PERS PE C T IV E
Access to Data from Industry-Sponsored Clinical Trials
Data Sharing, Year 1 — Access to Data from Industry-Sponsored Clinical Trials Brian L. Strom, M.D., M.P.H., Marc Buyse, Sc.D., John Hughes, B.Sc., and Bartha M. Knoppers, Ph.D.
T
here has been considerable interest of late in increasing the transparency of clinical trials, including increasing access to the raw data from trials sponsored by the pharmaceutical industry.1 Since May 2013, investigators have been able to request access to deidentified patient-level data from clinical trials sponsored by GlaxoSmithKline,2 subject to review and oversight by an independent review panel (https:// clinicalstudydatarequest.com/ Default.aspx). As the members of this panel, we now have more than 12 months of An audio interview with Dr. Strom is experience with this available at NEJM.org initiative — and can report that it has been a productive and successful first step. The system was launched on May 7, 2013, and initially included approximately 200 clinical trials that had been started since January 1, 2007 — the date on which more consistent data standards began to be required for GlaxoSmithKline-sponsored trials. Additional studies were to be added to the system over time or by request. Data were to be requested through a publicly available website and were to include the raw data set, the analysis-ready data set, and the clinical study documents, but no original radiographs, electrocardiograms, or other images or test results, except by special request. Analyses were to be conducted by the investigators with the use of statistical software provided on a separate secure website, so that the data were not downloaded onto
2052
other computers. Investigators were responsible for obtaining any approvals required for their research (e.g., from institutional review boards). They were also required to sign a data-sharing agreement committing them to using the data only for the research purpose described in the accepted research proposal. Investigators were expected to disclose any conflicts of interest and to provide their plans for publishing the analyses in the medical literature. Approved applications are made public on the website. The independent review panel was charged with accepting or rejecting proposals on the basis of their scientific rationale and relevance to medical science or patient care. The panel also considered the qualifications of the investigators, the existence and management of any potential conflicts of interest, and the clarity of a plain-English lay summary provided as part of the proposal. GlaxoSmithKline currently has approximately 950 studies listed for request. In January 2014, the system transitioned into a multisponsor site, with 10 companies currently participating. With all these companies combined, there are currently more than 1200 studies listed on the site. Sponsors are building their lists of studies over time, with some like GlaxoSmithKline listing studies conducted over a 10-to-15-year period and others choosing to list only studies for newly approved products; the criteria used for listing studies are summarized
on the site. In recent months, the system has added a question asking investigators who claim that they do not need approval from an ethics committee or institutional review board to specify why they deem such approval unnecessary. The table shows the number of submitted research proposals at the various stages of the process. Between May 2013 and May 2014, a total of 36 research proposals were approved (or approved with conditions) by the independent review panel. Of these, 23 had progressed to a signed datasharing agreement. Each of these proposals included requests for data from 1 to 11 studies (average, 2 studies). Several categories of research were represented: studies of risk factors or biomarkers (six proposals), such as investigations examining the predictive value of blood–cell ratios for treatment outcomes or the effect of tumor size on survival; methodologic studies (five proposals), such as efforts to develop predictive toxicology tools or a risk model for data analysis; studies comparing treatment regimens (three proposals), including a meta-analysis of studies of epilepsy treatments and a study comparing the safety of anticoagulants; studies aimed at optimizing treatments (three proposals), for instance, by calibrating activated clotting time to avoid bleeding complications or honing a dose algorithm; patient-stratification efforts (three proposals), for example, to assess differential
n engl j med 371;22 nejm.org november 27, 2014
The New England Journal of Medicine Downloaded from nejm.org at NEW YORK MEDICAL COLLEGE on August 10, 2015. For personal use only. No other uses without permission. Copyright © 2014 Massachusetts Medical Society. All rights reserved.
PE R S PE C T IV E
Access to Data from Industry-Sponsored Clinical Trials
Experience with the Open-Access Clinical Trial System, May 7, 2013, through May 31, 2014.* Part of the Process
No. of Proposals
Submission
58
Requirements check In process
4
Withdrawn by the requestor
2
Did not meet requirements or further details were required
7
Met requirements
45
Review by independent review panel In process
6
Rejected or advised to resubmit
3
Approved or approved with conditions
36
Data-sharing agreement In process
12
Withdrawn by the requestor
1
Agreed and signed
23
Data preparation In process
10
Complete, with data available
13
Research project in process
13
* A total of 4 of the 37 proposals submitted since January 2014 have included requests for data from multiple sponsors. The requirements check is the check done by the sponsor of the study whose data are being requested to make sure the information is complete and that the proposal meets the requirements of this initiative and the sponsor’s requirements for informed consent. Proposals are then sent to the independent review panel.
