Key Messages Identified by Individual Speakers
• Data sharing does not have to be all or nothing; sharing of clinical study reports, including protocols, may be an intermediate step representing “low-hanging fruit” that would be relatively easy to implement now. However, even this approach brings its own challenges.
• Although notable progress has been made in sharing of placebo and comparator arm data, companies need to think about the boundaries of precompetitive space and what is gained and risked by sharing active arm data.
• Given their access to valuable data and their urgency to advance treatment alternatives, patients need to be engaged as partners in the clinical research process. When this is done successfully, much can be learned about the natural history of diseases and how best to match patients with promising treatment options.
• Clinical trials participants deserve to receive information from trials that will help them make health decisions, though decisions about what information to return to patients are the province of institutional review boards.
Throughout the workshop and during the closing discussion period, low-hanging fruit and priority actions were identified by speakers and other participants. Those key take-away points from the workshop are
gathered in this final chapter of the summary as a way to highlight and elaborate on next steps for advancing the sharing of clinical trials data.
As a possible intermediate step in the release of participant-level data, several participants raised the possibility of sharing clinical study reports, which are much more detailed reports than typically appear in a publication or even in an online publication. These reports, which generally contain the methodology, the subgroup analyses, the sensitivity analyses, and other detailed analyses of the data, are “low-hanging fruit” that could add to the information available in publications, without engendering the kinds of concerns raised when sharing patient-level data, said Jesse Berlin, Janssen Research & Development. Peter Doshi, Johns Hopkins School of Medicine, agreed that clinical study reports may be good initial targets for data sharing because they already exist for nearly all trials. Making them available would not be expensive, but would promote research integrity, medical knowledge, and public health (Rodwin and Abramson, 2012).
In fact, some trial organizers have already instituted this practice. Sachin Jain, chief medical information and innovation officer at Merck & Co., Inc., said that starting in July 2011, Merck has included the protocol and statistical analysis plan as part of its submission package to journals. Upon a journal’s acceptance of a manuscript for publication, Merck also provides the journal with the opportunity to post on its website the key sections of the protocol, including the objectives and hypotheses, patient inclusion and exclusion criteria, study design and procedures, efficacy and safety measures, the statistical analysis plan, and any amendments relating to those sections. “The response was terrific,” said Jain. “This is an idea whose time has come.” If external investigators need additional data at the patient level, Merck also has an initiative through which they can approach the company and ask for access. The data request goes to the product teams, who have the opportunity to review the request and ask questions.
Study protocols, which show how the investigators intended to run the trial, collect data, and perform analyses, provide far more detail than can be condensed into the methods section of a journal article, and often include important information from protocol amendments, such as changes to outcome measures. To emphasize the potential impact of
sharing study protocols on evaluation of trial results, Doshi cited a trial on the use of celecoxib (Celebrex) in the treatment of osteoarthritis and rheumatoid arthritis (Silverstein et al., 2000). Availability of the original study protocol allowed investigators to determine that the trial had been misreported, including the key claim that Celebrex was safer than alternative treatments (Hrachovec and Mora, 2001).
However, although the partial release of data may seem preferable to not having any data released, challenges with this model need to be considered. For example, partial release of data where trial organizers are allowed to decide which information to publicize creates the potential for selection bias.
Many of the data-sharing initiatives discussed during the workshop, such as the Datasphere Project, the Coalition Against Major Diseases and ePlacebo, involve the pooling of data from placebo or comparator arms of trials. The value of such initiatives was clearly acknowledged; for example, Janet Woodcock, director of the Food and Drug Administration’s (FDA’s) Center for Drug Evaluation and Research (CDER), indicated that by sharing the placebo group, one can decrease the numbers of people one needs to enroll in a clinical trial. However, added value of and barriers to sharing active treatment arm data was a notable topic of discussion. The successes that emerged for cardiovascular disease (see Box 2-2) were raised as motivation for making the sharing of these kinds of data a high-priority action.
Carolyn Compton, Critical Path Institute, talked about this issue in the context of defining the boundaries for precompetitive collaboration. Companies need to think about what they can collectively gain by pooling not only data, but expertise and other resources, “so that they all get something that none of them could acquire on their own,” she said. People will need to get past their anxieties over intellectual property (IP) in order to get more benefit out of money they have already spent and data they have already collected, both from successful and failed trials. Michael Cantor, Pfizer Inc., added that it is not just about the IP risk. There is also a lot of concern about the potential of uncovering previously undetected safety signals when one starts pooling data. Questions remain about what role the FDA could play in helping to sort out the
validity of safety concerns resulting from pooled analyses but, said Compton, there is the potential for enormous benefit to both the public and private sectors if people would be willing to contribute these kinds of data to a common source that was accessible to everyone.
In her closing statement, Josephine Briggs, National Institutes of Health, emphasized the need for clinical research organizations across all sectors to build systems that incorporate “patient-facing” pathways to data sharing, leveraging the urgency patients feel around advancing treatments, their access to other sources of valuable information, and the fact that they often have different tolerances for privacy risks as compared to the well public. During the closing discussion period and throughout the workshop, many participants talked about forming stronger partnerships with patients as a priority action for advancing clinical research. There was also discussion on how best to give back to patients who donate their time and personal information, and willingly take on the risks associated with novel treatments and protocols.
