Trish Groves: How research data sharing can save lives


Everyone’s been missing a trick.

The whole debate on sharing clinical study data has focused on transparency, reproducibility, and completing the evidence base for treatments. Yet public health emergencies such as the Ebola and MERS outbreaks provide a vitally important reason for sharing study data, usually before publication or even before submission to a journal, and ideally in a public repository. Not just from randomised controlled trials, but from case series and samples, lab testing studies, surveillance studies, viral sequencing, genomic work, and other epidemiological observational studies too.

During the Ebola crisis, researchers couldn’t or wouldn’t share data. Last week WHO held a consultation meeting in Geneva to tackle this. One big reason for withholding data was the mostly unfounded fear of having subsequent papers rejected by journals. But researchers capturing vital information in the field and in coordinating centres were too busy to write and submit those papers, and thus much time was lost before vital information could be disseminated. Did people die because of the Ingelfinger rule against prior publication?

There were also, of course, some commercial disincentives to early data sharing, with numerous competitors scrambling to test potential magic bullets. All on top of the usual logistical challenges, and—probably—fear that data sharing would allow others to do quick secondary analyses and beat them to publication.

Researchers have to share early data with WHO when absolutely required in a PHEIC (Public Health Emergency of International Importance, of which there have been only three so far—related to H1N1, polio, Ebola).

WHO then shares this information with member states. But authors’ fears about the Ingelfinger rule prevent them and WHO from making these data public. WHO and all at last week’s consultation meeting agree that this isn’t nearly good enough. And, anyway, there are many other kinds of public health emergency, eg the Haiti earthquake, in which it would be powerful to share study data quickly, widely, and sometimes publicly. This isn’t only about infectious disease outbreaks.

The WHO meeting was excellent and, rather amazingly, got a short summary posted within a few hours of winding up.

The four editors there, including me, drafted a statement whose final version will be out in the final WHO report in a few weeks. It will then have full attribution and, we hope, endorsement from a wide range of major medical and science journals. Our draft is largely captured in the WHO summary report, albeit without naming names or journals:

“It was unequivocally agreed by representatives from leading biomedical journals that public disclosure of important information of potential relevance to public health emergencies should not be delayed by publication timelines, and that pre-publication disclosure must not and will not prejudice journal publication. It was agreed that pre-publication information sharing should become the global norm in the context of public health emergencies. Researchers should take the responsibility to ensure that results—even when preliminary—are adequately robust and have undergone quality control, prior to public disclosure to enable an evidence-based dialogue with the media and communities.”

This may all seem a no-brainer, but the Ebola crisis has made it clear that there’s much unnecessary confusion and reticence out there. Journals need to give unequivocal support to early—often prepublication/presubmission—data sharing, and not just for clinical trials.

Trish Groves is head of research, The BMJ. Follow Trish on Twitter @trished

Competing interests: WHO booked and paid for my economy class return flights from London to Geneva. WHO also sent me a subsistence payment which I used to pay for one night at the airport hotel where the meeting took place, meals en route, and a taxi to Heathrow. The meeting came with two buffet lunches at the hotel and one evening reception with free local wine and a finger buffet.