Case text (Anonymised)
Author A has published approximately 150 original articles since ~1994, with ~100 on one particular topic. Since some of these events were up to 16 years ago, and there are no formal records from then relating to these studies, the only information we have is the memory of the editors of the affected journals in post at the time. According to their accounts, suspicions were aroused over the validity of the data, in particular the similarity between baseline data of some of the different studies. When author A was pressed to provide raw data, he stopped responding and stopped submitting papers to the specialty journals, switching to general journals where he continues to publish. The editor of one specialty journal raised concerns with the author’s institution (in another country) approximately 12 years ago; it responded saying it saw no reason to investigate further. A letter, published in one of the specialty journals in 2000 by an independent researcher, asked the question “Why are author A et al’s data so nice?”, pointing out that the probability of such results occurring by chance were infinitesimally small, but as far as we know there have been no formal investigations of author A’s work
Following an April 2010 editorial in one of the specialty journals about research fraud in general, that mentioned this particular author by name, a correspondent raised the lack of investigation into author A, stating that his update of a systematic review was being hampered by the (suspect) influence of author A’s work in this area. The current editor-in-chief of that journal contacted the current editors of seven other affected specialty journals, who until this point were largely unaware of the problem, or its extent, having not been in post at the time the papers were submitted to their journals. We have since been discussing the problem and possible courses of action. The points raised are:
(1) Regarding the older papers:
(i) the journals themselves do not have the ability to mount an investigation;
(ii) it is unlikely that an investigator, bona fide or not, will still have original data from the older studies;
(iii) it is unlikely that author A’s institution will be interested in investigating studies so old, and we think he might have moved universities since then;
(iv) currently we do not have any firm data of wrongdoing, just suspicions. Options for gathering more data include asking the original correspondent and the systematic reviewer to provide a more formal commentary, although we have not done that yet. Meanwhile, one of the editors has gathered data on all author A’s studies: there are 135 in which author A is the first author, reporting almost 12,000 randomised patients in 17 years. Most are with one of the same three co-authors. The largest group of papers (by topic) are all very similar in design, with very little variability in baseline placebo event rates, and generally similar results although the outcome measures differ and there are one or two ‘surprising’ (at best) findings. One particular drug features in 71 studies. Dropouts are hardly ever reported.
(2) Regarding the newer papers:
(i) these may be easier to investigate since the data should still exist;
(ii) we could contact the editors of the non-specialty journals (there are many, publishing just 1-2 of A’s articles each) to alert them but the problem of having only suspicions remains (compounded by the relatively large number of journals, each with a small number of papers);
(iii) we could ask a respected academic in author A’s country to make discreet enquiries of author A’s co-investigators, some of whom may not realise what is going on, or they may have concerns themselves. However, this could be a delicate situation for such a person.
We would welcome COPE’s advice on how best to proceed.
The Forum was unanimous in their opinion that this should really be resolved by the researcher’s institution. Although there is no hard evidence, it was suggested that all of the journals, as a team, approach the institution and ask the institution to conduct an investigation. It was felt that this would provide a more powerful case than a single editor on his own. Meanwhile, the editor should try and gather more evidence, perhaps by contacting the ethics committees who supposedly approved these studies. The editor may then be able to determine whether in fact the studies took place as reported. The Forum advised against informing the non-specialty journals at this point as there is no real evidence at this point, so it would be difficult for them to know what to do. The Forum also suggested that the editor should advise anybody doing a meta-analysis on this subject to include a sensitivity analysis to test the effects of including the data from these studies.
The group of editors-in-chief are planning on sending a letter to the author and the institutions. The delay has been in obtaining an independent analysis of the suspect works, which so far indicates that the likelihood of fabrication is very high.
Follow up (May 2011):
The editors are still planning on sending a letter to the author and the institutions but this has been delayed pending the analysis, which has now been submitted to journal A for publication! (It concludes that fabrication is almost certain.) Meanwhile, a separate publication scandal has distracted the editors’ attention recently.
We did not investigate ethics approval since we felt that should be for the institution to investigate. My main question is what to do with the analysis that has been submitted:
(i) process as usual and publish if accepted (after discussion with/approval by the publishers); that is, let the author/institution respond if they wish
(ii) present it to the suspect author (with the analyst’s permission) and remove from the publication process for the time being
(iii) present it to the suspect author’s institution(s) as per (ii)
(iv) present it to the suspect author and institution(s) as per (ii)
I have sought an independent statistical review of the analysis from a respected statistician but have not had a response.
There were various views on how to proceed in this difficult case. Some agreed with sending the analysis to the institution and the suspect author, at the same time, informing them that the intention is to publish the analysis in the journal, and then see what type of response this elicits. Others disagreed and recommended treating the analysis like any other submission (despite the fact that the review was commissioned by the journal and the editor hadn’t originally intended to publish it) — send it out for peer-review and publish it in the normal way. Most agreed that a copy of the analysis should be sent to the institution, either before or after peer-review. But the paper does not need to be peer-reviewed before it is sent to the institution.
The Forum also advised consulting the journal’s legal department before going ahead with publication.