The Future of Data Sharing in a Changing Landscape
In Chapters 3 through 5, the committee recommends strategies and practical approaches for the responsible sharing of clinical trial data, addressing the who, what, when, and how of data sharing in the current environment. This chapter envisions the future that would emerge if stakeholders committed to responsible data sharing, modified their work processes to facilitate it, and possessed the resources and tools necessary to do so. This chapter also looks at the remaining infrastructure, technological, workforce, and sustainability challenges to achieving this future and provides the committee’s recommendation for next steps in a path forward.
The committee intends this report to be the beginning and not the end of discussions about how to develop a responsible global ecosystem for the responsible sharing of clinical trial data. More public discussions about how best to address the challenges of data sharing will be needed, ideally informed by the ongoing experience of data sharing initiatives. To help guide such discussion, the committee articulates its vision for the future:
- There are more platforms for sharing clinical trial data, with different data access models and sufficient total capacity to meet demand. Stakeholders are able to identify the platform that is most
-
appropriate for their needs. The various initiatives are interoperable (e.g., data obtained from different platforms can easily be searched and combined to allow further analyses).
- A culture of sharing clinical trial data with effective incentives for sharing flourishes.
- Best practices for sharing clinical trial data are identified and modified in response to ongoing experience and feedback. The sharing of clinical trial data forms a “learning” ecosystem in which data on data sharing outcomes are routinely collected and continually used to improve how data sharing is conducted.
- There is adequate financial support for sharing clinical trial data, and costs are fairly allocated among stakeholders.
- Protections are in place to minimize the risks of data sharing (for example, threats to valid secondary analyses, participant privacy, intellectual property, and professional recognition) and to reduce disincentives for sharing.
Existing models described in previous chapters provide an initial foundation for building a global ecosystem for sharing clinical trial data. However, challenges need to be addressed if this future is to be realized. The following sections describe these infrastructure, technical, workforce, and sustainability challenges in greater detail, as well as the committee’s views on a structure for further collaboration that will accomplish important next steps into the future. The final section presents the committee’s recommendation for addressing these challenges and effecting this collaboration.
If sponsors and investigators are to implement the recommendations in Chapters 3, 4, and 5, repositories with the capacity to hold and manage the vast amounts of incoming data will have to be created. Investigators are not in a position to hold and manage data from trials for an extended period of time; therefore, without a place to easily store the data after trials have been completed, investigators will have difficulty complying with the committee’s recommendations for data sharing.
In addition to capacity, a data sharing infrastructure will need to be capable of managing data access according to the strategies laid out in Recommendation 3 in Chapter 5. As big data approaches become more widespread, newer technological solutions to data access may offer effective ways of achieving the benefits of sharing clinical trial data while mitigating its risks. These newer solutions are predicated on an approach to data query that differs from the traditional one with which most clini-
cal trialists are familiar. In the traditional approach, data are brought to the query. That is, if a data requester wants to run a query, the requester obtains a copy of the data, installs the data on his/her own computer, and runs the query on the downloaded data. Because the data requester now holds a copy of the data, the original data holder has effectively lost control over access to the data.
The converse approach is bringing the query to the data: the data requester submits the query to the machine where the data reside, the query is run on that remote machine, and the results are returned back to the requester. Queries can, of course, be complex computations and analyses, not just simple search and retrieval queries. In this model, data holders retain control of the data, and the requester never has a copy of or control over the data.
This basic idea of bringing the query to the data can be implemented through many different configurations of databases and query servers. Three example configurations are described below to illustrate their representative benefits and challenges; many variations on each are possible. The committee makes no recommendation on data query architectures for data sharing because detailed consideration of this topic is beyond its charge and expertise.
Local Data Stores
In the simplest model, every data holder hosts its own data on its own server. External data requesters are allowed to establish user accounts on that server, perhaps with one of the access control models discussed in Chapter 5. The requester then can view and analyze the data but cannot download a copy of the data to his/her own machine.
Variants of this model currently exist for sharing clinical trial data (e.g., Yale University Open Data Access [YODA], in which a third party, Yale University, acts as the data holder for Johnson & Johnson). However, this model is infeasible for widespread data sharing because it is prohibitively expensive and inefficient for multiple data holders to handle access control, data provision, and user account services. Moreover, data requesters wishing to query multiple databases—for example, to carry out a meta-analysis—must establish multiple user accounts and navigate multiple access policies and procedures. Even after obtaining access to multiple data sets, the requester could not merge them.
