Clinical trial data typically have been shrouded in secrecy, but major new initiatives are pulling back the veil and promising to provide greater access. The European Medicines Agency (EMA) plans to publish summary trial data from all new approved drugs, and provide data from individual patients upon request, starting in 2014. In the United States, the Food and Drug Administration (FDA) is considering making trial results pooled by drug class available for data-mining projects. Meanwhile, pharmaceutical industry organizations have set forth their own guidelines for greater data sharing, and individual companies such as GlaxoSmithKline have begun to fulfill data requests through a panel of independent academic researchers. Some nonprofit and public groups already share standardized, pooled trial data or are developing guidelines for data sharing.
Increased access to clinical trial data brings both benefits and risks, according to two articles in the October 21 New England Journal of Medicine, one written by EMA regulators and the other by experts in health policy at the Harvard School of Public Health, Boston. On the one hand, it could advance research more quickly, improve trial designs, provide an external check on drug safety and effectiveness, and bolster public trust in regulators’ decisions. On the flip side, open data runs the risk of breaching patient confidentiality and could be used to conduct flawed analyses that may spread misconceptions about drugs, experts said.
Indeed, while stakeholders agree on the need for greater openness, they differ on how data should be shared. Many patient advocates want free access with no restrictions other than the removal of information that might identify patients. Pharmaceutical companies and some nonprofit research groups, however, favor a model of controlled access, where only qualified researchers could access data for specific purposes. Some academic and industry researchers suggest the EMA proposal does not have adequate safeguards against data misuse. The issue may be decided in court. The EMA already faces lawsuits from companies unhappy about past data disclosures, and some observers expect the new policy to spur additional legal action. It is unclear what the landscape will look like when the dust finally settles.
A Sea Change at EMA
Historically, clinical trial data has been considered commercial confidential information, and the EMA routinely turned down requests for documents, noted Hans-Georg Eichler, the organization’s senior medical officer. Public pressure to disclose trial data has mounted over the last decade, however, particularly in light of safety issues discovered in approved drugs such as Vioxx and Avandia, both of which were withdrawn from the market for heightening the risk of heart disease. In 2007 the Nordic Cochrane Centre, an independent health research group, lodged a complaint with the European Ombudsman about the lack of access to trial data held by the EMA (see Gøtzsche and Jørgensen, 2011). The Ombudsman ruled in 2010 that public health concerns regarding drugs on the market take precedence over commercial confidentiality, and ordered that the data be disclosed.
This ruling signaled the turning of the tide, and led to a 2010 EMA policy allowing the release of raw, redacted trial data upon request. This in turn spurred a lawsuit by pharmaceutical companies AbbVie Inc., North Chicago, Illinois, and InterMune Inc., Brisbane, California, seeking to block disclosures about their drugs Humira and Esbriet, respectively, to competitor companies (see, e.g., Bloomberg news story). That lawsuit, which has the support of the European (EFPIA) and U.S. (PhRMA) pharmaceutical trade organizations, is still pending. In the meantime, the European General Court issued a temporary injunction halting release of the documents (see Burrill Report news story).
The 2010 policy created practical challenges for the EMA, Eichler told Alzforum. The agency has been swamped with requests for data that have strained resources, as preparing disclosures and redacting personal information from trial data requires a lot of time. To streamline the process, the EMA proposes to proactively publish summaries of trial data on its website, where they will be freely accessible, while it continues to provide redacted patient-level data only upon request. Over the course of the last year, the agency hashed out a new draft policy spelling this out. The process started with a workshop in November 2012 at which interested parties—including advocacy groups, industry, academia, and regulatory agencies—aired their views. The EMA put out the draft policy for review and is now revising it based on the feedback received. It plans to finalize the policy before the end of the year and have it take effect January 1, 2014. No one knows how the current lawsuit might affect implementation, however.
In the new policy, the EMA distinguishes between three types of data, Eichler said. Category 1 comprises commercial confidential information, such as details about drug formulations. These types of data will remain confidential under the new policy. Category 2 consists of trial summary data, which will now be published. Category 3 encompasses any data containing identifying information, such as names, addresses, or birthdates. These data will be available upon request after identifying information has been redacted in adherence with accepted guidelines (see Hrynaszkiewicz et al., 2010). Requesters must identify themselves and agree to use the data appropriately, which includes preserving patient anonymity, keeping data confidential, and making public all analyses performed on the data. In addition, requesters are encouraged but not required to share their analysis plan with the EMA.
The Rosy Side of Data Sharing
Eichler sees two main benefits to opening access to clinical trial data. First, all interested parties will be able to see for themselves the safety and effectiveness data that the agency used to make decisions about approving drugs. “We’re hoping that by increasing transparency, in the long term we will increase trust in the regulatory system,” Eichler said. Because the EMA makes decisions on behalf of and for the benefit of society, society has the right to see the raw data used in this process, he added. Second, analyses of the rich trove of trial data could give rise to additional insights into disease and advance the pace of research.
