Data manipulation and institute’s internal review

Case number: 
17-02

Case text (Anonymised)

A journal received an enquiry from a reader stating that they had found some discrepancies in the spectra published in the electronic supporting information for a published paper. They suggested that the discrepancies would be consistent with the spectra being manually ‘cleaned’. If this were true, the characterisation and purity of the compounds reported in the paper would be called into question.

The editor checked the spectra in close detail and verified that the discrepancies that the reader had identified were a reasonable cause for concern. The editor also checked the author’s related papers in the journal and identified a total of four papers that were affected by similar discrepancies in the spectra. When the editor contacted the lead author to discuss the concerns, they explained that ‘cleaning’ spectra to remove impurity peaks was not a practice that was carried out by their research group, and they did not believe that it had occurred in this instance. However, the researcher who had carried out the analysis had now left the group and the original data files where no longer available.

As a comparison with the original data files could not be made, the journal approached an independent expert to obtain a second opinion on the evidence available in the published spectra. The expert confirmed that there was clear evidence that the spectra had been altered and that this could be consistent with an attempt to overestimate the yields for the reported reactions.

Following this, the journal contacted the director of the institute to request their assistance in determining whether the spectra had in fact been altered. The director consulted with the lead author and the head of their facility. They confirmed that it was not possible to locate the original data due to a limitation of their archival system. They stated that their internal review had not found any ‘intentional altering of the spectra’. They stated that on that basis, the papers should not be suspected and should be allowed to stand.

This recommendation runs contrary to the evidence that we believe can be seen in the spectra, but in the absence of the original data files it is difficult to make a conclusive judgement.

Question(s) for the COPE Forum
• What action should the editor now take to resolve this matter? The journal is considering two options:
—accept the research institute’s recommendation that without evidence to prove deliberate manipulation of the data no further action should be taken.
—publish an expression of concern notice on each of the affected articles stating that discrepancies in the spectra were identified, the institute was asked to investigate, but that the original data were not available and they found no evidence of deliberate manipulation of the spectra.
 

 

Advice: 

The Forum acknowledged it is difficult for the editor to decide on whether to accept the institution’s conclusions on the case or if in fact the journal should do more and work on their own investigation, involving the editorial board and/or their independent expert.

The Forum questioned the type of investigation the institution carried out. If it was a thorough research misconduct investigation, the journal should be able to rely on the results of that investigation as this usually involves multiple levels of investigation, an enquiry, with a faculty board reviewing all of the data that are then made available to the journal. However, if the journal received a relatively rapid response from the institution, then perhaps the internal review is not very reliable.

The Forum asked if the journal had a data availability policy—does the journal require the data from a study to be made available on request? The real issue is why the original data were not available. The lack of the original data is a serious concern. The minimum requirement of an institution is to curate and preserve the data, and it would be expected that any reputable institution would normally comply with data being available for a period of time after the end of the research (usually about 5 years). Hence this a failure of the institution. This alone could be grounds to retract the paper or publish an Expression of Concern.

If the editor is confident that there is a problem with the paper, and confident in the advice of their experts, then the journal should consider publishing an Expression of Concern, detailing the facts of the case, and pointing out the discrepancies between the findings of the institution and what the editor believes.

If the journal has a post-publication comments section, another suggestion was to encourage the reader to post their concerns, giving the authors a chance to respond as well as allowing more participation from readers. This would also allow for more transparency of the issue.

Resolution: 
On-going
Year: