Skip to main content

Secondary Analysis of Individual Patient-Level Clinical Trial Data: A Primer

  • 132 Accesses

Part of the Munich Studies on Innovation and Competition book series (MSIC,volume 16)

Abstract

Proponents of access to data argued that secondary analyses of clinical trial data—especially individual patient-level data—can generate knowledge far beyond original research hypotheses and the benefit-risk profile of investigational products. The chapter explores this proposition. Admittedly, such task goes beyond a legal inquiry. However, without a detailed understanding of the role of secondary IPD analysis in medical research and drug R&D, arguments—both de lege lata and de lege ferenda—regarding the applicable legal framework lack a substantive basis. The overview of potential implications of secondary IPD analysis presented here is by no means exhaustive. Instead, insights drawn from the general medical literature are systematised to inform and illustrate the subsequent legal analysis.

Keywords

  • Clinical trial data
  • Confirmatory analysis
  • Data-driven drug R&D
  • Exploratory analysis
  • Extrapolation
  • Meta-analysis
  • Primary analysis
  • Predictive modelling
  • Secondary analysis
  • Subgroup analysis

The research for this chapter is based on a literature review conducted within the databases of medical journals NEJM, JAMA, BMJ, PLoS Med, Ann Intern Med, Lancet and JAMA Intern Med. In March 2018, a workshop with researchers at the Medical University of Vienna was conducted. The author is especially thankful to Professor Franz Koenig at the Medical University of Vienna, Dr. Sarah Nevitt at the University of Liverpool and Dr. med. Arnoud Templeton at Claraspital Basel for valuable insights and clarifications provided in the course of research.

This is a preview of subscription content, access via your institution.

Buying options

Chapter
USD   29.95
Price excludes VAT (USA)
  • DOI: 10.1007/978-3-030-86778-2_3
  • Chapter length: 26 pages
  • Instant PDF download
  • Readable on all devices
  • Own it forever
  • Exclusive offer for individuals only
  • Tax calculation will be finalised during checkout
eBook
USD   119.00
Price excludes VAT (USA)
  • ISBN: 978-3-030-86778-2
  • Instant PDF download
  • Readable on all devices
  • Own it forever
  • Exclusive offer for individuals only
  • Tax calculation will be finalised during checkout
Softcover Book
USD   159.99
Price excludes VAT (USA)
Hardcover Book
USD   159.99
Price excludes VAT (USA)

Notes

  1. 1.

    ICH (5 Feb 1998) ICH Harmonised tripartite guideline. Statistical principles for clinical trials. E9, pp. 2–5. EMA (2 Mar 2016) External guidance on the implementation of the European Medicines Agency policy on the publication of clinical data for medicinal products for human use. EMA/90915/2016, p. 8.

  2. 2.

    ICH (5 Feb 1998) ICH Harmonised tripartite guideline. Statistical principles for clinical trials. E9, pp. 2–5. See also CONSORT. Glossary. http://www.consort-statement.org/resources/glossary#H. Accessed 26 Mar 2021.

  3. 3.

    The effect size refers to a quantitative measure that provides the statistical description of the study hypothesis and is derived from the outcome variable. Laake and Breien (2015), p. 114.

  4. 4.

    See National Institutes of Health, Biomarkers Definition Working Group (2001), p. 91 (defining a clinical endpoint as ‘a characteristic or variable that reflects how a patient feels, functions or survives’, and a surrogate endpoint ‘as a biomarker intended to act as a clinical endpoint’). For instance, endpoints used in cancer clinical trials often include survival, tumour response rate, progression-free survival and patient-reported endpoints such as quality of life. George et al. (2016), pp. 4–5.

  5. 5.

    ICH (5 Feb 1998) ICH Harmonised tripartite guideline. Statistical principles for clinical trials. E9, p. 5. The main results of a clinical trial are based on the primary endpoints as the most relevant characteristics for the disease and intervention under examination. See Huque and Röhmel (2010), p. 3; Moyé (2003), p. 114.

  6. 6.

    ICH (5 Feb 1998) ICH Harmonised tripartite guideline. Statistical principles for clinical trials. E9, p. 6. Secondary endpoints characterise additional benefits of the treatment under investigation. See Huque and Röhmel (2010), p. 3.

  7. 7.

    Goffin (2009), pp. 13–14.

