Transparency of Regulatory Data across the European Medicines Agency, Health Canada, and US Food and Drug Administration

Abstract Based on an analysis of relevant laws and policies, regulator data portals, and information requests, we find that clinical data, including clinical study reports, submitted to the European Medicines Agency and Health Canada to support approval of medicines are routinely made publicly available.

H istorically, sponsors and regulatory agencies have kept confidential much of the clinical data generated to support the approval and continued monitoring of small molecule and biologic drugs (i.e., regulatory data). Notable cases, such as rofecoxib (Vioxx) and gabapentin (Neurontin), revealed how treating data as confidential can conceal important information about drug safety and efficacy, as well as improper research practices, including the selective publication and reporting of trial results. 1 Providing access to these data may not only help prevent unscrupulous research practices, but also advance our understanding of the safety and effectiveness of medical products. 2 Over the past two decades numerous initiatives have been launched to make regulatory data, including clinical study reports (CSRs) and individual patient-level data (IPD), publicly available. 3 For instance, there are data sharing initiatives organized by industry (e.g. ClinicalStudyDataRequest. com [CSDR]) and independent organizations (e.g. Yale University Open Data Access [YODA] Project). 4 While these efforts have advanced transparency for many medical products, 5 they have not gained traction industry-wide and remain constrained by their lack of authority to require companies to make data publicly available. 6 In contrast, regulatory agencies, as gatekeepers of market authorization, are best positioned to disclose data, particularly clinical data submitted for regulatory review.
Several laws and policies -some recent -autho-to make clinical data of drugs publicly available, either proactively or reactively in response to information requests. These data (Box 1), such as clinical summaries, CSRs, individual case safety reports, and other information sponsors must submit to regulators, contain substantially more information than published articles and can be used to more comprehensively ascertain the risks and benefits of drugs, assess regulatory decisions, and inform clinical decision making. 7 While regulatory data have been relied on for many studies, including systematic reviews and meta-analyses, 8 there is substantial opportunity to increase use of regulatory data for secondary research. 9 For example, one survey found that only 3% of authors of Cochrane Reviews obtained data from regulatory agencies. 10 Raising awareness of the scope of clinical data made available across the EMA, HC, and FDA, and the most efficient ways to obtain it, may increase the use and utility of clinical research data for patients, clinicians, and researchers.
In this analysis, we provide an overview of the key laws and policies governing disclosure of data at the EMA, HC, and FDA, including how each regulator defines certain data as "confidential commercial information" (CCI) that may be kept secret. Based on a review of agency data sharing portals, published research on information requests, and our own parallel information requests, we then compare the accessibility and comprehensiveness of data proactively disclosed and made available upon request by the EMA, HC, and FDA. Lastly, we propose ways for these regulatory bodies to enhance transparency.

Transparency Laws and Policies European Medicines Agency
Policy 0043 and Policy 0070 govern the EMA's approach to providing access to regulatory data for drugs and biologics ( Table 1). Policy 0043, adopted in 2010, authorizes the EMA to release data reactively. 11 Policy 0070, adopted in 2014, authorizes the EMA to proactively publish data on an online data sharing portal. 12 The scope of policy 0043 is expansive, providing access to any document originated, received, or held by the EMA. The scope of Policy 0070 is narrower, applying only to data submitted under the central marketing authorization procedure after January 1, 2015. EMA plans to implement Policy 0070 in two phases. Phase 1 publishes clinical reports, which include clinical overviews, clinical summaries, CSRs, along with several appendices to the CSRs, including protocol and protocol amendments, sample case report forms (CRFs), and statistical analysis plans (SAPs). Phase 2 will publish IPD. However, in 2018 EMA temporarily suspended its proactive publication of data, citing the disruption and resource constraints caused by the United Kingdom's withdrawal from the European Union. 13 EMA's proactive and reactive disclosure policies take a similar position on CCI, generally considering information contained in clinical reports not as CCI, unless disclosure undermines the competitive position of the information's owner. This position faced several challenges in court but was recently validated by the European Court of Justice. The Court found that clinical reports were not covered by a general presumption of confidentiality and that market authorization holders must meet a high standard to qualify any data included in the reports as CCI, by establishing that disclosure poses the risk of concrete harm to their commercial interests. 14 Additional details regarding data accessibility and the timeline for data publication are available in Table 1.

