- Email: [email protected]
Perspectives on clinical trial data transparency and disclosure
Contemporary Clinical Trials 39 (2014) 28–33
Contents lists available at ScienceDirect
Contemporary Clinical Trials journal homepage: www.elsevier.com/locate/conclintrial
Perspectives on clinical trial data transparency and disclosure Demissie Alemayehu ⁎, Richard J. Anziano, Marcia Levenstein Pfizer Inc., United States
a r t i c l e
i n f o
Article history: Received 27 May 2014 Received in revised form 2 July 2014 Accepted 3 July 2014 Available online 11 July 2014 Keywords: Clinical trial Data sharing Transparency Patient privacy
a b s t r a c t The increased demand for transparency and disclosure of data from clinical trials sponsored by pharmaceutical companies poses considerable challenges and opportunities from a statistical perspective. A central issue is the need to protect patient privacy and adhere to Good Clinical and Statistical Practices, while ensuring access to patient-level data from clinical trials to the wider research community. This paper offers options to navigate this dilemma and balance competing priorities, with emphasis on the role of good clinical and statistical practices as proven safeguards for scientific integrity, the importance of adopting best practices for reporting of data from secondary analyses, and the need for optimal collaboration among stakeholders to facilitate data sharing. © 2014 Elsevier Inc. All rights reserved.
1. Introduction Drug development is a complex and costly process that includes the collection, analysis and reporting of data from human subjects under strict protocols. Pharmaceutical companies routinely submit clinical trial results, as well as data, to regulatory agencies for licensing and other promotional activities pertaining to a new drug. In addition, pertinent information on the risks and benefits of medicinal products is communicated by publishing results of such trials in medical journals or presenting them at professional meetings. In recent years, there have been ongoing discussions among various stakeholders on the need to enhance confidence in the reliability of data reported by sponsors of clinical trials [1–6]. Accordingly, several measures have been instituted to enhance transparency through the establishment of registries for clinical trials as well as posting of basic results from such trials in publicly accessible electronic formats [7,8]. With the heightened focus on evidence-based medicine and comparative effectiveness research, there is now growing demand by third parties for enhanced transparency and ⁎ Corresponding author at: Pfizer Inc., 235 East 42nd Street/9-11, New York, NY 10017. E-mail address: demissie.alemayehu@pfizer.com (D. Alemayehu).
http://dx.doi.org/10.1016/j.cct.2014.07.002 1551-7144 © 2014 Elsevier Inc. All rights reserved.
disclosure of clinical trial data, with a view to advancing the field of medicine, accelerating drug development and approval, and protecting public safety. One common theme has been the importance of making widely accessible patient-level data from clinical trials, as well as other aspects of the trial, for the purpose of validating claims of sponsors or executing post-hoc analyses to address other research objectives [9–18]. In a recent draft policy statement, the European Medicines Agency (EMA) highlighted the need for access to clinical trial data and the associated issues [19]. Key elements addressed in the draft policy statement include the significance of enabling public scrutiny and secondary analysis of clinical trials for valid scientific and public health objectives, without compromising patient privacy and informed consent and stifling innovation and investment in biopharmaceutical research and drug development. From the sponsors' perspective, data sharing presents both challenges and opportunities. Availability of patient-level data can help drug-developers to learn from the experiences of others, and use such data to inform trial design and hone their development programs. On the other hand, without proper mechanisms in place, the sharing of patient level data could have the potential to adversely impact public health, compromise patient privacy, and stifle innovation in drug development. The sponsors' views are encapsulated in a joint statement issued by
D. Alemayehu et al. / Contemporary Clinical Trials 39 (2014) 28–33
the European Federation of Pharmaceutical Industries and Associations (EFPIA) and the Pharmaceutical Research and Manufacturers of America (PhRMA) [20]. Incidentally, both EMA and EFPIA/PhRMA, while recognizing the scientific and public health value of data sharing, emphasize the importance of curtailing the risks of the potentially untoward effects of inappropriate secondary analyses and communication of results from such analyses. Viewed from a statistical standpoint, this calls for institutionalization and utilization of best practices to guard against known pitfalls of post-hoc analysis, and an effective management of operational issues associated with data sharing, including data standards, data quality and other infrastructural impediments. In this paper we offer options to address competing priorities of diverse stakeholders, and share statistical perspectives on clinical trial data transparency and disclosure, with special emphasis on the importance of adherence to good clinical and statistical practices as sine qua non for enhancing confidence and establishing trust in evidence-based medicine, and the need for effective collaboration among all parties concerned to tackle procedural and operational issues to optimize the value of shared data to advance medical science without stifling innovation and compromising patient privacy. The paper is organized as follows. In Section 2, we highlight the importance of Good Clinical Practice (GCP) in fostering transparency. Section 3 provided several statistical considerations that are pertinent to the issue. In Sections 4 and 5, we address the need for appropriate infrastructure to ensure optimal collaboration among stakeholders for effective clinical trial data sharing and disclosure. 2. Good clinical practice as a precondition for transparency When it comes to establishing trust in the integrity of clinical trial data, there is no substitute for strict adherence to the principles of Good Clinical Practice (GCP) [21], whose central tenets require that clinical trials be conducted, analyzed and reported with the highest ethical and scientific standards, including ensuring maximum protection of the rights of human subjects, integrity and reproducibility of data, and transparency of study conduct. Key elements of these principles, as enshrined in the International Conference on Harmonization (ICH) of GCP guidelines (E6) [22,23], include: • Adequate qualification of study personnel • Primacy of the rights, privacy and well-being of study subjects • Adherence to highest standards of scientific integrity and quality in the design, conduct, analysis and reporting of study • Adherence to the Declaration of Helsinki • State-of-the-art handling of data to ensure reproducibility. Of particular importance are the safeguards that need to be in place for managing the data, notably the processes for collecting, cleaning and recording the data and monitoring the study. The data management plan should have detailed data handling procedures, including timelines for key activities, database design and validation, monitoring guidelines, data flow and tracking, data entry procedures, query handling, database back-up, database lock, and data archiving and security. An essential requirement for transparency is the need to have
29
documentation for each trial activity, thereby creating a clear audit trail pertaining to actions of study personnel. The documentation effort should include data management Standard Operating Procedures (SOPs), which not only enhance transparency but also ensure consistency and proper communication in the conduct of a study that involves several staff members. Good Statistical Practice (GSP) is an integral component of GCP that is critical to establish trust in the reporting of data by sponsors as well as in the evidence generated by third parties [24]. The hallmarks of GSP include pre-specification of hypotheses and analytical strategy, adequate rationale for study size and power, use of sound statistical tools, implementation of high quality data standards and effective quality control (QC) and quality assurance (QA) plans, and interpretation of results with fair balance. GSP presupposes a data management plan that carefully defines data handling rules, edits checks and dictionaries, as well as processes for trial monitoring, ongoing data review, database release and audit trails. In addition to an analytical approach that is grounded in the application of state-of-the-art methods and procedures, it is a vital facet of GSP to incorporate a clearly specified programming QC/QA plan which includes programming specifications, design and implementation. 3. Statistical considerations with data sharing From statistical perspectives, the benefits of having access to patient-level data are inestimable, ranging from enhancing study design to assessing heterogeneity of treatment effects pooling information across various data sources. However, some of the issues that can impact effective data sharing have statistical import and require methodical approaches to mitigate the consequences. While problems of post-hoc analysis are very well known in the statistical community, other aspects of data sharing, such as de-identification of patient information to preserve patient privacy, or establishing standards and quality metrics for the data to be shared, are emerging areas that concern effective collaboration among stakeholders. For trials intended for new drug application (NDA) or other regulatory submissions, the International Conference on Harmonization (ICH) E9 guidelines [25] provide extensive directions on steps to be taken to ensure transparency, including study design, trial conduct, analytical considerations, evaluation of safety, and reporting of results. The guidelines also address issues that are pertinent to the overall clinical development, including study population, design options to avoid bias, and definitions of study endpoints. While ICH E9 principally focuses on the requirements for the primary analysis and reporting of data, these best practices should also be applied to secondary analyses from such trials. Indeed, the issues associated with secondary analyses tend to be even more complex and require additional measures to ensure the credibility of the findings. When the focus of the secondary analysis is to replicate the sponsor's primary findings, it is critical that the data analyst has thorough familiarity with the original study objectives, planned analytical strategy, and other aspects of the study conduct and data quality. Deviations from the planned analysis must be justified, and may only be acceptable