One Single Centralized Data Store
The opposite of having each data holder maintain its own server is to collect all clinical trial data worldwide into one central database. This
model benefits from economies of scale, and data requesters need submit their queries to only one database. However, a single global database of clinical trial results is unlikely to be adopted, given multiple global stakeholders, interests, and sensitivities as discussed in Chapter 3.
Federated Query Model
The federated query model combines the approach of bringing the query to the data with federated databases. Databases are federated when independent geographically dispersed databases are networked in such a way that they can respond to queries as if all the data were in a single virtual database. Thus, when data requesters submit a query to a federated query service, that query is routed to all databases participating in the federation. The provider of the query service may or may not be the “trusted intermediary” that adjudicates access control requests as discussed above. Federated query services can be purchased as a standalone technical service. Data holders maintain full control over their data, and neither the data requester nor the query service provider ever has direct access to the data. Federated query systems can protect against invasions of privacy, as discussed in Chapter 5. The U.S. Food and Drug Administration’s (FDA’s) Mini-Sentinel program is using a federated model to combine data sets for comparative effectiveness and outcomes research (Mini-Sentinel, 2014).
If all databases of clinical trial results were housed in a single global federated system that adopted a uniform technical approach to implementing access control, or even in a few federated systems, significant economies of scale and technical ease of data access would result. However, a large federated system could coexist with multiple trusted intermediaries purchasing query services from multiple providers.
A federated query approach can support access control models in which data requesters are able to query multiple databases simultaneously while data holders maintain control of their own data at all times. Yet while the technology and methods to implement a global federated data access system theoretically exist today, substantial challenges arise in practice. Different platforms making up the federated system need to be interoperable. Common data models and data exchange protocols that meet the needs of scientific analysts need to be defined and adopted. User authentication and authorization processes must be defined across different cultures, languages, and legal jurisdictions. Furthermore, secondary users may raise concerns that their data analyses will take more time and be more cumbersome than if they had all the data on their computers.
Conclusion: Currently there are insufficient platforms to store and manage clinical trial data, under a variety of access models, if all sponsors and investigators commit to data sharing.
Just because data are accessible does not mean they are usable. Data are usable only if an investigator can search and retrieve them, can make sense of them, and can analyze them within a single trial or combine them across multiple trials. Given the large volume of data anticipated from the sharing of clinical trial data, the data must be in a computable form amenable to automated methods of search, analysis, and visualization.
To ensure such computability, data cannot be shared only as document files (e.g., PDF, Word). Rather, data must be in electronic databases that clearly specify the meaning of the data so that the database can respond correctly to queries. If data are spread over more than one database, the meaning of the data must be compatible across databases; otherwise, queries cannot be executed at all, or are executable but elicit incorrect answers. In general, such compatibility requires the adoption of common data models that all results databases would either use or be compatible with.
The meaning of the data in a database is specified through two basic mechanisms. The first is the data model or database schema, which, like column titles in a data table, describes the contents of each data field and defines the kinds of queries to which the database can respond. For example, for a database with a table column titled “Body Weight,” all the cells under that column contain measures of weight, and this database can respond to queries about weight. A trial results database will have a data model that has the equivalent of many tables and table columns.1 To be most useful for purposes of scientific reuse of data, this data model must include tables and columns for elements of the trial protocol that investigators will want to query for when they search for relevant trials, such as intervention names, primary outcomes, and study population characteristics (Sim and Niland, 2012). This protocol data model should be robust enough to respond to the complexity of typical scientific queries—for example, “Find me trials of metformin without exercise and diet for prevention of diabetes mellitus. Then give me access to their analyzable data sets.” Although ClinicalTrials.gov contains study protocol information, its database currently is not robust enough to respond to such
________________
1 Relational databases follow a table structure, but there are other types of databases that are not based on tables. Relational databases are discussed here for illustrative purposes, recognizing that results databases may be of different types.
detailed queries (Tasneem et al., 2012). With the high level of investment required to enable sharing of clinical trial results, it is imperative that a sufficiently robust common protocol model be defined and adopted to ensure that descriptions of trials can be computationally searched, analyzed, and visualized across multiple databases. Leading protocol data models include CDISC (CDISC, 2014), the PICO ontology from the Cochrane Collaboration (Data.cochrane.org, 2014), and the Ontology of Clinical Research (Sim et al., 2013).