Health Canada
In March 2019, HC launched its Public Release of Clinical Information (PRCI) initiative, which goes beyond In this analysis, we provide an overview of the key laws and policies governing disclosure of data at the EMA, HC, and FDA, including how each regulator defines certain data as "confidential commercial information" (CCI) that may be kept secret. Based on a review of agency data sharing portals, published research on information requests, and our own parallel information requests, we then compare the accessibility and comprehensiveness of data proactively disclosed and made available upon request by the EMA, HC, and FDA. Lastly, we propose ways for these regulatory bodies to enhance transparency.
• Provide insight into the quality and limitations of studies, issues encountered during product development and testing, and unusual prescribing information

Clinical study report
Comprehensively describes the methodology and results of a clinical trial. It contains more detail regarding trial design, conduct, and results than is contained in published versions of the same trial. It has several appendices, which include protocol and protocol amendments, sample case report forms, and statistical analysis plans.
• Appraise a trial and synthesize evidence, including evaluation of product safety and effectiveness • Inform systematic reviews, meta-analyses, and risk of bias evaluations • Provide greater detail than journal publications, which are often subject to word-count restrictions, particularly on secondary effectiveness endpoints and safety • Determine improper research practices, such as datadredging, outcome switching, and selective reporting

Protocol and protocol amendments
Describe the original objectives, design, methods, statistical considerations and organization of a clinical trial, and include any subsequent protocol modifications. Both the protocol and later modifications are reviewed by an ethics committee.
• Assess study design and facilitate replication • Identify unscrupulous research practices, such as outcome switching and P-hacking Case report forms Questionnaires used by the sponsor of the clinical trial to record information about each trial participant.
• Ensure the accuracy of individual patient-level data

Statistical analysis plan
Provides a complete description of the planned methods for collection, analysis, interpretation, presentation, and organization of the data.
• Assess statistical methodology, including clinical trial power calculations, endpoint definitions, and proposed and conducted analyses • Appraise and replicate statistical methods

Individual case safety report
Provides detailed information related to a suspected adverse reaction to a medicinal product that occurs in a single patient at a specific point of time.
• Identify an increase in the reporting of a known adverse event, a new serious adverse event not listed in a medicines label, or a new drug-drug interaction

Individual patientlevel data
Individual data recorded for each participant in a clinical study, such as age, gender, race, efficacy and safety outcomes, laboratory results, etc.
• Appraise a trial and synthesize evidence, including for evaluation of product safety and effectiveness and for use in systematic reviews and meta-analyses • Classify all adverse events when performing a meta-analysis focused on safety  0070 by proactively releasing data for not  only approved, unapproved, and withdrawn drug and  biologic submissions but also Class III and IV medical device applications (Table 1). 15 The clinical data made available is similar in scope to EMA Policy 0070, with HC publishing clinical reports. HC also intends to make clinical reports available upon request for medical products that had a final regulatory decision prior to March 2019, similar to Policy 0043. However, HC will not release IPD under the PRCI initiative. HC plans to phase in the proactive release of clinical reports over four years, beginning in 2019.
HC construes CCI (known as "confidential business information" in Canada) narrowly, protecting only clinical information not used by the applicant to support the proposed conditions of use or clinical information that describes tests, methods, or assays used exclusively by the manufacturer, and then only with adequate justification. 16 Going forward, HC aims to publish data within 120 days after issuance of a final regulatory decision or after an information request is lodged. Unlike the EMA, data posted in response to information requests are made available on HC's online portal, requests are not limited to the citizens of the nation's regulator (i.e. Canadians), and registration is not required to access data.

