HIDDEN DATA Putting GlaxoSmithKline to the test over Blockbuster paroxetine is no stranger to headlines. The drug is now back centre stage as requests for clinical data from one of its trials are testing its manufacturer’s commitment to full transparency. Peter Doshi reports

hen the Journal of the American In 2012, GSK agreed to pay $3bn in a fraud Academy of Child and Adolescent settlement with the United States government. Psychiatry (JAACAP) published In a statement connected with the lawsuit, the in 2001,1 its editors Department of Justice declared that “the center- could have had no idea that the piece of GSK’s efforts to market Paxil for child- Wpaper would spark a controversy, not only about hood depression was the GSK funded Study the use of the antidepressant paroxetine in chil- 329,” about which the published JAACAP dren but also about secrecy in clinical trials. It is “article distorted the study results and gave a controversy that rages to this day and that goes the false impression that the study’s findings to the heart of recent campaigns to gain access to were primarily positive, when they were, in drug companies’ trial data. al paper concluded that “paroxetine is generally fact, primarily negative.”9 By most accounts, GlaxoSmithKline is lead- well tolerated and effective for major depression Jureidini and colleagues have led a long ing the pack in its efforts to liberate access to its in adolescents.”1 campaign to compel the journal to correct or data. It was the first major pharma- Jureidini was subsequently contracted to retract the article, which was authored by both ceutical company to sign up to the international provide expert advice as part of a class action academics and GSK employees.10 Earlier this AllTrials petition calling for all trials to be lawsuit against GSK in 2004. Through this legal year, Jureidini presented GSK’s chief executive, registered with the full methods and the results action, some internal company documents were , with a final plea to help correct reported.2 Whereas companies like AbbVie and released into the public domain, and Jureidini the scientific record. “Your corporation has so InterMune have lodged lawsuits aiming to block and colleagues reported that study 329 had an far failed to take responsibility for a published access to clinical trial data,3 GSK has forged additional six secondary outcomes specified in report that has harmed young patients who ahead with a new website enabling third party the protocol.6 Paroxetine was not more effective were prescribed paroxetine on the basis of this access to deidentified participant level data than on any of these outcomes either. misleading article. As the CEO of GSK, you have “because it is the right thing to do, both scien- the opportunity to correct the scientific record. I tifically and for society.”4 GSK’s website states Troubled history respectfully urge you to do so,” Jureidini wrote that five requests have been approved up to 20 Paroxetine was a blockbuster antidepressant, (see data supplement on bmj.com ). September. None has been rejected. One is under known by its trade names Paxil in the United But GSK defended the integrity of the 2001 review. States and Seroxat in the United Kingdom, publication. “GSK does not agree that the article But one group’s request for data is testing the and was widely prescribed “off is false, fraudulent or mislead- The drug came under limits of GSK’s commitment to full transparency. label” for use in children and ing,” John E Kraus, head of Jon Jureidini, clinical professor of psychiatry at adolescents. The drug came heightened attention medical governance, wrote to the University of Adelaide, is leading a team to under heightened attention in in the early 2000s, Jureidini. reanalyse and republish the results of GSK’s the early 2000s, after a decade after a decade of rising Jureidini has responded by study 329—a randomised, double blind, placebo of rising antidepressant use antidepressant use assembling a team to reana- controlled trial of paroxetine for the treatment of among youths,7 over concerns among youths lyse and republish study 329. depression in adolescents. For over a decade, about a link between paroxetine In July they publicly declared Jureidini has been critical of how the study was and suicidality in children. their intention to produce a reported in JAACAP in 2001. In 2003, Jureidini In 2003, the UK Committee on Safety of new journal report of study 329, written in and Tonkin wrote to JAACAP: “We believe that Medicines recommended that paroxetine accordance with the BMJ endorsed restor- the Keller et al study shows evidence of distorted not be used in children and adolescents for ing invisible and abandoned trials (RIAT) and unbalanced reporting that seems to have the treatment of depressive illness because initiative, which calls for third party authors evaded the scrutiny of your editorial process.”5 of concerns about an increased risk of self harm to publish or republish unpublished and misre- They noted that “on neither of [the study’s two and potentially suicidal behaviour. And in 2004, ported clinical trials.11 The team’s starting place primary outcome] measures did paroxetine dif- the US Food and Drug Administration placed a is a trove of over 6000 pages from a previously fer significantly from placebo”—yet the Keller et boxed warning, its most serious type of warning, internal clinical study report written by Smith- on all , stating that they increase Kline Beecham in 1998 that was forced into the ЖЖRead about the BMJ’s open data the risk of suicidal thinking and suicidal behav- public domain as a condition of a consent order campaign at bmj.com/open-data iour in these age groups.8 GlaxoSmithKline agreed to in the settlement of

