Skip to Main Content

Sharing data after a clinical trial has been completed seems like it should be a slam dunk, a win for many stakeholders, including the general public. Instead, such data sharing is still something of a hot-button issue, with critics questioning the capabilities and motives of those requesting the data, doubting the utility of replication analyses, and speculating that spurious safety findings would receive unwarranted attention and disrupt patient care.

This week, the National Academies of Science, Engineering, and Medicine are convening the workshop “Sharing Clinical Trial Data: Challenges and a Way Forward” just shy of five years after the Institute of Medicine released its seminal report, “Sharing Clinical Trial Data: Maximizing Benefits, Minimizing Risk.”


During this time, the scientific culture regarding data sharing has shifted. Just last week, the National Institutes of Health requested public comments on its draft “Policy for Data Management and Sharing.” In 2018, the International Committee of Medical Journal Editors began requiring data-sharing plans for clinical trials as a condition for publication in member journals. And platforms such as, Project Data Sphere, and BioLINCC have emerged or grown. These platforms use a variety of different governance structures and models for data access, developed both with and without the support of industry or government.

The Yale Open Data Access (YODA) Project, which two of us (J.S.R. and H.M.K.) co-direct, launched in 2011 and formed a partnership with Johnson & Johnson in 2014. This five-year partnership offers an opportunity to reflect on some of the questions about sharing clinical trial data that may inform ongoing and future efforts.

Early experience of the YODA Project

The partnership between the YODA Project and Johnson & Johnson, whose trials make up 95% of those available via the platform, facilitates access to the company’s clinical trial data for pharmaceutical, medical device, and consumer products across a range of therapeutic areas.


The YODA Project uses a controlled access model. It requires investigators seeking to use the data to register and submit proposals that sufficiently detail their planned methods. Once reviewed by the YODA Project team, the proposals are then publicly posted. Once approved, investigators gain virtual access to a secure data platform to conduct their analyses.

Any investigator is eligible to request data, including those employed by industry. Because the controlled-access model makes information as publicly accessible as possible, including clinical trial metadata, prespecified analysis plans, and the results of all data-sharing requests, the process promotes transparency as well as the responsible conduct and reporting of research.

By using a controlled-access model and a board to appraise data requests, the YODA Project prevents data redistribution, protects against patient re-identification, and ensures good stewardship of clinical research data.

While the majority of trials made available through the YODA Project were completed at least 18 months ago, a special pathway was established with Johnson & Johnson to request Phase 1 trials and those that have been recently completed, should the research question be of timely public health importance.

By working with Johnson & Johnson to reactively de-identify and prepare trials for external use upon request, the YODA Project works more efficiently, such that more than 75% of trials listed as being available have been used by external investigators, far higher than the 15% reported for other data sharing platforms.

There are, of course, challenges to sharing clinical trial data, including how to make older trial data available in a contemporary technology format for use, the need for systematic adoption of data format standards across sponsors to enable meta-analyses, and the development of a sustainable model that covers the cost of data sharing. The YODA Project is currently directly supported by Johnson & Johnson and other companies using the platform to share clinical trial data. Other platforms, such as Vivli, have experimented with charging users modest fees in addition to charging sponsors.

But the largest challenges to achieving the promise of data sharing are broadening awareness of data availability and fostering expertise in using data from clinical trials. The skills needed for analyzing clinical trial data, particularly for meta-analysis, are considerable. The primary reason that requests for data that have been approved by the YODA Project do not result in a completed project is external investigators’ realization that their expertise was insufficient to perform the proposed analyses.

What’s been learned from the YODA Project?

As awareness of the availability of shareable clinical trial data has grown, use of these data and findings from them have also grown. As of this month, 350 clinical trials are shared through the YODA Project, with additional trials commonly becoming available. We have received and approved 117 requests for projects that used, on average, three trials per request. Among the completed projects, 31 manuscripts have been submitted for publication to peer-reviewed journals, 26 of which have been published.