  8. 8.

    Moyé (2003), p. 70. Hypotheses based on exploratory endpoints are often evaluated in the subsequent trials. ibid 156. See also Huque and Röhmel (2010), p. 3.

  9. 9.

    Huque and Röhmel (2010), p. 3.

  10. 10.

    ICH (5 Feb 1998) ICH Harmonised tripartite guideline. Statistical principles for clinical trials. E9, p. 4.

  11. 11.

    Zarin and Tse (2016).

  12. 12.

    EMA (21 Mar 2019) European Medicines Agency policy on publication of clinical data for medicinal products for human use. Policy/0070 [hereinafter EMA publication policy 0070], p. 3. Sometimes literature refers to CSRs as summary data. The present analysis follows the approach of the EU Clinical Trials Regulation and distinguishes between a CSR and a summary of a CSR; therefore, complete CSRs are considered non-summary data. Depending on the practice of a regulatory authority, CSRs can be treated separately from IPD. For instance, the EMA makes such distinction. See EMA publication policy 0070, p. 3. According to the International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use, patient data listings (including individual efficacy response data) constitute an appendix to a CSR that might or might not need to be submitted together with a CSR but should be provided upon a regulatory authority’s request. ICH (30 Nov 1995) ICH Harmonised tripartite guideline. Structure and content of clinical study reports. E3, para 16.

  13. 13.

    EMA (2 Mar 2016) External guidance on the implementation of the European Medicines Agency policy on the publication of clinical data for medicinal products for human use. EMA/90915/2016, p. 9.

  14. 14.

    ICH (10 Jun 1996) ICH Harmonised tripartite guideline. Guideline for good clinical practice. E6, para 1.51.

  15. 15.

    Institute of Medicine of the National Academies (2015), pp. 98–99; Zarin and Tse (2016).

  16. 16.

    The trial protocol and statistical analysis plan are also known as ‘meta-data’. Institute of Medicine of the National Academies (2015), pp. 92–93.

  17. 17.

    CSRs are prepared following the structure specified in Directive 2001/83/EC, annex I, part I, module 5 that, in turn, follows the common format of the International Conference on Harmonisation of Technical Requirements for Registration of Pharmaceuticals for Human Use. See Reg 726/2004/EC, art 6. On the CSR contents particularities, see generally EMA (23 Jun 2004) Note for guidance on the inclusion of appendices to clinical study reports in marketing authorisation applications. CHMP/EWP/2998/03; ICH (30 Nov 1995) ICH Harmonised tripartite guideline. Structure and content of clinical study reports. E3.

  18. 18.

    See Chap. 4 at Sect. 4.3.1.1.

  19. 19.

    Reg 536/2014/EU, art 37(4) and annex IV.

  20. 20.

    Doshi and Jefferson (2013). See EMA (20 Oct 2016) Opening up clinical data on new medicines. EMA provides public access to clinical reports. EMA/650519/2016 (reporting that the first CSRs related to two medicines released by the EMA on 20 October 2016 comprised ‘approximately 260,000 pages of information’). See also Sudlow et al. (2016) (noting that a ‘typical CSR (including all appendices) can be in excess of 1000 pages’).

  21. 21.

    Institute of Medicine of the National Academies (2015), p. 7 (emphasis added).

  22. 22.

    Wang et al. (2009), p. 1056.

  23. 23.

    Trocky and Brandt (2009), p. 198.

  24. 24.

    Reg 536/2014/EU, rec 1, art 3(a).

  25. 25.

    Reg 536/2014/EU, rec 1, art 6(1)(b)(i) third indent.

  26. 26.

    Reg 536/2014/EU, art 2 (2)(1)(a) (emphasis added).

  27. 27.

    British Pharmacological Society. Pharmacology skills for drug discovery. https://thebiologist.rsb.org.uk/images/Pharmacology_Skills_for_Drug_Discovery.pdf. Accessed 26 Mar 2021.

  28. 28.

    ibid.

  29. 29.

    Dalrymple (2003), p. 35 (emphasising that scientific knowledge is systematic, testable and verifiable).

  30. 30.

    Implications of secondary IPD analysis for cumulativeness of medical research and drug R&D are discussed more in detail in Chap. 8 at Sects. 8.2.3.1 and 8.2.3.2.

  31. 31.

