Sabine Kleinert, executive editor The Lancet
A summary of cases presented to COPE over 6.5 years up to September 2004 won't
necessarily provide solutions, but will help to raise issues and pose the
questions that need to be asked in these areas.
Year | No of cases | "Evidence of misconduct" | "Probably no misconduct" | Not applicable |
1997 | 16 | 11 | 0 | 5 |
1998 | 33 | 30 | 2 | 1 |
1999 | 27 | 20 | 3 | 4 |
2000 | 32 | 26 | 6 | 0 |
2001 | 39 | 30 | 9 | 0 |
2002 | 18 | 14 | 4 | 0 |
2003 | 22 | 15 | 5 | 2 |
2004 | 25* | 17 | 7 | 1 |
Total | 212 | 163 | 36 | 13 |
In most of the 212 cases (n=163), there is
evidence of misconduct; in 36 cases, there is probably none.
Common causes of research misconduct |
|
Duplicate/redundant/salami publication | 58 |
Authorship issues before or after publication | 26 |
No ethics approval | 25 |
No or inadequate informed consent | 22 |
Falsification or fabrication | 19 |
Plagiarism | 17 |
Unethical research or clinical malpractice | 15 |
Conflict of interest | 8 |
Reviewer misconduct | 6 |
Editor misconduct | 3 |
Miscellaneous | 39 |
Two thirds of the cases relate to submitted papers rather than published papers.
This is encouraging, because it means that taking action can avoid unethical
publication.
Redundant publication
Redundant publication means either salami publishing or significant overlap of
data. But the crucial issue is transparency. If the authors explain in their
covering letter that the primary outcome has already been published elsewhere,
and that this is the secondary outcome, then it's up to the editor to decide if
s/he wants to publish.
But in many cases editors are not told, and find out only when the reviewer
picks it up, or worse, when someone writes in after the article has been
published.
Is overlap worse than salami publishing? How much overlap is too
much? Those are questions for reviewers to tackle.
Is it ever legitimate—for example, if it's targeted at a different
readership? If two different outcomes are being reported, the clinician
has to know the whole picture. It then becomes a matter of transparency: readers
can then judge for themselves if it's justified to have two papers
on different outcomes or a follow up paper on a main results paper.
Is it acceptable to report mortality data in one paper and disability data in
another? That's a matter of clinical judgement. But there are studies where hard
outcome data are important for a general readership, but some sort of sub-study on markers would be more relevant to a specialised
readership.
Does it matter? It does if there is a great deal of overlap, because
systematic reviews will suddenly count double the number of patients.
An interesting viewpoint was published recently in the BMJ,
suggesting that duplicate submissions encouraged competition among editors to
reduce time to publication (BMJ
2005; 330: 305-7).
Duplicate publication/submission is often revealed by reviewers, who happen
to have seen it elsewhere or reviewed it for another journal.
Can it be incidental? It's very rare. When it is uncovered by the
reviewer or the editor, and only then does the author respond, that is
suspicious in and of itself. It should be the author openly declaring it if
there has been a genuine mistake.
What about duplicate publication in other languages? It happens, and is
acceptable if the article is cross referenced and the original journal grants
copyright.
Is it enough to withdraw the paper with the authors' agreement, when the
editor finds out? No. Arguably, no harm has been done, but clearly it's a case
of authorial malpractice. The authors might have done it before, so the
institution should be informed. If it's published, both journals will have
to publish a notice of duplicate publication.
Authorship issues
These are very difficult to deal with. If an author has already submitted it
elsewhere, it should not be submitted to another journal until a definitive
accept or reject decision has been made.
What happens when an article is published in a main journal but then published
again in a supplement? An example is an article published in the American
Journal of Cardiology, which was then published again in a supplement, paid
for by the sponsoring company. Only the title was changed and readers were not
alerted to its prior publication.
That's clearly wrong. But transparency is key: if that second
publication had acknowledged the first, it would have been acceptable.
There is some evidence to suggest that authors repeat this offence.