In addition to a common protocol data model, data standardization is required at the level of the study variables across trials (also termed the “data dictionary”). For example, if one trial collects body weight in kilograms, while another collects it as a categorical variable (e.g., 0-50 lb, 51-100 lb, etc.), while yet another collects only body mass index, scientific integration is greatly hampered if not impossible. If sharing of clinical trial data is to be useful for meta-analysis and large-scale data mining, trial protocols should ideally use common data elements for eligibility criteria and baseline and outcome variables. These data elements also should be indexed to standard clinical vocabularies (e.g., SNOMED) as appropriate. There are many common data element initiatives worldwide, from funding agencies (e.g., PROMIS and PhenX from the U.S. National Institutes of Health [NIH]), professional societies (e.g., the American Heart Association and American College of Cardiology Foundation), research collaboratives (e.g., PCORnet), industry collaboratives (e.g., CDISC), and nonprofit organizations (e.g., COMET). The adoption of common data elements has to date been slow to occur, in part because trialists are not aware of these initiatives (HIMSS, 2009) and in part because of a lack of incentives and clear value for doing so. Scientific value will accrue from data sharing only if investigators can easily access results data and can query, align, compute on, and visualize what will no doubt be a large amount of complex, heterogeneous data.
Conclusion: Current data sharing platforms are not consistently discoverable, searchable, and interoperable. Special attention is needed to the development and adoption of common protocol data models and common data elements to ensure the capacity for meaningful computation across disparate trials and databases.
An adequate workforce trained in the operational and technical aspects of data sharing is essential to meet the goals of responsible sharing of clinical trial data. As outlined in the 2012 report of the Advisory Committee to the NIH Director, there is a growing gap between the supply
of trained quantitative research scientists at all levels (e.g., M.D., Ph.D., M.S., and B.S.) and the growing demand, stimulated in part by the recent explosion of big data in basic, translational, and clinical research (NIH, 2012a). In addition, the data on NIH-funded researchers indicate that the clinical trial and more generally the clinical research workforce is aging, and a new generation needs to be trained.
NIH’s training mission has traditionally been focused on training doctoral-level researchers, with limited ability to train at the support staff level or to provide non-Ph.D.-level graduate training, such as that for a master’s degree in clinical research or clinical trials (IOM, 2012, 2013b,c; NCATS, 2013; NIH, 2012b; Zerhouni, 2005). The need for the latter research support workforce is coupled with another key goal of the Clinical Translational Science Award (CTSA) program—training the scientific workforce needed for the translational sciences (NCATS, 2013). NIH’s recent funding of CTSAs at more than 60 U.S. institutions offers an opportunity for training at various levels, as well as for short courses to address specific technical needs (NIH, 2014). Thus, the CTSA program could require CTSA institutions to provide support and training for designing clinical trials with an eye toward data sharing and its implementation. CTSA institutions could provide technical support and an infrastructure for sharing clinical trial data and also develop and disseminate best practices for data sharing among the CTSA consortium and partner institutions. As the academic homes for advancing innovative clinical and translational research, the leaders of National Center for Advancing Translational Sciences (NCATS) and CTSA institutions have a unique opportunity to create a professional culture conducive to sharing clinical trial data.
Conclusion: Training for the sharing of clinical trial data needs to be part of the overall mission of funders of research training programs.
Other stakeholders can and should contribute to workforce development as well. Large pharmaceutical, device, and biotechnology companies, as well as smaller industry sponsors when feasible, can contribute valuable hands-on, state-of-the-art training in data sharing. Foundations that sponsor medical research and training also can enlarge the scope of their programs to include data sharing. And international bodies that fund training of the clinical trial workforce can make training researchers in data sharing a core component of their initiatives. Making more clinical trial data available for analyses will yield few gains if too few data scientists are adequately trained to turn these data into knowledge.
Conclusion: A workforce with the skills and knowledge to manage the operational and technical aspects of data sharing needs to be developed.