    Institute of Medicine of the National Academies (2015), p. 24.

  32. 32.

    Data from confirmatory trials can be used to generate and explore new hypotheses that would need to be confirmed in the subsequent studies. See ICH (5 Feb 1998) ICH Harmonised tripartite guideline. Statistical principles for clinical trials. E9, p. 4; Wang and Bakhai (2006), p. 4.

  33. 33.

    Grady and Hearst (2007), p. 211.

  34. 34.

    Koenig et al. (2015), p. 10; Institute of Medicine of the National Academies (2015), p. 104.

  35. 35.

    The issue of reproducibility in clinical trials is discussed in more detailed in Chap. 6 at Sect. 6.4.2.

  36. 36.

    Menikoff J (7 Mar 2017) Letter to the ICMJE secretariat. http://www.icmje.org/news-and-editorials/menikoff_icmje_questions_20170307.pdf. Accessed 26 Mar 2021.

  37. 37.

    Exploratory analysis is sometimes referred to as analysis suggested by data. CONSORT. Glossary. http://www.consort-statement.org/resources/glossary#E. Accessed 26 Mar 2021.

  38. 38.

    Institute of Medicine of the National Academies (2015), p. 20. See also Lauer (2010), p. 91 (noting that clinical trials ‘can function as rich sources of observational data, useful for exploring questions that go beyond their original hypotheses’); Browner et al. (2007), p. 61 (referring to trial data as ‘a fertile source of potential research questions for future studies’).

  39. 39.

    Selby et al. (2018), p. 285.

  40. 40.

    Menikoff J (7 Mar 2017) Letter to the ICMJE Secretariat. http://www.icmje.org/news-and-editorials/menikoff_icmje_questions_20170307.pdf. Accessed 26 May 2021.

  41. 41.

    USFDA (4 Jun 2013) Availability of masked and de-identified non-summary safety and efficacy data—request for comments. Fed. Reg. 78(107), p. 33422. https://www.govinfo.gov/content/pkg/FR-2013-06-04/pdf/2013-13083.pdf. Accessed 26 Mar 2021.

  42. 42.

    Institute of Medicine of the National Academies (2015), p. 20.

  43. 43.

    ibid pp. 164–167.

  44. 44.

    See e.g. Selby et al. (2018), p. 285; Gustafsson et al. (2010), pp. 938–939 (discussing the example where secondary analysis of historical trial data generated two new hypotheses—one was successfully confirmed in the subsequent research and the other one could not be validated).

  45. 45.

    Strom et al. (2016), p. 1608 (reporting that only three out of 177 requests for IPD submitted through the ClinicalStudyDataRequest.com portal between 7 May 2013 and 14 November 2015 were filed for confirmatory analysis purposes; the rest of the proposals were directed at new research questions).

  46. 46.

    ICH (5 Feb 1998) ICH Harmonised tripartite guideline. Statistical principles for clinical trials. E9, p. 27.

  47. 47.

    Cleophas et al. (2006), pp. 149–150.

  48. 48.

    ICH (5 Feb 1998) ICH Harmonised tripartite guideline. Statistical principles for clinical trials. E9, pp. 26–27.

  49. 49.

    Meinert (2012), p. 454.

  50. 50.

    ibid.

  51. 51.

    Song and Bachmann (2016).

  52. 52.

    ibid.

  53. 53.

    Selby et al. (2018), p. 285.

  54. 54.

    Brody (2016), p. 88. See also ICH (5 Feb 1998) ICH Harmonised tripartite guideline. Statistical principles for clinical trials. E9, p. 29 (noting that the results of exploratory analysis ‘should be interpreted cautiously [and] any conclusion of treatment efficacy (or lack thereof) or safety based solely on exploratory subgroup analyses are unlikely to be accepted’).

  55. 55.

    Biltaji et al. (2017), p. 2338; Cleophas et al. (2006), pp. 149–150 (noting that ‘[i]f such subgroups are identified, the exploratory nature of the regression analysis should be emphasized and the subgroup issue should be further assessed in subsequent independent and prospective data-sets’).

  56. 56.

    Koenig et al. (2015), p. 23.

  57. 57.

    Song and Bachmann (2016).

  58. 58.

    ICH (5 Feb 1998) ICH Harmonised tripartite guideline. Statistical principles for clinical trials. E9, p. 34.

