The data sharing statement journals require
Journals that follow the International Committee of Medical Journal Editors require a manuscript reporting a clinical trial to state whether and how the trial's individual participant data will be shared, and require trials that begin enrolling from 2019 to record a data sharing plan when they are registered. The statement and the plan commit a sponsor long before the manuscript is written, and journals differ in how much more than a statement they ask for.
What the statement and the plan require
The data sharing requirements of the International Committee of Medical Journal Editors are a condition for considering publication of clinical trial reports in member journals. Manuscripts that report the results of clinical trials and are submitted on or after 1 July 2018 to journals following the committee's recommendations must contain a data sharing statement. In the statement, trial sponsors indicate their readiness and willingness to share individual participant data. Clinical trials that begin enrolling participants on or after 1 January 2019 must include a data sharing plan in the trial's registration. If the plan changes after registration, the change should be reflected in the statement submitted and published with the manuscript and updated in the registry record.
The committee expects the data sharing statement and the data sharing plan to say:
- whether individual de-identified participant data, including data dictionaries, will be shared
- which specific data will be shared
- whether additional related documents, such as the study protocol or the statistical analysis plan, will be available
- how the data will be accessible
- when the data will become available and for how long
- by what access criteria the data will be shared, such as with whom, for what types of analyses and by what mechanism
For a trial that started after 1 January 2019, the submission has to include a data sharing plan covering all the statement's items, along with any deviation from the plan, so publishing in a highly reputable journal requires a plan for data sharing.
How the requirement came about
In the 20 years to 2017, significant strides were made toward best practices for publishing clinical trial results, and by 2017 their latest focal point was data sharing. In January 2016 the committee published a proposal to establish the sharing of deidentified participant data from clinical trials as the norm. The editor-in-chief and the deputy editor of the New England Journal of Medicine quickly pointed to two possible unintended and negative consequences of such sharing. The first was that individual participant data might be used in systematic analyses without a thorough understanding of the protocols that generated them; the second, which generated the hashtag #researchparasites, was their suggestion that some researchers felt that "research parasites", meaning data scientists, would emerge from such broad sharing. Yet data scientists had been working with large data sets for years, especially in genomics. Given the chance by a contributor to Forbes to clarify his opinion, the editor-in-chief did not retract the comment, but in May 2016 he wrote an editorial supporting the committee's recommendations on data sharing, provided appropriate systems were in place to release trial data within six months of publication in his journal. He granted that data scientists can improve human health, while stating that participant data from clinical trials are among the highest-quality data in medicine and should be used responsibly.
After an active and turbulent public discussion of the proposal, the committee announced its two requirements on sharing participant data from interventional trials in June 2017 by one account and in July 2017 by another. It recognised that substantial challenges still kept clinical trialists and their sponsors from releasing participant data, and so required data sharing statements to be included in manuscripts submitted to member journals. At the end of 2017 the committee updated its recommendations, and its guidelines for clinical trial registration came to include the new data sharing policy.
In 2017 the committee acknowledged that several issues remained unresolved: transparent processes for data requests, appropriate scholarly credit for those who acquired the data, and data archiving. One commentator held in 2017 that guidance on these issues would need to be provided soon for clinical trial manuscripts to meet the 2018 deadline. In 2019 the full effect of the requirements, and their resolution with other legal provisions, were still to be settled, especially for the protection of trial participants' personal information and of sponsors' commercially confidential information.
Where registries record the plan
In line with the committee's requirements, the WHO International Clinical Trials Registry Platform expanded its Trial Registration Data Set by four new data elements, including a new field for the participant data sharing statement. ClinicalTrials.gov added a participant data sharing field to its protocol registration data element definitions for new trial registrations. For an interventional study, the registry expects a "Yes" or "No" to its question Plan to Share IPD, IPD being individual participant data. The response is optional in the Protocol Registration and Results System, but the committee requires it as part of the registration information for interventional studies. Describing the plan is not required to register a trial in ClinicalTrials.gov, but the response is visible to the public in the registration record and its history of changes. The plan, with all changes to it, will be reflected in the data sharing statement when the trial's results are submitted to a journal.
In 2018 EudraCT, the EU/EEA clinical trials register, did not yet have fields for the data sharing plan, which were being planned, while ClinicalTrials.gov already had them. In 2019 EudraCT still had no dedicated field for the data sharing statement. Regulation EU 536/2014 specifies that EudraCT is to be replaced by the Clinical Trials Information System for all EU/EEA-relevant trials. In 2019 the new system's implementation had been delayed by technical issues expected to be resolved by late 2020, and in the meantime, with no data sharing field in EudraCT, it was not clear how sponsors of trials performed in the EU/EEA would comply with the committee's data sharing requirements.