To assess the benefits and burdens of sharing clinical trial data, it is essential to have sound estimates of the costs of different data sharing models. The only cost information in a peer-reviewed publication that the committee could identify is contained in a paper (Wilhelm et al., 2014) that breaks down the costs of data sharing in the Alzheimer’s Disease Neuroimaging Initiative (ADNI), a longitudinal study, into four components:
- Infrastructure and administration costs include the data repository, storage and curation of data, review of informed consent forms, and management of material transfer agreements.
- Standardization costs include efforts to organize data and make them understandable to others.
- Human resources costs encompass building and maintaining the infrastructure, providing data access, and responding to queries from potential users. The committee notes that responding to queries from potential secondary users is a challenge because staff that worked on a clinical trial are assigned to new projects after the trial’s completion or leave the sponsor, data coordinating center, or institution that carried out the study. Responding to such queries is an important aspect of responsible sharing of clinical trial data: If there is no one to answer questions about the data from secondary users, the scientific usefulness of the shared data will be compromised. The committee also notes that the use of an Independent Review Board to determine access to clinical trial data may add further costs.
- There are opportunity costs associated with not carrying out new research or new analyses of existing data.
The authors of this paper report that ADNI investigators estimated that data sharing would account for 10-15 percent of the total costs of the program and require about 15 percent of investigators’ time (Wilhelm et al., 2014). The authors note problems with determining the costs of data sharing: “Meanwhile, there are few benchmarks by which to ascertain the costs of data sharing and as yet no prospectively derived metrics by which to reliably estimate the categorical costs.”
Although this paper is helpful, the committee notes that it has several limitations. The authors examined only one study, a longitudinal neuroimaging study that collected large amounts of complex imaging data. Raw data from longitudinal neuroimaging studies may be much more extensive than complete analyzable data sets from clinical trials and thus more expensive to prepare, curate, and share. Costs were estimated by the study investigators without direct measurement, verification, or
audit. The cost estimates did not include the costs of preparing the clinical study report (CSR) for a clinical trial that would not be submitted to regulatory agencies or manually redacting the CSRs from legacy trials for participant identifiers or commercially confidential information (Scott, 2013; Shoulson, 2014). The committee notes also that the figures in this paper may overestimate future data sharing costs if data collection templates and procedures are revised with data sharing in mind, as discussed further below.
Wilhelm and colleagues (2014, p. 1202) comment on inequities in the current business model for data sharing:
The cost categories described are borne by those researchers who originally collected the data, with little if any cost to data users. Therefore, cost recovery in data sharing is needed and justifiable, especially because the current funding milieu provides limited support for data sharing.
The committee heard testimony about the current distribution of the costs of data sharing. Representatives of pharmaceutical companies that are carrying out or planning the sharing of clinical trial data testified that they are now paying all the costs of sharing data from trials they sponsor. In their view, other stakeholders should also contribute, and public sponsors should provide financial support for sharing data from trials they fund (Kuntz, 2013; Scott, 2013). In addition, the committee heard testimony from a biotechnology firm that for small companies with no revenue stream the cost of sharing clinical trial data would be prohibitive and a serious disincentive to investors (Moch, 2014). However, the committee also heard the counterargument from investors that if data sharing is carried out with appropriate controls to minimize invalid analyses (e.g., through review of data requests and analysis plans by independent review boards; see Chapter 5), investors will have greater confidence in promising trials and invest accordingly (Leff, 2014). The committee notes that small funders account for a significant proportion of new therapeutic discoveries. According to the Small Business Administration, 42 percent of new drug approvals in 2012 were granted to emerging sponsors (FDA, 2013). The committee notes further that there are precedents for reducing fees for small companies; the FDA, for example, charges small companies reduced application fees for new products. Public funders of clinical research such as NIH, whose budget has been declining in real dollars, may be reluctant to fund the sharing of clinical trial data if doing so would further reduce the funds available for new research grants. Finally, from a global perspective, the committee notes that the costs of data sharing may be prohibitive for clinical trialists in low-resource countries.
From the above findings, the committee concludes that the current business model for sharing clinical trial data is not sustainable. Further-
more, the current model for funding the sharing of clinical trial data does not distribute the costs equitably among the various stakeholders that participate in and benefit from such data sharing.
The committee is mindful that it was not constituted to address the issue of the cost of sharing clinical trial data and how to distribute those costs. In keeping with its charge to “outline strategies and suggest practical approaches to facilitate responsible data sharing,” however, the committee presents the following conceptual framework regarding the costs of responsible sharing of clinical trial data.