30
D. Alemayehu et al. / Contemporary Clinical Trials 39 (2014) 28–33
when the intent is to assess the robustness of the sponsor findings to departure from assumptions about models or data (e.g., missing data handling, model specification for primary analyses or data exclusion criteria). However, while the idea of sharing data to independently replicate results is appealing in principle, this may not always be feasible or may even be counterproductive. Conventions pertaining to the initial analysis may have evolved over time and other relevant aspects of the analysis plan may not be reproducible. This is especially true with relatively older historical data in which such important parameters as study endpoints may have changed and the original technology platform for data processing may be outdated. It may also be the case that the data shared in electronic formats does not reflect the source data accurately. Further, in situations where the objective is to pool data across trials conducted by different sponsors and at different time frames, additional complexity may be introduced due to heterogeneity in outcome measures, study characteristics, medical practices, and other aspects of the study designs. Therefore, a reanalysis of such shared data may have the potential to lead to erroneous conclusions, and to compromise public health. In situations where the purpose of data sharing is for confirmatory or exploratory analyses that are outside the scope of the original protocol objectives, best practices should be applied to ensure the integrity of the analyses. For confirmatory analysis, in which the goal is to affirm a pre-specified hypothesis, the methodology for analysis must be clearly stated, including all model assumptions and any sensitivity analysis that may be executed. The results should also be interpreted with caution, in consideration of the fact that the analysis was motivated by preconceived conjecture or prior findings. For exploratory analysis, in which the intent is to generate hypothesis or identify subgroups of interest, the methodology may be limited to descriptive statistics or other approaches that do not heavily rely on traditional inferential statistics [26]. In all cases, extreme caution should be exercised in interpreting the results of such analyses. Wang et al. [27] give a detailed discussion of the requirements for reporting of subgroup analyses in clinical trials. In many cases, the purpose of data sharing may be to perform pooled or meta-analyses, combining data from different trials. Due to the inaccessibility of patient-level data, meta-analysis is customarily performed on the basis of summary statistics compiled from the literature. However, analysis based on aggregate data has several limitations, including inability to adjust for covariate imbalance or to perform extensive sensitivity analyses. In addition reliance on published aggregate data may introduce the so-called publication bias, since there may be a tendency by researchers to publish predominantly positive results. Accordingly, the availability of readily accessible data at the patient level has the potential to mitigate some of the limitations of the standard meta-analytic approaches that rely on summary statistics. Nonetheless, even when patient-level data are available, the analysts should still take appropriate statistical measures to handle heterogeneity, and to ensure combinability of the data from disparate sources [28]. Personal data privacy is another challenge to be contended within data sharing. The U.S. Department of Health and Human Services (HHS) has recently issued the so-called Privacy Rule to implement the requirement of the Health Insurance Portability
and Accountability Act of 1996 (HIPAA) [29]. The Privacy Rule provides several standards to safeguard individuals' health information, while facilitating access to health information “to promote high quality health care and to protect the public's health and well being”. In the United Kingdom, the Department of Health has established the so-called “NHS Confidentiality Code of Practice” relating to the confidentiality of patient information, and the circumstances under which de-identified patient data could be used. The guidance, along with the Data Protection Act of 1998, addresses important legal and ethical aspects of the protection of personal data [30]. Despite the availability of such standards, the practical implementation requires extreme care and caution. The conventional approach of handling the privacy requirements involves removal of personally identifiable information, including recoding identifiers and removing free text verbatim terms, and destroying any linkage between the shared dataset and the original dataset to ensure that re-identification of study participants is not feasible. However, recent studies have shown that the approach may not always be sufficient to preclude re-identification of individuals. In one study, for example, a review of the literature concerning re-identification attacks suggested a fairly large percentage of success [31]. In rare disease research, anonymization of data may be even more difficult to achieve, since the overall pool of patients concerned may be relatively small. Most rare diseases tend to affect vulnerable populations, and the risk of inappropriate data sharing may be relatively more significant to such patients. Failure to protect the data adequately can result in misuse by third parties, including insurance companies or employers, for nefarious purposes. In addition, similar problems may need to be addressed when the data involves rare events that may again require stringent measures to protect patient privacy. Evidently, stringent measures can be put in place to reinforce de-identification of subject-level data. However, these may come with major statistical drawbacks, limiting the ability to link clinical trial data with other data sources to perform pooled analyses, or losing other critical information pertinent to the hypotheses of interest. In addition, the absence of individual level covariates may constrain efforts to implement models intended to adjust for important confounders. As there is no standard method for addressing these issues, variation in the specific implementation across multiple sponsors may increase the complexity of performing such analyses. 4. Infrastructural requirements Effective data sharing among disparate stakeholders requires well-developed infrastructure that ensures security and enables easy transfer and aggregation of data. In particular, if the data is to be accessed remotely, there should be adequate system security, with reliable password protection, firewall configuration, network safety measures, encryption, and other system controls. Use of reliable data warehouses can simplify data aggregation and ensure data integrity. This is particularly needed when the goal is to synthesize data from different sources. This involves converting the shared data into a standard data model, e.g., Clinical Data Interchange Standards Consortium (CDISC), so that cross-study analyses could be executed. Data warehousing
D. Alemayehu et al. / Contemporary Clinical Trials 39 (2014) 28–33
systems, however, present their own security issues. In particular, if the primary reason of the data warehouse is to enable exploratory analysis to generate hypotheses pooling data across trials, often done using modern data mining techniques, the need to protect patient privacy may limit the aggregation of the data for the intended purpose. When encryption is used to transmit data over an insecure medium, it may not be possible to implement varying levels of access. Further, data warehouses may be easy targets of hackers, as the data stored in the warehouses may often be a focus of attack. As the volume of data increases, traditional data warehouses may need to be replaced by modern platforms that permit scalability and ensure greater data security. New analytic and data warehousing platforms, such as Hadoop Distributed File System (HDFS), have not yet made inroads into clinical data processing. However, the new technologies may not have the attractive features of more familiar relational database management system (RDBMS) for data sharing use in the immediate future. The infrastructure should also permit accessibility of documents pertinent to the study in question. This typically includes study protocols, analysis plans, data convention documents, study reports and other publications. Incidentally, alternative approaches are currently being followed to provide access to participant level clinical trials data [32,33]. These approaches have been developed to address the potential pitfalls described earlier. One model is provided through SAS and addresses many of these concerns by providing data access through a portal, rather than by transferring data to researchers. The platform used provides several capabilities, including: a data request review step; secure access to data with the ability to limit extracts from the system; access to analytical programs including SAS and R; and access to relevant documents necessary to understand the study design, analytical methods, and data properties. Access is granted subject to a data sharing agreement, which presupposes a written request from a third party researcher comprising a research proposal and requiring prior review by the data generator. This approach is being used by individual companies, and by multiple companies through a multi-sponsor platform. The multi-sponsor platform facilitates research requests that require data from more than one sponsor. Another model is the Yale University Open Data Access (YODA) Project, which involves a partnership between Yale and at least two pharmaceutical companies [34]. Researchers with a research agenda are required to apply to the program to request access to clinical trial data. Once the request is approved, the researchers will get access to detailed information to the level of the individual patient, while the patients' privacy is protected.
