Data sharing is all over academic news now. We had Research Parasites, a noxious species of scientists who want to analyse others’ published data without granting its “owners” co-authorships and a certain control over the interpretations. Then there is a major battle between patients and clinicians about the release of the original data from the so-called PACE trial, originally published in The Lancet, which analysed medical efficiency and economic costs of different therapies for chronic fatigue syndrome/ myalgic encephalomyelitis (CFS/ME). Since the PACE study came out in 2011, the patients, but also a number of academic scientists, remained unconvinced of the published therapy recommendations and suspected a misinterpretation of data. The authors felt harassed and even threatened by the patients’ incessant demands. The relevant research institutions, the Queen Mary University London and the King’s College London, took the side of their clinicians and refused the release of data, using as argument the allegedly inappropriate nature of such requests and the privacy rights of trial participants.
Importantly, the data sharing requests always concerned anonymised patient data, where names and any other personal information of the trial participants was specifically deleted, to avoid any even approximate identification and breach of privacy. Yet even then, several attempts of patients as well as academics, to obtain the anonymised PACE trial data were converted by the universities from academic inquiries into the bureaucratic Freedom of Information Act (FOIA) requests, which were then repeatedly rejected. At the same time, some of the original PACE authors have been apparently somewhat critical of their original interpretations.
In 2012, a cost-effectiveness analysis of the PACE trial therapies was published in the open access journal PLOS One, where the authors by default had to agree “to make freely available any materials and information described in their publication that may be reasonably requested by others for the purpose of academic, non-commercial research”. James Coyne, professor of Health Psychology at the Dutch University Medical Center in Groningen, has since used this clause to demand the release of the published PACE data (Coyne is also an academic editor at PLOS One and writes a PLOS blog). His request was once again converted into a FOIA and turned down by King’s College London as being “vexatious” (just as Queen Mary University did before). The official letter to Coyne read:
“The university considers that there is a lack of value or serious purpose to your request. The university also considers that there is improper motive behind the request. The university considers that this request has caused and could further cause harassment and distress to staff”.
Nevertheless, PLOS One has issued an editorial notification saying:
“we are seeking further expert advice on the analyses reported in the article, and we will evaluate how the request for the data from this study relates to the policy that applies to the publication”.
Coyne, it seems, brings it in his blog post to the point:
“No one forced Peter White [lead author of PACE study,- LS] and colleagues to publish in an open access journal committed to data sharing, but by doing so they incurred an obligation. So, they should simply turn over the data”.
More recently, four tenured US-American biomedical scientists have approached Queen Mary University about the PACE data release. The university’s rejection of their FOIA was much more polite and cited as reason the “explicit guarantees to participants” regarding their data confidentiality.
The PACE trial case might become a key test for the Open Data movement in academia.
Most recently, an essay was published in Nature, authored by Dorothy Bishop, professor of Developmental Neuropsychology at the University of Oxford, and Stephan Lewandowsky, professor in Cognitive Psychology at the University of Bristol. The essay’s title was: “Research integrity: Don’t let transparency damage science”, many took it as general criticism of open science and data sharing. Except that it wasn’t. As I learned, the misleading title was actually designed by the Nature editors, who were possibly keen on creating some controversy and click-baiting or, who knows, maybe they even acted out of general editorial distrust to open data sharing. The Nature editors even employed a cartoon which seemingly compared data sharing advocates with treacherous knife-stabbers.
In fact, as both Bishop and Lewandowsky assured me, they both are strong proponents of data sharing (see this blog post by Bishop). Lewandowsky is also one of the authors of “The Peer Reviewers’ Openness Initiative” and had his own experience of “constantly being accused of not being transparent”.
The Nature essay briefly mentioned PACE trial, and in fact it also generally addressed the concerns of harassment and reputation damage, which were used as excuse by the PACE researchers and their universities as to not to release the trial’s data.
There apparently two main problems with any possible PACE data release. One is the complete breakdown of any reasonable communications between researchers and their critics, and the other is the uncertainty about what kind of consent the PACE trial participants originally gave to the PACE researchers regarding handling of their data.
Bishop comments that in the case of PACE that
“the discussion is infected by a complete breakdown of trust between CFS/ME researchers and their critics. This goes back years before PACE and includes attempts to stop PACE happening at all. Given that history, and others of our right-hand column ‘red flags‘ being ticked by some critics, it is not surprising that the PACE investigators doubt that they will get a fair evaluation from those demanding the data”.
The Oxford neuropsychologist then explained that, given to word limitation, she and Lewandosky were not able to elaborate on their suggestions on how to avoid such a “stalemate situation”, one such solution being:
“setting up a kind of ombudsman , i.e. an ‘honest broker’ who would be able to do statistical analyses from a neutral perspective. […] I don’t think the journals could do it – I’m not talking about someone to decide if data should be released, but rather a group who could agree what analyses needed doing and do them competently, in a buffer zone between researcher and critics. I am in very early discussions about a possibility of having some kind of unit with statistics expertise that would be dedicated to this, but we’d have to persuade someone to fund it. It’s possible that the uproar around PACE might just help this happen, as medical profession is concerned that nobody wants to research this topic in current atmosphere”.
In this regard, Simon Wessely, professor of Psychological Medicine at the King’s College London, indicated over Twitter that “a robust trusted independent system” is being set up at his university “because at the moment there is no trust”. This supposedly with the purpose of assuring proper confidentiality of the trial participants’ data.