First, responsible sharing of clinical trial data benefits the public and multiple stakeholders. Data sharing is a public good, as is the original research, leading to additional scientific knowledge regarding the effectiveness and safety of therapies, diagnostic tests, and the delivery of health care. In addition to patients and their physicians, other stakeholders benefit from this additional knowledge. These stakeholders include payers for health care (public insurance and health care payers at the state and local levels, private insurers, and employers that cover health insurance costs) that determine reimbursement based on evidence regarding the benefits and risks of therapies, as well as organizations that establish clinical practice guidelines (professional organizations and government agencies) (IOM, 2011).
Second, as a matter of fairness, those who benefit from responsible sharing of clinical trial data, including the users of shared data, should bear some of the costs of sharing. There is policy precedent for charging user fees to obtain access to data collected by others. The Health Information Technology for Economic and Clinical Health (HITECH) Act explicitly states that organizations that share data are entitled to reasonable cost recovery (infrastructure costs plus the marginal cost of delivering the data) (Evans, 2014). Any user fees will have to include provision for researchers from resource-poor areas, where the burden of disease and potential for benefit may be disproportionately great. Additional sources of funding for responsible sharing of clinical trial data may be identified; for example, private philanthropies may be interested in funding the development of infrastructure and the conduct of pilot projects for responsible sharing of clinical trial data or subsidizing data access in resource-poor countries.
Third, policies on equitably sharing the costs of responsible sharing of clinical trial data among stakeholders should be based on accurate information on the costs of data sharing for various kinds of clinical trials. Such cost data do not currently exist and would best be collected by impartial accounting and economics experts. Furthermore, there are no estimates of the potential savings that may result from sharing clinical trial data—which may include lower costs for secondary research
conducted using shared data sets—for future clinical trials as a result of information gleaned from previous trials, or for health care because of lower use of ineffective or less effective therapies and reduced complications resulting from better safety data.
Fourth, the costs of responsible sharing of clinical trial data will decrease in the future if data collection and management are designed to facilitate data sharing. As an example, the Open Science Framework developed by the Center for Open Science makes transparent virtually all the metadata and data required for sharing on an ongoing, real-time basis as the research is being conducted (Open Science Framework, 2014). A shareable data set, with audit trails, is created in real time. Thus, the same technological platform used for both real-time data management and the conduct of research can be used for long-term data storage. After a trial has been completed, the study data, metadata, and relevant study documents can be made accessible almost immediately with minimal additional effort. In addition, as discussed above, another innovation with the potential to reduce the cost of data sharing is disease-specific standardized data elements and outcomes, as are currently being developed through the CDISC initiative, PCORnet, and other collaborative research enterprises. The lesson to be learned is that technological and procedural innovations that improve the quality of clinical trials can also reduce the costs of data sharing relative to current study procedures and data systems.
Conclusion: Currently, the costs of data sharing are borne by a small subset of sponsors, funders, and clinical trialists; for data sharing to be sustainable, its costs will need to be equitably distributed across both data generators and users.
Conclusion: A market/landscape analysis of the costs of sharing clinical trial data and an economic analysis of sustainable and equitable funding options would provide an evidence base to facilitate the development of sustainable and equitable models for responsible data sharing.
STRUCTURE FOR COLLABORATIVE NExT STEPS
Sharing clinical trial data is a global health priority that is gathering momentum. The commissioning of the present study was intended to further the dialogue and begin to build a stronger foundation for a robust data sharing culture. But such data sharing is still nascent, and the challenges to achieving the vision outlined in this chapter are formidable. The committee has proposed guiding principles that need to be balanced in responsible sharing of clinical trial data and made recommendations
addressing a number of the challenges to data sharing. To attempt to suggest how all specific issues should be resolved would, however, be presumptuous, imprudent, and beyond the committee’s expertise and charge.
Several models for sharing clinical trial data exist today, and more can be expected in the future. Sponsors will try different approaches, and the outcomes of these approaches will provide useful information on what does and does not work in various contexts. New issues and challenges are likely to emerge as more experience is gained with sharing clinical trial data and as clinical trials themselves change. Thus, the sharing of clinical trial data will evolve in ways that cannot be predicted today.
