December 2016

A Critical Juncture for Clinical Data Sharing
By Frank Rockhold, PhD
For The Record
Vol. 28 No. 12 P. 5

Data sharing and transparency are vital to a brighter future for clinical research and health care in general. However, there are hurdles to clear. The question at the center of the open-science discussion is not whether data should be shared, but rather how responsible methods for doing so can be ushered in.

Data sharing can help produce valuable new insights, establish the credibility of clinical research, honor the contributions of participating patients to the greatest extent possible, and improve the design of future trials.

Tremendous progress has been made in the public disclosure of clinical research data. This progress is not a sudden phenomenon, but rather the latest point on a trajectory that began in 2004 with a move by medical journal editors to counter publication bias. Greater sharing of individual patient data is a logical step in this process.

Answering the Question of What Gets Shared
Two platforms for data disclosure initially emerged: The Yale University Open Data Access (YODA) Project and a GlaxoSmithKline website (now known as Clinical Study Data Request, or CSDR). Today, the CSDR includes more than 3,000 trials listed from 13 industry sponsors. While some sponsors wanted the option to deny requests in cases of potential conflicts of interest or competitive risk, this safety net has not yet been used. As sponsors gain experience, they may gain confidence that this option is unnecessary.

Another platform focused on cancer research, the widely used Project Data Sphere, only discloses trial control groups.

In a recent initiative, the Duke Clinical Research Institute (DCRI) developed SOAR (Supporting Open Access for Researchers), a collaborative effort between academia and industry to open clinical research data for the benefit of the broader research community.

In the SOAR initiative, the available longitudinal dataset includes all cardiac catheterization records for 50,000 patients treated between 1969 and 2015 at Duke University Medical Center. The analysis file contains 100,000 cardiac catheterization procedures. An independent review committee reviews each proposal.

A DCRI study of the use of open-access platforms found that although 3,000 trials were available to investigators, access to only 15.5% had been requested. The study examined three open access platforms: ClinicalStudyDataRequest.com, YODA, and SOAR. Most proposals did not focus on validating the primary results of the trial. Instead, they suggested secondary uses such as epidemiological studies, and subgroup analyses.

The reasons for underutilization of data from open-access platforms may include lack of knowledge their existence, poor funding for analyses, and the length of time needed to prepare and submit publications. Difficulties in conducting analyses of data from trials that used different structures and standards for data and metadata may also be slowing additional studies.

Changing Attitudes
A key overarching factor in the challenge of data sharing is an unwillingness to accept change. Because data have been kept hidden for most of the history of clinical trials, it will take a mammoth effort to change course. Several concerns have been raised about data sharing, including privacy, consent, intellectual property, costs, logistics, infrastructure, data standards, the risk of free-riding researchers, and potentially erroneous conclusions. Many cannot be eliminated but all can be at least mitigated and managed.

The main issues cited as barriers center on patient privacy, commercial and intellectual property loss, and a lack of data standards. Despite patient privacy being complex, many viable solutions that successfully deidentify data have been introduced. The concept of intellectual capital has been more difficult to define and remains an issue. Industry may be concerned that commercially sensitive information could be uncovered, while academic investigators are fearful that future research, publications, and grant funding could be compromised by the early release of data. While far from resolved, the impact of these issues is being ameliorated by independent review panels.

To address challenges involving a central platform to locate interoperable datasets, an effort sponsored by The Multi-Regional Clinical Trials Center of Brigham and Women's Hospital and Harvard, Wellcome Trust, and others, has focused on developing a common neutral platform for the housing and disclosing of clinical trial data. Many of the technical and governance challenges are being tackled by this collaboration.

In the end, for patients to benefit from the combined knowledge of the massive information trove of clinical trials, the industry needs a focused, interoperable source. This effort is the best attempt thus far to accomplish the task.

What's Ahead
The clinical research enterprise must build on what exists today to devise a one-stop shop for clinical trial data sharing. A single system that has the full support of sponsors, researchers, and other stakeholders could simplify the process of obtaining data, producing more secondary analyses and new studies using existing data. Such a system could form the basis for sharing preclinical, real-world epidemiological and other types of data and create a coordinated approach to realize value for patients.

— Frank Rockhold, PhD, is a professor of biostatistics and bioinformatics at the Duke Clinical Research Institute in Durham, North Carolina.