systems for investigating research misconduct. We realise that one of the problems is that journals care solely about the trustworthiness of what they publish, while institutions are expected to determine if an individual researcher has committed misconduct. These are different questions. As we note “it is possible for research reports to be misleading or untrustworthy and therefore to require correction or retraction even when the authors/researchers are not considered to have committed research misconduct by their institution.” We’re not saying that institutions should stop trying to investigate misconduct, but we’d like to see a mechanism by which journals and institutions could work together to assess the soundness of a research report. We envisage this as a much swifter (and cheaper) process than a full investigation. Also, because the guilt or innocence of individuals is not at stake, it needn’t be confidential, and the findings can be freely shared. Then, readers can be promptly informed if a publication shouldn’t be relied upon.

Is this workable? We’d welcome comments on the preprint.

Liz Wager PhD is a freelance medical writer, editor, and trainer. She is co-editor-in-chief of Research Integrity & Peer Review. She was the chair of the Committee on Publication Ethics (COPE) (2009-2012).

Competing interests: The author has no further relevant interests to declare.