What satisfies the requirement and what journals ask beyond it
The committee calls its initial requirements minimum requirements: they do not yet mandate data sharing itself, only statements relating to data sharing, but it warns investigators that editors may take data sharing statements into consideration when making editorial decisions. If the authors are not prepared to share their data, a short statement such as "Data will not be shared" should satisfy the requirements, although the response may affect the editorial decision. Examples of possible statements that fulfil the requirements are available in the committee's editorial, on its website and in a table in its updated recommendations. The final decision on how to deal with a submitted manuscript and its data sharing statement rests with the editor of each journal.
The committee also notes that some journals already maintain, or may choose to adopt, more stringent requirements for data sharing. In March 2019 some member journals, such as PLoS and the British Medical Journal, already required data sharing as a condition of publication, while other member journals had not yet taken a position and were expected to do so in the future. By 2019 many editors following the committee's requirements appeared to grant authors a grace period and did not insist on a clear "Yes" to data sharing as a condition of publication. Even so, some member journals maintain, or may adopt as a condition of publication, stringent requirements not only for a positive statement but also for the trial protocol and data analysis plan, with their amendments, to be submitted with the manuscript. Submitting these documents is meant to avoid selective reporting, endpoint switching and spin, the distorted interpretation of statistically non-significant results, which persist despite editorial advice and the existence of trial registries.
In 2019 several companies reported that JAMA editors had asked for an unredacted trial protocol and statistical analysis plan as part of their review of submitted manuscripts, although redacted documents will probably be made public when a manuscript is published. Another company, after submitting a manuscript to the New England Journal of Medicine with proposed redactions to the protocol and statistical analysis plan, was asked to unredact all commercially confidential information, was allowed to keep personal protected data redacted, and was asked to justify every redaction.
The committee's guidelines do not make data sharing mandatory for non-interventional studies, such as epidemiological registers, but some journals ask for the data set for observational studies, especially where the analysis is tricky or open to different interpretation. One speaker on data transparency holds that data sharing for non-interventional studies is something to start thinking about, because it enhances the credibility of the submitted data. In the same speaker's view, a final conclusion that one intervention is better than another cannot be drawn from a non-randomised study, so where such a study suggests that one intervention may be better, submitting the individual patient data, how they were managed and how they were statistically evaluated is a good way to strengthen the data.
What the plan commits a sponsor to
The data sharing commitments are expected to be made by sponsors before the trial has started, and they have wide and long-term implications for drug development and the product's life cycle, so the sponsor's upper management needs to be aware of them. The requirements concern medical writers working on regulatory and medical communication documents, the data managers and statisticians who collate and process participant data, and the other stakeholders who plan, implement and report clinical trials.
The requirements prescribe neither when the participant data become available nor for how long; each sponsor decides this when it registers the trial. A sponsor that states at registration that it is not willing to share data may face ramifications if the compound is out-licensed or partnered in a co-development agreement: the choice could affect a decision to in-license the compound, the publication strategy, who is responsible for changing the registry's "No" to "Yes", and who provides the rationale for the change when the manuscript is submitted. Initial decisions on data sharing will very likely lead to further discussions between the sponsor and the in-licensing company or co-development partner.
Speakers at a 2019 drug-information meeting made clear that sponsors need well-defined internal processes with clear responsibilities for the data sharing plan and for evaluating data sharing proposals from external researchers. The internal stakeholders responsible could include clinical trial disclosure, therapeutics, regulatory, legal, intellectual property and patents, and publication planning. More than four years can pass between the initial trial registration with its data sharing statement and submission of the manuscript.
Whether a strong policy produces shareable data
Naudet and colleagues assessed the effectiveness of the data sharing policies of The BMJ and PLOS Medicine by gathering data from 37 published randomised controlled trials and reanalysing their primary outcomes. Where reanalysis was possible, it mostly yielded results similar to the original analysis, although it used data at a mature analytical stage. Data availability was not optimal even in these two journals with a strong data sharing policy, but the 46% data sharing rate observed was higher than elsewhere in the biomedical literature. Problems in contacting corresponding authors, lack of resources to prepare the data sets, and heterogeneity in data sharing practices are barriers to overcome. The study showed that the data sharing policy the committee recommends can be implemented, even if not optimally. Few journals had a strong data sharing policy in 2018, so the potential to reanalyse data from trials published in specialty journals was questionable, and further similar studies are needed to improve confidence in publications.