5. Shared responsibilities To ensure proper use of clinical trial data in the service of public health, both access requesters and sponsors should be held to the same standard of excellence. As such, the value of secondary use of clinical data can be optimized only when there is a realization by all parties concerned that the entire venture is a shared responsibility.
31
It is critical that requesters of access to clinical data have the required expertise and sophistication to adhere to accepted standards for GCP and GSP. First and foremost, they need to be thoroughly familiar with the original research objectives, study protocol, data standards and analytical strategies. The requesters would need to provide a well-formulated research plan that includes the scientific objectives, data and studies required to address the hypotheses of interest, analysis strategy, and publication plans. They should also have adequate infrastructure to access and analyze data, and trained programming and statistical resources to execute the analyses. Further, in the spirit of transparency, they should have open communication with the sponsor, and be prepared to share and discuss their approaches and findings with the sponsor. There should be requirement that secondary researchers provide research protocols, analysis plans, and communication plans as part of the data access request. It is also essential to develop mechanisms for assuring that this information is posted publicly and that the results are published for broader dissemination regardless of whether the results are successful or unsuccessful in proving the research hypothesis. This is an invaluable opportunity for medical journals and other institutions to spearhead so that transparency standards are established for secondary researchers mirroring the ones that exist for clinical trial sponsors. On the other hand, the sponsor should have a process in place specifying the requirements for data access, including the steps to be taken for making the requests, the review and approval procedures, and any other prerequisites that the requester needs to fulfill. The sponsor should also be prepared to make available all relevant documents to the requester in a controlled access environment, including protocols with any amendments, annotated case report form, statistical analysis plans, data conventions and specifications, anonymized raw as well as analysis-ready data, and all pertinent study publications and reports. The sponsor should have a well-established and reliable medium to provide access to raw and derived data, and provide resources to support the requests. It may also be essential to have dedicated personnel to help requesters navigate the data. Some sponsors have already taken steps to facilitate data sharing using a site to request access to patient level data and supporting documents [35]. The approval process should be transparent, with well established criteria to assess the scientific rationale, the adequacy of requested data to address the research hypothesis, and the qualification of personnel to execute the desired analyses. To evaluate the scientific merit of a request, it may be appropriate to have an independent review panel consisting of members with diverse backgrounds, including those from the medical, statistical as well as ethical professions. In addition, attention should be paid to measures that will be taken to ensure patient privacy, and to mange any conflict of interest. For the purpose of full disclosure, it is important to have a mechanism that permits access to the list of all requests approved and rejected. In some situations it may be necessary to have data sharing agreements between the sponsor and the requester to ensure that data will be used only for the intended use, and that patient privacy and confidentiality would be respected. The agreement should also have provisions for the requester to inform the sponsor and regulatory agencies about any significant safety related findings.
32
D. Alemayehu et al. / Contemporary Clinical Trials 39 (2014) 28–33
6. Conclusion There is now a general consensus about the need for increased access to patient level data from clinical trials to promote public health, advance medical science and share experience among drug developers, without compromising patient privacy and stifling innovation. In this paper, we have addressed the overarching reasons for data sharing, and the safeguards that must be instituted to ensure protection of patient privacy, the integrity of scientific research, and promotion of investment in biomedical research. A central argument from the perspectives of the sponsors is that the need for access to clinical data should balance legitimate scientific objectives against regulatory, logistical and innovation concerns. There are compelling arguments that any effort to implement a data sharing or disclosure initiative should have at its core the patients' privacy and informed consent, as that is likely to violate their trust and discourage their participation in future trials. While substantial progress has been made in managing the competing priorities of researchers who wish to get access to patient-level data, and those of the sponsors of clinical trials, there are still several operational and logistical issues that need to be resolved. The relative roles and responsibilities of the various stakeholders need to be clarified, and the issue of data warehousing and security should be addressed. In this regard, effective collaboration among various stakeholders is essential to ensure a reliable infrastructure that facilitates data sharing while ensuring the integrity of the scientific research. The effort should also serve to foster investment in biomedical research, rather than to stifle innovation. With proper use and implementation, data sharing can certainly help to enhance collaboration between sponsors and external researchers, thereby accelerating biomedical research. Indeed, this would be an essential vehicle in building confidence in the integrity of clinical research sponsored by pharmaceutical companies, and in ensuring improved access to medicines through better understanding of the risks and benefits of treatments by patients. Since data sharing involves costs, consideration should be given to alternative approaches for managing them. The costs are typically associated with establishing the infrastructure, standardization of data, and human resources. To the extent possible, recent advances in technology pertaining to data management, processing and warehousing should be leveraged. Retrospective data standardization is generally prohibitively expensive; so, a concerted effort should be made to establish a framework for data standards for prospective use by all entities conducting clinical trials. Notably, the future use of the Accelerated Clinical Data Interchange Standards Consortium (CDISC) standards, increasingly expected by regulatory agencies, will help drive uniformity of data across sponsors. Participating companies may also incur additional cost to provide technical and other study related support to users of the shared data. To minimize the burden on individual companies, a collaborative platform may help provide the necessary support in a cost-effective manner. There should also be a requirement for cost sharing, including the costs of data mapping and infrastructure maintenance, among all stakeholders. In addition, there are opportunity costs that are associated with the effort that could have been used for other