  59. 59.

    A predictor is a variable (e.g. age, sex, smoking history, the intake of supplements) that can predict a health outcome (e.g. heart attack or quality of life). By analysing the associations between different variables predicting a health outcome, investigators can draw implications regarding causality. Cummings et al. (2007), p. 7.

  60. 60.

    Newman et al. (2007b), pp. 137–138.

  61. 61.

    ibid.

  62. 62.

    ibid.

  63. 63.

    ibid p. 138.

  64. 64.

    ICH (5 Feb 1998) ICH Harmonised tripartite guideline. Statistical principles for clinical trials. E9, p. 27. See also Cleophas et al. (2006), p. 141 (noting that, ‘when studying interactions, the results of the regression analysis are more valid when complemented by additional exploratory analyses within relevant subgroups of patients or within strata defined by the covariates’).

  65. 65.

    ibid. Other common purposes of secondary IPD analysis include meta-analysis, survival analysis and testing new analytical methods.

  66. 66.

    USFDA (4 Jun 2013) Availability of masked and de-identified non-summary safety and efficacy data—request for comments. 78(107) Fed. Reg, p. 33423. https://www.govinfo.gov/content/pkg/FR-2013-06-04/pdf/2013-13083.pdf. Accessed 26 Mar 2021. See also USFDA (2011) Advancing regulatory science at FDA, p. 12 (stating that clinical trial data can be leveraged ‘to develop quantitative models and measures of disease progression’). https://www.fda.gov/media/81109/download. Accessed 26 Mar 2021. Stewart and Tierney (2002), p. 89 (noting that IPD can be used in the exploratory analysis for constructing or validating prognostic indices).

  67. 67.

    Tierney et al. (2015), p. 1330.

  68. 68.

    ibid.

  69. 69.

    ibid. As observed by Tierney et al., ‘IPD meta-analyses have played a role in the selection of participants, and in the conduct, analysis, and interpretation of trials, particularly in response to subgroup or prognostic factor analyses’. Tierney et al. (2015), p. 1332.

  70. 70.

    ICH (5 Feb 1998) ICH Harmonised tripartite guideline. Statistical principles for clinical trials. E9, p. 34. See also Cooper and Patall (2009), pp. 165–166; Cochrane Methods Group. About IPD meta-analyses. https://methods.cochrane.org/ipdma/about-ipd-meta-analyses. Accessed 26 Mar 2021.

  71. 71.

    Egger (1997), p. 1371.

  72. 72.

    Engberg (2008), pp. 258–265; Haidich (2010), p. 29.

  73. 73.

    Mulrow (1994), p. 597.

  74. 74.

    Ioannidis (2004), p. 522.

  75. 75.

    Tierney et al. (2015), p. 1325.

  76. 76.

    Cochrane Methods Group. About IPD meta-analysis. https://methods.cochrane.org/ipdma/about-ipd-meta-analyses. Accessed 26 Mar 2021. See generally Tierney et al. (2015); Simmonds et al. (2015); Stewart and Parmar (1993).

  77. 77.

    Cooper and Patall (2009), p. 167; Nevitt et al. (2017).

  78. 78.

    Riley et al. (2010).

  79. 79.

    Naci et al. (2015), p. 58.

  80. 80.

    Cooper and Patall (2009), p. 167.

  81. 81.

    Institute of Medicine of the National Academies (2015), p. 12.

  82. 82.

    Tierney et al. (2015), p. 1332 (noting that, even though IPD meta-analysis reports usually focus on providing recommendations for clinical practice, they can play ‘an equally important role in informing subsequent clinical research’); Institute of Medicine of the National Academies (2015), pp. 61–62 (with further references). See also Stewart et al. (2011), p. 18:2 (definining situations where IPD analysis can be particularly important, such as conducting complex types of analyses (e.g. multivariate analysis) and exploring interactions between the intervention and patient-level characteristics).

  83. 83.

    Nevitt et al. (2017).

  84. 84.

    Council of Europe (2012), para 6.C.20.2.

  85. 85.

    Bath and Gray (2009), p. 25 (stating that, ‘[u]nfortunately, if summary meta-analyses are complicated by missing trial data, this problem is magnified in analyses based on individual patient data’).

  86. 86.

