We received a paper reporting a trial. There has only been one previous trial of this intervention in this condition that we know of (which was also done by these investigators). There were substantial issues with the reporting of that trial but the end result, as reported by them, favoured the intervention.
The trial we received, presumably approved after that result had come out, had the complication that most patients also received another treatment, and on an intention to treat (ITT) analysis of all patients, those given the intervention did no better and there was increased mortality in the intervention arm. The only positive outcome was from a per-protocol subgroup analysis of patients who did not have the other treatment (which they say is the only group comparable to the previous trial, and hence shows that the first trial was correct).
Although the trial was investigator led, it seemed to us that the authors were trying very hard to make something positive out of this actually rather worrying result. We sent the paper for review, including to a statistician; the reviewers raised a number of issues about the interpretation (eg, the overemphasis on subgroup analysis) and the analysis and reporting.
We felt that this was an important trial that needed to be reported, mainly because of the excess mortality in the intervention arm, but we had the rather odd situation that the authors wanted to emphasise the positive, and the need for further trials of this intervention, whereas the reviewers and editors saw the paper as delivering a negative message and feel actually that the paper will be the death knell for this treatment.
We therefore rejected the paper but offered to see a revised version if it was written more in line with our concerns.
The authors revised and the paper was re-reviewed. The paper was felt to be more balanced, but not yet completely satisfactory (ie, there was still too much emphasis on the positive result in one subgroup and not enough on the mortality).
A further issue then arose in that a reviewer spotted (on re-review) that three of the authors were noted as being on the DSMB for this trial. In their author contributions all are listed as having been involved in “analysing the data” and one, X, as “supervising the statistical analysis”
We asked the authors about these points and they replied:
"(1) We are fully aware that it is unusual for members of the DSMB to be listed as authors, as independence is obviously important for such boards. In our case, the DSMB’s independence was not affected for the following reasons:
(a) Members of the DSMB worked for the entire study period (ie, between 2003 and 2008) completely independent and without any promises or expectation that they would be credited later by a coauthorship.
(b) My personal decision to include three of the four DSMB members in the list of authors was made a significant time after the final database lock. This decision was long after completion of the clinical study and its analysis. It credited three members of the board who made some significant advisory contributions to the present manuscript. Only for this reason they were included as coauthors, and it was quite unexpected for them. This decision from December 2008 has in no way influenced their independence and objectivity at the time when the study was running.
(2) As to the contribution of X (one of the members of the DSMB), we have to admit a simple language problem. In our use of the word “supervision”, the word meant that he took a final comprehensive look at our data analysis before the paper was submitted for publication. Importantly, he never supervised (like an academic supervisor) data analysis at any time point before database lock and processing of the data by the clinical research organization. We will change the terminology accordingly.”
We subsequently found that X was also an author on the previous trial.
Finally, the authors did not declare initially any competing interest but after we enquired specifically, they declared that the corresponding author “holds a patent on the use of t[he intervention] for treatment of [the condition]”.
Our concerns overall therefore were that this paper not only reports the outcomes in a way that is not appropriate, but also the composition of the DSMB and the presence of some DSMB members as authors means that the trial may not have had adequate independent oversight.
We felt we had two possible options with regard to publication:
(1) We reject the paper because it was inappropriately conducted and not appropriately reported.
(2) We publish the trial after further revision to ensure it is reported appropriately and publish alongside it an editorial that lays out our concerns with the conduct of the trial, but notwithstanding those, our reasons why we think it should be published.
We also discussed whether we needed to raise the issue of the DSMB with the authors’ institution.
We discussed the paper with our internal ethics board and they unanimously agreed we should reject the paper (mainly because of the concerns over the DSMB) and inform the authors' institution. We have as yet heard nothing from the institution.
We are bringing this to COPE as this paper raised a number of serious issues we had not come across before. We would appreciate the Forum’s opinion on whether we handled this correctly.
Some of the members of the Forum suggested that perhaps the journal should have a formal policy that DSMBs should be independent and not involved in the study in any way. The Forum questioned whether these authors fulfil the criteria for authorship, as outlined in the ICJME guidelines. One opinion was that perhaps the paper should not have been rejected until the outcome of the investigation was known. However, most agreed that rejecting the paper was the correct decision and the editor might consider contacting the ethics committee who approved the study if no response is received from the authors’ institution.
The editor reported the case to the author’s institution but no response has been obtained. The editor is pursuing the case.