Managing suspected research misconduct

Abstract

Authors, editors, and systematic reviewers should protect the public from unsound research data

In this week's BMJ, Ian Roberts, Richard Smith, and Stephen Evans describe the worrying story of Dr Julio Cruz.¹ Cruz, a previously highly regarded medical researcher and clinician, committed suicide two years ago. Three of his publications about the use of high dose mannitol in head injury have recently been called into question. Furthermore, his coauthors and the editors of the journals in which the three papers were first published have failed to respond adequately to concerns raised about the integrity of the data in these papers.

These events have several important consequences. Many doctors base key treatment decisions on the results of published randomised trials. If some or all of Cruz's data on high dose mannitol are false, then doctors will be providing their critically ill patients with uncertain and possibly harmful treatment. In doing so, those doctors will also deny their patients other treatments that are based on reliable evidence.

The failure to retract unsound data also has long term consequences as the data become integrated into reviews, meta-analyses, and guidelines. These syntheses of primary research affect the practice of clinicians worldwide, and in turn affect even larger numbers of patients. In addition, unless and until the veracity of Cruz's data is formally proved or disproved, there is a risk that further research will not be conducted in this area. This again denies practising clinicians the possibility of access to the sound data they need.

Concerns about the authenticity of biomedical research data are increasingly being publicised. Recent high profile examples include publications by Jon Sudbo² and Hwang Woo-suk,³ both of whom have had their results discredited or retracted. Importantly, however undesirable the publication of unsound data is, the consequences of such publication are made far worse by the subsequent failure of the people involved to react appropriately to valid concerns and correct the scientific record where necessary.

Part of the difficulty in dealing with Cruz's data relates to his tragic death before the start of any formal investigation. As with at least one other prominent case of alleged misconduct,⁴ he belonged to no institution that could be charged with undertaking the necessary investigation. In the wake of Cruz's death, any reasonable person would assume that the responsibility for the disputed publications rested with the other investigators whose names appeared alongside his on the original papers. These individuals are coauthors of the published papers as defined by the International Committee of Medical Journal Editors,⁵ or at least contributors,⁶ and have themselves denied they were “gift authors.” Where they are unable to verify the findings with which their names are associated, they have a clear obligation, in our view, to take all necessary steps to correct the record. To date they have failed in this duty.

Several groups already provide guidelines on how editors should react if research misconduct, including publication of false data, is suspected. These groups include medical journals,⁷,⁸ the Committee on Publication Ethics,⁹ and the World Association of Medical Editors.¹⁰ In broad terms, editors are advised to discuss the situation with the authors involved. If this discussion does not produce a satisfactory result the situation should be referred to an appropriate higher authority—perhaps the authors' academic institution or funding body. If, after an appropriate investigation, it is shown that false data have been published, the data should be retracted. Of note, the Committee on Publication Ethics makes particular mention that editors should react promptly to alert readers in situations where inaccuracies or misleading statements may have been published.

It is clear from communications between Roberts and John Jane, editor of the Journal of Neurosurgery (which published one of the disputed Cruz papers), that Jane doubted the veracity of the Cruz data. However, an accompanying editorial in that journal by Marshall only alluded to general problems with single centre research studies. Jane did not inform readers that he did not trust the Cruz data and indeed suspected that it was fabricated.¹,¹¹

Despite making considerable efforts, both Roberts and the BMJ Clinical Evidence editorial team have had great difficulty in contacting Michael Apuzzo, editor of Neurosurgery, in connection with the Cruz data. Neurosurgery published the two other disputed Cruz papers, and Apuzzo seems reluctant to address the serious criticisms of his journal's content. In failing to alert readers promptly to concerns about the Cruz data published in their journals, Jane and Apuzzo have created confusion. The position ought to be made clear in the interests of patients around the world.

It has been left to Roberts, in his capacity as coordinating editor of the Cochrane injuries group, and his colleagues to investigate and attempt to resolve the tangle of claims and counter claims that surround Cruz's data, and then to bring them out into the open. Their investigation of the three disputed papers should act as a model to which future systematic reviewers could usefully aspire. Systematic reviews, with or without meta-analyses, are appropriately replacing single clinical trials as agents that change and shape clinical practice. Consequently, researchers, such as Roberts, and organisations, such as the Cochrane Collaboration, that produce reviews have a growing responsibility to ensure that the data they summarise are valid. If done consistently, this assessment would become another important check on dubious information remaining as part of the evidential record on which global health care is based. However, efforts such as Roberts' do not affect the primary responsibility of journal editors to investigate thoroughly for themselves when serious suspicions of research misconduct are brought to their attention, and when doubts cannot be dispelled to make appropriate amendment to the scientific record.⁹

Bodies do exist in some countries that either directly undertake investigations into research misconduct (for example, the Danish Committees on Scientific Dishonesty (http://fist.dk/site/english/councils-commissions-committees/the-danish-committees-on-scientific-dishonesty)) or indirectly support investigation by institutions (for example, the US Office of Research Integrity (http://ori.dhhs.gov) and the recently launched UK Panel for Research Integrity (http://www.ukrio.org)). However, international awareness about research misconduct, though growing, is still low. Roberts and colleagues propose the creation of an international group with responsibility for dealing with research misconduct.

In the case of Cruz's data, more immediate steps must be taken to protect patients with head injuries. The editors and coauthors involved in the disputed Cruz papers must act quickly either to confirm the veracity of what they have published or to withdraw it. In the interim, clinicians treating patients with head injuries should approach these data with caution. Finally, in the absence of any patently sound evidence for its effectiveness, the research community should be encouraged to subject high dose mannitol to further research.