Many observers agree, and hail the new policy. “I think this is a great idea. It will open up opportunities for data mining that can generate new ideas and hypotheses,” said Merit Cudkowicz at Massachusetts General Hospital, Boston. “Patients participate in studies to help advance the field; their data should be shared as openly as possible.” As co-chair of the North East ALS consortium (NEALS), Cudkowicz has championed disclosure of data from amyotrophic lateral sclerosis trials. Likewise, Peter Doshi, an expert in health policy at Johns Hopkins University School of Medicine, Baltimore, lauds the EMA’s move toward open access. “I think it’s a fantastic policy. It states in very clear terms that clinical trial data is not commercial confidential information. It allows for a mechanism for the responsible dissemination and sharing of participant-level data, and binds the receiving party to very reasonable standards for use.”
Moreover, open data will allow independent researchers to verify the safety and effectiveness of new drugs, said Michelle Mello at the Harvard School of Public Health. Mello is first author on the second NEJM paper. Independent analyses of summary data from the Vioxx and Avandia trials helped reveal safety concerns, she noted. “This demonstrated that we can use the research community as a check to confirm what regulators believe is true about the safety of a drug.”
At the same time, many researchers worry about unintended consequences of the EMA policy. No matter how carefully patient identities are masked in released data, it may be possible to re-identify trial participants. This has been shown to be relatively easy to do in some genetic studies, for example (see Jan 2013 news story). Although the EMA policy requires data requesters to protect patient privacy, the agency has no mechanism for enforcing this requirement, said Mark Barnes, a partner at Ropes & Gray LLP and faculty member at Harvard Law School. He is a co-author on the NEJM article with Mello. Ideally, groups that receive the data should be liable to fines and penalties for misuse, Barnes suggested.
Industry and independent researchers voiced the common concern that trial data could be misinterpreted. “These datasets are in many cases quite complex and large. If one were to do exploratory data mining without using appropriate statistical methodology, one could arrive at an incorrect or misleading set of conclusions,” said Jonathan Sorof, who leads implementation of the data-sharing policy at the pharmaceutical company F. Hoffmann-La Roche. Furthermore, once a question about a drug’s safety or effectiveness enters the public mind, it often takes on a life of its own, he added. Mello agreed. “Even if the trial sponsor later rebuts the study by pointing out the shoddy methods, you can’t unring the bell in terms of how the public perceives the drug. There could be damage to public health and business interests that can’t be undone,” she said. Flawed research has led to serious public consequences before, noted Martha Brumfield, who leads the Critical Path Institute, an applied research organization that facilitates early drug development. For example, a single study erroneously linking vaccines to autism sowed widespread distrust of childhood vaccinations that allowed diseases such as whooping cough to resurge, even though the study was eventually withdrawn.
Many researchers would like to see a more rigorous vetting of requests for patient-level trial data before the EMA releases it. “Researchers should submit a proposed analysis plan and some indication of the capabilities they have to conduct those analyses,” Brumfield suggested. Others agree. “Data sharing could be a positive good for everybody. But it has to be done in a more careful way than what the EMA has outlined to date,” Barnes said. Pharmaceutical companies express similar misgivings. “Industry has concerns with several aspects of the draft policy, the implementation of which, in its current form, we believe would yield no benefit for patients and would even conflict with a number of public health imperatives,” Justin McCarthy at Pfizer wrote to Alzforum. McCarthy suggested that data sharing could endanger patient privacy, lead to misinterpretation of clinical trial results, and weaken incentives for companies to invest in biomedical research due to disclosure of their confidential information to competitors.
Eichler agrees that poor-quality analyses are a risk, but notes that the ratio of false “safety signals” (i.e., adverse events due to the drug) to true signals in trial data is less than one in 10. “If there are a handful of additional false signals, that is the price of transparency,” he told Alzforum. Moreover, flawed analyses have always been present in the literature, and greater data availability may help correct misinformation faster, Eichler added. Doshi agreed. “The more eyes we have on the data, the quicker the bugs can be worked out,” he said.
Eichler pointed out that the EMA policy is not final, and that many of the finer details about how data will be shared must still be worked out. “We will have to see how we can balance all these interests that are at stake in the best interests of public health,” Eichler said.
FDA Toys With Change
Other groups are taking different approaches. The FDA recently invited comments on the idea of publishing pooled data from trials of particular classes of drugs. Not only would patient information be removed, drug identities would also be masked. These data would be useless for checking safety and effectiveness of products. Instead, they would facilitate big-data projects that seek to increase the understanding of specific diseases, for example, identifying biomarkers that track with clinical improvement (see Jun 2013 news story).