US Food and Drug Administration
In January 2018, FDA launched a new pilot program to proactively publish CSRs of pivotal studies for nine recently-approved novel drugs, including trial protocols, protocol amendments, and SAPs ( Table 1). 17 However, FDA announced in March 2020 that it was ending the pilot, with only a single CSR having been made available. 18 The Freedom of Information Act (FOIA), enacted in 1966, requires federal agencies, including the FDA, to disclose records upon request by the public, unless records fall under one of nine specific exemptions protecting interests such as CCI and personal privacy. 19 FOIA is the only mechanism to access certain types of regulatory data, such as CSRs, for medical products approved by the FDA. 20 The FDA has defined CCI as "valuable data or information which is used in one's business and is of a type customarily held in strict confidence or regarded as privileged and not disclosed to any member of the public by the person to whom it belongs." 21 However, even if information meets the definition of CCI, FDA may have discretion to release it if there is a compelling public interest in disclosure. 22

Assessing Proactive Disclosure of Regulatory Data
To assess the EMA and HC's proactive publication of clinical data under Policy 0070 and the PRCI initiative, respectively, we systematically searched each agency's online data sharing portal, 23 documenting for each data release through April 2021: the type of medical product (drug, biologic, medical device, or vaccine); regulatory procedure (initial marketing authorization application (MAA) or post-authorization application); regulatory decision (approved, unapproved, or withdrawn); regulatory decision date; public release date; and time from regulatory decision to release. EMA and HC generally release the same categories of data across medical product types and regulatory procedures. Thus, to compare the data released across agencies, we randomly selected a single product for which both EMA and HC had made clinical data available, and then characterized the category of data released, number of pages, presence of redactions, reason for redactions provided (protected personal data (PPD) or CCI), and described the information redacted.

Assessing Reactive Disclosure of Regulatory Data
To assess EMA, HC, and FDA's reactive data disclosure processes, we first reviewed the literature to identify studies on information requests submitted to EMA, HC, or FDA. To better understand the scope of data made available by EMA, HC, and FDA in response to information requests, we submitted a FOIA request to FDA in 2014 for a wide range of clinical data and regulatory records for Gilead's Hepatitis C drugs sofosbuvir (Sovaldi) and ledipasvir/sofosbuvir (Harvoni) (Appendix Box 1), as well as a parallel request to EMA in 2016 under Policy 0043 and to HC in 2020 under the PRCI initiative. Sovaldi and Harvoni were selected as illustrative test cases because of their novelty, the safety issues that were being evaluated at the time we initiated this work, their use in the treatment of Hepatitis C, a major global health problem, and high cost. 24 We documented the date of each request milestone (e.g. initial request filed, appeal filed) and the date of each document production. For each production, we then characterized the category of data and the number of pages made available. We then described the information that had been redacted and compared the number of pages and redactions in CSRs of phase 2 and 3 trials produced by EMA, HC, and FDA.
Since HC releases information in response to requests on the same online data sharing portal as the information it proactively discloses, we followed the  of a person to whose business or affairs the information relates, means -subject to the regulations -business information: a. That is not publicly available b. In respect of which the person has taken measures that are reasonable in the circumstances to ensure that it remains not publicly available, and c. That has actual or potential economic value to the person or their competitors because it is not publicly available and its disclosure would result in a material financial loss to the person or a material financial gain to their competitors. 41 Valuable data or information which is used in one's business and is of a type customarily held in strict confidence or regarded as privileged and not disclosed to any member of the public by the person to whom it belongs. 42 Not specified same methodology (described above) for extracting information on HC's reactive disclosure as we did for the agency's proactive disclosure.

Findings: Proactive Disclosure of Regulatory Data
Between 2016 and April 2021, EMA proactively released data for 123 unique medical products (Appendix Table 1), including 81 drugs, 38 biologics, and 4 vaccines ( Table 2). Data supporting 147 regulatory procedures reviewed by the EMA between 2015 and 2021 were made publicly available, including 95 initial MAAs and 52 post-authorization applications; of which 135 were approved and 12 were withdrawn. Between 2019 and April 2021, HC proactively released data for 73 unique medical products, including 45 drugs, 23 biologics, 3 vaccines, and 2 medical devices. HC disclosed data supporting 62 initial MAAs and 13 post-authorization applications; all 75 were approved between 2016 and 2021. In 2018, FDA proactively disclosed data supporting the initial MAA of 1 drug that was approved in 2018. EMA, HC, and FDA took a median of 511 (interquartile range 416-574), 150 (interquartile range 122-204), and 33 days, respectively, after each agency's regulatory decision to release data. At the time of initial data collection (April 30, 2020), no data supporting the same MAA were proactively released by both EMA and HC. However, data supporting the same initial MAA for 4 medicines were made available proactively by EMA and reactively by HC. After reviewing HC's data releases and confirming the agency typically makes available equivalent categories of information proactively and reactively, we randomly selected daratumumab (Darzalex) among those 4 medicines to compare the information shared across EMA and HC. EMA and HC released 28 and 29 documents, respectively, for Darzalex, including all of the same categories of data: a clinical overview; summaries of biopharmaceutical and pharmacological studies, summaries of safety and efficacy, a biomarker technical report, and a population pharmacokinetic report; for each of its 5 clinical trials: a CSR (4 full, 1 synoptic), protocol and protocol amendments, sample CRFs, and SAPs for 3 of the trials ( Table 3). There were no substantive discrepancies in the released data. However, EMA disclosed a more updated clinical overview, while HC posted an additional clinical overview addendum describing Canadian treatment approaches for multiple myeloma. The full CSRs, protocols, including those with amendments, CRFs, and SAPs had a median length of 1612, 108, 115, and 76 pages, respectively, for both EMA and HC. Redactions were comparable and minimal across EMA and HC: PPD and CCI were provided as the reasons for redactions in 3 and 15 documents, respectively. Names of report investigators or subject ID numbers were the most common redactions.