endeavors. It is therefore worthwhile to factor this aspect in the planning stage to realize the full potential of data sharing [36]. It should be underscored that when it comes to guaranteeing transparency, there is no substitute for good clinical and statistical principles. Trust in the value of data from clinical trials, whether the analysis is performed by sponsors or other researchers, can be established if all the parties adhere to best practices for the collection, analysis and reporting of data. Furthermore, it is imperative that secondary data analysts be held to the same standards of excellence as sponsors to ensure the scientific validity of their findings. Acknowledgments The authors would like to thank Roma Tretiak and Caroline Stockwell of Pfizer Inc. for their helpful comments. References [1] Korieth K. Clinical trial transparency efforts multiply. CenterWatch, 12; 2005 1–13. [2] Godlee F. An international standard for disclosure of clinical trial information. BMJ 2006;332:1107–8. [3] Irwin RS. Clinical trial registration promotes patient protection and benefit, advances the trust of everyone, and is required. Chest 2007;131:639–41. [4] O'Halloran R. Transparency in disclosure of clinical trial information. The Write Stuff, 15; 2006 15–7. [5] Doshi P, Jefferson T, Del Mar C. The imperative to share clinical study reports: recommendations from the Tamiflu experience. PLoS Med 2012;9(4):e1001201. http://dx.doi.org/10.1371/journal.pmed.1001201. [6] Viergever RF, Ghersi D. The quality of registration of clinical trials. PLoS One 2011;6(2):e14701. http://dx.doi.org/10.1371/journal.pone.0014701. [7] Fact sheet, ClinicalTrials.gov. U.S. National Library of Medicine; May 3, 2011. http://www.nlm.nih.gov/pubs/factsheets/clintrial.html; July 2 2014. [8] Food and Drug Administration Amendments Act of 2007 (FDAAA 801). http://www.gpo.gov/fdsys/pkg/PLAW-110publ85/pdf/PLAW-110publ85. pdf#page=82; February 24 2014. [9] Berlin JA, Morris S, Rockhold F, Askie L, Ghersi D, Waldstreicher J. Bumps and bridges on the road to responsible sharing of clinical trial data. Clin Trials February 2014;11:15–8. [10] Mello MM, Francer JK, Wilenzick M, Teden P, Bierer BE, Barnes M. Preparing for responsible sharing of clinical trial data. N Engl J Med 2013;369:1651–8. [11] Eichler H-G, Abadie E, Breckenridge A, Leufkens H, Rasi G. Open clinical trial data for all? A view from regulators. PLoS Med 2012;9(4): e1001202. http://dx.doi.org/10.1371/journal.pmed.1001202. [12] Ross JS, Lehman R, Gross CP. The importance of clinical trial data sharing: toward more open science. Circ Cardiovasc Qual Outcomes 2012;5(2):238–40. [13] Gøtzsche PC. Strengthening and opening up health research by sharing our raw data. Circ Cardiovasc Qual Outcomes 2012;5:236–7. [14] Kirwan JR. Making original data from clinical studies available for alternative analysis. J Rheumatol 1997;24:822–5. [15] Vickers AJ. Whose data set is it anyway? Sharing raw data from randomized trials. Trials 2006;7:15. [16] Hrynaszkiewicz I, Altman DG. Towards agreement on best practice for publishing raw clinical trial data. Trials 2009;10:17. [17] Institute of Medicine. Ensuring the integrity, accessibility, and stewardship of research data in the digital age. Washington, DC: National Academy Press; 2009. [18] Walport M, Brest P. Sharing research data to improve public health. Lancet 2011;377:537–53. [19] European Medicines Agency. Publication and access to clinical trial data. http://www.ema.europa.eu/docs/en_GB/document_library/Other/2013/06/ WC500144730.pdf; February 24 2014. [20] European Federation of Pharmaceutical Industries and Associations (EFPIA), Pharmaceutical Research and Manufacturers of America (PhRMA). Principles for responsible clinical trial data sharing. http:// www.phrma.org/phrmapedia/responsible-clinical-trial-data-sharing# sthash.je1WkJR5.dpuf; February 24 2014. [21] Van Dongen AJ. Good clinical practice, a transparent way of life. A review. Comput Med Imaging Graph 2001;25:213–6.
D. Alemayehu et al. / Contemporary Clinical Trials 39 (2014) 28–33 [22] EMEA ICH Topic E 6 (R1): guideline for good clinical practice. http:// ethikkommission.meduniwien.ac.at/fileadmin/ethik/media/dokumente/ rechtsgrundlagen/GCP.pdf; February 24 2014. [23] Food Drug Administration (FDA). International conference on harmonization, good clinical practice: consolidated guidelines. Fed Regist 1997;62:25692–709. [24] PSI Statisticians in the Pharmaceutical Industry. Guidelines for standard operating procedures for good statistical practice in clinical research. Drug Inf J 1998;32:665–82. [25] ICH E9, statistical principles for clinical trials. http://www.ich.org/fileadmin/ Public_Web_Site/ICH_Products/Guidelines/Efficacy/E9/Step4/E9_Guideline. pdf; February 24 2014. [26] Foster JC, Taylor JMC, Ruberg SJ. Subgroup identification from randomized clinical trial data. Stat Med 2011;30:2867–80. [27] Wang R, Lagakos SW, Ware JH, Hunter DJ, Drazen JM. Statistics in medicine—reporting of subgroup analyses in clinical trials. N Engl J Med 2007;357(21):2189–94. [28] Pocock SJ, Assmann SE, Enos LE, Kasten LE. Subgroup analysis, covariate adjustment and baseline comparisons in clinical trial reporting: current practice and problems. Stat Med 2002;21(19):2917–30. [29] US Department of Health and Human Services. Summary of the HIPAA privacy rule. http://www.hhs.gov/ocr/privacy/hipaa/understanding/ summary/privacysummary.pdf; February 24 2014.
33
[30] United Kingdom Department of Health, Confidentiality Unit, IPU. NHS Confidentiality Code of Practice; 2003. http://webarchive.nationalarchives. gov.uk/20130124004052/http://www.dh.gov.uk/prod_consum_dh/groups/ dh_digitalassets/@dh/@en/documents/digitalasset/dh_4069254.pdf; July 2 2014. [31] Gymrek M, McGuire AL, Golan D, Halperin E, Erlich Y. Identifying personal genomes by surname inference. Science 2013;339:321–4. [32] http://www.pfizer.com/news/press-release/press-release-detail/pfizer_ expands_clinical_trial_data_access_policy_and_launches_data_access_portal; February 24 2014. [33] Nisen P, Rockhold F. Access to patient-level data from GlaxoSmithKline clinical trials. N Engl J Med 2013;369:475–8. [34] http://www.bloomberg.com/news/print/2014-01-30/j-j-sets-drug-datafree-in-yoda-collaboration-with-yale.html; February 24 2014. [35] https://www.clinicalstudydatarequest.com/Default.aspx; February 24 2014. [36] Wilhelm EE, Oster E, Shoulson I. Approaches and costs for sharing clinical research data. JAMA 2014;311(12):1201–2.