BMJ | 16 NOVEMBER 2013 | VOLUME 347 15 HIDDEN DATA

STORY OF STUDY 329

1994-98 July 2001: 26 August 2004 2 July 2012: 26 April 2013 SmithKline Beecham Keller et al publish study GSK signs consent US Department of Jureidini writes to conducts study 329, a 329 in the Journal of the order with the New York Justice announced GSK chief executive study of 275 adolescents 24 November 1998: American Academy of State Attorney General, that GSK “agreed to requesting his help in with depression Date of SmithKline Child and Adolescent agreeing to pay $2.5m plead guilty and to pay retracting the Keller et 1  billion to resolve al JAACAP article Beecham’s internal Psychiatry and publicly post clinical its criminal and civil clinical study report for study reports for GSK liability arising from the study 329, which was sponsored trials of company’s unlawful made public by the 2004 paroxetine in children and promotion of certain consent order adolescents prescription drugs, its failure to report certain safety data, and its civil liability for alleged false price reporting practices”

a 2004 lawsuit with the New York State Attorney ers. Complete CRFs are available to regulatory Witty’s phrase “all the data” sounds straight- General. 12 (SmithKline Beecham and Glaxo Well- authorities for audit and for them to assure the forward, but the company’s 11 October response come merged in 2000 forming GSK.) The pages integrity of the data sets and CSRs.” However, to Jureidini implied that “all the data” would not include a report of the trial, the study protocol, last week the company suggested a phone call include case report forms from study 329—or, it statistical analysis plan, blank case report forms, with Jureidini, “to explore with you how we can would seem, any other study. Nor, it seems, is and numerous data tables, which Jureidini’s help with this.” GSK going to “publish” any of the deidentifi ed team will use for its analysis. patient level data on its new website, if “pub- But GSK’s public posting of its internal report Commitment to transparency lish” means to make something public and freely on study 329 is incomplete, lacking an unknown GSK’s wavering responses contrast with the accessible. number of pages containing original case report many upbeat public proclamations by its execu- forms from Appendix H. Jureidini and colleagues tives about the company’s commitment to trans- Caveats have therefore asked GSK for access to the parency. “Increasing transparency about our A more careful reading of GSK’s stance is that it deidentifi ed case report forms and the corre- research is a critical area we’ve been pursuing at believes in what it calls a “closed-access system,” 4 sponding deidentifi ed electronic participant GlaxoSmithKline for almost a decade,” Shannon in which only approved researchers are permit- level data, “so that we can GSK’s wavering wrote in a September editorial ted to query (but not download) data in preap- restore the publication of trial in the Huffi ngton Post .13 “We’ve proved ways. To gain access to GSK’s participant 329 in a fair, complete and pub- responses contrast with also launched a new website level data, requestors must fi rst submit and have licly transparent way.” the many upbeat public allowing scientists to request their analysis plan approved by an “independent “With regard to the electronic proclamations by its access to the very detailed, review panel” and sign a data sharing agreement. database,” James Shannon, executives about the anonymised patient-level data A sample agreement posted on GSK’s website GSK’s chief medical officer, company’s commitment sitting behind the results of our indicates that researchers