Findings: Reactive Disclosure of Regulatory Data
We identified two studies on information requests to EMA under Policy 0043, no studies on information requests to HC under the PRCI initiative, one study on FOIA requests to FDA, and one study examining CSRs released in response to information requests to EMA and FDA, among other sources. One study determined that of the 457 information requests EMA received between 2010 and 2012, 66% were granted, 27% were denied, and 7% were pending. 25 Requests were processed in a median of 26 days. In a case series of 12 information requests filed between 2011 and 2015, the EMA released a wide variety of regulatory data, including CSRs, regulatory comments, meeting and decision records, periodic safety update reports, correspondence, and postmarket data and took a median of 301 days to process the requests. 26 A study examining 78 CSRs, including 11 obtained from information requests to EMA and FDA, found that key appendices of CSRs, such as protocols and case report forms were frequently omitted. 27 Based on a study of FOIA requests to FDA between 2008 and 2017, FDA fully or partially granted 72% of requests. 28 FDA processed one-fifth of requests in 20 days but took more than 61 days to process two-thirds of requests.
HC released data for 55 unique medical products in response to 70 processed requests between February 2019 and April 2021, including for 23 drugs, 6 biologics, 6 vaccines, and 20 medical devices. HC took a median of 132 (interquartile range 103-167) days to process requests and published the same categories of regulatory data reactively as it had proactively.
In response to our parallel requests for information supporting approval of Sovaldi and Harvoni, FDA and HC released substantially more regulatory data than EMA, including clinical overviews, summaries and integrated summaries of safety and efficacy, clinical study protocols and amendments, and narratives of deaths and serious adverse events. Information only made available by FDA include individual safety case reports; records and correspondence related to product labeling, safety concerns, pediatric studies, expedited approval pathway designations, and postmarket study requirements and commitments; safety update reports; site initiation visit reports; and IPD, albeit heavily redacted (Appendix Table 2). Information only produced by HC include sample case report forms and statistical analysis plans (Appendix Table 3).   Information only made available by EMA include periodic safety update reports and pharmacovigilance risk assessment committee reports (Appendix Table 4).
All three agencies made available CSRs for phase 1, 2, and 3 trials. Comparing the productions of CSRs of phase 2 and 3 clinical trials, EMA released CSRs corresponding to more phase 2 and 3 trials than FDA and HC (25 vs. 18 vs. 24); however, of the 16 CSRs produced for the same clinical trials by all three regulators, 11 released by FDA and HC were longer, 4 were the same, and 1 released by EMA was longer (Appendix Table 5). Among the 16 CSRs, those produced by FDA had an average of 473 pages compared to 464 and 179 pages for those produced by HC and EMA, respectively. EMA, HC, and FDA redactions of the CSRs were minimal, most commonly redacting names and contact information of study investigators and administrators and subject ID numbers. EMA and FDA also commonly redacted information about the drug manufacturer and supplier, citing CCI, while HC redacted narratives of deaths and serious adverse events, citing PPD. Time from initial request to final data production was 918 and 968 days for FDA and EMA, respectively (Appendix Table 6). While HC processed the information request much faster, publicly posting the data packages for Harvoni and Sovaldi 155 and 351 days, respectively, after the initial request. The request to FDA required considerably more resources, including multiple appeals and court filings. Additionally, FDA waived processing fees; however, the agency normally charges fees unless it is shown disclosure of the requested information is in the public interest. 29 Last, HC organizes its data productions into categories, making the information much easier to navigate and process.

