Data-sharing proposal has pros and cons

LAST month the International Committee of Medical Journal Editors (ICMJE) floated its proposal to require authors of papers reporting clinical trials to share with others the deidentified data from the individual patients who took part in the study.

The ICMJE published this as 14 simultaneous editorials, in journals as diverse as the New England Journal of Medicine, Ugeskrift for Laeger and the Ethiopian Journal of Health Sciences.

At first sight, and probably even at second sight, the idea has merit. It aligns with the NHMRC Statement on Data Sharing that says “research data should be made available for use by … other researchers unless this is prevented by ethical, privacy or confidentiality matters”.

Sharing individual patient data, on which the results and conclusions in the original paper were based, will allow independent confirmation of the findings. As the ICMJE writes: “Sharing data will increase confidence and trust in the conclusions drawn from clinical trials”.

• Related: MJA InSight — Concerns about access to trial data
• Related: MJA — Medical specialists and pharmaceutical industry-sponsored research: a survey of the Australian experience

What the ICMJE does not spell out, but is germane, is that this greater transparency will give some protection against scientific fraud – although not of course if the data themselves are fraudulent.

Another advantage will be the opportunity for others to dig deeper into the raw data, looking for outcomes and associations that the original authors had not thought of and did not have journal space to present.

The ICMJE suggests that in such cases the secondary and primary authors should be encouraged to collaborate, giving due credit for the efforts of both.

At present, investigators performing meta-analyses often want detail that was not included in a publication. It might be data they suspect were collected on a subgroup, relevant to the outcome measure they are pooling but peripheral to the focus of the primary authors. They can ask those authors for data. Often no reply is forthcoming. A mandated data repository could be invaluable in these situations.

The issue of protecting the rights of the original investigators will not always be straightforward. What if they intended to go back to further analyse their material for a subsequent paper (most medical journals have strict word limits for each)? An injustice would be done if a requirement to make the source material available let a competitor “raid” their data. After all, academic survival does depend on publication outputs and primacy in getting into print on a particular topic.

The ICMJE recognises that this is a complex area, suggests that agreements will need to be struck between sharer and recipient, and is asking for input from the research community.

At least as important will be questions of additional costs if data need to be made available in ways that will readily make sense to others wanting to access them.

Ethics committees will need to do a little more work to assess the prospective data-sharing plans the editors are calling for. Authors and sponsors for multicentre clinical trials may face the additional difficulty that source data need to be stored separately at each site. Presumably the ICMJE is only thinking about making available the amalgamated data that has been extracted from these, but it is not quite so explicit.

There is also the question of how deep into the raw data it is feasible for an external researcher to actually delve.

Imagine a study whose primary data source is MRI scans. Measurements made on these by the original investigators could clearly be lodged in a repository. Similarly, text reports by radiologists and pathologists are easy to store, and would be expected to be filed with the source data. However, the scans themselves could not be stored.

The final issue we think needs to be considered is how completely data will need to be deidentified – and again the costs that this will add to a study. Should date of birth and other identifying demographics be redacted from the individual patient data? We think they should – all data should be rendered non-identifiable.

Despite these concerns, there seems to be value in what the ICMJE proposes. In any case, the Editors control this space by having decided to mandate it for papers they publish. However, the ICMJE is asking for feedback by 18 April 2016 at www.icmje.org.

We hope they get a lot.

Professor Neville Yeomans and Dr Sianna Panagiotopoulos are, respectively, Director and Manager of Research at Austin Health, Melbourne, with affiliation to the University of Melbourne’s Department of Medicine.