New science data-sharing rules are two scoops of disappointment

After more than a year of deliberation, editors of some of the world’s leading biomedical journals have come up with a declaration on data sharing destined to usher in a glorious world of transparency and rigor in science.

No, sorry, scratch that. What they’ve come up with is a policy with a couple of weak sticks and no carrots.

In an editorial this week, the International Committee of Medical Journal Editors (ICMJE) — which counts among its members the editors of such hard-hitting publications as the New England Journal of Medicine, the Journal of the American Medical Association, and the BMJ — said that, starting in July 2018, members must require a data-sharing statement in all submitted manuscripts. But the committee stops short of requiring such sharing, saying only that editors “may” take into account authors’ plans for data sharing when they’re deciding whether to publish a paper.

The upshot: A paper published in one of these journals may very well look no different a year from now than it does today, apart from a sentence that says something like, “The authors will share their data upon request, subject to various conditions” — and that often means they won’t share it at all. Which is unfortunate from the organization perhaps best placed to change standards in science publishing: The ICMJE has 11 official members, but many hundreds more follow the group’s recommendations.

We’re a bit underwhelmed. And we’re apparently not the only ones. “I am disappointed that they did not seize the opportunity to create an imperative to share,” Dr. Harlan Krumholz, a cardiology professor at Yale who helps run a data-sharing platform called YODA, told STAT. “They could have provided a runway — and an opportunity to appeal in exceptional circumstances. Still, they are committed to the goal, which is good.”

The ICMJE editorialists couch their cautious approach to data sharing as an “ethical obligation” to the people who agree to participate in clinical trials — and who, by doing so, “put themselves at risk.”

That’s true, and important, but it’s not the whole story. A major impetus for data sharing is the belief (well-founded, we believe) that giving outside researchers access to results is critical to improving the rigor and reproducibility of science.

But the document inexplicably is silent on reproducibility. The word does not appear at all.

Failure to making data sharing mandatory for acceptance of papers raises the prospect that some authors might decide to ignore the hint. In fact, some scientists might reasonably interpret the wishy-washy policy as saying: The sexier your findings, the less important sharing data is — which not only won’t promote sharing but will exacerbate an already significant problem with science publishing.

The plan also will require, as a condition of consideration for future publication, that any clinical trial that starts enrolling subjects on or after Jan. 1, 2019, includes a data-sharing plan when registering the study.

“We envision a global research community in which sharing deidentified data becomes the norm,” the 16 authors wrote. “Working toward this vision will help maximize the knowledge gained from the efforts and sacrifices of clinical trial participants.”

But the journal editors really don’t do much to promote that vision or to challenge the status quo. And that’s a shame, especially because editors know full well how well their requirements can guilt — sorry, encourage — reluctant researchers to do the right thing. When they made registration of clinical trial protocols a requirement of publication in 2005, the number of registered trials nearly doubled within just one month.

Still, Brian Nosek, a psychologist at the University of Virginia and head of the Center for Open Science, which promotes data sharing, told STAT that the new policy was an “important first step.”

“Incremental, iterative change provides journals and editors with opportunity to gain experience with data sharing in practice, and identify pitfalls and challenges to address before moving to more stringent requirements,” Nosek said.

However, he added, the lack of a mandate for sharing likely will prove an attractive loophole for researchers. “I suspect that many authors will simply say ‘data available upon request’ and not actually prepare their data for sharing or post to an independent repository,” Nosek told us. “There is good evidence that this will be ineffective for data sharing. Promises to share data at some point in the future are often broken promises because people lose the data or have high burden to effectively curate the data for sharing after the fact.”

One way around this problem would be for the journals to start awarding badges to authors who share their data — an approach that appears to be effective, he noted.

Here’s another idea: Since patients are the ones taking the bodily risks here, maybe they should start boycotting all clinical trials whose investigators refuse to share their data. After all, if data sharing truly is an ethical obligation, bringing it about demands calls to action, not articles of faith.