Comments on the ICMJE’s Proposals for Sharing Clinical Trial Data

Displaying 26 - 30 of 320 comments
  • Trish Hepburn
    Nottingham Clinical Trials Unit, University of Nottingham
    Role(s): Clinical trialist
    Date Submitted: Monday, April 18, 2016 - 09:41

    Requirement To Share Data
    • I agree with this general approach
    Comments: We support the principle of data sharing and agree that only the data underlying the results presented should be available. However, we are concerned at the lack of detail contained within the proposal. Data sharing requires both expertise and resource which may not be available eg due to funding restrictions, and that in turn will impact on publication of high quality studies. De-identification of participant data can be complex, especially when studies are in rare disease areas, which analyse rare events and where there are few participants per centre. Where data is collected and published sequentially (eg clinical end points, sub studies, long term follow-up), the original de-identification procedures may not remain robust. There is also the risk that the use of de-identified IPD will not allow reproduction of published results, thus leading to time-consuming correspondence for authors. There are practical considerations about what data can be shared given varied regulations between territories. There is no indication about whether data should be shared unconditionally, or whether controlled systems are acceptable. If sharing is unconditional consideration should be given to the motives of requesters and potentially a distinction made between commercial and non-commercial organisations.

    6 Month Time Frame
    • I agree with this general approach
    Comments: This needs to be considered alongside the time and resource required to obtain, de-identify the data and provide supporting documentation. Analyses in some publications may be performed by different researchers using different datasets – eg clinical and Health Economic end points may be analysed separately. Such datasets would have to be brought together. De-identification is not a standard procedure and needs careful planning to ensure that participants cannot be identified at any point. There is no detail about the requirement for supporting documentation, but it is assumed that this would be necessary and that standard study documentation may not suffice. Many smaller research teams may just not have the expertise or the capacity to provide this and having such a short time restriction on such provision also has the potential to delay secondary and exploratory analyses and therefore the publication of important and relevant additional results. In addition, since systems are not yet in place to hold such data, consideration would have to be given to the time it would take for institutions to both investigate such systems to ensure they were compliant with legislation for data storage and to become familiar with using these systems.

    Require a Data Sharing Plan
    • I agree with this general approach
    Comments: We agree that the intent for data sharing should be documented prospectively. Further detail would be required as to what the editors would accept as a ‘data sharing plan’. There should be flexibility over where details of data sharing intent is held. Where registries do not include an item on data sharing, consideration should be given to documentation such an undertaking to share in the protocol publication or even as part of a specific agreement with the journal.

    Providing Credit
    • I agree with this general approach
    Comments: Credit should always be given to the team who collected the data and indeed the trial participants. Different scenarios will require different levels of credit eg the provision of control arm data to enable a sample size calculation to design a study vs the use of a data set to investigate a hypothesis generated by a different group of researchers. There should also be a statement about the level of input the original authors have had to the publication. For example, have they collaborated in the new research, were they aware of the research, do they agree with the findings, if the publication is questioning their original research, have they had opportunity to respond etc.

    Other Comments: Care needs to be taken to ensure that good research is not “lost” through an inability to fully meet any new guideline regarding data sharing eg through a lack of resource or due to restrictions through institutional, national or international regulations. There is the potential that free provision of IPD data could decrease number of participants consenting to be in clinical trials which could slow down research and result in a less generalizable participant population. It is not known whether funders will finance the additional resource required to service data sharing requirements. There will be on-going costs for provision of this data and supporting documentation and also a requirement for on-going support to respond to queries and additional findings. Resource may also be required to ‘police’ any system to ensure that data is being used according to stated purposes and is not being used for the purpose originally intended (ie the purpose ethical approval was given for). There needs to be controls placed on publications arising from shared data. It should be registered prospectively (before receipt of data) to avoid “data dredging”. There also needs to be consideration of whether data can only be shared from the original source.

    Position:
    Comments:
  • Paula Anne Newman-Casey
    University of Michigan
    Role(s):
    • Researcher
    • Clinician
    Date Submitted: Monday, April 18, 2016 - 09:37

    Requirement To Share Data
    Comments: While such sharing sounds good in principle, it may have a chilling effect on participant enrollment in trials, especially in small scale studies. The assumption that the public always wants transparency runs counter to the logic proposed in the recent U.S. Notice of Proposed Rule Making regarding the Common Rule that people want more complete autonomy regarding their samples and information. Deidentified data is one thing in a multi-site trial enrolling thousands of people. It’s very different in a single site trial that aims to enroll 20 people with a special condition. Participants could well fear that insurance companies, for example, would mine trials in certain zip codes to change rates they’re charging. GINA does not offer protection for things like that. Without serious research to validate that enrollments will not be harmed, this policy could hurt, rather than promote, scientific discovery. In order to allow such research to be carried out, a slowly phased in approach would be more appropriate.