The FDA proposal has met with less resistance than the EMA policy, but nonetheless has some critics. Many researchers think aggregated data could be useful, but wonder if it is practical for the FDA to undertake the job. To properly standardize and aggregate data from numerous trials would require an enormous amount of resources and might prove a distraction from the agency’s core mission, Sorof noted. There is also a legal question as to whether the FDA can release any trial data, which in the United States is considered the property of the drug sponsors.
Industry and Independent Initiatives
Nonprofit groups such as the Critical Path Institute already have made strides in the arena of data sharing. This organization runs the C-Path Online Data Repository (CODR), which combines standardized placebo data donated by several pharmaceutical partners. This resource already has been used for big-data projects, such as developing a tool to simulate trials in order to choose the best design (see Jul 2013 news story). Government-funded groups such as the Alzheimer’s Disease Cooperative Study (ADCS) routinely share data. Paul Aisen at the University of California, San Diego, who heads ADCS, notes that data from all trials is freely available on its website starting six months after publication. ADCS also shares data with other initiatives such as CODR. “Industry working on Alzheimer’s drugs have embraced the ideal of sharing knowledge. The extent of precompetitive collaboration is enormous, and it’s made a very big difference,” Aisen told Alzforum. Prize4Life, a Cambridge, Massachusetts, nonprofit dedicated to finding a cure for ALS, created PRO-ACT, the largest open-access ALS clinical trials database, with more than 8,500 patient records (see Dec 2012 news story). PRO-ACT incorporates data from NEALS and several pharmaceutical companies.
Industry has developed its own set of guidelines for sharing trial data. Their ground rules specify that requests for patient-level data be handled by independent scientific review boards that evaluate the merits of the proposed analysis plan and the qualifications of the requester. GlaxoSmithKline has led in this area, publishing its own policy, establishing an online site for data requests, and convening an academic panel to review requests for patient-level data. Roche has a similar policy and has now joined the GSK effort, Sorof told Alzforum. In addition to granting controlled access to patient-level data, both companies will freely provide clinical study reports. “What is new about the policy is that, for the first time, the documents we submit to health authorities for regulatory review will now be part of what we will fulfill upon request. We believe those summary documents should be available to anyone who has an interest in seeing them,” Sorof said. Other companies are expected to follow suit.
Belying the trend toward openness, a recent study found that almost one-third of large clinical trials remain unpublished five years after completion (see Jones et al., 2013). Several academic groups and journals, including the Cochrane Collaboration, PLoS, and the British Medical Journal, have launched the AllTrials initiative, an online petition to make all trial results public. Others believe there must be standard rules for data-sharing across the spectrum of research and industry. “If we could get some sort of guidance that the general research community would agree to, then I think we could manage a lot of our concerns moving forward,” Brumfield said. Some groups are working on this. The Institute of Medicine, a nonprofit organization based in Washington, D.C., held a 2012 workshop on this topic and is now developing recommendations for responsible data sharing, with the interim report due in January 2014.
While the details of how data will be shared remain hazy, all agree that it will become ever more public in the coming years. “We’re transitioning from a world where data secrecy was the norm, to a world where data availability and sharing is the norm,” Doshi said.—Madolyn Bowman Rogers
- Research and Recreational Genetics Collide to Breach Privacy
- Building a Roadmap for Shared Combination Drug Trials
- AD Trial Simulation Tool Receives Regulators’ Blessings
- Chicago—ALS Database Opens for Business
- Gøtzsche PC, Jørgensen AW. Opening up data at the European Medicines Agency. BMJ. 2011;342:d2686. PubMed.
- Hrynaszkiewicz I, Norton ML, Vickers AJ, Altman DG. Preparing raw clinical data for publication: guidance for journal editors, authors, and peer reviewers. Trials. 2010;11:9. PubMed.
- Jones CW, Handler L, Crowell KE, Keil LG, Weaver MA, Platts-Mills TF. Non-publication of large randomized clinical trials: cross sectional analysis. BMJ. 2013;347:f6104. PubMed.
- European Ombudsman
- 2010 EMA policy
- Bloomberg news story
- Burrill Report news story
- draft policy
- data from amyotrophic lateral sclerosis trials
- invited comments
- C-Path Online Data Repository
- Alzheimer’s Disease Cooperative Study
- online site
- Institute of Medicine
- 2012 workshop
- recommendations for responsible data sharing
- Eichler HG, Pétavy F, Pignatti F, Rasi G. Access to patient-level trial data--a boon to drug developers. N Engl J Med. 2013 Oct 24;369(17):1577-9. PubMed.
- 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 Oct 24;369(17):1651-8. PubMed.