Models for Sharing Trial Data Abound, but With Little Consensus

Marcia Frellick

August 03, 2016

Authors of four perspective articles published online today in the New England Journal of Medicine agree that sharing patient-level clinical trial data will advance medicine, but they disagree on the best guidelines for sharing responsibly and effectively.

One expert tells Medscape Medical News, however, that none of the models goes far enough and says the obligation to share the data promptly is part of the public contract for research.

As previously reported by Medscape Medical News, the International Committee of Medical Journal Editors (ICMJE) proposed new rules January 20 that would require authors to agree to share deidentified patient-level data from clinical trial results for their manuscripts to be considered for publication.

Such data include tables, figures, and appendices or supplementary material, and authors would have to share them no later than 6 months after publication, according to the ICMJE proposal.

The proposal was published simultaneously in 14 ICMJE member journals, including the Annals of Internal Medicine, New England Journal of Medicine, BMJ, and Lancet.

Since then, the debate has raged about how to balance requests for the data with advancing science, protecting patients' privacy, avoiding redundant future studies, and crediting the work of the original authors.

YODA Project Uses Intermediary for Data Use Requests

In one of the perspective articles published this week, Harlan M. Krumholz, MD, from the Department of Medicine at Yale University in New Haven, Connecticut, and Joanne Waldstreicher, MD, chief medical officer of Johnson & Johnson, New Brunswick, New Jersey, point to The Yale Open Data Access (YODA) Project, which was developed in 2011 and uses a "trusted intermediary" approach, as a successful model for data sharing.

Under this model, participating original authors transfer full jurisdiction over the data to the project and applicants seeking data submit proposals directly to the YODA Project. Before releasing data, YODA Project scientists conduct a blinded review to ensure that the scientific purpose is clearly described by an applicant, "that the data requested will be used to create or materially enhance generalizable scientific or medical knowledge to inform science and public health, and that the proposed research can be reasonably addressed using the requested data."

The YODA Project partners with Johnson & Johnson, but the company cannot veto proposals. The system was designed "to ensure cooperation with the data owner in promoting secure, responsible access, but independence with regard to data-sharing decisions," Dr Krumholz and Dr Waldstreicher write.

The data are made available for analysis within a secure platform, but they can be released to investigators if the YODA Project and Johnson & Johnson, with input from an advisory committee, decide the request is justified.

Within the first year of the debut of the project's data-request website in October 2014, Johnson & Johnson companies made data from 123 trials publicly available for researchers worldwide, according to the authors.

Also in that first year, 29 requests for data were received, and all were approved.

Group Says Authors Should Get to Keep Data for 2 Years

Meanwhile, writing in a separate perspective article, a group called the International Consortium of Investigators for Fairness in Trial Data Sharing says it does not support the ICMJE recommendations as written and suggest three main changes. The writing committee for the paper included P. J. Devereaux, MD, PhD, and four other researchers, all from McMaster University in Hamilton, Ontario, Canada.

First, they suggest extending the deadline for release of individual patient data to 2 years after publication of results, instead of 6 months as the ICMJE proposed, with an additional 6 months for every year it took to complete the study, up to a maximum 5 years.

The extra time, they contend, acknowledges the considerable work and time researchers invest in major trials, and would give them a reasonable amount of time to explore their own data and conduct secondary analyses.

Without an extended deadline, the current ICJME proposal may encourage researchers to delay publishing their original results until they can get their secondary studies ready, the authors warn.

The second recommendation, meant to build confidence in published data, is to have an independent statistician confirm analyses before publication of an article.

The third is for those who were not involved in the trial but who want access to the data to compensate the original investigators for their efforts and for costs of making data available.


To help put the ICJME recommendations into action, a group of clinical researchers, including Manesh R. Patel, MD, associate professor of medicine and director of interventional cardiology and cardiac catheterization labs across the Duke University Health System, Durham, North Carolina, formed the Academic Research Organization Consortium for Continuing Evaluation of Scientific Studies–Cardiovascular (ACCESS CV).

Similar to the consortium's plan, ACCESS CV's proposal, described in yet another perspective article, also recommends the original authors maintain exclusive access to clinical trial data for 2 years after publication.

Although they support sharing data, Dr Patel and colleagues note that protections must be in place.

