On Jan. 10, the comment period closed on the draft NIH Policy for Data Management and Sharing, and in their comments several stakeholders identified flaws and shortcomings they would like to see addressed in the final policy.
As organization called Public Responsibility in Medicine and Research (PRIM&R) stated that it supports the NIH's stated policy position that sharing data resulting from taxpayer-funded research enhances the value of that research, advances the pace of scientific discovery, and maximizes the contributions of human research subjects. “To that end, we encourage the agency to explicitly mandate that researchers they fund, in both the pre-clinical and clinical research settings, share their data unless the agency determines that there is a compelling scientific, ethical, and/or logistical reason not to do so.”
PRIM&R also cautioned the NIH that deidentification, which is currently discussed in the supplemental draft guidance, is “not a viable privacy risk mitigation strategy given that it is no longer possible to guarantee that data will remain permanently deidentified. At the very least, this fact should be appropriately communicated to grantees, oversight bodies, and other relevant stakeholders in both the final policy itself as well as any supplemental draft guidance the NIH develops.”
PRIM&R suggests the agency offer specific guidance on the ethical issues involved in data- sharing for the research oversight community. Such guidance should help institutional review boards (IRBs) and others ensure that participants are adequately informed of the limits of deidentification and include clear recommendations for how both the facts about data sharing and its inherent risks should be conveyed during the informed consent process.
Last November, in an e-mail exchange with Healthcare Innovation, Jeff Smith, AMIA’s vice president of public policy, explained why the organization was disappointed with the draft policy. “The plan delegates a good deal of responsibility to Institutes and Centers, but this is really a continuation of the status quo and an inadequate position from which to coordinate and lead – as the NIH data-sharing policy should do,” he said, adding that NIH’s approach will actually lead to much more variation in data-sharing plans across the entire NIH portfolio. “This is where a ‘check-the-box’ mentality will come into play,” Smith said. “Data sharing will be seen as a red-tape exercise without much (if any) benefit from sharing, assuming the data sharing across projects is even remotely consistent and coordinated.”
As AMIA presented its full comments to NIH this month, it recommended that the NIH finalize an NIH-wide data management and sharing plan over the course of three years that positions NIH Institutes, Centers, and Offices to develop their own requirements, subject to approval by the NIH Office of Data Science Strategy and the Office of Science Policy.
AMIA also encouraged the NIH to take a stronger leadership position in establishing guardrails for Institutes, Centers, and Offices by requiring them to (1) factor the quality of grantees’ plans into the overall impact score through a peer-review process for those grants that are supported at high levels or focused on programmatic priorities; (2) identify and incentivize deposition of scientific data in endorsed depositories and knowledgebases; and (3) Establish graduated Plan requirements based on funding levels, subject to the aforementioned NIH review.
AMIA also recommended the NIH establish a funding policy for data management and sharing activities that earmarks a percentage (at least 5 percent) of a grant award for such activities, rather than merely allow for such activities to be included in NIH budget requests.
Scoring data-sharing plan during grant review process
Founded in 2018, Vivli is a nonprofit organization that manages the world’s largest clinical trial data-sharing platform. It provides a single point of search and request to participant-level data from over 4,700 trials representing 2.2 million participants from 109 countries.
Vivli said the NIH plan as currently drafted is significantly weakened by choosing “deliberate flexibility” over a robust and clear mandate for clinical trial data sharing. “Typically, flexibility in the conduct of science is a benefit; however, in this instance this approach significantly weakens our accountability to participants.”
Vivli recommends the following be mandated elements within the data-sharing plan with respect to clinical trials data rather than ‘flexible’ elements managed at the discretion of the investigator:
• The current proposal leaves open the timeframe for when data would be made available to users at the discretion of researchers. Vivli recommends that NIH-funded clinical trials require reporting of individual participant-level data (IPD) to an approved repository within a reasonable time period.
• The current proposal does not bind clinical trial proposals to declare a particular trial repository in the data-sharing plan. Vivli recommends that NIH establish clear standards, criteria and best practices for clinical trial data-sharing repositories, maintain a list of these approved repositories, promote awareness among researchers of this list, and require investigators to declare which approved repository they will be using.
• For clinical trial proposals, Vivli recommends that NIH institute a requirement that demonstrates a rigorous search of prior relevant summary and IPD results in the research plan section. This would ensure that duplicative trials are not initiated, and that researchers are respecting participants contributions by leveraging them to the fullest.
In conclusion, Vivli notes that perhaps the single most impactful change to the current draft policy would be to score the data-sharing plan during the grant review process and ensure that this score impacts the funding decision. “We have waited for 15 years for this important update to the NIH’s data-sharing policy. As this new policy lacks any effective mandate for sharing of clinical trial data, it in effect relinquishes NIH’s responsibility to the research community, researchers and patients. This incremental proposal if enacted would signal to researchers that clinical trial data sharing is a voluntary endeavor, which breaks trust with trial participants’ strong desire to share. We can do better.”
Different needs for pragmatic clinical trials
A group of researchers involved in pragmatic clinical trials embedded in healthcare systems co-signed a letter with their opinions for NIH. They included investigators and leadership from the National Institutes of Health (NIH) Health Care Systems Research Collaboratory, participants in the National Academy of Medicine (NAM) Clinical Effectiveness Research Innovation Collaborative of the Leadership Consortium for Value and Science-Driven Health Care, and leaders of the Health Care Systems Research Network (HSCRN).
Their letter applauded the policy and the requirement that all research funded by the NIH provide a data management and sharing plan, but they offered additional information regarding different types of research and acceptable mechanisms for data sharing, They recommend:
• Acknowledge in the policy that simple removal of explicit identifiers may be insufficient to protect the needs of stakeholders. Prior to sharing research data, investigators may need to remove or alter data elements that could enable re- identification via linkage.
• Examine and acknowledge the unique data-sharing concerns of other stakeholders, including secondary subjects, who may include health care providers or organizations delivering care to research participants, family members of research participants, or members of other identifiable vulnerable classes.
• Add information regarding different acceptable data-sharing mechanisms to the policy. Indicate that when using data enclaves or other restricted-access data environments, although the data itself cannot be shared, the specific resources and the technical tools used to create and analyze research datasets can be shared.
• Develop mechanisms to link data sets to data generators and track data re-use.