are expected to run wrote to Jureidini on 11 October, to transparency clinical trials. This will mean only preapproved analyses: “GSK and Researcher “I would ask that you do indeed independent researchers, agree that GSK will provide the Researcher with submit an analysis plan via the website and sign with a fresh perspective, can conduct further access to patient level data from the GSK-spon- a data sharing agreement.” Jureidini had initially research which could advance medical science sored clinical studies listed in Exhibit A for the rejected submitting an analysis plan arguing that and improve patient care.” sole purpose of analysis according to Researcher’s “such a plan is irrelevant when restoring a pub- Transparency is “at the core of how we work,” approved research plan (the “Analysis”) attached lication where our primary focus is the original Shannon explains. “People have asked me, as Exhibit B and for no other purpose.”15 analysis plan drawn up and implemented by ‘what if a new side eff ect comes to light for one GSK suggests that such a system is necessary to your own statisticians.” Nonetheless, Jureidini of your medicines? Or what if a scientist discov- protect the privacy of research participants. It is complied, and submitted an analysis plan— ers that you made a mistake in your research?’ not enough to simply remove personally identifi a- mostly a direct copy and paste from the protocol My answer back is ‘why wouldn’t we want that ble information from the participant level dataset. contained in the company’s 1998 clinical study to happen? Isn’t it better that we know? There is “It may be possible to combine deidentifi ed data report—placing GSK’s independent review panel always the potential for us to fi nd a better way with other information to identify individuals. To in the unusual position of refereeing the appro- to do things.’”13 minimize any such risk, our approach will be to priateness of the manufacturer’s own analysis Last month, Witty took it one step further in a provide access to anonymous patient-level data plan. television interview with the US Fox News . “We’re on a password-protected website that has controls As for the case report forms, GSK initially not simply going to publish data on trials still to in place to prevent data from being downloaded rejected Jureidini’s request, explaining, “We do come, but we’re going to go back, and we’re going or transferred.”4 not publicly disclose Case Report Forms (CRFs) to publish all the data for all the trials that have This concern is underscored in an editorial, and we do not provide them to other research- been done since the company was formed.”14 published in the Lancet, coauthored by Patrick

16 BMJ | 16 NOVEMBER 2013 | VOLUME 347 HIDDEN DATA

3 May 2013 13 June 2013 25 June 2013 15 July 2013 28 October 2013 GSK responds that BMJ publishes the GSK expresses support Jureidini publicly Jureidini and “GSK does not agree restoring abandoned for RIAT, stating announces his colleagues formally that the article is and invisible trials that “by making the team’s intent to submit a request for false, fraudulent or (RIAT) declaration 11 Clinical Study Reports republish study  in deidentified electronic misleading” available we are very accordance with the participant level data happy for others to RIAT initiative through GSK’s website. publish on the records The result of their if they wish to and if 10 October 2013 request is pending journals consider the Andrew Witty tells review by GSK’s work to be of scientific US Fox News: “We’re independent review merit” going to publish all the panel data for all the trials that have been done since the company was formed”