    Clinical Study Data Request. Metrics. Published proposals. https://www.clinicalstudydatarequest.com/Metrics/Published-Proposals.aspx. Accessed 26 Mar 2021.

  87. 87.

    British Pharmacological Society. Pharmacology skills for drug discovery. https://thebiologist.rsb.org.uk/images/Pharmacology_Skills_for_Drug_Discovery.pdf. Accessed 26 Mar 2021.

  88. 88.

    Vallance and Smart (2006), p. 7.

  89. 89.

    See e.g. Gallin et al. (2018), p. 649 ff; British Pharmacological Society. Pharmacology skills for drug discovery. https://thebiologist.rsb.org.uk/images/Pharmacology_Skills_for_Drug_Discovery.pdf. Accessed 26 Mar 2021.

  90. 90.

    Merrill (2015), p. 3.

  91. 91.

    Strom (2005), pp. 3–5.

  92. 92.

    Newman et al. (2007a), p. 113.

  93. 93.

    Meinert (2012), p. 69.

  94. 94.

    Reg 536/2014/EU, art (2)(1) (emphasis added).

  95. 95.

    Cummings et al. (2007), p. 21.

  96. 96.

    Merrill (2015), p. 3.

  97. 97.

    Institute of Medicine of the National Academies (2015), p. 31.

  98. 98.

    CIOMS (2005), p. 67.

  99. 99.

    Meinert (2012), pp. 424–425.

  100. 100.

    Declaration of Helsinki, para (6).

  101. 101.

    CIOMS (2005), p. 259 (emphasis added).

  102. 102.

    Porth (2011), p. xix.

  103. 103.

    ibid.

  104. 104.

    Berlin et al. (2014).

  105. 105.

    Braun and Anderson (2007), p. 2.

  106. 106.

    National Institutes of Health, Biomarkers Definition Working Group (2001), p. 91. Biomarkers can be applied inter alia to indicate a disease prognosis or predict and monitor clinical response to medical intervention. ibid. See also Schulte and Mazzuckelli (1991), p. 239 (defining a biomarker as ‘an indicator that signals events in biological systems or samples […] generally taken to be any biochemical, genetic, or immunologic indicator that can be measured in a biological specimen’).

  107. 107.

    Laterza and Zhao (2016), p. 28.

  108. 108.

    ibid pp. 28–29.

  109. 109.

    In the early stages of drug development, biomarkers can be applied, for instance, in the development of the proof of concept and safety studies. In later stages, they can aid in selecting the dose, understanding clinical efficacy in a subset of the study population or identifying risks. See Singh et al. (2016), p. 202; Wnek et al. (2016), p. 143; Ray (2016), p. 1 ff; USFDA (2011) Advancing regulatory science at FDA, p. 12. https://www.fda.gov/media/81109/download. Accessed 26 Mar 2021. On the types of clinical biomarkers and their role in drug development, see e.g. Lee (2016), p. 47.

  110. 110.

    European Commission (25 Oct 2013) Use of ‘-omics’ technologies in the development of personalised medicine. SWD(2013) 436 final, pp. 11–12; Singh et al. (2016), p. 202.

  111. 111.

    Strimbu and Tavel (2010).

  112. 112.

    Davis et al. (2016), p. 17 (with further references).

  113. 113.

    Data-mining can be applied in identifying, selecting and prioritising potential disease target, diagnostic or prognostic markers. See Yang et al. (2009), pp. 150–151; Kilicoglu (2018).

  114. 114.

    For instance, machine learning can be applied to the subgroup analysis. See e.g. Helal (2016), p. 561; Lipkovich et al. (2018).

  115. 115.

    Clinical research informatics utilises methods of informatics to develop and manage medical knowledge and information. American Medical Informatics Association. Clinical research informatics. http://AMIA.org/applications-informatics/clinical-research-informatics. Accessed 26 Mar 2021.

  116. 116.

    Biomedical informatics is defined as ‘the interdisciplinary field that studies and pursues the effective uses of biomedical data, information, and knowledge for scientific inquiry, problem solving and decision making, motivated by efforts to improve human health’. Kulikowski et al. (2012), p. 933. Bioinformatic methods apply algorithms to analyse biological data, including DNA sequence, RNA expression and cells proteins. See Altman (2012), p. 994.

  117. 117.