They say use of the data by researchers who are inexperienced or unfamiliar with a data set could point to nonexistent discrepancies in their interpretation of the data and those of the original researchers, which could alarm and confuse the public, and thereby delay scientific advances instead of accelerating them.

Also, they write, "Unplanned exploratory analyses from a publicly shared database may be numerous and redundant."

Patients' privacy must also be protected, they emphasize. "Most patients in a deidentified database can be identified on the basis of their birth date, sex, and ZIP Code," especially those with rare disease and the very elderly.

After the 2-year window, requests for the data would go to a group comprising "ACCESS CV members who were not involved in the trial, along with the trial's principal investigator, a trial statistician, and a member of the data and safety monitoring board."

The group would then approve those that are deemed feasible, nonduplicative, and guided by qualified investigators. All requests and decisions would be posted on an ACCESS CV Web portal, "ideally within 60 days."

"An unresolved issue is the question of how to provide meaningful academic credit (typically authorship) to the team that designed and conducted the trial, given that trials are usually enormous undertakings requiring several years of work," the group writes.

Costs of Data Sharing Should be Included in Trial Expenses

In a fourth paper, US Sen. Elizabeth Warren, JD (D-Mass.), applauded the ICJME proposal, particularly its requirement that data sharing be a condition of publication in major medical journals.

Among other advantages, she says, data sharing could help address conflict of interest concerns when trials are supported by industry.

"By making trial results available for independent scrutiny by outside reviewers," she writes, "data sharing makes it less likely that trial sponsors can buy the analysis and results they want."

She adds that more negative, null, or inconclusive results should be published to honor the ethical standards of human subjects research and help hold industry sponsors accountable.

Sen. Warren notes that costs for data sharing should be built into any grants and contracts researchers negotiate with trial sponsors and treated as standard, necessary costs of completing a trial.

Proposed Models Are "Timid"

Vinay Prasad, MD, MPH, a hematologist-oncologist and assistant professor of medicine at the Oregon Health and Sciences University in Portland, called the extensions for data exclusivity and the need for gatekeepers and protections for original authors "timid," and said data must be more aggressively shared than the models suggest.

He also said widely sharing the data can help correct and clarify inadequate reporting and eliminate bias in some clinical trials and allow other scientists to "explore ideas that the primary authors could not even imagine."

"There is a tension between a researcher's desire to hog or hoard data for their career glory, and patients' and society's desire to maximally leverage the data for the greater good," he told Medscape Medical News. "In these articles, some are fear mongering that sharing will lead to false analyses. That is an absurd objection. Does sharing [Surveillance, Epidemiology and End Results (SEER)] data or Medicare data lead to rampant false analyses? These are data sets that are available to many and used without restrictions. There are no gatekeepers here, and maybe sometimes a bad analysis is published, but the scientific record corrects itself, and the sky is not falling."

He calls the assumption that data sharing will remove the incentive to conduct trials, "totally false," and makes this analogy: "Imagine if a contractor for a federal bridge says after finishing the bridge, which was paid for by the US government, that only the contractor can use the bridge for 2 years before others are allowed to use the bridge because that is part of the incentive to build it.

"Researchers are paid by the [National Institutes of Health] to run the trial. That is the incentive. You got salary support, so you could do this work, and not see patients.... Hoarding data for years to milk secondary publications is not part of the deal."

Dr Prasad also sees no need for "gatekeeping" of the data. "Current gatekeepers are journal editors and peer reviewers," he notes, "and these have been thought sufficient for people who 'create' data, and have been sufficient for thousands of papers using publicly available data, yet are somehow inadequate for people using trial data — such a bizarre distinction. Why don't people submit requests before data mining Medicare or SEER data? They don't. They just purchase the data, or download it for free, respectively, and last I checked, there are many such papers that are disputed, but that is part of the scientific process."

Dr Prasad says he is not sure what model will ultimately work best, but at least one option should be unfettered data sharing, which he describes as posting deidentified data alongside the primary article.

That way, "Anyone with Excel can run any analysis they wish, and we will rely on all the usual gatekeepers (journal editors/reviewers) to police claims."

"I would support a variety of pilot programs to see what works best," he added.

Full conflict-of-interest information is available on the journal website. Dr Prasad has disclosed no relevant financial relationships.

N Engl J Med. Published online August 3, 2016.

For more news, join us on Facebook and Twitter


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.