Update 19.02.2016: Wessely has shared his views on data sharing and patient privacy in UK, which I refer in a comment below.
The King’s College’s press release from December 18th 2015 stated that they are
“currently seeking further ethical and scientific advice, as well as the advice of patients, on how best to provide independent decisions about appropriate access to relevant data”
Remarkably however, and bringing us back to the “research parasitism” debate, the PACE data was so far apparently available to any scientific collaborator.
This one can deduce from the press release by King’s College researchers:
“We have supplied requested, anonymised data to independent scientists, as part of normal research collaboration”.
In academia, collaboration generally equals co-authorship. This would mean, any PACE collaborator requiring the original trial data would in some way involve the original PACE authors in any follow-up research or evaluation. Whatever their collaborative effort may entail beyond data sharing, it would likely provide the PACE data “owners” co-authorships on all collaborative research publications. These automatic co-authorships (or their equivalent) were roughly what the “research parasites” editorial in the New England Journal of Medicine (NEJM) insisted upon.
I asked Bishop and Lewandowsky for their insights. Why can anonymised patients’ data be readily shared with any new collaborators, whom the original participants never even heard of back then when they signed their consent, yet it cannot be shared with any non-collaborating academics, who are interested in their own independent analysis of the published data?
It seems, the issue is basically the legalese of what exactly the patient participants of the PACE trial have signed their consent to.
Lewandowsky explained that in his field of psychology research, participant consent must include open data:
“The sharing must be governed by legal and ethical considerations: For example, some of my data can be found here, but you will note that some of those require an application. This is because under UK law, participants must declare their explicit consent to sharing for data to be freely available. If that consent was not given (in my case, because the studies were conducted either before this need was recognized or they were collected in Australia where laws are different), then data cannot be completely open. Likewise, there may be confidentiality issues involved, which I discuss here”.
“For medical data, things are far more complicated and nuanced”.
Bishop seemed quite concerned by the possibility that clinical research data may be shared with selected collaborators only, and not with a wider research community:
“If there are inconsistencies in how data-sharing is implemented, this would be problematic. We need to avoid situations where researchers are sole gate-keepers of access to data and refuse access to bona fide researchers who are critical of their work (assuming there are no ethical or legal reasons for refusal). The current data sharing guidelines of the MRC [British Medical Research Council, which also funded the PACE trial- LS] appear to be devised to handle this situation, by requiring independent oversight of requests for data, and having an appeal process when data requests are denied.”
However, in regard to studies without explicit patient participants’ consent to open data sharing, MRC guidelines are unclear on whether the data sharing with non-collaborating researchers is covered by the standard patients’ consent:
“For established studies, the extent to which ethics committees will approve broad consent for future new uses may be constrained by the consent(s) given by participants in the past, which may be rather vague or restrictive (e.g. “…not shared outside the research team”). For new sweeps of existing studies, and new studies, it should be reasonably straightforward to establish consent for sharing such that participants’ data are used to maximal good effect.”
Bishop then explained:
“In brief: for many years the consent form said something like “My data will be kept in a locked filing cabinet and not be available to anyone other than the research team. Then people realised that this was over-restrictive. There has been gradual change to encourage consent that allows for data-sharing.
The other thing is that trial participants would not, in general, give a blanket approval for their data to be used for research. They would have been given an information sheet that specified not just who would see the data, but what purposes it would be put to. More information here”
So how should the situation with the lack of patients’ original consent be resolved?
Like Bishop, MRC guidelines recommend “independent advice and oversight”:
“The access governance process is subject to independent oversight of (a) external requests for data (and related materials, such human tissue and DNA, if appropriate); and (b) new investigation requests from within the study team for uses that are not already approved by MRC (or other relevant funder or sponsor) as part of the study’s research programme”.
Indeed, this is what apparently has been lacking in case of the PACE trial, where the two supposedly impartial universities, Queen Mary University and King’s College, have so far busied themselves by defending their clinical scientists from “harassment and distress” as well blocking all “vexatious” requests of data sharing from independent non-collaborating researchers.
It is probable that a functional solution to sharing of PACE data might be found, likely with the help of some kind of “intermediary honest broker“ proposed by Bishop. If not, at least the PLOS One publication might have to be retracted, due to the lack of patient consent to data sharing. However, even if the anonymised PACE data will be finally shared also with non-collaborating academics for a purpose of precisely specified analysis, it is far from sure that all CFS/ME patients will be satisfied then. All actors are, to put it carefully, somewhat distrustful towards each other.
Update 8.02.2016, 20:20 Apparently, the original patient participant consent form of the PACE trial has been made public already
Update 16.02.2016: The anonymised patient data from the FINE trial on chronic fatigue syndrome therapies efficiencies, which is occasionally referred to as PACE sister trial, has been made publicly available as it was published in PLOS One in December 2015:
“Data Availability: The authors have prepared a dataset that fulfills requirements in terms of anonymity and confidentiality of trial participants, and which contains only those variables which are relevant to the present study. Data are available as Supplementary Information”.
Update 7.03.2016. As James Coyne informed me, PLOS One has issued an editorial note demanding that the PACE trial authors release anonymised patient data. Excerpt:
“We have now carefully assessed the study and sought advice from two editorial board members, who have provided guidance on the data necessary to replicate the cost-effectiveness analyses reported in the article, and thus we have established which data we would expect the authors to share in the context of the analyses presented in this PLOS ONEarticle. We have contacted the authors to request the release of the data, which include individual patient-level data underlying tables in the article”.