Letters

RESEARCH LETTER Table. Analytic Goal of Proposed Studies to Open Access Platform Use of Open Access Platforms for Data Review Panels With Representative Title Examplesa Concerns over in clinical trial reporting have stimulated No. of 1 Proposed Representative Example calls for more open data sharing. In response, multiple phar- Analytic Goal Studies of Proposed Study Title maceutical companies have created mechanisms for investi- Secondary Analysis of Treatment Effect (n = 50) gators to access patient-level clinical trials data. However, if Predictors 18 Can Machine Learning Algorithms Using General and how these shared clinical trial data are being used is of response Laboratories Predict Clinical Remission and Mucosal Healing for Patients on Vedolizumab? unknown. Subgroup 10 Low-dose vs Standard-dose Unfractionated analysis Heparin for Percutaneous Coronary Intervention Methods | We evaluated how many clinical trials were publicly in Acute Coronary Syndromes Treated With in Patients With Peripheral available to investigators through 3 open access platforms: Arterial Disease ClinicalStudyDataRequest.com, the Yale University Open Safety/adverse 11 Review of Fever and Febrile Convulsion Rates effect prediction in Children After Trivalent Influenza Vaccine Data Access Project (YODA), and the Supporting Open Access Pharmacokinetic 6 Development of Population Pharmacokinetic/ for Researchers (SOAR) initiative.2-4 Sponsors depositing data evaluation Pharmacodynamic Models for Pazopanib Within in these platforms include the EuroTARGET Project Effect on 2 The Effects of Maraviroc vs Efavirenz in GlaxoSmithKline, Astellas, biomarker Combination With / Editorial page 1238 Boehringer Ingelheim, Eisai, on the CD4/CD8 Ratio in Treatment-Naïve HIV-Infected Individuals Eli Lilly, Novartis, Roche, Off-target 2 Efficacy of Belimumab on Mucocutaneous Sanofi, Takeda, Union Chimique Belge, ViiV Healthcare, treatment Involvement in Patients With Systemic Lupus Johnson & Johnson, Medtronic, and Bristol-Myers Squibb. effect Erythematosus Duration 1 Statistical Methods for Estimating the Duration Company policies on what trials are shared vary and are of response of Protection of Malaria Vaccines: Secondary available online, but most include all trials within certain Analysis of Data From the RTSS/AS02A Trial in Mozambique date ranges after regulatory review and publication of Secondary Disease State Analysis (n = 31) results. Predictors 11 Incidence Rate and Risk Factors Associated With Investigators submit a research proposal to the plat- of disease Hepatocellular Carcinoma Development forms, which is first reviewed for the availability of the trial(s) progression in Patients With Chronic Hepatitis C or severity and Severe Thrombocytopenia: the Lesson requested and completeness of the application. Next, the pro- From ENABLE Studies posal is reviewed by a panel comprising independent experts Complications 6 Relationship of Activated Clotting Time of disease and Bleeding-Related Outcomes or members of the platform. The panel then rejects or ap- or procedure in the FUTURA/OASIS 8 Trial proves the proposal based on scientific merit and adequacy of Alternate outcome 10 Developing New Outcome Measures for SLE the research design to achieve scientific objectives. A data shar- measures/disease by Deconvoluting Available Data Sets classification ing agreement is then created. Each platform requires inves- schemes tigators to report any resulting publications. Disease 4 Cytokine Patterns in Antiviral Disease Vaccines: We reviewed all proposals with data use agreements since characterization A Knowledge-Driven Approach inception of each platform (first in 2013) and December 31, 2015 Other Proposed Studies (n = 32) 4 Predictors of Placebo Response in Pediatric (obtained from ClinicalStudyDataRequest.com and the YODA or placebo effect Patients With Generalized Anxiety Disorder website or directly from SOAR), the characteristics of ac- Reevaluation of 5 COPERNICUS Reanalysis cepted proposals, and reported publications. We classified the primary endpoint in study population main objective of the analysis based on review of the analytic Meta-analysis 12 Assessing the Comparative Efficacy of plan and study design. of treatment Therapeutics for Chronic Obstructive Pulmonary effect Disease: A Meta-analysis of Randomized Trials Statistical 11 Assessing and Reporting Heterogeneity of Results | A total of 3255 clinical trials were available in the plat- or trial Treatment Effect in Randomized Clinical Trials forms, most of which were phase 3 (44.3% overall), followed methodology by phase 1 (23.6%), phase 2 (18.3%), phase 4 (12.7%), and other Abbreviations: COPERNICUS, prospective randomized cumulative (1.1%). Of the 234 proposals submitted, 177 had been pro- survival; ENABLE, therapy for hepatitis C virus–related thrombocyto- penia; EuroTARGET,targeted therapy in renal cell cancer: genetic and tumor cessed and met initial requirements. The review panel re- related biomarkers for response and toxicity; FUTURA/OASIS 8, fondaparinux trial jected 12 of 177 proposals (6.8%), 10 were under review, and 4 with unfractionated heparin during revascularization in acute coronary syndromes; had been withdrawn. Of the 154 proposals approved, data shar- HIV, human immunodeficiency virus; SLE, systemic lupus erythematosus. ing agreements were completed for 113, including requests from a Obtained from ClinicalStudyDataRequest.com and the Yale University Open 17 countries, most from the United States (n = 61; 54.0%). Most Data Access Project website or directly from Supporting Open Access for Researchers initiative. studies were not directly funded (n = 77; 68.1%).