Enhancing Regulatory Data Transparency in the 2020s
Over the past decade, EMA and HC have greatly expanded the public availability of regulatory data, while FDA has lagged behind by not proactively publishing clinical reports. EMA and HC's routine publication of clinical reports, including clinical overviews and summaries, CSRs, protocols, sample CRFs, and SAPs, which just a decade ago had largely been treated by regulators as CCI, represents a paradigm shift in clinical trial transparency. HC's more recent PRCI initiative goes beyond EMA Policy 0070 in proactively posting clinical reports supporting medical device applications, not just drug and biologic MAAs. HC also offers the most efficient source of regulatory data, typically posting clinical reports about 5 months after a regulatory decision, nearly a year quicker than EMA, and processing most information requests in less than 5 months, about 150 days quicker than EMA takes to release requested regulatory data that are complex or voluminous. 30 EMA and FDA required nearly three years to process our request for comprehensive regulatory data supporting the approvals of Sovaldi and Harvoni, substantially longer than EMA had previously taken to process requests for CSRs. 31 While EMA and FDA processing times vary widely, 32 our findings suggest there may be an opportunity for EMA and FDA to increase their processing speed to align with HC. The CSRs released upon request by FDA and HC were more than double the length of those produced by EMA, suggesting FDA and HC may source more comprehensive CSRs. A similar request to FDA may require greater resources than to EMA and HC. Nonetheless, use of EMA or FDA's information request processes may be necessary to gain access to other types of regulatory data, such as sponsorregulator correspondence. Consistent with findings from other recent studies, redactions across the agencies were mostly minor (e.g., primarily researcher and participant identifying information) and generally did not impede interpretation of the evidence. 33 Two notable exceptions are FDA and HC's complete redaction of IPD and narratives of serious adverse events and deaths, respectively.
Our study has several limitations. First, HC's start date, the date when HC initiates the process to prepare information for public release, was used to calculate HC's processing time for information requests, not the actual date information requests were submitted, which is not publicly available. HC may take several weeks, and in some cases months, to begin processing requests, particularly when a request requires clarification or where paper records need to be digitized. Therefore, HC may take moderately longer than a median of 132 days to process requests. However, it is unlikely to impact the study's finding that HC is the most efficient source of clinical reports, given EMA required a median of 301 days to process a series of comparable information requests and HC also completed our requests for information on Sovaldi and Harvoni in about one-third the time that EMA and FDA required. 34 Second, parts of the study were based on analyses of case studies, including Sovaldi, Harvoni, and Darzalex, which represent just a few of the many medical products for which data has been made publicly available. Third, the length of CSRs was used to compare the scope of data made available reactively by each agency; specific differences in the content of CSRs were not examined. Last, the study was limited to transparency of regulatory data, which comprises just a portion of data generated in clinical trials. Given the progress government agencies have made toward transparency of regulatory data, non-governmental led data sharing initiatives, such as the YODA Project, might consider shifting their focus and resources toward advancing data transparency of trials not submitted to regulators (e.g., academic trials).
While clinical reports and other categories of regulatory data supporting MAAs of drugs, biologics, and medical devices have been made accessible, there are several ways regulators could enhance transparency over the next decade. First, EMA and FDA should mitigate or remove barriers to accessing clinical reports, including citizenship requirements and processing costs associated with information requests, particularly for those requesting information in the public interest. Second, to improve access to information, EMA and FDA should post requested clinical data on a public online portal and increase their processing speed to more closely align with HC. To reduce delays, regulators should consider hiring more personnel and devoting greater resources to responding to requests. Third, EMA and FDA should extend access to data supporting medical devices. Last, while EMA remains committed to make available IPD in Phase 2 of Policy 0070, FDA and HC also should implement policies to share IPD, which would provide the raw data necessary to conduct secondary analyses. While sharing IPD poses increased risks of patient reidentification, data sharing initiatives, such as the YODA Project and CSDR, have demonstrated that through use of a "trusted intermediary," regulatory agencies could share IPD with researchers to maximize the use and utility of the data while still protecting patient privacy. 