    Translational bioinformatics is defined as ‘the development of storage, analytic, and interpretive methods to optimize the transformation of increasingly voluminous biomedical data into proactive, predictive, preventative, and participatory health’. Butte (2008), p. 709.

  118. 118.

    Besides clinical trial data, data-driven drug discovery can integrate and utilise other types of health-related data such as genomic, epigenetic, transcriptomic and proteomic data. Xia (2017), p. 1709; Altman (2012), p. 994; Butte and Ito (2012), p. 949.

  119. 119.

    Mayo et al. (2017) (noting that machine learning approaches ‘can be used to leverage the wide range of data element categories contained in [data resource systems] to identify unanticipated interactions and dependencies that should be considered in the RCT design’).

  120. 120.

    Jones and Warren (2006), p. 253.

  121. 121.

    Butte and Ito (2012), p. 950 (with further references).

  122. 122.

    USFDA (2011) Advancing regulatory science at FDA, p. 9. https://www.fda.gov/media/81109/download. Accessed 26 Mar 2021.

  123. 123.

    Biltaji et al. (2017), p. 2337; Elwood (2017), p. 254.

  124. 124.

    Hughes et al. (2011), p. 1239.

  125. 125.

    Yang et al. (2009), p. 147.

  126. 126.

    This method is also known as a ‘system approach’ to drug discovery, whereby potential targets are selected by studying a disease using data gathered in earlier clinical trials. Yang et al. (2009), p. 147.

  127. 127.

    Examples include the collaborative projects implemented by Pfizer and IBM Watson, Sanofi and Exscientia, Genentech and GNS Healthcare.

  128. 128.

    Examples include Sanofi’s ‘Translational Medicines for Patients’ program, Novartis’ Institute for Biomedical research and Pfizer’s ‘Precision Medicine Analytics Ecosystem’ initiative.

  129. 129.

    The platform provides access to trial protocols, case report forms and comparator arm data from cancer clinical trials that can be used as a quasi-comparator arm in future studies or in the decease progression models. See Project Data Sphere. https://www.projectdatasphere.org/projectdatasphere/html/about. Accessed 26 Mar 2021.

  130. 130.

    Pharma firms pool and share cancer trial data (2014).

  131. 131.

    ibid.

  132. 132.

    Critical Path Institute (10 Jul 2013) U.S. Food and Drug Administration and European Medicines Agency reach landmark decisions on Critical Path Institute’s clinical trial simulation tool for Alzheimer’s Disease. https://c-path.org/wp-content/uploads/2014/03/US-FDA-EMA-agency-reach-landmark-decisions-C-Path-clinical-trial-simulation-tool-for-alzheimers-disease.pdf. Accessed 26 Mar 2021.

  133. 133.

    Fleming (2018), p. 56.

  134. 134.

    Sellwood et al. (2018), p. 2027. Notably, AI-based methods have been used in medicinal chemistry since the 1960s. ibid 2025. See also WHO, IUPHAR, CIOMS (2012), p. 33 (concluding that ‘high expectations of innovation models that involve combinatorial chemistry, high-throughput screening, rational drug design, pharmacogenomics, bioinformatics and disease and pathway modelling have not been met despite the high level of investment’).

  135. 135.

    Sellwood et al. (2018).

  136. 136.

    For an overview, see Clayton et al. (2017).

  137. 137.

    Jones et al. (2013).

  138. 138.

    Cleophas et al. (2006), p. 205.

  139. 139.

    Tierney et al. (2015), p. 1331.

  140. 140.

    Sutton et al. (2007), p. 2496 (pointing out that, ‘if patient level covariates are identified as explaining heterogeneity, designing future studies controlling such effects through study design and pooling sub-grouped data defined by such covariate would seem sensible’).

  141. 141.

    ibid.

  142. 142.

    ibid.

  143. 143.

    Tierney et al. (2015), p. 1325.

  144. 144.

    USFDA (2011) Advancing regulatory science at FDA, p. 9. https://www.fda.gov/media/81109/download. Accessed 26 Mar 2021.

  145. 145.