effects in different ethnic groups or to stratify patients’ responses; and miscellaneous other studies, including one proposal to evaluate the relationships between influenza vaccination and lifestyle factors, one to assess the effectiveness of paroxetine in treating depression in adolescents, and one to attempt to elucidate the expected frequency of severe adverse events before a new trial, to be conducted in the developing world, was designed. There were requests for 97 GlaxoSmithKline studies that were not included in the system. Of these, 20 turned out to be general inquiries that did not require data access through the system;
of the others, 3 are still undergoing feasibility assessment, and 74 have been approved. There is considerable interest in open access to clinical trial data on the part of both the research community and the wider public. Limited but important evidence suggests that reanalyses of previously published data can lead to conclusions about the types of patients who should receive a given treatment that differ from the conclusions drawn by the original investigators.1 Many government funding agencies require investigators whose trials they support financially to share their data. The system created by Glaxo SmithKline was pathbreaking in
allowing access to individual patient–level, analyzable data from clinical trials sponsored by the pharmaceutical industry. In this system, access to the data could be denied by the company before committee review only if it was considered logistically impractical to provide it. The independent review panel’s remit was to make data available for responsible use. The committee did not see its role as performing a scientific review of the proposals as a funding agency would do; access to the data was not intended to be inherently limited, so the committee believed that a rigorous assessment of the scientific merit of each project should be left for later peer review. Hence, the default decision was to permit access to the data. However, the committee was also tasked with screening for “‘rogue’ reanalysis by nonexperts or by analysts who have conflicts of interest,”1 analyses by teams lacking the skills to perform such analyses, and analyses without accompanying plans for disclosing the results publicly. Although none of the data provided have yet led to peer-reviewed publications, the approach we have used comes very close to fulfilling all the criteria for reanalyses that have recently been set forth by Christakis and Zimmerman.3 A number of other companies have since joined this system, which naturally permits easier aggregation of data from multiple companies. Some companies have elected to join other dataaccess systems.4 When manufacturers came to the independent review panel seeking to participate using alternative approaches — suggesting, for instance, that the company be allowed to con-
n engl j med 371;22 nejm.org november 27, 2014
2053
The New England Journal of Medicine Downloaded from nejm.org at NEW YORK MEDICAL COLLEGE on August 10, 2015. For personal use only. No other uses without permission. Copyright © 2014 Massachusetts Medical Society. All rights reserved.
PERS PE C T IV E
Access to Data from Industry-Sponsored Clinical Trials
duct a scientific review of each proposal before passing it along to the committee — we declined, believing that such an approach would compromise our independence. The system described here will undoubtedly evolve considerably over time. Management of the proposals is likely to be turned over to an independent organization, a development that should make it more appealing for other manufacturers to join and should enhance public confidence in the process. We hope that academic researchers will be adding their data to such systems in the future.
Some of us believe that all clinical trial data should eventually be put into the public domain for unconditional, universal access, but we see the current initiative as an important, swiftly implemented, and successful first step. The number of data requests submitted in the first 12 months after the system was launched (58 requests) is a testament to the eagerness of the medical community to reap the benefits of data sharing. Disclosure forms provided by the authors are available with the full text of this article at NEJM.org. From Rutgers Biomedical and Health Sciences, Newark, NJ (B.L.S.); Hasselt Univer-
sity, Hasselt, and the International Drug Development Institute, Louvain-la-Neuve — both in Belgium (M.B.); the United Kingdom Clinical Research Collaboration Board, London (J.H.); and the Center of Genomics and Policy, McGill University, Montreal (B.M.K.). This article was published on October 15, 2014, at NEJM.org. 1. Ebrahim S, Sohani ZN, Montoya L, et al. Reanalyses of randomized clinical trial data. JAMA 2014;312:1024-32. 2. Nisen P, Rockhold F. Access to patientlevel data from GlaxoSmithKline clinical trials. N Engl J Med 2013;369:475-8. 3. Christakis DA, Zimmerman FJ. Rethinking reanalysis. JAMA 2013;310:2499-500. 4. Krumholz HM, Ross JS, Gross CP, et al. A historic moment for open science: the Yale University Open Data Access project and Medtronic. Ann Intern Med 2013;158:910-1. DOI: 10.1056/NEJMp1411794
Copyright © 2014 Massachusetts Medical Society.