Vallance, GSK president of pharmaceuticals to GSK’s follow-up of patients who were in Study sharing: “responsible.” The Institute of Medicine research and development, and Iain Chalmers, 329? For instance, were those who became sui- has named its ongoing consensus study on the one of the founders of the Cochrane Collabora- cidal or violent on Paxil subsequently advised of topic “Strategies for Responsible Sharing of tion and now coordinator of the James Lind the possible role of the drug in their dangerous Clinical Trial Data” and a recent essay in the New Initiative. 16 They write that “the protection of and distressing feelings/actions and counselled England Journal of Medicine entitled Preparing privacy is vital in IPD [individual participant that it may be better for them to avoid SSRIs [selec- for Responsible Sharing of Clinical Trial Data, data] analyses,” but point out that the process of tive serotonin reuptake inhibitors] in future?” warns that “poor-quality analyses can harm deidentifi cation can be carried out so thoroughly Perhaps most concerning, Jureidini explained rather than advance public health.”17 as to render the scientifi c value of the data use- to GSK that his team has concerns about how less. Deidentify the data insuffi ciently, and trial some of the adverse event data were reported in Irony of study 329 participants may be re-identifi ed, violating their the 1998 clinical study report and therefore needs There is a certain irony in the story of study 329 privacy. Vallance and Chalmers posit that a “con- the additional requested data. and its 2001 publication in JAACAP . In a letter trolled system” of restricted access off ers a pos- Recently, GSK has soft ened its position and to Jureidini, GSK explained that the JAACAP pub- sible solution to the reidentifi cation problem.16 seems willing to discuss the possibility of rethink- lication “was subjected to peer review on three It is therefore unsurprising that GSK initially ing its previous position. “I recognize, however, occasions” and “accurately refl ects the honestly- refused Jureidini and colleagues access to the case that you believe you need to see the CRFs and I held views of the clinical investigator authors.” report forms on grounds of protecting the privacy would like to explore with you how we can help But a more pertinent question is whether the of trial participants. “The content of Appendix H with this,” Shannon wrote. “Please could we published article accurately refl ects the trial. is not posted on our website because it contains arrange a telephone call to discuss this more fully In their most recent letter to GSK, Jureidini information (such as names) that can be used to and agree a way forward?” and his colleagues reiterate their need for the readily identify the patients concerned.” Jureidini has declined the telephone call, and study 329 case report forms and their intention But Jureidini and company are challenging requested keeping the interactions by email. “I to analyse study 329 “following the original ana- GSK. “As a group we do not accept your argument would be grateful if you could indicate what in lytic plan.” As such, Jureidini’s team’s eff orts to about patient confi dentiality . . . The blank case your opinion is the safest way of getting all CRFs independently analyse and publish the results report forms (CRFs) in the Clinical Study Reports from study to me, suitably de-identified, but of study 329 can be viewed as perhaps the most (CSRs) make it clear that the only patient identi- otherwise complete with narrative elements “responsible” of all analyses—and one that it fi ers in any CRF not contained in the CSRs were intact.” seems may yet overturn the JAACAP publication initials. Redacting initials is the work of minutes.” that GSK continues to defend. Jureidini adds that reidentifi cation of patients “is “Responsible” data sharing Peter Doshi associate editor, BMJ, London WC1H 9JR, UK not our intention. We think anyone in our group Jureidini’s quest to access the complete partici- [email protected] attempting to do this would do signifi cant dam- pant level data for study 329 highlights some of Competing interests: I am the fi rst author of the RIAT declaration, which was co-authored by David Healy, who is age to the data access cause. We are happy to sign the anxieties surrounding disclosure of clinical part of Jureidini’s team. I am providing the Jureidini team with agreements that there will be no eff ort to identify trial data. unpaid advice on the RIAT process. I am also a member of a Cochrane review of neuraminidase inhibitors that is based in anyone and that the de-identifi ed CRFs will not be Looming large are concerns about misuse of part on clinical study reports provided by GSK for . shared with anyone outside the 329 group, with data. “We believe that there are public health I initiated an inquiry in 2012 that resulted in an additional access limited to two to three designated individu- risks if the proposed analyses are not scientifi - 38 781 pages from the clinical study reports of study 329 and eight other studies to be publicly posted on GSK’s website. als within our group.” cally robust and give rise to erroneous concerns Provenance and peer review: Commissioned; not externally Jureidini also questioned GSK’s commitment about safety or false hopes of a potential benefi t peer reviewed. to its patients: “noting the concern you expressed for patients,” GSK has declared. 4 This fear of References can be found in the version on bmj.com. in your letter for the wellbeing of patients who misleading analyses is embodied in an adjective Cite this as: BMJ 2013;347:f6754 participate in clinical trials, can we enquire as gaining popularity among discussions over data

BMJ | 16 NOVEMBER 2013 | VOLUME 347 17