Approaches to data sharing are likely to change and improve if stakeholders learn from experience and new approaches to clinical trials are introduced. Chapter 5 describes how pioneers in sharing clinical trial data, such as YODA and the European Medicines Agency (EMA), have revised their policies and procedures in response to feedback from stakeholders and early experience with data sharing. The committee proposes in that chapter (under Recommendation 3) that organizations sharing clinical trial data “learn from experience by collecting data on the outcomes of data sharing policies, procedures, and technical approaches (including the benefits, risks, and costs), and share information and lessons learned with clinical trial sponsors, the public, and other organizations sharing clinical trial data.”
In this chapter, the committee has further developed the idea of improving the sharing of clinical trial data by drawing on experience in other areas of biomedical research and health care. The sharing of data from biobanks and genomic sequencing projects offers several insights. The U.K. Biobank has advocated “adaptive governance” for biobanks, characterized by willingness to adapt to unforeseen or emerging issues, flexibility, and nimbleness (Laurie and Sethi, 2013; O’Doherty et al., 2011a). Scholars who have consulted for biobanks regarding their governance also have advocated for adaptive governance (Kaye, 2014; O’Doherty et al., 2011b). Another aspect of adaptive governance in the context of responsible sharing of clinical trial data is using discretion to modify timelines for data sharing in exceptional circumstances, as discussed in Chapter 5.
Quality improvement in health care and industry builds on data-driven improvements. A “learning health care system” improves the quality of health care and reduces costs (IOM, 2013a). In the continuous quality improvement model, an opportunity for improvement is identified through outcome metrics, a broad-based team suggests how to improve the activity or process, and the impact of the intervention is tracked through ongoing monitoring of the metric, leading to a cycle of further
improvements. “Learning” organizations have additional characteristics that facilitate improvement, including effective leadership, a culture that prizes improvement, and an emphasis on taking advantage of digital and Internet-based technology (IOM, 2013a). These organizational characteristics complement those advocated by proponents of adaptive governance.
In addition to individual funders and trusted intermediaries, the committee has considered the ecosystem of responsible sharing of clinical trial data as a whole. Although individual sponsors and trusted intermediaries can do a great deal within their own organizations to make the sharing of clinical trial data more responsible, effective, and efficient, other challenges can be addressed only in collaboration with other institutions.
First, different funders and intermediaries can share outcome data for different data sharing models with each other and the public (Strom et al., 2014). Sharing these outcomes will give individual organizations incentives to improve, develop common metrics, share lessons learned, and consider how to address common challenges.
Second, some challenges can be met only from a broader perspective than the standpoint of an individual funder or intermediary. Earlier in this report, the committee discussed how common data elements, interoperability, federated models for sharing data, a data set identification system, and sustainable and equitable business models would make the sharing of clinical trial data more useful and likely reduce costs as well. None of these elements can be developed by a single sponsor or trusted intermediary.
Third, the ecosystem for sharing clinical trial data consists of many types of organizations. In Chapter 3, the committee recommends that disease advocacy organizations, regulatory and research oversight agencies, Institutional Review Boards or Research Ethics Committees, research institutions and universities, medical journals, and membership and professional societies take certain steps to promote responsible sharing of clinical trial data. Many challenges will best be addressed through collaborative efforts involving different types of institutions. Chapter 3 presents the committee’s analysis of the need for academic institutions and funders of clinical trials to provide incentives for investigators to share clinical trial data. One important need is to develop a way of tracking secondary analyses by other investigators that use a clinical trial data set so the original clinical trial and its investigators can receive appropriate professional recognition. Connecting a shared data set to subsequent publications of other investigators is a problem that other fields of science also are addressing (NRC, 2012). Still another problem requiring collaborative effort is how academic institutions should give appropriate professional recognition to a researcher who produces clinical trial data sets that other investigators use for secondary research. Universities might benefit from
discussing with each other and with pretenure faculty and secondary users of data how to document and assess the scholarly contribution due to data that are shared.