    6 Month Time Frame
    Comments: The six month timeframe is inadequate for sharing deidentified individual patient data. Some analyses may be completed and prepared for publication before other analyses may have even begun. Imposing a short deadline like this may appear to try to encourage more rapid dissemination of results, but it may have the reverse effect, whereby authors will hesitate to publish any results until all the related analyses have been completed, so that this timeframe doesn’t impinge on their own or collaborators’ ability to get the full benefit of their years of research. Either lengthening the standard timeframe to two years, or correlating the timeframe in some fashion to the length of the trial and the size or complexity of the data sets would be more appropriate.

    Require a Data Sharing Plan
    Comments: The inclusion of data sharing plans as part of clinical trial registration sounds reasonable. However, few models for such plans exist. Even in the NIH’s currently proposed protocol templates data sharing section pp. 45, 46, there is no actual data sharing model provided. Only after ICMJE or other authorities can draft a reasonable number of model plans that are easily understandable, easily followed, and appropriately tailored to a variety of research and data types, complete with the metadata criteria that would be needed, will it be appropriate to require such public exposure.

    Providing Credit
    Comments: Those investigators who would like to use the data should find a mechanism to help subsidize the cost of the research it took to produce the data.

    Other Comments: This policy has been created by a few with little if any budgetary responsibility for the costs of its implementation. At a minimum, sunset provisions for data availability/ maintenance are needed to avoid an ever increasing tax. Researchers, universities, and libraries all need to be able to cull their collections periodically. A standard minimum for data availability (10 years?) may work for those that have centralized resources, but even such an approach ignores differential costs when data is collected via specialized equipment with its own formats, situations where data isn’t all centrally curated, or data storage technology becomes obsolete etc. The costs this imposes on research need to be measured and weighed. The one year timeline from proposal announcement to implementation based on trial enrollment start date is inadequate. Studies that have IRB approval but are not yet enrolling when the policy applies would have to amend their consents, clogging the IRB system unnecessarily. Researchers and IRBs will struggle with developing informed consents adequate to explain data sharing plans. Developing appropriate templates will take significant effort. The policy should be written to apply only to trials that are granted IRB approval 2 years after any policy is announced.

    Position:
    Comments:
  • Anastassia Negrouk
    EORTC
    Role(s):
    • Clinical trialist
    • Clinical trial sponsor / funder
    Date Submitted: Monday, April 18, 2016 - 09:04

    Requirement To Share Data
    • I agree with this general approach
    Comments: EORTC agrees with the statement that there is an ethical obligation to responsibly share data. Therefore, EORTC has established its publically available policy on data sharing for external research projects (http://www.eortc.org/wp-content/uploads/2011/11/POL008-v-2.3-NS.pdf). Similarly, EORTC welcomed provisions put in place in the framework of the EU clinical trials regulation aiming to increase transparency and data sharing. EORTC notes that the scope of requirements proposed by International Committee of Medical Journal Editors goes beyond EU and also concerns other types of trials and therefore is of a real added value to already existing legal incentives. However, aside this general support to the initiative and the document itself, EORTC would like to raise a number of aspects that need to be further clarified and adjusted. The draft document is not entirely clear about the scope of proposed data sharing. Does it aim to exclusively enable other researchers and/or organizations to verify and reproduce results of the research published or does it also have an ambition to stimulate data sharing for the purpose of new research purposes? Clearly stimulating data sharing for future research purposes is essential.

    6 Month Time Frame
    • I agree with this general approach
    Comments: EORTC appreciated the understanding of the practical aspects involved in the implementation of these requirements and the provision to allow for 1 year after the implementation to enable researchers to adapt patient information and to proceed with Ethical approvals. However, this document is not consistent in the way to describe when this 1 year period starts. EORTC would suggest as stipulated in the beginning of the document that requirements would apply “for clinical trials that begin to enroll participants beginning 1 year after the ICMJE adopts its data-sharing requirements…”. EORTC would also like to suggest to add “… unless such a sharing was specifically prohibited by either Ethical committees or any other competent body”. Indeed, EORTC sometimes faces Ethical committees or authorities that do not allow any data sharing, unless data are fully anonymized or prohibit data sharing with some types of stakeholders. EORTC would also like to mention that many Ethical committees suggest allowing patients to make a choice whether they wish to have their data shared. Therefore, the one shall be conscious of the fact that patient choices may impact the capacity the full data set (as opposed to the sub-set of patients having agreed on sharing).