Contents lists available at ScienceDirect
Contemporary Clinical Trials journal homepage: www.elsevier.com/locate/conclintrial
Perspectives on clinical trial data transparency and disclosure Demissie Alemayehu ⁎, Richard J. Anziano, Marcia Levenstein Pfizer Inc., United States
a r t i c l e
i n f o
Article history: Received 27 May 2014 Received in revised form 2 July 2014 Accepted 3 July 2014 Available online 11 July 2014 Keywords: Clinical trial Data sharing Transparency Patient privacy
a b s t r a c t The increased demand for transparency and disclosure of data from clinical trials sponsored by pharmaceutical companies poses considerable challenges and opportunities from a statistical perspective. A central issue is the need to protect patient privacy and adhere to Good Clinical and Statistical Practices, while ensuring access to patient-level data from clinical trials to the wider research community. This paper offers options to navigate this dilemma and balance competing priorities, with emphasis on the role of good clinical and statistical practices as proven safeguards for scientific integrity, the importance of adopting best practices for reporting of data from secondary analyses, and the need for optimal collaboration among stakeholders to facilitate data sharing. © 2014 Elsevier Inc. All rights reserved.
1. Introduction Drug development is a complex and costly process that includes the collection, analysis and reporting of data from human subjects under strict protocols. Pharmaceutical companies routinely submit clinical trial results, as well as data, to regulatory agencies for licensing and other promotional activities pertaining to a new drug. In addition, pertinent information on the risks and benefits of medicinal products is communicated by publishing results of such trials in medical journals or presenting them at professional meetings. In recent years, there have been ongoing discussions among various stakeholders on the need to enhance confidence in the reliability of data reported by sponsors of clinical trials [1–6]. Accordingly, several measures have been instituted to enhance transparency through the establishment of registries for clinical trials as well as posting of basic results from such trials in publicly accessible electronic formats [7,8]. With the heightened focus on evidence-based medicine and comparative effectiveness research, there is now growing demand by third parties for enhanced transparency and ⁎ Corresponding author at: Pfizer Inc., 235 East 42nd Street/9-11, New York, NY 10017. E-mail address: demissie.alemayehu@pfizer.com (D. Alemayehu).
http://dx.doi.org/10.1016/j.cct.2014.07.002 1551-7144 © 2014 Elsevier Inc. All rights reserved.
disclosure of clinical trial data, with a view to advancing the field of medicine, accelerating drug development and approval, and protecting public safety. One common theme has been the importance of making widely accessible patient-level data from clinical trials, as well as other aspects of the trial, for the purpose of validating claims of sponsors or executing post-hoc analyses to address other research objectives [9–18]. In a recent draft policy statement, the European Medicines Agency (EMA) highlighted the need for access to clinical trial data and the associated issues [19]. Key elements addressed in the draft policy statement include the significance of enabling public scrutiny and secondary analysis of clinical trials for valid scientific and public health objectives, without compromising patient privacy and informed consent and stifling innovation and investment in biopharmaceutical research and drug development. From the sponsors' perspective, data sharing presents both challenges and opportunities. Availability of patient-level data can help drug-developers to learn from the experiences of others, and use such data to inform trial design and hone their development programs. On the other hand, without proper mechanisms in place, the sharing of patient level data could have the potential to adversely impact public health, compromise patient privacy, and stifle innovation in drug development. The sponsors' views are encapsulated in a joint statement issued by
D. Alemayehu et al. / Contemporary Clinical Trials 39 (2014) 28–33
the European Federation of Pharmaceutical Industries and Associations (EFPIA) and the Pharmaceutical Research and Manufacturers of America (PhRMA) [20]. Incidentally, both EMA and EFPIA/PhRMA, while recognizing the scientific and public health value of data sharing, emphasize the importance of curtailing the risks of the potentially untoward effects of inappropriate secondary analyses and communication of results from such analyses. Viewed from a statistical standpoint, this calls for institutionalization and utilization of best practices to guard against known pitfalls of post-hoc analysis, and an effective management of operational issues associated with data sharing, including data standards, data quality and other infrastructural impediments. In this paper we offer options to address competing priorities of diverse stakeholders, and share statistical perspectives on clinical trial data transparency and disclosure, with special emphasis on the importance of adherence to good clinical and statistical practices as sine qua non for enhancing confidence and establishing trust in evidence-based medicine, and the need for effective collaboration among all parties concerned to tackle procedural and operational issues to optimize the value of shared data to advance medical science without stifling innovation and compromising patient privacy. The paper is organized as follows. In Section 2, we highlight the importance of Good Clinical Practice (GCP) in fostering transparency. Section 3 provided several statistical considerations that are pertinent to the issue. In Sections 4 and 5, we address the need for appropriate infrastructure to ensure optimal collaboration among stakeholders for effective clinical trial data sharing and disclosure. 2. Good clinical practice as a precondition for transparency When it comes to establishing trust in the integrity of clinical trial data, there is no substitute for strict adherence to the principles of Good Clinical Practice (GCP) [21], whose central tenets require that clinical trials be conducted, analyzed and reported with the highest ethical and scientific standards, including ensuring maximum protection of the rights of human subjects, integrity and reproducibility of data, and transparency of study conduct. Key elements of these principles, as enshrined in the International Conference on Harmonization (ICH) of GCP guidelines (E6) [22,23], include: • Adequate qualification of study personnel • Primacy of the rights, privacy and well-being of study subjects • Adherence to highest standards of scientific integrity and quality in the design, conduct, analysis and reporting of study • Adherence to the Declaration of Helsinki • State-of-the-art handling of data to ensure reproducibility. Of particular importance are the safeguards that need to be in place for managing the data, notably the processes for collecting, cleaning and recording the data and monitoring the study. The data management plan should have detailed data handling procedures, including timelines for key activities, database design and validation, monitoring guidelines, data flow and tracking, data entry procedures, query handling, database back-up, database lock, and data archiving and security. An essential requirement for transparency is the need to have
29
documentation for each trial activity, thereby creating a clear audit trail pertaining to actions of study personnel. The documentation effort should include data management Standard Operating Procedures (SOPs), which not only enhance transparency but also ensure consistency and proper communication in the conduct of a study that involves several staff members. Good Statistical Practice (GSP) is an integral component of GCP that is critical to establish trust in the reporting of data by sponsors as well as in the evidence generated by third parties [24]. The hallmarks of GSP include pre-specification of hypotheses and analytical strategy, adequate rationale for study size and power, use of sound statistical tools, implementation of high quality data standards and effective quality control (QC) and quality assurance (QA) plans, and interpretation of results with fair balance. GSP presupposes a data management plan that carefully defines data handling rules, edits checks and dictionaries, as well as processes for trial monitoring, ongoing data review, database release and audit trails. In addition to an analytical approach that is grounded in the application of state-of-the-art methods and procedures, it is a vital facet of GSP to incorporate a clearly specified programming QC/QA plan which includes programming specifications, design and implementation. 3. Statistical considerations with data sharing From statistical perspectives, the benefits of having access to patient-level data are inestimable, ranging from enhancing study design to assessing heterogeneity of treatment effects pooling information across various data sources. However, some of the issues that can impact effective data sharing have statistical import and require methodical approaches to mitigate the consequences. While problems of post-hoc analysis are very well known in the statistical community, other aspects of data sharing, such as de-identification of patient information to preserve patient privacy, or establishing standards and quality metrics for the data to be shared, are emerging areas that concern effective collaboration among stakeholders. For trials intended for new drug application (NDA) or other regulatory submissions, the International Conference on Harmonization (ICH) E9 guidelines [25] provide extensive directions on steps to be taken to ensure transparency, including study design, trial conduct, analytical considerations, evaluation of safety, and reporting of results. The guidelines also address issues that are pertinent to the overall clinical development, including study population, design options to avoid bias, and definitions of study endpoints. While ICH E9 principally focuses on the requirements for the primary analysis and reporting of data, these best practices should also be applied to secondary analyses from such trials. Indeed, the issues associated with secondary analyses tend to be even more complex and require additional measures to ensure the credibility of the findings. When the focus of the secondary analysis is to replicate the sponsor's primary findings, it is critical that the data analyst has thorough familiarity with the original study objectives, planned analytical strategy, and other aspects of the study conduct and data quality. Deviations from the planned analysis must be justified, and may only be acceptable