Ebola Virus Disease in West Africa — Clinical Manifestations and Management Daniel S. Chertow, M.D., M.P.H., Christian Kleine, M.D., Jeffrey K. Edwards, M.D., M.P.H., Roberto Scaini, M.D., Ruggero Giuliani, M.D., and Armand Sprecher, M.D., M.P.H.
I
n resource-limited areas, isolation of the sick from the population at large has been the cornerstone of control of Ebola virus disease (EVD) since the virus was discovered in 1976.1 Although this strategy by itself may be effective in controlling small outbreaks in remote settings, it has offered little hope to infected people and their families in the absence of medical care. In the current West African outbreak, infection control and clinical management efforts are necessarily being implemented on a larger scale than in any previous outbreak, and it is therefore appropriate to reassess traditional efforts at disease management. Having cared for more than 700 patients with EVD between August 23 and October 4, 2014, in the largest
2054
Ebola treatment unit in Monrovia, Liberia (see diagrams), we believe that our cumulative clinical observations support a rational approach to EVD management in resource-limited settings. Early symptoms of EVD include high fever (temperature of up to 40°C), malaise, fatigue, and body aches (see table).2,3 The fever persists, and by day 3 to 5 of illness, gastrointestinal symptoms typically begin, with epigastric pain, nausea, vomiting, and diarrhea. Patients routinely presented to our facility after 2 or 3 days of severe vomiting or diarrhea, during which they posed a substantial risk to their communities and had a high probability of testing positive for Ebola virus in blood by polymerase chain reaction (PCR). Although some patients tested
positive on PCR within 24 hours after symptom onset, we found that a negative test result could not be relied on to rule out disease until 72 hours after symptoms began. Of the patients who tested positive for Ebola, none that we were aware of had contracted disease from an infected contact during the early febrile phase of illness. No ancillary testing was available in our facility. We observed that recurrent episodes of emesis resulted in an inability to tolerate oral intake. Large volumes of watery diarrhea estimated at 5 or more liters per day (a manifestation not unlike that of cholera) presented suddenly, persisted for up to 7 days or (rarely) longer, and gradually tapered off. Associated signs and symptoms included asthenia,
n engl j med 371;22 nejm.org november 27, 2014
The New England Journal of Medicine Downloaded from nejm.org at NEW YORK MEDICAL COLLEGE on August 10, 2015. For personal use only. No other uses without permission. Copyright © 2014 Massachusetts Medical Society. All rights reserved.
Access to Data from Industry-Sponsored Clinical Trials
Data Sharing, Year 1 — Access to Data from Industry-Sponsored Clinical Trials Brian L. Strom, M.D., M.P.H., Marc Buyse, Sc.D., John Hughes, B.Sc., and Bartha M. Knoppers, Ph.D.
T
here has been considerable interest of late in increasing the transparency of clinical trials, including increasing access to the raw data from trials sponsored by the pharmaceutical industry.1 Since May 2013, investigators have been able to request access to deidentified patient-level data from clinical trials sponsored by GlaxoSmithKline,2 subject to review and oversight by an independent review panel (https:// clinicalstudydatarequest.com/ Default.aspx). As the members of this panel, we now have more than 12 months of An audio interview with Dr. Strom is experience with this available at NEJM.org initiative — and can report that it has been a productive and successful first step. The system was launched on May 7, 2013, and initially included approximately 200 clinical trials that had been started since January 1, 2007 — the date on which more consistent data standards began to be required for GlaxoSmithKline-sponsored trials. Additional studies were to be added to the system over time or by request. Data were to be requested through a publicly available website and were to include the raw data set, the analysis-ready data set, and the clinical study documents, but no original radiographs, electrocardiograms, or other images or test results, except by special request. Analyses were to be conducted by the investigators with the use of statistical software provided on a separate secure website, so that the data were not downloaded onto
2052
other computers. Investigators were responsible for obtaining any approvals required for their research (e.g., from institutional review boards). They were also required to sign a data-sharing agreement committing them to using the data only for the research purpose described in the accepted research proposal. Investigators were expected to disclose any conflicts of interest and to provide their plans for publishing the analyses in the medical literature. Approved applications are made public on the website. The independent review panel was charged with accepting or rejecting proposals on the basis of their scientific rationale and relevance to medical science or patient care. The panel also considered the qualifications of the investigators, the existence and management of any potential conflicts of interest, and the clarity of a plain-English lay summary provided as part of the proposal. GlaxoSmithKline currently has approximately 950 studies listed for request. In January 2014, the system transitioned into a multisponsor site, with 10 companies currently participating. With all these companies combined, there are currently more than 1200 studies listed on the site. Sponsors are building their lists of studies over time, with some like GlaxoSmithKline listing studies conducted over a 10-to-15-year period and others choosing to list only studies for newly approved products; the criteria used for listing studies are summarized