    Require a Data Sharing Plan
    • I agree with this general approach
    Comments: It is also not very clear if proposal aims to enforce peer to peer data sharing or if it aims to mandate making de-identified IPD public. In case the intention is to make IPDs public, EORTC would have serious concerns about patient confidentiality in certain cases. Indeed, IPD we hold are essentially made of data which are at the “extremely high risk of disclosure” since it frequently includes genomic data or time to event data, known for their specific risk of re-identification from de-identified sets. Methods to limit these risks may indeed exist / be developed at least for some types of data, but will likely require a lot of cumbersome data preparation. The above methods may still not be sufficient to protect data such as genetic data, as already mentioned above, or patient reported outcome data. Indeed, sharing data originating from Patient-Reported-Outcome-Measures (PROMs), generally used to assess Quality of Life, would imply that questionnaire answers from the patient are made public. This contravenes the general policy of PROMs where the patient is informed that individual responses will not be disclosed to ensure unbiased responses. Note that some of these answers may include narratives (in native languages). See also below.

    Providing Credit
    • I agree with this general approach

    Other Comments: Further to the above, in our view, making such datasets directly publically available, gives no possibility to guarantee data are not being misused for unethical purposes. Currently the EORTC model is one of “user agreement for off-site use”, where assess is given directly to the person/organization who will be in charge of the analysis. Last, but not the least, the term of de-identified data is controversial. Indeed, currently and unfortunately, there is no recognized data de-identification standards that would be universally accepted. Accepted practices sensibly differ between different regions and even within EU. Therefore, it is more appropriate to speak in terms of degree of de-identification which shall be adapted to the type of data and intended use. EORTC would propose to introduce this risk-based notion in the introduction of the proposal and to systematically replace “de-identified IPD” by “appropriately de-identified IPD” In conclusion, EORTC welcomes the initiative, but urges the International Committee of Medical Journal Editors to appropriately address the above issues. The final framework shall remain flexible to stimulate sharing while protecting patient confidentiality without duplicating or making more cumbersome already existing methods, frameworks or procedures when proven to fit the purpose.

    Position:
    Comments:
  • Jennifer Darragh
    Johns Hopkins University, Welch Medical Library
    Role(s): Other
      • Data Services Librarian
    Date Submitted: Monday, April 18, 2016 - 08:53

    Requirement To Share Data
    • I agree with this general approach
    Comments: I would recommend that best practices for de-identification be established as well as referenced on the ICJME Website as a resource to authors.

    6 Month Time Frame
    Comments: Six months is too short. For those collecting longitudinal data or those who have funding for several years could be generating many papers on the data. How do you preserve first publishing rights?

    Require a Data Sharing Plan
    • I agree with this general approach
    Comments: Would this be a data sharing plan specific to ICMJE or would a mandated data management plan approved by a funding agency be sufficient? As in, would researchers be able to attach an existing plan. If ICMJE has its own, would there be some sort of template provided?

    Providing Credit
    • I agree with this general approach
    Comments: As many hands are involved in the collection of data, a proper metadata schema to collect project staff names and roles should be applied. Training for proper metadata application should be recommended. For example, as they are further developed point to http://metadatacenter.org/ or https://biocaddie.org/group/working-group/working-group-3-descriptive-metadata-datasets

    Other Comments: Developing best practices, recommendations and partnering with medical libraries to develop workflows for data deposit are key to ensuring compliance with data sharing. In addition, as others have noted – individual-patient level data, even with direct identifiers removed, still needs to be protected. Archives with proven protocols to vet secondary research requests and enforce security restrictions should be recommended as points of deposit.

    Position:
    Comments:
  • Manesh Patel
    Duke University - Duke Clinical Research Institute
    Role(s): Clinical trialist
    Date Submitted: Monday, April 18, 2016 - 08:40

    Requirement To Share Data
    • I agree with this general approach
    Comments: I am working with a group of clinical researchers (co-lead by C. Michael Gibson) to put an official response together for peer review publication. We agree with transparency and data sharing. However we have concerns if there is not an orderly manner in which this is done. Specifically, our two main concerns revolve around the potential for inaccurate and erroneous results from unstructured non-specified analysis and analysis without hypothesis. We worry about the multiplicity of testing and a variety of potential errors that we highlight in our response. We are also concerned about patient privacy and methods to get access to the data.

    6 Month Time Frame
    Comments: We do not agree with the 6 month timeframe for a variety of reasons which importantly include actually providing planned, per-specified analysis to be conducted. We will also highlight a variety of actual practical and operational issues that have to determined for this to occur.

    Require a Data Sharing Plan
    • I agree with this general approach
    Comments: We agree that a plan should be shared. However, we think a lot more work and consideration has to go into understanding how the access and data will be shared, specifically around governance, plans for updating queries for the data, and improving and sharing the knowledge gained. We

    Providing Credit
    Comments: We are not sure what the alternative means of sharing credit would be beyond collaborating and improving scientific output. We look forward to this conversation.

    Other Comments: We have formed a group called ACCESS CV - Academic Research Organization Consortium for Continuing Evaluation of Scientific Studies - Cardiovascular. We will aim to provide a published document that describes our proposal and recommendations for data sharing.

    Position:
    Comments:

Pages