30
D. Alemayehu et al. / Contemporary Clinical Trials 39 (2014) 28–33
when the intent is to assess the robustness of the sponsor findings to departure from assumptions about models or data (e.g., missing data handling, model specification for primary analyses or data exclusion criteria). However, while the idea of sharing data to independently replicate results is appealing in principle, this may not always be feasible or may even be counterproductive. Conventions pertaining to the initial analysis may have evolved over time and other relevant aspects of the analysis plan may not be reproducible. This is especially true with relatively older historical data in which such important parameters as study endpoints may have changed and the original technology platform for data processing may be outdated. It may also be the case that the data shared in electronic formats does not reflect the source data accurately. Further, in situations where the objective is to pool data across trials conducted by different sponsors and at different time frames, additional complexity may be introduced due to heterogeneity in outcome measures, study characteristics, medical practices, and other aspects of the study designs. Therefore, a reanalysis of such shared data may have the potential to lead to erroneous conclusions, and to compromise public health. In situations where the purpose of data sharing is for confirmatory or exploratory analyses that are outside the scope of the original protocol objectives, best practices should be applied to ensure the integrity of the analyses. For confirmatory analysis, in which the goal is to affirm a pre-specified hypothesis, the methodology for analysis must be clearly stated, including all model assumptions and any sensitivity analysis that may be executed. The results should also be interpreted with caution, in consideration of the fact that the analysis was motivated by preconceived conjecture or prior findings. For exploratory analysis, in which the intent is to generate hypothesis or identify subgroups of interest, the methodology may be limited to descriptive statistics or other approaches that do not heavily rely on traditional inferential statistics [26]. In all cases, extreme caution should be exercised in interpreting the results of such analyses. Wang et al. [27] give a detailed discussion of the requirements for reporting of subgroup analyses in clinical trials. In many cases, the purpose of data sharing may be to perform pooled or meta-analyses, combining data from different trials. Due to the inaccessibility of patient-level data, meta-analysis is customarily performed on the basis of summary statistics compiled from the literature. However, analysis based on aggregate data has several limitations, including inability to adjust for covariate imbalance or to perform extensive sensitivity analyses. In addition reliance on published aggregate data may introduce the so-called publication bias, since there may be a tendency by researchers to publish predominantly positive results. Accordingly, the availability of readily accessible data at the patient level has the potential to mitigate some of the limitations of the standard meta-analytic approaches that rely on summary statistics. Nonetheless, even when patient-level data are available, the analysts should still take appropriate statistical measures to handle heterogeneity, and to ensure combinability of the data from disparate sources [28]. Personal data privacy is another challenge to be contended within data sharing. The U.S. Department of Health and Human Services (HHS) has recently issued the so-called Privacy Rule to implement the requirement of the Health Insurance Portability
and Accountability Act of 1996 (HIPAA) [29]. The Privacy Rule provides several standards to safeguard individuals' health information, while facilitating access to health information “to promote high quality health care and to protect the public's health and well being”. In the United Kingdom, the Department of Health has established the so-called “NHS Confidentiality Code of Practice” relating to the confidentiality of patient information, and the circumstances under which de-identified patient data could be used. The guidance, along with the Data Protection Act of 1998, addresses important legal and ethical aspects of the protection of personal data [30]. Despite the availability of such standards, the practical implementation requires extreme care and caution. The conventional approach of handling the privacy requirements involves removal of personally identifiable information, including recoding identifiers and removing free text verbatim terms, and destroying any linkage between the shared dataset and the original dataset to ensure that re-identification of study participants is not feasible. However, recent studies have shown that the approach may not always be sufficient to preclude re-identification of individuals. In one study, for example, a review of the literature concerning re-identification attacks suggested a fairly large percentage of success [31]. In rare disease research, anonymization of data may be even more difficult to achieve, since the overall pool of patients concerned may be relatively small. Most rare diseases tend to affect vulnerable populations, and the risk of inappropriate data sharing may be relatively more significant to such patients. Failure to protect the data adequately can result in misuse by third parties, including insurance companies or employers, for nefarious purposes. In addition, similar problems may need to be addressed when the data involves rare events that may again require stringent measures to protect patient privacy. Evidently, stringent measures can be put in place to reinforce de-identification of subject-level data. However, these may come with major statistical drawbacks, limiting the ability to link clinical trial data with other data sources to perform pooled analyses, or losing other critical information pertinent to the hypotheses of interest. In addition, the absence of individual level covariates may constrain efforts to implement models intended to adjust for important confounders. As there is no standard method for addressing these issues, variation in the specific implementation across multiple sponsors may increase the complexity of performing such analyses. 4. Infrastructural requirements Effective data sharing among disparate stakeholders requires well-developed infrastructure that ensures security and enables easy transfer and aggregation of data. In particular, if the data is to be accessed remotely, there should be adequate system security, with reliable password protection, firewall configuration, network safety measures, encryption, and other system controls. Use of reliable data warehouses can simplify data aggregation and ensure data integrity. This is particularly needed when the goal is to synthesize data from different sources. This involves converting the shared data into a standard data model, e.g., Clinical Data Interchange Standards Consortium (CDISC), so that cross-study analyses could be executed. Data warehousing
D. Alemayehu et al. / Contemporary Clinical Trials 39 (2014) 28–33
systems, however, present their own security issues. In particular, if the primary reason of the data warehouse is to enable exploratory analysis to generate hypotheses pooling data across trials, often done using modern data mining techniques, the need to protect patient privacy may limit the aggregation of the data for the intended purpose. When encryption is used to transmit data over an insecure medium, it may not be possible to implement varying levels of access. Further, data warehouses may be easy targets of hackers, as the data stored in the warehouses may often be a focus of attack. As the volume of data increases, traditional data warehouses may need to be replaced by modern platforms that permit scalability and ensure greater data security. New analytic and data warehousing platforms, such as Hadoop Distributed File System (HDFS), have not yet made inroads into clinical data processing. However, the new technologies may not have the attractive features of more familiar relational database management system (RDBMS) for data sharing use in the immediate future. The infrastructure should also permit accessibility of documents pertinent to the study in question. This typically includes study protocols, analysis plans, data convention documents, study reports and other publications. Incidentally, alternative approaches are currently being followed to provide access to participant level clinical trials data [32,33]. These approaches have been developed to address the potential pitfalls described earlier. One model is provided through SAS and addresses many of these concerns by providing data access through a portal, rather than by transferring data to researchers. The platform used provides several capabilities, including: a data request review step; secure access to data with the ability to limit extracts from the system; access to analytical programs including SAS and R; and access to relevant documents necessary to understand the study design, analytical methods, and data properties. Access is granted subject to a data sharing agreement, which presupposes a written request from a third party researcher comprising a research proposal and requiring prior review by the data generator. This approach is being used by individual companies, and by multiple companies through a multi-sponsor platform. The multi-sponsor platform facilitates research requests that require data from more than one sponsor. Another model is the Yale University Open Data Access (YODA) Project, which involves a partnership between Yale and at least two pharmaceutical companies [34]. Researchers with a research agenda are required to apply to the program to request access to clinical trial data. Once the request is approved, the researchers will get access to detailed information to the level of the individual patient, while the patients' privacy is protected.