35 The administrative and redaction costs incurred by industry and regulatory authorities represent a substantial obstacle to the future of clinical data sharing programs, exemplified by the ongoing, albeit temporary, suspension of Policy 0070 and the restriction of information requests to EMA to EU citizens and legal residents in 2018. To address this challenge, governments and institutions, such as the European Union, should give greater priority and funding to regulatory data sharing programs. Concurrently, efforts to improve the efficiency of data sharing initiatives should be considered. For instance, multiregional disclosure requirements with varying anonymization standards for CCI and PPD increase costs. Regional regulators may consider a harmonized approach for clinical report disclosure to help reduce inefficiencies. In fact, HC has already begun accepting clinical information previously published under Policy 0070 for its PRCI initiative. Also, FDA recently announced that sharing of CSRs in harmony with international regulators is a long-term goal. 36 However, an important benefit of a multiregional approach is that it encourages greater transparency by incentivizing sponsors to anonymize data in accordance with the standards least amenable to CCI claims. For example, we found that names of drug manufacturers and suppliers were often redacted by EMA and FDA, but HC did not consider this information as CCI, and thus pharmaceutical sponsors may be encouraged to not devote resources to redact such information in a multiregional approach, recognizing the data will become public under the PRCI initiative. Therefore, it is important that measures, such as enforcement of HC's narrower definition of CCI, are taken to prevent harmonization of disclosure standards toward less transparency.
In summary, regulatory data pertinent to public health and clinical medicine that were used to support the approval of medicines and medical devices are now available proactively or in response to information requests. Over the next decade, regulatory agencies should make IPD available, and additional resources might be needed to ensure the long-term viability of regulatory data sharing programs and to encourage researchers to take advantage of the data that is -for now -more available than ever before.
In summary, regulatory data pertinent to public health and clinical medicine that were used to support the approval of medicines and medical devices are now available proactively or in response to information requests. Over the next decade, regulatory agencies should make IPD available, and additional resources might be needed to ensure the long-term viability of regulatory data sharing programs and to encourage researchers to take advantage of the data that is -for now -more available than ever before.
Autologous CD34+ cells transfected with retroviral vector containing adenosine deaminase gene (Strimvelis) Oxycodone Hydrochloride (OxyContin and Oxyneo)   Clinical overview; summary of biopharmaceutical studies and associated analytical methods; summary of clinical pharmacology studies; summary of clinical efficacy; summary of clinical safety; comparative bioavailability and bioequivalence study reports; plasma protein binding study reports; reports of hepatic metabolism and drug interaction studies; reports of studies using other human biomaterials; healthy subject PK and initial tolerability study reports; patient PK and initial tolerability study reports; intrinsic factor PK study reports; extrinsic factor PK study reports; population PK study reports; healthy subject PD and PK/PD study reports; patient PD and PK/PD study reports; study reports of controlled clinical studies pertinent to the claimed indication; integrated summary of safety; integrated summary of efficacy, and other study reports. Most study reports include a synopsis, report body, protocol and protocol amendments, sample case report forms, and statistical analysis plans, and all study reports of Phase 2 and 3 trials also included narratives of SAEs (albeit completely redacted) Clinical overview; summary of biopharmaceutical studies and associated analytical methods; summary of clinical pharmacology studies; summary of clinical efficacy; summary of clinical safety; comparative bioavailability and bioequivalence study reports; plasma protein binding study reports; reports of hepatic metabolism and drug interaction studies; reports of studies using other human biomaterials; healthy subject PK and initial tolerability study reports; patient PK and initial tolerability study reports; intrinsic factor PK study reports; extrinsic factor PK study reports; population PK study reports; healthy subject PD and PK/PD study reports; patient PD and PK/PD study reports; study reports of controlled clinical studies pertinent to the claimed indication; integrated summary of safety; integrated summary of efficacy, and other study reports. Most study reports include a synopsis, report body, protocol and protocol amendments, sample case report forms, and statistical analysis plans, and several study reports of Phase 2 and 3 trials also included narratives of SAEs (albeit completely redacted). Appendix