Sharing Clinical Research Data: Perspectives on an IOM Workshop

Abstract

Few people would argue that individuals in need of diagnostics and therapies would benefit from the sharing of clinical research data. Government research agencies, pharmaceutical companies, and academic investigators continually amass large amounts of such data. Sharing information both within and across institutions has great potential to advance science and public health. Unfortunately, little is shared in a useful manner. The Institute of Medicine of the National Academy of Sciences held a public workshop October 4–5, 2012 (http:// www.iom.edu/Activities/Research/SharingClinicalResearch Data.aspx), focused on the benefits of sharing clinical research data. Stakeholders from government, industry, academia, and advocacy participated in presentations and discussions about barriers to sharing data and concrete ways to overcome those barriers. The workshop also welcomed the unique perspective of patients who often find themselves frustrated and left out of the very studies that are designed for their benefit. The workshop began by articulating that ‘‘data sharing’’ means different things to different people in different situations. A variety of continua were described in a search for precise language that would enable clarity of potential solutions. The workshop’s main focus was on participant-level data sharing. However, various levels of aggregate data were also considered, as participants realized that a productive dialogue must take into account several continua. Clinical trial data are usually publically reported on a summary level, but this summary does not always accurately reflect the underlying participant-level data. A variety of issues can arise as the trial progresses, resulting in inaccuracies, or at least misrepresentations. For example, structural changes to studies— such as going from four groups to two—result in major discrepancies. Some participants in the workshop suggested that transparency in the transition of data from participantlevel data to aggregate data could aid in decision-making in both policy and therapeutic development. Another apparent continuum defines with whom to share data. Clinical research data could be shared among the original researchers only, or with approved researchers, or any researchers, or anyone at all. Although there is potentially more risk to individual’s privacy the more broadly participant-level data are shared, there is also potentially greater benefit if more eyes are on the data. Finally, another important continuum includes levels of data sharing. Data can be discovered, used, and accessed, with a variety of mechanisms allowing those levels of sharing. If data are discoverable, then this finding could be as simple as ‘‘they exist.’’ If data are used, then they might be part of an aggregate data set, but the researcher still may not be accessing the individual record. Accessing the participant-level data might include actually transferring the data or, conversely, leaving the data in a central store but accessing them nonetheless. Many scientific success stories illustrate what can be accomplished through pooling data from clinical trials. For example, a workshop presenter described the pooled data from drug therapy studies that enabled the discovery of aspirin as an effective means to decrease the risk of heart attacks. This information may be taken for granted now, but at the time it was a breakthrough precipitated by data sharing. Liberating data across trials has the potential to help answer important clinical questions. In another example, large pharmaceutical companies shared clinical data in the Biomarkers Consortium. The Consortium then was able to identify the biomarker adiponectin as a robust predictor of patient response to a certain class of diabetes drugs. The project demonstrated that crosscompany collaboration is a powerful approach to biomarker qualification, suggesting that data sharing might define a precompetitive space that would accelerate drug development. Similarly, NEWMEDS (Novel Methods Leading to New Medications in Depression and Schizophrenia) is one of the largest ever academic–industry research collaborations, with data from 91 clinical studies on depression and schizophrenia. This consortium has identified ways to design significantly more cost-effective studies. Along with the many benefits that data sharing can provide, a variety of disincentives and misalignment of incentives affect all stakeholders. Patients are understandably concerned about confidentiality, security, and privacy. As long as academic institutions build some of their credit systems on authorship, academic researchers may not see a good reason to share data, particularly if doing so results in being one of many authors. Further, data sharing would allow others to opine about or even analyze data, and, in the words of one participant: ‘‘Scientists don’t want to be scooped by their own