5. Shared responsibilities To ensure proper use of clinical trial data in the service of public health, both access requesters and sponsors should be held to the same standard of excellence. As such, the value of secondary use of clinical data can be optimized only when there is a realization by all parties concerned that the entire venture is a shared responsibility.
31
It is critical that requesters of access to clinical data have the required expertise and sophistication to adhere to accepted standards for GCP and GSP. First and foremost, they need to be thoroughly familiar with the original research objectives, study protocol, data standards and analytical strategies. The requesters would need to provide a well-formulated research plan that includes the scientific objectives, data and studies required to address the hypotheses of interest, analysis strategy, and publication plans. They should also have adequate infrastructure to access and analyze data, and trained programming and statistical resources to execute the analyses. Further, in the spirit of transparency, they should have open communication with the sponsor, and be prepared to share and discuss their approaches and findings with the sponsor. There should be requirement that secondary researchers provide research protocols, analysis plans, and communication plans as part of the data access request. It is also essential to develop mechanisms for assuring that this information is posted publicly and that the results are published for broader dissemination regardless of whether the results are successful or unsuccessful in proving the research hypothesis. This is an invaluable opportunity for medical journals and other institutions to spearhead so that transparency standards are established for secondary researchers mirroring the ones that exist for clinical trial sponsors. On the other hand, the sponsor should have a process in place specifying the requirements for data access, including the steps to be taken for making the requests, the review and approval procedures, and any other prerequisites that the requester needs to fulfill. The sponsor should also be prepared to make available all relevant documents to the requester in a controlled access environment, including protocols with any amendments, annotated case report form, statistical analysis plans, data conventions and specifications, anonymized raw as well as analysis-ready data, and all pertinent study publications and reports. The sponsor should have a well-established and reliable medium to provide access to raw and derived data, and provide resources to support the requests. It may also be essential to have dedicated personnel to help requesters navigate the data. Some sponsors have already taken steps to facilitate data sharing using a site to request access to patient level data and supporting documents [35]. The approval process should be transparent, with well established criteria to assess the scientific rationale, the adequacy of requested data to address the research hypothesis, and the qualification of personnel to execute the desired analyses. To evaluate the scientific merit of a request, it may be appropriate to have an independent review panel consisting of members with diverse backgrounds, including those from the medical, statistical as well as ethical professions. In addition, attention should be paid to measures that will be taken to ensure patient privacy, and to mange any conflict of interest. For the purpose of full disclosure, it is important to have a mechanism that permits access to the list of all requests approved and rejected. In some situations it may be necessary to have data sharing agreements between the sponsor and the requester to ensure that data will be used only for the intended use, and that patient privacy and confidentiality would be respected. The agreement should also have provisions for the requester to inform the sponsor and regulatory agencies about any significant safety related findings.
32
D. Alemayehu et al. / Contemporary Clinical Trials 39 (2014) 28–33
6. Conclusion There is now a general consensus about the need for increased access to patient level data from clinical trials to promote public health, advance medical science and share experience among drug developers, without compromising patient privacy and stifling innovation. In this paper, we have addressed the overarching reasons for data sharing, and the safeguards that must be instituted to ensure protection of patient privacy, the integrity of scientific research, and promotion of investment in biomedical research. A central argument from the perspectives of the sponsors is that the need for access to clinical data should balance legitimate scientific objectives against regulatory, logistical and innovation concerns. There are compelling arguments that any effort to implement a data sharing or disclosure initiative should have at its core the patients' privacy and informed consent, as that is likely to violate their trust and discourage their participation in future trials. While substantial progress has been made in managing the competing priorities of researchers who wish to get access to patient-level data, and those of the sponsors of clinical trials, there are still several operational and logistical issues that need to be resolved. The relative roles and responsibilities of the various stakeholders need to be clarified, and the issue of data warehousing and security should be addressed. In this regard, effective collaboration among various stakeholders is essential to ensure a reliable infrastructure that facilitates data sharing while ensuring the integrity of the scientific research. The effort should also serve to foster investment in biomedical research, rather than to stifle innovation. With proper use and implementation, data sharing can certainly help to enhance collaboration between sponsors and external researchers, thereby accelerating biomedical research. Indeed, this would be an essential vehicle in building confidence in the integrity of clinical research sponsored by pharmaceutical companies, and in ensuring improved access to medicines through better understanding of the risks and benefits of treatments by patients. Since data sharing involves costs, consideration should be given to alternative approaches for managing them. The costs are typically associated with establishing the infrastructure, standardization of data, and human resources. To the extent possible, recent advances in technology pertaining to data management, processing and warehousing should be leveraged. Retrospective data standardization is generally prohibitively expensive; so, a concerted effort should be made to establish a framework for data standards for prospective use by all entities conducting clinical trials. Notably, the future use of the Accelerated Clinical Data Interchange Standards Consortium (CDISC) standards, increasingly expected by regulatory agencies, will help drive uniformity of data across sponsors. Participating companies may also incur additional cost to provide technical and other study related support to users of the shared data. To minimize the burden on individual companies, a collaborative platform may help provide the necessary support in a cost-effective manner. There should also be a requirement for cost sharing, including the costs of data mapping and infrastructure maintenance, among all stakeholders. In addition, there are opportunity costs that are associated with the effort that could have been used for other