This report articulates guiding principles and high-level recommendations to guide responsible sharing of clinical trial data. Several early adopters have established proof of principle that the sharing of clinical trial data can be accomplished. For responsible sharing of clinical trial data to become pervasive, sustained, and rooted as a professional norm, however, much additional work will need to be done. Many interrelated issues need to be resolved, and as changes occur in how clinical trials are designed and carried out, new issues and challenges undoubtedly will arise. Discussion among stakeholders and the exchange of ideas and empirical evaluations of different data sharing models will help best practices, incentives, and areas of agreement to emerge. To some extent, such discussion is already occurring. However, establishing forums for the discussion of issues and experiences among a broad range of stakeholders with varied interests can catalyze implementation. Because responsible sharing of clinical trial data is so multifaceted, people working on one aspect of data sharing need to be aware of how their work interacts with work on other aspects. The committee recommends that a combination of public, nonprofit, and industry funders, similar to the sponsors of this project, take the lead in convening these stakeholders. However, to ensure broad representation of stakeholder interests, including those of participants and investigators, it would be desirable for members of the convening body not to have a direct stake in clinical trials as sponsors, funders, or investigators. Ideally, the convener should be regarded as impartial and trusted by the multiple stakeholders who have countervailing interests in the sharing of clinical trial data. In the United Kingdom, for example, the Nuffield Council on Bioethics, the Academy of Medical Sciences, and the Medical Research Council each have issued several consensus reports on science and research policy (EAGDA, 2014); similar impartial, trusted bodies in other countries (such as the Rathenau Instituut in the Netherlands and the Canadian Academy of Health Sciences) could also play this role. Collaborations between convening organizations in different countries could help assure a global perspective. After a few initial meetings, it is likely that stakeholders most committed to responsible sharing of clinical trial data would agree on whether some ongoing forum or forums were desirable and if so, how they might be convened.
Recommendation 4: The sponsors of this study should take the lead, together with or via a trusted impartial organization(s), to convene a multistakeholder body with global reach and broad representation to address, in an ongoing process, the key infrastructure, technological, sustainability, and workforce challenges associated with the sharing of clinical trial data.
CDISC (Clinical Data Interchange Standards Consortium). 2014. CDISC vision and mission. http://www.cdisc.org/CDISC-Vision-and-Mission (accessed October 17, 2014).
Data.cochrane.org. 2014. PICO ontology. http://data.cochrane.org/ontologies/pico (accessed October 17, 2014).
EAGDA (Expert Advisory Group on Data Access). 2014. Establishing incentives and changing cultures to support data access. http://www.wellcome.ac.uk/stellent/groups/corporatesite/@msh_peda/documents/web_document/wtp056495.pdf (accessed December 19, 2014).
Evans, B. J. 2014. Sustainable access to data for postmarketing medical product safety surveillance under the amended HIPAA Privacy Rule. Health Matrix Clevel 24:11-47.
FDA (U.S. Food and Drug Administration). 2013. Section 1128 of the Food and Drug Administration Safety and Innovation Act (FDASIA):Small business report to Congress. http://www.fda.gov/downloads/RegulatoryInformation/Legislation/FederalFoodDrugandCosmeticActFDCAct/SignificantAmendmentstotheFDCAct/FDASIA/UCM360058.pdf (accessed November 5, 2014).
HIMSS (Healthcare Information and Management Systems Society). 2009. Defining and testing EMR usability: Principles and proposed methods of EMR usability evaluation and rating. http://www.himss.org/files/HIMSSorg/content/files/himss_definingandtestingemrusability.pdf (accessed October 17, 2014).
IOM (Institute of Medicine). 2011. Clinical practice guidelines we can trust. Washington, DC: The National Academies Press.
IOM. 2012. Envisioning a transformed clinical trials enterprise in the United States: Establishing an agenda for 2020: workshop summary. Washington, DC: The National Academies Press.
IOM. 2013a. Best care at lower cost: The path to continuously learning health care in America. Washington, DC: The National Academies Press.
IOM. 2013b. The CTSA program at NIH: Opportunities for advancing clinical and translational research. Edited by A. I. Leshner, S. F. Terry, A. M. Schultz, and C. T. Liverman. Washington, DC: The National Academies Press.
IOM. 2013c. Sharing clinical research data: workshop summary. Washington, DC: The National Academies Press.
Kaye, J. 2014. Adaptive governance. Paper presented at IOM Committee on Strategies for Responsible Sharing of Clinical Trial Data: Meeting Four, May 5-7, Washington, DC.
Kuntz, R. 2013. Yale University Open Data Access Project: A model for dissemination and independent analysis of clinical trial program data. Paper presented at IOM Committee on Strategies for Responsible Sharing of Clinical Trial Data: Meeting One, October 22-23, Washington, DC.