on the site. In recent months, the system has added a question asking investigators who claim that they do not need approval from an ethics committee or institutional review board to specify why they deem such approval unnecessary. The table shows the number of submitted research proposals at the various stages of the process. Between May 2013 and May 2014, a total of 36 research proposals were approved (or approved with conditions) by the independent review panel. Of these, 23 had progressed to a signed datasharing agreement. Each of these proposals included requests for data from 1 to 11 studies (average, 2 studies). Several categories of research were represented: studies of risk factors or biomarkers (six proposals), such as investigations examining the predictive value of blood–cell ratios for treatment outcomes or the effect of tumor size on survival; methodologic studies (five proposals), such as efforts to develop predictive toxicology tools or a risk model for data analysis; studies comparing treatment regimens (three proposals), including a meta-analysis of studies of epilepsy treatments and a study comparing the safety of anticoagulants; studies aimed at optimizing treatments (three proposals), for instance, by calibrating activated clotting time to avoid bleeding complications or honing a dose algorithm; patient-stratification efforts (three proposals), for example, to assess differential
n engl j med 371;22 nejm.org november 27, 2014
The New England Journal of Medicine Downloaded from nejm.org at NEW YORK MEDICAL COLLEGE on August 10, 2015. For personal use only. No other uses without permission. Copyright © 2014 Massachusetts Medical Society. All rights reserved.
PE R S PE C T IV E
Access to Data from Industry-Sponsored Clinical Trials
Experience with the Open-Access Clinical Trial System, May 7, 2013, through May 31, 2014.* Part of the Process
No. of Proposals
Submission
58
Requirements check In process
4
Withdrawn by the requestor
2
Did not meet requirements or further details were required
7
Met requirements
45
Review by independent review panel In process
6
Rejected or advised to resubmit
3
Approved or approved with conditions
36
Data-sharing agreement In process
12
Withdrawn by the requestor
1
Agreed and signed
23
Data preparation In process
10
Complete, with data available
13
Research project in process
13
* A total of 4 of the 37 proposals submitted since January 2014 have included requests for data from multiple sponsors. The requirements check is the check done by the sponsor of the study whose data are being requested to make sure the information is complete and that the proposal meets the requirements of this initiative and the sponsor’s requirements for informed consent. Proposals are then sent to the independent review panel.
effects in different ethnic groups or to stratify patients’ responses; and miscellaneous other studies, including one proposal to evaluate the relationships between influenza vaccination and lifestyle factors, one to assess the effectiveness of paroxetine in treating depression in adolescents, and one to attempt to elucidate the expected frequency of severe adverse events before a new trial, to be conducted in the developing world, was designed. There were requests for 97 GlaxoSmithKline studies that were not included in the system. Of these, 20 turned out to be general inquiries that did not require data access through the system;
of the others, 3 are still undergoing feasibility assessment, and 74 have been approved. There is considerable interest in open access to clinical trial data on the part of both the research community and the wider public. Limited but important evidence suggests that reanalyses of previously published data can lead to conclusions about the types of patients who should receive a given treatment that differ from the conclusions drawn by the original investigators.1 Many government funding agencies require investigators whose trials they support financially to share their data. The system created by Glaxo SmithKline was pathbreaking in
allowing access to individual patient–level, analyzable data from clinical trials sponsored by the pharmaceutical industry. In this system, access to the data could be denied by the company before committee review only if it was considered logistically impractical to provide it. The independent review panel’s remit was to make data available for responsible use. The committee did not see its role as performing a scientific review of the proposals as a funding agency would do; access to the data was not intended to be inherently limited, so the committee believed that a rigorous assessment of the scientific merit of each project should be left for later peer review. Hence, the default decision was to permit access to the data. However, the committee was also tasked with screening for “‘rogue’ reanalysis by nonexperts or by analysts who have conflicts of interest,”1 analyses by teams lacking the skills to perform such analyses, and analyses without accompanying plans for disclosing the results publicly. Although none of the data provided have yet led to peer-reviewed publications, the approach we have used comes very close to fulfilling all the criteria for reanalyses that have recently been set forth by Christakis and Zimmerman.3 A number of other companies have since joined this system, which naturally permits easier aggregation of data from multiple companies. Some companies have elected to join other dataaccess systems.4 When manufacturers came to the independent review panel seeking to participate using alternative approaches — suggesting, for instance, that the company be allowed to con-
n engl j med 371;22 nejm.org november 27, 2014
2053
The New England Journal of Medicine Downloaded from nejm.org at NEW YORK MEDICAL COLLEGE on August 10, 2015. For personal use only. No other uses without permission. Copyright © 2014 Massachusetts Medical Society. All rights reserved.
