Annual Meeting Reports

Ethics Clinic: The What, Where, When, and How of Data Sharing

There are at least two compelling reasons to promote data sharing among scientists. The first is to speed the advance of science. Science typically advances in small increments rather than through single blockbuster studies. Someone is certain to replicate the study to make sure you’re right. Data sharing can accelerate the time it takes to progress from one incremental advance to the next. Enabling scientists to use their own data that are already collected to reproduce findings reduces the time and expense to gain confidence about a particular discovery. However, many scientists are reluctant to share their data. If data sharing is important to science, do peer-reviewed journals have a role in promoting it?

That and other questions about data sharing resonated with participants in this session, which was developed by the CSE Editorial Policy Committee. To prepare participants for lively discussion of two vexing cases related to data-sharing issues, the moderators provided a quick summary of the current climate in scientific research and publishing.

Existing data-sharing policies of scientific journals are designed to enable scientists to reproduce the experimental results of their colleagues. Specific policies differ among journals and depend in large part on the scientific discipline. Barriers to sharing are many and include academic competition, ethical issues, and commercial concerns.

For example, researchers who discover a molecule or compound that could lead to a viable product and financial reward for themselves or their research sponsor may be reluctant to share proprietary information. In research involving humans, ethical concerns about privacy and ownership of patient information create other roadblocks to the free passage of data. When scientific discoveries might be used for terrorist activities or war, public-safety and nationalsecurity concerns may make researchers reluctant to share their data.

Sometimes, researchers simply want to exhaust their own data before providing them to others or may fear academic failure should their results prove to be irreproducible.

Funding agencies and sponsors can set the tone for data sharing. Researchers may be obligated to share, depending on the rules set by those who pay for a study. The National Science Foundation (NSF) expects investigators to share data, samples, and supporting materials with other researchers and encourages the sharing of software and other inventions that facilitate future work. The National Institutes of Health (NIH) also support the sharing of final research data to advance the improvement of human health and expect researchers who they support to share data. Yet, both NSF and NIH recognize the legitimate interests of scientists and acknowledge the complexities generated by issues of confidentiality, costs of transmitting data, and researchers’ legitimate desires for priority in using the data they worked hard to assemble. In contrast, commercial sponsors may ask their researchers not to share data to protect proprietary interests.

A number of scientific journals have begun to hone their data-sharing policies. For example, Blood expects its authors to include in their publications the data, algorithms, and other information that are central to the study and make them accessible to others. Annals of Internal Medicine does not require authors to provide data but does require them to state in their articles whether and under what conditions the authors will allow access to the protocol, statistical code, and data.

Given the complexities of data sharing, the ethics clinic was designed to get participants thinking about the issue rather than to provide definitive answers. The clinic focused on two cases, one involving transgenic mice and the other involving two research groups, one of which wanted the other’s data.

The group achieved some consensus on the following issues:

  • Many journals are not equipped to be data archivists, but they should promote processes to make data available. When dealing with data-sharing issues, journals should try to ignore interpersonal politics and emotional issues among researchers and focus on the facts.
  • Journals have a responsibility to make authors, readers, funding organizations, and academic institutions aware of any journal requirements, policies, or expectations related to data sharing.
  • When a journal becomes aware an author is not cooperating with pre-specified requirements set by either the funding organization or the journal itself, it should make the author’s institution aware of the issue.
  • Journals should make their data-sharing policies clear to researchers so they will know what is expected before they submit their manuscripts.