    A notable example is the clinical trial simulation tool for Alzheimer’s Disease developed by the Critical Path Institute’s (C-Path), which the EMA found to be ‘suitable […] for use in drug development as a longitudinal model for describing changes in cognition in patients with mild and moderate [Alzheimer’s Disease], and for use in assisting in trial designs in mild and moderate [Alzheimer’s Disease], as defined by the context of use’. EMA (19 Sep 2013) Qualification opinion of a novel data driven model of disease progression and trial evaluation in mild and moderate Alzheimer’s disease. EMA/CHMP/SAWP/567188/2013, p. 50. The model can estimate inter alia a quantitative rationale for the selection of inclusion criteria and compare the results of the post hoc analyses with historical controls to reduce the risk of false positives. ibid p. 2. The model was developed based on the de-identified IPD provided by the research-based pharmaceutical companies. Critical Path Institute (10 Jul 2013) US Food and Drug Administration and European Medicines Agency Reach Landmark Decisions on Critical Path Institute’s Clinical Trial Simulation Tool for Alzheimer’s Disease, pp. 1–2. https://c-path.org/wp-content/uploads/2014/03/US-FDA-EMA-agency-reach-landmark-decisions-C-Path-clinical-trial-simulation-tool-for-alzheimers-disease.pdf. Accessed 26 Mar 2021.

  146. 146.

    See e.g. Ioannidis and Lau (2001); Golder et al. (2016); Prayle et al. (2012); Law et al. (2011); Zarin et al. (2011).

  147. 147.

    For the statistics, see e.g. Harrison (2016), pp. 817–818.

  148. 148.

    Nightingale and Mahdi (2006), p. 81 (with further references). See also British Pharmacological Society. Pharmacology skills for drug discovery. https://thebiologist.rsb.org.uk/images/Pharmacology_Skills_for_Drug_Discovery.pdf. Accessed 26 Mar 2021; Multi-Regional Clinical Trials Center at Harvard University (2014) Overview of data disclosure initiatives: current and ongoing data transparency activities in the pharmaceutical industry (observing that ‘access to data from discontinued programs, including early, exploratory trials, could inform future research and potentially reduce the risk to subjects in new clinical trials’). https://www.regulations.gov/comment/FDA-2013-N-0271-0031. Accessed 26 Mar 2021.

  149. 149.

    A failure to confirm a hypothesis should be distinguished from ‘uninformative’ results. The latter means that a trial might have had inadequate power but does not necessarily imply the lack of association between the treatment and the intended health benefit. See Altman and Bland (1995), p. 485.

  150. 150.

    See e.g. Gustafsson et al. (2010), p. 938 (discussing the examples of cardiovascular trials). In this regard, the Restored Trials Initiative should be mentioned. See Doshi et al. (2013) (proposing a system for independent analysis of clinical study reports on abandoned and non-reported trials. Even though secondary analyses under the initiative intended to follow the analyses specified in the original trial protocols, as acknowledge by Professor Doshi, exploratory analysis of IPD from abandoned and non-reported trials can be highly valuable. E-mail correspondence of 14 Jul 2018 (on file with the author).

  151. 151.

    USFDA (4 Jun 2013) Availability of masked and de-identified non-summary safety and efficacy data—request for comments. 78(107) Fed. Reg, p. 33422. https://www.govinfo.gov/content/pkg/FR-2013-06-04/pdf/2013-13083.pdf. Accessed 26 Mar 2021. In particular, the FDA envisages that the analysis of aggregated safety and effectiveness data can be used to address ‘key hurdles in drug development[,] identify potentially valid endpoints for clinical trials, understand the predictive value of preclinical models, clarify how medical products work in different diseases, and inform development of novel clinical designs and endpoints to the benefit of patients’.

  152. 152.

    ibid.

  153. 153.

    USFDA (2011) Advancing regulatory science at FDA, p. 12. https://www.fda.gov/media/81109/download. Accessed 26 Mar 2021. The specific mandate to promote public health innovation by advancing regulatory science is vested in the USFDA by the Administration Safety and Innovation Act (FDASIA). FDASIA, Pub. L. 112-144, sec 1124.

  154. 154.

    USFDA (2011) Advancing regulatory science at FDA, p. 12. https://www.fda.gov/media/81109/download. Accessed 26 Mar 2021.

  155. 155.

    USFDA (4 Jun 2013) Availability of masked and de-identified non-summary safety and efficacy data—request for comments. 78(107) Fed. Reg, p. 33422. https://www.govinfo.gov/content/pkg/FR-2013-06-04/pdf/2013-13083.pdf. Accessed 26 Mar 2021. In this regard, the FDA’s approach differs in principle from that of the EMA. See Koenig et al. (2015), pp. 10–11. On the EMA’s data analysis practice, see also Chap. 6 at Sect. 6.5.1.2.