Learning from Patients
Jay “Marty” Tenenbaum, founder and chair of Cancer Commons, echoed Briggs’ sentiments. Health care is witnessing the dawn of personalized genomic medicine, but not enough information exists to make informed clinical choices on the basis of genomic data. For example, asserted Tenenbaum, cancer is not just a handful of diseases, but hundreds or thousands of diseases depending on the particular molecular drivers of a person’s tumor. Clinical trials using small cohorts of volunteers are not enough; 95 percent of patients will not be treated in clinical trials. Tenenbaum suggested implementing the Institute of Medicine’s vision of a rapid learning community in cancer (NRC, 2010) as the driving application behind personalized genomic medicine. Genomic and clinical information should be captured “from every cancer patient, from every conceivable source, from clinical trials, from electronic health records, and from the patients themselves, because they are the ones who have the natural instincts and incentives to share.” Thousands of experiments occur every day as oncologists try to extend the lives of their patients and
researchers need to be capturing these longitudinal data. This information could be used to create molecular models of cancer subtypes. As additional data are developed following treatments for those subtypes, models can be updated and subtypes can be split or modified, if necessary.
To realize this vision, patients will need to donate their data, which will increasingly include genomic information as such data are placed in patient medical records, Tenenbaum added. Information about potential improvements in treatments will create incentives to participate for both patients and physicians, as will transparency on what happens with the data and assurances that patients participating in research will be the first to benefit from resulting treatment advances. Patients’ data could be available online, which will “unleash an ecosystem of third-party applications,” according to Tenenbaum, offering value “that we can’t even anticipate but that will have a dramatic impact on health.”
Janet Woodcock, CDER, agreed with Tenenbaum that the Internet and social media could be used to learn much more about patient populations and the natural history of diseases, particularly those that are rare and therefore difficult to study. “It is a tragedy that [information about] most people’s course of illness is not used to further treatment of that illness,” she said. However, in contrast to Tenenbaum, she advocated for conducting research in a systematic manner, as can be done with clinical trials. She mentioned the I-SPY trial, which is being used to study biomarkers for predicting response to cancer therapies. Moving that kind of initiative out into the community, for example, by inviting people who have cancer to get their tumors genotyped, could enable the design of large-scale trials in which people could be matched to the drugs that are most likely to have effects on that tumor, and then the trial could be adapted over time. “Everyone agrees that sharing clinical data is a public good … and that we really need to figure out how to leverage these data as much as possible because [they’re] so precious,” said Woodcock.
In the context of incentivizing patients to donate their clinical data, Cindy Geoghegan, principal at Patient and Partners, asked about the extent to which participant-level data should be made available to the participants in a clinical trial, whether in a de-identified form or more personally, so that, for example, incidental findings are conveyed to participants.
Harlan Krumholz, Yale University School of Medicine, had previously reminded the workshop participants that to restore trust in clinical trials, investigators need to listen to people so as not to seem “self-serving even as we are trying to promote society’s best interest.” With that in mind, principles could be adopted that would govern the return of information to research participants. Many trialists do not make enough effort to let participants know the results of the trial and what the implications are for society, let alone for them as individual patients, he said. It is easier to end a trial and “not have an individual interaction with the participants to make sure that they understood what had accrued as a result.” People who are persistent often can get information about their assignment in a trial, but most trials do not provide the information as a service and in an easy-to-understand way.
Deborah Zarin, National Library of Medicine, observed that such decisions are the province of the institutional review board. Each trial would have its own guidelines, which would be laid out during the informed consent process. Rob Califf, Duke University Medical Center, agreed that there could be a standard and that it could be articulated fairly easily. “In general, there should be an obligation to inform participants in clinical trials about the result of the trial,” he said. In most cases, participants can be informed about whether they got a treatment or placebo after the trial concludes. But making sure that a participant can put his or her individual assignment in the context of the results of a trial will take time and effort.
The Need for Leadership
Sharon Terry, president and chief executive officer of Genetic Alliance, concluded the workshop by observing that far too few people are participating in clinical research. When fewer than 10 percent of the population is engaged, either through participation in trials or through citizen science, the system is broken, she said. We need to find ways to involve many more people in biomedical research. Disease-specific and cross-disease advocacy or interest groups already exist that could foster this involvement. Industry has demonstrated its willingness to participate and increase its sharing of data. Researchers in specific areas have moved toward models of cooperation and data sharing.
Terry reminded the group that it is a mistake to refer to patients as “them” and to researchers as “us.” Everyone is either a patient or a po-
tential future patient. “When we say we need to communicate with them, that’s us.” Public engagement in clinical trials has been lacking but as involvement becomes more commonplace, the resulting changes could be abrupt and dramatic.
What is needed, she concluded, is leadership from each of the stakeholders involved in clinical research. Each group needs to ask how it is impeding the flow of information and the conduct of research and take steps to remove those barriers. Each needs to move forward, said Terry, “for the sake of all the people who need us to do this quickly, efficiently, and carefully.”