Why data sharing should be the expected normBMJ 2015; 350 doi: https://doi.org/10.1136/bmj.h599 (Published 05 February 2015) Cite this as: BMJ 2015;350:h599
- Harlan M Krumholz, professor of medicine
- 1Section of Cardiovascular Medicine and Robert Wood Johnson Foundation Clinical Scholars Program, Department of Internal Medicine, Yale University School of Medicine, New Haven, CT 06510, USA
- 2 Department of Health Policy and Management, Yale School of Public Health, New Haven
- 3Center for Outcomes Research and Evaluation, Yale-New Haven Hospital, New Haven
The Institute of Medicine (IOM), a venerable American institution that seeks to provide authoritative recommendations to decision makers and the public, released a report last month on Sharing Clinical Trial Data.1 The report is a welcome codification of guiding principles and frameworks. It reinforces many arguments for data sharing and urges that stakeholders “should foster a culture in which data sharing is the expected norm.”1 The IOM joins many other organizations, including drug companies,2 3 the European Medicines Agency,4 the National Institutes of Health,5 and the Bill and Melinda Gates Foundation,6 in making clear that study reporting and data sharing in medical research are imperative and the questions ahead are how, not whether.
The general support for data sharing in the IOM document will be familiar to The BMJ’s readers, as this journal has long advocated the importance of transparency in scientific research and the need to report results and share data.7 The report provides what should be easily achievable recommendations for the research community. The committee called for data sharing only for new trials, though it urged consideration of sharing older studies. It stated that 18 months could elapse after the completion of a trial before investigators were asked to share. The committee also determined that sharing raw data would be too burdensome and that such data should be made available only on a case by case basis. The report states that summary level results should be publicly reported within 12 months of study completion, including lay summaries to trial participants.
Further to go
Those of us who are eager for a change in the scientific culture remain impatient for even more progressive action plans. The harms of not reporting results and not sharing data are well established.8 9 But people who continue to caution about the harms of data sharing have yet to provide persuasive examples where data sharing has harmed the public.
Among other problems, the absence of data sharing creates the important problem of information asymmetry among researchers. Those with the data can assert authority over interpretation and diminish the ability of others to debate key points, muffling scientific discourse. Confirmation bias can lead a group of investigators to see their data through a particular lens, while others could reasonably interpret the data differently and provide another perspective. Even independent expert reviewers come to different conclusions when provided with the same individual patient level data.10
One of the peculiarities of clinical research is the challenge of replicating a study. A tenet of the scientific method is reproducibility. And yet, for clinical trials, reproducibility is often precluded by cost or logistic, ethical, or business reasons. As a result, direct replication by repeating the experiment in a different laboratory is not an option. This constraint makes it even more important that independent scientists can study the raw data.
Challenges surrounding replication are not unique to medicine. It is not feasible to independently repeat experiments conducted with the Hubble telescope or the Hadron collider. Instead, astronomy and physics researchers conduct an experiment and then share the results widely for scientists to study and produce more knowledge and insight. Such collaboration strengthens the trust in research, expands the volume of research that can derive from single experiments, and includes investigators who otherwise would be unable to fund such experiments. Unfortunately, what occurs every day in medical research is akin to a few astronomers with access to the most powerful telescope interpreting for us what they saw without allowing us to look for ourselves.
Signs of progress
Fortunately, in medicine, the remedies to the current culture are coming. Industry is already making voluntary major steps to share data. Concurrent with the release of the IOM report, Johnson and Johnson, with the Yale Open Data Access (YODA) project, agreed to release its medical device trial data going forward, the first commitment to a broad release of that type of data.11 Johnson and Johnson, along with other major companies, is already releasing data from drug trials.2 3 There is some variation among companies and many data remain out of public view, but there is notable progress. The United States National Institutes of Health responded to concerns about non-publication and is receiving public comment on a rule requiring reporting of its studies.5 The agency’s next step should be to promote sharing of raw data. The Gates Foundation, in the most audacious show of leadership, stated that it will soon require data sharing at the time of publication.6
Yet it will take more to transform the current culture and we need to move quickly.8 9 All trials should be reported promptly after study completion and, if possible, published in peer reviewed journals. And any data generated in the course of medical research that might help fight disease and alleviate suffering should be shared as soon as possible. To delay knowledge generation by restricting access to such data should be considered counter to medical research’s code of responsible and civil conduct. Solving the health problems that patients face must prevail over the self serving desire to treat data as hoarded treasure.
Yes, there are challenges. We need financial support for data sharing in grants. Academics must find ways to credit those who produced the data. Companies need time to get their products approved. And there are others. But time wasted in reporting and sharing can cost lives, and we need to push forward. We should heed the implicit message that the Gates Foundation is sending: solving the world’s health problems is more important than protecting the opportunity for researchers to have exclusive access to data. It’s time for us to create a collaborative scientific culture in which investments in trials are leveraged through sharing with the broader community and to move past the reasons for why doing so is hard.
Cite this as: BMJ 2015;350:h599
Competing interests: I have read and understood BMJ policy on declaration of interests and declare I receive research grants from Medtronic and from Johnson and Johnson, through Yale University, to develop methods of clinical trial data sharing and with Medtronic and the Food and Drug Administration to develop methods of device surveillance. I chair a cardiac scientific advisory board for UnitedHealth and am supported by grant U01 HL105270-05 from the National Heart, Lung, and Blood Institute.
Provenance and peer review: Commissioned; not externally peer reviewed.