Sometimes the whistle is blown on duplicate publication in a letter to the
editor, and sometimes by someone with a vested interest. It could be malicious
and may be untrue. The editor would always have to investigate to find out.
When and how should editors get involved in author issues before
publication? Authors may declare what their role has been, but ultimately it's the authors who have to agree who is an author and who isn't.
And it's their responsibility. If they can't agree, they have to
seek the help of their institution.
If an author has been deliberately or inadvertently missed off the list, the
record must be corrected. This is an obligation in health publications.
The issue of ghost authors is a thorny one. "Vanishing authors" are often
medical writers or the drug company, and they don't appear on the
paper. Gift authorship is also a well known issue.
There are huge cultural issues, and in some countries it's accepted
practice that the head of a department has to be on every paper, regardless of
whether they merit it.
There may be disagreement about data interpretation, and we have had to publish
two different discussions on a paper, because the authors could not agree as to
how these data should be interpreted.
Unethical research
There is a grey area. Is a new surgical technique research or slightly changed
practice? What is audit? What is normal practice, and is ethics
approval needed for it? Authors claim not, but what is presented sometimes
does not appear to be normal practice.
Does no ethical approval automatically mean that the study is unethical?
No it does not, in the same way that ethics approval does not automatically mean
that a study is ethical. Editors have a duty to judge papers for themselves.
Some journals do request evidence of ethics approval.
Informed consent
Is the consent truly informed? The authors often say that written informed
consent has been obtained, but have the patients really been told about the
risks? If that question arises, a very low threshold of suspicion is
warranted, and a copy or a translated version of the consent form should be
requested.
This can be done even after acceptance, if there is any doubt at all. We have
done this, after seeing a consent form, which was not what we would truly regard
as a consent form.
Clinical malpractice
When malpractice is suspected, who should editors contact?
- Always challenge the authors first.
- Notify the institution if the response is unsatisfactory.
- When authors are in private practice or head up their institutions, approach the
licensing/regulatory body of the country concerned. - It's perfectly legitimate for two editors to talk about a case and act
together.
Editors do have a duty to pursue suspected misconduct, as outlined in the
editors' code.
What about defamation? Never write to an author's institution making
a direct allegation of malpractice. Rather say that this particular problem has
arisen, and that you have had this response from the authors. You are not
satisfied with it, for the following reasons, and would that person look into
it? It's then the duty of the institution to look into it. Whether
they do it is another matter.
Fabrication/falsification
Fabrication/falsification is another very difficult area for editors. When
does it start? It is a deleted outlier? It's a gradational process, and often
editors only have a vague suspicion, because the data look too good to be true,
for example. Or there's only a single author on a
randomised controlled trial. These should prompt warning bells, but there is
still no solid proof.
Falsification is seldom picked up by reviewers, although sometimes the
statistician will detect it. Figures are one area where reviewers are better at
uncovering it. Editors have the right to ask for raw data, but analysing this
can be very time consuming and difficult. And what happens when the data have
disappeared or the authors were only asked to keep their raw data for a specific
period of time?
Plagiarism
Extent is important. Is it one sentence or whole paragraphs? Has this been
done before? Take care over co-authors, because one or two might
have done this, but the other four named authors might have no idea that this
has happened. Write separately to all of them.
Can it be unintentional? Sometimes there's a language problem, or phrases from
another paper get used. It's easy to do, and this needs to be
borne in mind.
A reviewer for a specialty journal discovered that the review paper had been
plagiarised from his own published syllabus. The authors were very apologetic,
and could not explain it. They made a very good case, and the paper was rejected
without further action. However, the editor was contacted six months later by
the same reviewer saying that the same paper had been submitted to another
journal and nothing had been changed.
Common difficulties for editors
Pursuing research misconduct is time consuming to do this, and often replies
from the authors and the institutions are simply not forthcoming. Or the
institution agrees to investigate, but it's barely adequate. Or there may
be no institution.
Then editors can only publish notices of concern or letters, or express generic
concerns in editorials.