Should data from clinical trials be available to all? Or just for certain investigations? Who decides? And how would that work? These, and a host of other questions, were considered at a recent Stanford Health Policy Forum moderated by Keith Humphreys, PhD.
The issue of data sharing in clinical trials is timely: Beginning in July 2018, most medical journals will require papers based on clinical trials to include a data sharing statement that clarifies how and whether data will be shared, although data sharing is not yet mandated.
This requirement goes too far, said speaker PJ Devereaux, MD, PhD, a clinical trialist at McMaster University in Ontario, shown in the center of the photo above.
“I want to be clear I am not opposed to data sharing,” Devereaux said. “I have reservations about what has been proposed.”
He said he worries that data sharing will take time away from the many other demands facing clinical trialists, ultimately leading them to conduct fewer or slower trials. In addition, the current academic system that rewards researchers based on publications doesn’t incentivize sharing.
Data sharing is beneficial, but in the case of clinical trials, there needs to be a process and a set of guidelines in place that ensure everyone is comfortable and to prevent unintended consequences, said speaker Nigam Shah, MBBS, PhD, a Stanford biomedical data specialist.
“The typical conversation about data sharing ignores one basic tenet of sharing that our moms would teach us, which is trust,” Shah said. “The idea of putting stuff out there for anyone to take is a bit utopian.”
Devereaux said he worries about data dredging, also known as p-hacking, which is the practice of scouring data for relationships that may not correspond with any particular hypothesis. Or, the data may be used by someone — or a company — who disagrees with the conclusions of the original paper as an effort to undermine it.
Yet if the data is out there for everyone, Shah said, fact-checkers will be able to identify who is correct and the wisdom of the crowd should triumph.
One oft-repeated benefit of data sharing is that it could make it possible to make new discoveries using existing data, saving resources and advancing science. And, additional analyses could pick up on potential mistakes.
Devereaux said he isn’t convinced. “A lot of people think the cure for cancer exists in the secondary analysis. It doesn’t.” And mistakes should be caught before publication, he said.
What effect could data sharing have on trial enrollment, Humphreys asked. Many potential participants already have concerns about privacy and those concerns are multiplied in trials related to stigmatized issues such as addiction or HIV.
For example, Devereaux said, if patient data is compromised after he has shared it with someone else, he will feel responsible. “I owe it to [the participants] to make sure they are protected.” Any new system that mandates data sharing should include measures to keep data secure, he said.
As debate about the issue advances, Shah said he hopes the conversation will shift from “why” to “how” to best share data.
Previously: Dean Lloyd Minor hosts discussion on health care data sharing and leadership, Big Data in Biomedicine panel examines data sharing difficulties and The future of clinical trials considered during Dean’s Lecture Series address
Photo by Becky Bach