At the same time, after the upfront investment in retroactively preparing data to be shared and in establishing a data sharing platform, including a server for secure access and analysis, costs have declined. All Johnson & Johnson clinical trials are now initiated with the knowledge that data will be shared, and all materials are prepared prospectively, making it possible for the YODA Project to leverage efficiencies as multiple investigators make use of the same shared data.

Several of the projects using Johnson & Johnson data completed through the YODA Project illustrate the benefits of data sharing. Data have been used for a systematic review and meta-analysis of therapies used to treat type 2 diabetes mellitus; for a World Health Organization report on the use of bedaquiline for the treatment of multidrug-resistant tuberculosis; for development and validation of machine-learning models to predict remission of Crohn’s disease; and methodological research to better understand placebo response among patients with schizophrenia and non-treatment related population risk among patients being treated with antipsychotics. As part of a broad collaboration with the NIH and academic investigators known as the Open Translational Science in Schizophrenia (OPTICS) Project, YODA-distributed data have been used to pursue a number of projects, including modeling dose-response relationships and treatment effect heterogeneity of therapies for schizophrenia.

Equally important, there have been no breaches of patient privacy, nor have there been publications of spurious safety findings that received unwarranted attention or disrupted patient care. To our knowledge, no data have been used for commercial or litigious purposes. And while there have been few replication analyses, those that were done have provided valuable insights and suggestions for improving data sharing efforts.

Instead, what we have observed has been the conduct and dissemination of numerous studies that might not otherwise have been feasible to pursue, generating knowledge and contributing to the clinical and scientific communities.

Our early experience suggests that data-sharing efforts through the YODA Project have been worthwhile, and we expect all data sharing platforms to have similarly positive experiences to share. Investigators have generated knowledge through the secondary analysis of already collected data, and built collaborations.

The culture of clinical research is changing, such that there are now expectations that researchers will share data, even when it isn’t required. And many major funders, including the NIH, are increasingly requiring that data be shared and providing resources to do so.

While the initial time, effort, and resources invested in these platforms was considerable, much has been learned and gains are being made, suggesting that sharing research data is one of the surest paths toward honoring patient participation in research and ensuring the data have the greatest impact on medicine and science.

Joseph S. Ross, M.D., is a professor of medicine and of public health at the Yale School of Medicine, a member of the Center for Outcomes Research and Evaluation at Yale-New Haven Health System, and co-director of the Yale Open Data Access (YODA) Project. Joanne Waldstreicher, M.D., is the chief medical officer at Johnson & Johnson. Harlan M. Krumholz, M.D., is professor of medicine and of public health at the Yale School of Medicine, director of the Center for Outcomes Research and Evaluation at Yale-New Haven Health System, and co-director of the YODA Project.

  • Clinical trials contribute to information and progress in treating and preventing diseases. initial and foremost participants will facilitate others by contributive to medical information and rising public health. Further, a participant doesn’t ought to be a patient diagnosed with a particular sickness or pathological state as some clinical trials, specializing in safety, can embody healthy volunteers.

    • Thank you for your comment. I would appreciate it if anyone with ideas on how we can encourage more reanalysis papers in our medical journals to submit their ideas to this blog. Ask your colleagues for ideas. It would be wonderful if medical journal editors could give us their thoughts. I know the problem, not the solution.

  • Reanalysis of data of peer reviewed published research is a common practice by economists. These papers are frequently published in their economic journals. As we know, even elite health policy and medical journals reviewers make errors. Consequently these journals publish papers that give misleading and even incorrect conclusions. Given the complexity of healthcare policy research, this can be especially true for policy studies. According to a colleague, who is a member of National Academy of Medicine, it is uncommon for medical journals to publish reanalysis papers. I am curious which peer reviewed journals published the 26 reanalysis studies that the authors mention.
    My questions to the authors and the audience: Given the large amount of work that is required to publish a reanalysis paper, especially policy studies, and the low acceptance of these papers in good medical journals, how do we incentivize researchers to reanalyze published studies? Most researchers want to advance their career by publishing in high impact journals. And they do need to pay their bills. How can medical and health policy journal editors follow the example of their economic colleagues? Incentives matter if we want to advance evidence-based healthcare policy and healthcare practice.

Comments are closed.