jama.com (Reprinted) JAMA March 22/29, 2016 Volume 315, Number 12 1283

Copyright 2016 American Medical Association. All rights reserved.

Downloaded From: https://jamanetwork.com/ on 09/26/2021 Letters

The median number of trials requested by each proposal was Disclosures: Dr Navar reports receiving research funding 2 (range, 1-59; interquartile range, 1-6). Most proposals requested from Sanofi and Regeneron. Dr Pencina reports that Duke University receives funding for the SOAR Initiative from Bristol-Myers Squibb. Dr Peterson reports trials from a single sponsor; 29 proposals (8%) requested data receiving grant funding from the American College of Cardiology, American Heart from trials conducted by more than 1 sponsor. Only 505 unique Association, and Janssen; and consulting fees from Bayer, Boehringer Ingelheim, trials (15.5% of available trials) had ever been requested, includ- Merck, Valeant, Sanofi, Astra Zeneca, Janssen, Regeneron, and Genentech and ing 356 phase 3 trials (24.7% of phase 3 trials available). that Duke University receives funding for the SOAR Initiative from Bristol-Myers Squibb. No other disclosures were reported. Analytic goals of these proposals varied, including secondary Disclaimer: Dr Peterson, an associate editor for JAMA, was not involved in the analyses of a trial’s treat-ment effect (n = 50; 44.3%) and analy- editorial review of or the decision to publish this article. ses of the disease state itself (n = 31; 27.4%) (Table). Validation 1. Krumholz HM, Peterson ED. Open access to clinical trials data. JAMA.2014; of the study primary end point was rare (n = 5; 4.4%). Only 1 pro- 312(10):1002-1003. 5 posal led to publication. 2. The YODA Project. The YODA Project website. http://yoda.yale.edu. Accessed January 26, 2016. Discussion | Although more than 3000 trials are available to in- 3. ClinicalStudyDataRequest.com. ClinicalStudyDataRequest.com website. vestigators through open data platforms, only 15.5% had been https://clinicalstudydatarequest.com/. Accessed January 26, 2016. requested by a limited number of researchers. Most propos- 4. Pencina MJ, Louzao DM, McCourt BJ, et al. Supporting open access to clinical als focused on nonprespecified subgroups or predictors of re- trial data for researchers: the Duke Clinical Research Institute-Bristol-Myers Squibb Supporting Open Access to Researchers Initiative. Am Heart J. 2016;172:64-69. sponse rather than validation of study results. 5. Le Noury J, Nardo JM, Healy D, et al. Restoring Study 329: efficacy and harms Reasons for underutilization of clinical trials data may in- of and in treatment of major depression in adolescence. clude lack of knowledge about these resources, possibly due to BMJ. 2015;351:h4320. lack of publication of results from proposals, or lack of funding to support analyses. Incentive for validation studies may be lim- COMMENT & RESPONSE ited as confirmatory reanalyses are less likely to be published. However, the 1 publication using these data was a validation Intensive Care Unit Admission and Mortality study that found contradictory results from the initial article.5 Among Medicare Beneficiaries With Pneumonia This study has limitations. We focused on 3 platforms and To the Editor The study by Dr Valley and colleagues1 on the as- were unable to obtain data from individual companies that sociation between intensive care unit (ICU) admission and share data. Details on rejected studies were unavailable. Cer- mortality among Medicare beneficiaries hospitalized