Laurie, G., and N. Sethi. 2013. Towards principles-based approaches to governance of health-related research using personal data. European Journal of Risk Regulation 4(1):43-57.
Leff, J. 2014. Strategies and practical approaches for responsible sharing of clinical trial data: Perspectives of trial sponsors and investors. Paper presented at IOM Committee on Strategies for Responsible Sharing of Clinical Trial Data: Meeting Four, May 5-7, Washington, DC.
Mini-Sentinel. 2014. welcome to Mini-Sentinel. http://mini-sentinel.org (accessed October 17, 2014).
Moch, K. I. 2014. Clinical trial data sharing from the pre-revenue biotech perspective: Addressing the impact of limited resources on the implementation of data sharing technologies. Paper presented at IOM Committee on Strategies for Responsible Sharing of Clinical Trial Data: Meeting Two, February 3-4, Washington, DC.
NCATS (National Center for Advancing Translational Sciences). 2013. Scholar and research programs. http://www.ncats.nih.gov/research/cts/ctsa/training/programs/scholar-trainee.html (accessed September 30, 2014).
NIH (U.S. National Institutes of Health). 2012a. Data and informatics working group: Draft report to the advisory committee to the director. Bethesda, MD: NIH.
NIH. 2012b. Request for information (RFI):Enhancing community-engaged research through the Clinical and Translational Science Awards (CTSA) program. http://grants.nih.gov/grants/guide/notice-files/NOT-TR-13-001.html (accessed September 30, 2014).
NIH. 2014. Clinical and Translational Science Awards. http://www.ncats.nih.gov/research/cts/ctsa/ctsa.html (accessed October 24, 2014).
NRC (National Research Council). 2012. For attribution—developing data attribution and citation practices and standards: Summary of an international workshop. Washington, DC: The National Academies Press.
O’Doherty, K. C., M. M. Burgess, K. Edwards, R. P. Gallagher, A. K. Hawkins, J. Kaye, V. McCaffrey, and D. E. Winickoff. 2011a. From consent to institutions: Designing adaptive governance for genomic biobanks. Social Science & Medicine 73(3):367-374.
O’Doherty, K. C., M. M. Burgess, K. Edwards, R. P. Gallagher, A. K. Hawkins, J. Kaye, V. McCaffrey, and D. E. Winickoff. 2011b. From consent to institutions: Designing adaptive governance for genomic biobanks. Social Science & Medicine 73(3):367-374.
Open Science Framework. 2014. Project management with collaborators, project sharing with the public. https://osf.io (accessed October 24, 2014).
Scott, J. 2013. Access to anonymised patient level data from GSK clinical trials. Paper presented at IOM Committee on Strategies for Responsible Sharing of Clinical Trial Data: Meeting One, October 22-23, Washington, DC.
Shoulson, I. 2014. Cultural and financial incentives for data sharing: Recognition and promotion. Paper presented at IOM Committee on Strategies for Responsible Sharing of Clinical Trial Data: Meeting Two, February 3-4, Washington, DC.
Sim, I., and J. C. Niland. 2012. Study protocol representation. In Clinical research informatics, edited by R. L. Richesson and J. E. Andrews. New York: Springer. Pp. 155-174.
Sim, I., S. W. Tu, S. Carini, H. P. Lehmann, B. H. Pollock, M. Peleg, and K. M. Wittkowski. 2013. The Ontology of Clinical Research (OCRe):An informatics foundation for the science of clinical research. Journal of Biomedical Informatics 52:78-91.
Strom, B. L., M. Buyse, J. Hughes, and B. M. Knoppers. 2014. Data sharing, year 1—access to data from industry-sponsored clinical trials. New England Journal of Medicine 371: 2052-2054.
Tasneem, A., L. Aberle, H. Ananth, S. Chakraborty, K. Chiswell, B. J. McCourt, and R. Pietrobon. 2012. The database for Aggregate Analysis of ClinicalTrials.gov (AACT) and subsequent regrouping by clinical specialty. PLoS ONE 7(3):e33677.
Wilhelm, E. E., E. Oster, and I. Shoulson. 2014. Approaches and costs for sharing clinical research data. Journal of the American Medical Association 311(12):1201-1202.
Zerhouni, E. A. 2005. Translational and clinical science—time for a new vision. New England Journal of Medicine 353(15):1621-1623.