endeavors. It is therefore worthwhile to factor this aspect in the planning stage to realize the full potential of data sharing [36]. It should be underscored that when it comes to guaranteeing transparency, there is no substitute for good clinical and statistical principles. Trust in the value of data from clinical trials, whether the analysis is performed by sponsors or other researchers, can be established if all the parties adhere to best practices for the collection, analysis and reporting of data. Furthermore, it is imperative that secondary data analysts be held to the same standards of excellence as sponsors to ensure the scientific validity of their findings. Acknowledgments The authors would like to thank Roma Tretiak and Caroline Stockwell of Pfizer Inc. for their helpful comments. References [1] Korieth K. Clinical trial transparency efforts multiply. CenterWatch, 12; 2005 1–13. [2] Godlee F. An international standard for disclosure of clinical trial information. BMJ 2006;332:1107–8. [3] Irwin RS. Clinical trial registration promotes patient protection and benefit, advances the trust of everyone, and is required. Chest 2007;131:639–41. [4] O'Halloran R. Transparency in disclosure of clinical trial information. The Write Stuff, 15; 2006 15–7. [5] Doshi P, Jefferson T, Del Mar C. The imperative to share clinical study reports: recommendations from the Tamiflu experience. PLoS Med 2012;9(4):e1001201. http://dx.doi.org/10.1371/journal.pmed.1001201. [6] Viergever RF, Ghersi D. The quality of registration of clinical trials. PLoS One 2011;6(2):e14701. http://dx.doi.org/10.1371/journal.pone.0014701. [7] Fact sheet, ClinicalTrials.gov. U.S. National Library of Medicine; May 3, 2011. http://www.nlm.nih.gov/pubs/factsheets/clintrial.html; July 2 2014. [8] Food and Drug Administration Amendments Act of 2007 (FDAAA 801). http://www.gpo.gov/fdsys/pkg/PLAW-110publ85/pdf/PLAW-110publ85. pdf#page=82; February 24 2014. [9] Berlin JA, Morris S, Rockhold F, Askie L, Ghersi D, Waldstreicher J. Bumps and bridges on the road to responsible sharing of clinical trial data. Clin Trials February 2014;11:15–8. [10] Mello MM, Francer JK, Wilenzick M, Teden P, Bierer BE, Barnes M. Preparing for responsible sharing of clinical trial data. N Engl J Med 2013;369:1651–8. [11] Eichler H-G, Abadie E, Breckenridge A, Leufkens H, Rasi G. Open clinical trial data for all? A view from regulators. PLoS Med 2012;9(4): e1001202. http://dx.doi.org/10.1371/journal.pmed.1001202. [12] Ross JS, Lehman R, Gross CP. The importance of clinical trial data sharing: toward more open science. Circ Cardiovasc Qual Outcomes 2012;5(2):238–40. [13] Gøtzsche PC. Strengthening and opening up health research by sharing our raw data. Circ Cardiovasc Qual Outcomes 2012;5:236–7. [14] Kirwan JR. Making original data from clinical studies available for alternative analysis. J Rheumatol 1997;24:822–5. [15] Vickers AJ. Whose data set is it anyway? Sharing raw data from randomized trials. Trials 2006;7:15. [16] Hrynaszkiewicz I, Altman DG. Towards agreement on best practice for publishing raw clinical trial data. Trials 2009;10:17. [17] Institute of Medicine. Ensuring the integrity, accessibility, and stewardship of research data in the digital age. Washington, DC: National Academy Press; 2009. [18] Walport M, Brest P. Sharing research data to improve public health. Lancet 2011;377:537–53. [19] European Medicines Agency. Publication and access to clinical trial data. http://www.ema.europa.eu/docs/en_GB/document_library/Other/2013/06/ WC500144730.pdf; February 24 2014. [20] European Federation of Pharmaceutical Industries and Associations (EFPIA), Pharmaceutical Research and Manufacturers of America (PhRMA). Principles for responsible clinical trial data sharing. http:// www.phrma.org/phrmapedia/responsible-clinical-trial-data-sharing# sthash.je1WkJR5.dpuf; February 24 2014. [21] Van Dongen AJ. Good clinical practice, a transparent way of life. A review. Comput Med Imaging Graph 2001;25:213–6.
D. Alemayehu et al. / Contemporary Clinical Trials 39 (2014) 28–33 [22] EMEA ICH Topic E 6 (R1): guideline for good clinical practice. http:// ethikkommission.meduniwien.ac.at/fileadmin/ethik/media/dokumente/ rechtsgrundlagen/GCP.pdf; February 24 2014. [23] Food Drug Administration (FDA). International conference on harmonization, good clinical practice: consolidated guidelines. Fed Regist 1997;62:25692–709. [24] PSI Statisticians in the Pharmaceutical Industry. Guidelines for standard operating procedures for good statistical practice in clinical research. Drug Inf J 1998;32:665–82. [25] ICH E9, statistical principles for clinical trials. http://www.ich.org/fileadmin/ Public_Web_Site/ICH_Products/Guidelines/Efficacy/E9/Step4/E9_Guideline. pdf; February 24 2014. [26] Foster JC, Taylor JMC, Ruberg SJ. Subgroup identification from randomized clinical trial data. Stat Med 2011;30:2867–80. [27] Wang R, Lagakos SW, Ware JH, Hunter DJ, Drazen JM. Statistics in medicine—reporting of subgroup analyses in clinical trials. N Engl J Med 2007;357(21):2189–94. [28] Pocock SJ, Assmann SE, Enos LE, Kasten LE. Subgroup analysis, covariate adjustment and baseline comparisons in clinical trial reporting: current practice and problems. Stat Med 2002;21(19):2917–30. [29] US Department of Health and Human Services. Summary of the HIPAA privacy rule. http://www.hhs.gov/ocr/privacy/hipaa/understanding/ summary/privacysummary.pdf; February 24 2014.
33
[30] United Kingdom Department of Health, Confidentiality Unit, IPU. NHS Confidentiality Code of Practice; 2003. http://webarchive.nationalarchives. gov.uk/20130124004052/http://www.dh.gov.uk/prod_consum_dh/groups/ dh_digitalassets/@dh/@en/documents/digitalasset/dh_4069254.pdf; July 2 2014. [31] Gymrek M, McGuire AL, Golan D, Halperin E, Erlich Y. Identifying personal genomes by surname inference. Science 2013;339:321–4. [32] http://www.pfizer.com/news/press-release/press-release-detail/pfizer_ expands_clinical_trial_data_access_policy_and_launches_data_access_portal; February 24 2014. [33] Nisen P, Rockhold F. Access to patient-level data from GlaxoSmithKline clinical trials. N Engl J Med 2013;369:475–8. [34] http://www.bloomberg.com/news/print/2014-01-30/j-j-sets-drug-datafree-in-yoda-collaboration-with-yale.html; February 24 2014. [35] https://www.clinicalstudydatarequest.com/Default.aspx; February 24 2014. [36] Wilhelm EE, Oster E, Shoulson I. Approaches and costs for sharing clinical research data. JAMA 2014;311(12):1201–2.