with tain information in proposals may be incomplete, such as fund- pneumonia addresses a topic with potentially important im- ing or specific analytic plans. There may be a lag between pub- plications for clinical practice and public policy. Further evalu- lication of a report and posting on the platform website, and ation could yield more robust and targeted results. we did not independently search for publications. First, early institution of noninvasive ventilation may im- Early use of platforms designed to provide access to indi- prove outcomes after mild respiratory failure in some patients,2 vidual patient data, developed to increase transparency of clini- and it is commonly managed outside ICUs. The code for con- cal trial data, has been limited. Availability of shared clinical tinuous positive airway pressure (CPAP) could identify a sub- trial data should be promoted and use of individual patient data population that derived benefit from noninvasive support. for validation studies encouraged. Thus, the analysis should adjust for the capability of general wards to provide noninvasive ventilation. Ann Marie Navar, MD, PhD Second, differential distance used as the instrumental vari- Michael J. Pencina, PhD able in the analysis does not allow generalization of the results Jennifer A. Rymer, MD to the entire Medicare population. If the sample included not Darcy M. Louzao, PhD just fee-for-service Medicare beneficiaries but also the Medicare Eric D. Peterson, MD, MPH patients enrolled in a Medicare Advantage Health Maintenance Organization (HMO) plan, the proximity to non-HMO hospitals Author Affiliations: Duke Clinical Research Institute, Duke University Medical with high ICU admission volume might not influence the prob- Center, Durham, North Carolina (Navar, Pencina, Rymer, Louzao, Peterson); ability of being admitted to an ICU as out-of-network hospitals Associate Editor, JAMA (Peterson). cannot be used without prior authorization. Moreover, non– Correction: This article was corrected for errors in the conflict of interest disclosures on June 14, 2016. critically ill patients may have time to plan where to be treated, and proximity might not influence selection.3 Although the in- Corresponding Author: Ann Marie Navar, MD, PhD, Duke Clinical Research Institute, Duke University Medical Center, 2400 Pratt St, Ste 7521, Durham, NC strumental variable method is appealing conceptually, the use 27705 ([email protected]). of this approach may introduce more bias than it corrects.4 Author Contributions: Dr Navar had full access to all of the data in the study In assessing the intrinsic value of ICUs against noninten- and takes responsibility for the integrity of the data and the accuracy of the data sive care wards, the influence of other covariates should be con- analysis. Study concept and design: Navar, Peterson. sidered, such as specific treatments, intensity of monitoring, Acquisition, analysis, or interpretation of data: Navar, Pencina, Rymer, Louzao, and staff to patient ratios. Peterson. Drafting of the manuscript: Navar, Pencina, Rymer, Louzao, Peterson. Andrea Rossi Zadra, MD Critical revision of the manuscript for important intellectual content: Navar, Pencina, Rymer, Louzao, Peterson. Nicoletta Barzaghi, MD Study supervision: Peterson. Cinzia Di Novi, PhD

1284 JAMA March 22/29, 2016 Volume 315, Number 12 (Reprinted) jama.com

Copyright 2016 American Medical Association. All rights reserved.

Downloaded From: https://jamanetwork.com/ on 09/26/2021