PERS PE C T IV E
Access to Data from Industry-Sponsored Clinical Trials
duct a scientific review of each proposal before passing it along to the committee — we declined, believing that such an approach would compromise our independence. The system described here will undoubtedly evolve considerably over time. Management of the proposals is likely to be turned over to an independent organization, a development that should make it more appealing for other manufacturers to join and should enhance public confidence in the process. We hope that academic researchers will be adding their data to such systems in the future.
Some of us believe that all clinical trial data should eventually be put into the public domain for unconditional, universal access, but we see the current initiative as an important, swiftly implemented, and successful first step. The number of data requests submitted in the first 12 months after the system was launched (58 requests) is a testament to the eagerness of the medical community to reap the benefits of data sharing. Disclosure forms provided by the authors are available with the full text of this article at NEJM.org. From Rutgers Biomedical and Health Sciences, Newark, NJ (B.L.S.); Hasselt Univer-
sity, Hasselt, and the International Drug Development Institute, Louvain-la-Neuve — both in Belgium (M.B.); the United Kingdom Clinical Research Collaboration Board, London (J.H.); and the Center of Genomics and Policy, McGill University, Montreal (B.M.K.). This article was published on October 15, 2014, at NEJM.org. 1. Ebrahim S, Sohani ZN, Montoya L, et al. Reanalyses of randomized clinical trial data. JAMA 2014;312:1024-32. 2. Nisen P, Rockhold F. Access to patientlevel data from GlaxoSmithKline clinical trials. N Engl J Med 2013;369:475-8. 3. Christakis DA, Zimmerman FJ. Rethinking reanalysis. JAMA 2013;310:2499-500. 4. Krumholz HM, Ross JS, Gross CP, et al. A historic moment for open science: the Yale University Open Data Access project and Medtronic. Ann Intern Med 2013;158:910-1. DOI: 10.1056/NEJMp1411794
Copyright © 2014 Massachusetts Medical Society.
Ebola Virus Disease in West Africa — Clinical Manifestations and Management Daniel S. Chertow, M.D., M.P.H., Christian Kleine, M.D., Jeffrey K. Edwards, M.D., M.P.H., Roberto Scaini, M.D., Ruggero Giuliani, M.D., and Armand Sprecher, M.D., M.P.H.
I
n resource-limited areas, isolation of the sick from the population at large has been the cornerstone of control of Ebola virus disease (EVD) since the virus was discovered in 1976.1 Although this strategy by itself may be effective in controlling small outbreaks in remote settings, it has offered little hope to infected people and their families in the absence of medical care. In the current West African outbreak, infection control and clinical management efforts are necessarily being implemented on a larger scale than in any previous outbreak, and it is therefore appropriate to reassess traditional efforts at disease management. Having cared for more than 700 patients with EVD between August 23 and October 4, 2014, in the largest
2054
Ebola treatment unit in Monrovia, Liberia (see diagrams), we believe that our cumulative clinical observations support a rational approach to EVD management in resource-limited settings. Early symptoms of EVD include high fever (temperature of up to 40°C), malaise, fatigue, and body aches (see table).2,3 The fever persists, and by day 3 to 5 of illness, gastrointestinal symptoms typically begin, with epigastric pain, nausea, vomiting, and diarrhea. Patients routinely presented to our facility after 2 or 3 days of severe vomiting or diarrhea, during which they posed a substantial risk to their communities and had a high probability of testing positive for Ebola virus in blood by polymerase chain reaction (PCR). Although some patients tested
positive on PCR within 24 hours after symptom onset, we found that a negative test result could not be relied on to rule out disease until 72 hours after symptoms began. Of the patients who tested positive for Ebola, none that we were aware of had contracted disease from an infected contact during the early febrile phase of illness. No ancillary testing was available in our facility. We observed that recurrent episodes of emesis resulted in an inability to tolerate oral intake. Large volumes of watery diarrhea estimated at 5 or more liters per day (a manifestation not unlike that of cholera) presented suddenly, persisted for up to 7 days or (rarely) longer, and gradually tapered off. Associated signs and symptoms included asthenia,
n engl j med 371;22 nejm.org november 27, 2014
The New England Journal of Medicine Downloaded from nejm.org at NEW YORK MEDICAL COLLEGE on August 10, 2015. For personal use only. No other uses without permission. Copyright © 2014 Massachusetts Medical Society. All rights reserved.