  156. 156.

    In November 2016, the EMA held a workshop intending to identify the possibilities and challenges of how the potential of ‘big data’ could be leveraged to support drug R&D and regulatory decision-making. See EMA (2017) Identifying opportunities for ‘big data’ in medicines development and Regulatory Science. Report from a workshop held by EMA on 14–15 November 2016. EMA/740359/2016.

  157. 157.

    ibid.

  158. 158.

    At least, not at the time of writing. For the background information and related documents, see Workshop on identifying opportunities for ‘big data’ in medicines development and regulatory science. https://www.ema.europa.eu/en/events/workshop-identifying-opportunities-big-data-medicines-development-regulatory-science. Accessed 26 Mar 2021.

  159. 159.

    EMA (9 Oct 2017) Reflection paper on the use of extrapolation in the development of medicines for paediatrics. EMA/199678/2016, p. 3 (emphasis added).

  160. 160.

    EMA (19 Mar 2013) Concept paper on extrapolation of efficacy and safety in medicine development. EMA/129698/2012, p. 2.

  161. 161.

    EMA (9 Oct 2017) Reflection paper on the use of extrapolation in the development of medicines for paediatrics. EMA/199678/2016, p. 4 (emphasis added).

  162. 162.

    ibid p. 6 (pointing out that no new data needs to be generated to confirm the relationship between a treatment and efficacy if such relationship is well established and quantified and can be extrapolated to the target population, and if no gap in knowledge remains). Extrapolation can be of crucial importance for paediatric studies, where the extrapolation of data from adults to children could provide a basis for regulatory decision making, e.g. when planning and developing paediatric investigation plans. ibid pp. 9–10.

  163. 163.

    ibid p. 12.

  164. 164.

    ibid p. 6.

  165. 165.

    ibid p. 7.

  166. 166.

    ibid pp. 8, 12.

  167. 167.

    EMA (19 Mar 2013) Concept paper on extrapolation of efficacy and safety in medicine development. EMA/129698/2012, p. 4.

  168. 168.

    ibid (emphasis added).

  169. 169.

    ibid.

  170. 170.

    Atholl J, Pugatch MP and Taylor D (2013) Clinical trials and data transparency – the public interest case. A briefing paper, p. 10. http://www.pugatch-consilium.com/reports/PC_Report_Clinical_Trials_OnLine_Version.pdf. Accessed 26 Mar 2021.

  171. 171.

    EMA (30 Apr 2013) Advice to the European Medicines Agency from the Clinical Trial Advisory Group on Legal Aspects (CTAG5) – final advice, lines 89–90. https://www.ema.europa.eu/en/documents/other/ctag5-advice-european-medicines-agency-clinical-trial-advisory-group-legal-aspects-final-advice_en.pdf. Accessed 26 Mar 2021. Notably, this statement seems to contradict the very purpose of the EMA policy for data publication, namely, to promote drug research and innovation through secondary data analysis. See EMA publication policy 0070, pp. 3–4.

  172. 172.

    EMA publication policy 0070, annex 2.

  173. 173.

    See e.g. Cleophas et al. (2006), p. 205; Biltaji et al. (2017), p. 2338; Mayo et al. (2017).

References

Download references

Author information

Authors and Affiliations

Authors

Rights and permissions

Reprints and Permissions

Copyright information

© 2021 The Author(s), under exclusive license to Springer Nature Switzerland AG

About this chapter

Verify currency and authenticity via CrossMark

Cite this chapter

Kim, D. (2021). Secondary Analysis of Individual Patient-Level Clinical Trial Data: A Primer. In: Access to Non-Summary Clinical Trial Data for Research Purposes Under EU Law. Munich Studies on Innovation and Competition, vol 16. Springer, Cham. https://doi.org/10.1007/978-3-030-86778-2_3

Download citation

  • DOI: https://doi.org/10.1007/978-3-030-86778-2_3

  • Published:

  • Publisher Name: Springer, Cham

  • Print ISBN: 978-3-030-86777-5

  • Online ISBN: 978-3-030-86778-2

  • eBook Packages: Law and CriminologyLaw and Criminology (R0)