Niels Vermeer

Pharmacovigilance of biologicals dynamics in post-approval safety learning

Niels Sebastiaan Vermeer The studies presented in this thesis have been conducted under the umbrella of the Regulatory Science collaboration between the Dutch Medicines Evaluation Board (CBG-MEB) and the Utrecht Institute for Pharma- ceutical Sciences (UIPS). The CBG-MEB is dedicated to ensure that licensed medicinal products during their whole life-cycle have a positive benefit-risk. This role requires intensive collaboration with academic and clinical partners in order to develop new assessment and decision-making methods, to engage with the clinic and to strengthen regulatory science. This PhD thesis aims to go beyond its scientific merits as such by delivering science, learning and insight to promote public health.

Printing of this thesis was kindly supported by the Medicines Evaluation Board (CBG-MEB), the Utrecht Insti- tute for Pharmaceutical Sciences (UIPS), the Royal Dutch Pharmacist Association (KNMP), and the Nederlands Bijwerkingen Fonds [NBF, Dutch side effects fund].

ISBN: 978-94-6169-732-5 Layout and Printing: Optima Grafische Communicatie (www.ogc.nl) Cover illustration: Flock.2 by Anna Hepler (reprinted with permission)

Copyright © 2015 by NS Vermeer. The copyright of the articles that have been published has been transferred to the respective journals. Pharmacovigilance of biologicals dynamics in post-approval safety learning

Farmacovigilantie van biologicals dynamiek in kennisverwerving na registratie (met een samenvatting in het Nederlands)

Proefschrift

ter verkrijging van de graad van doctor aan de Universiteit Utrecht op gezag van de rector magnificus, prof. dr. G.J. van der Zwaan, ingevolge het besluit van het college voor promoties in het openbaar te verdedigen op woensdag 14 oktober 2015 des middags te 4.15 uur

door

Niels Sebastiaan Vermeer geboren op 25 juli 1985 te Den Helder Promotoren: Prof. dr. H.G.M. Leufkens Prof. dr. A.C.G. Egberts

Copromotoren: Dr. M.L. De Bruin Dr. S.M.J.M. Straus Table of contents

Chapter 1 | General introduction 7

Chapter 2 | The detection of adverse effects of biologicals after approval 21 2.1 Drug-induced Progressive Multifocal Leukoencephalopathy (PML): 23 lessons learned from contrasting natalizumab and rituximab Clinical Pharmacology & Therapeutics 2015 (in press) 2.2 Analytical interference of monoclonal antibodies in frequently used 43 immunoassays: an in vitro study

Chapter 3 | Challenges in the detection of manufacturing source-specific risks 55 3.1 Traceability of biopharmaceuticals in spontaneous reporting systems: 57 a cross-sectional study in the FAERS and EudraVigilance databases Drug Safety 2013; 36(8): 617-25 3.2 The effect of exposure misclassification in spontaneous ADR reports 75 on the time to detection of product-specific risks for biologicals: a simulation study 3.3 Traceability of biologicals: present challenges in pharmacovigilance 93 Expert Opinion on Drug Safety 2015; 14(1): 63-72

Chapter 4 | Dynamics in knowledge accrual on identified uncertainties 109 4.1 Risk management plans as tool for proactive pharmacovigilance: 111 a cohort study of new approved drugs in Europe Clinical Pharmacology & Therapeutics 2014; 96(6): 723-31 4.2 Cancer risks of medicines approved in the European Union: 133 what is known at licensing and what is solved post-marketing?

Chapter 5 | General discussion 161

Appendices | Summary 181 Nederlandse samenvatting 189 Acknowledgements/Dankwoord 197 List of co-authors 201 List of publications 205 About the author 207

1 General introduction

General introduction 9

Introduction

No pharmacologically active drug is without the risk of adverse effects. Ever since humans started using plant or otherwise nature-derived substances for the treatment and alleviation 1 of disease, adverse effects must have existed. The term “pharmacovigilance” (from phar- makon [Greek] and vigilare [Latin]) was first used in writing in 1969 [1] to describe the science and activities related to the detection, assessment, understanding and prevention of adverse effects [2]. Yet, pharmacovigilance has been practiced for many centuries. For example, Hippocrates (460 – 370 BC) recorded his observation that the use of hellebore for its laxative properties could in certain patients induce life-threatening convulsions [3]. Until today, such observations of the “intuitively unusual” by healthcare professionals or patients themselves are essential to pharmacovigilance, though the practice of pharmacovigilance and pharmaceutical landscape have changed dramatically. National schemes for the reporting of suspected adverse drug reactions (ADRs) have been established across Europe and other parts of the world in the 1960s [4,5], and have been refined with and supplemented by new methodologies and regulatory trajectories over the last decades, including quantitative signal detection methods, pharmacoepidemiological risk evaluation studies, pharmacovigilance planning and risk communication and mitigation strategies [6-10]. Nowadays, a wide array of medicines for numerous diseases is available, which have become imperative to human health. At the same time, adverse effects constitute an increasing public health burden. Recent European studies found that 2.4 - 6.5% of all hospital admissions and 0.05 - 0.5% of all in-hospital mortality are directly related to ADRs [11-14], corresponding to 2 - 5.4 million hospitalizations and 42,000 - 419,000 deaths in Europe each year [14]. Pharmacovigilance is therefore an important component of our current health care system, though its practice has become increasingly challenging with the use of more complex treatments. Biologicals are an increasingly important group of medicines that pose specific challenges in pharmacovigilance, because of their unique characteristics. According to EU law, biologicals are medicinal products in which the active substance is “a substance that is produced by or extracted from a biological source (…)” [15]. This definition distinguishes biologicals from medicines produced by chemical synthesis like acetylsalicylic acid (Aspirin), also referred to as “small-molecule drugs”. In other literature definitions, biologicals have been described as “pharmaceutical products consisting of (glyco)proteins” [16], or “products produced by modern biotechnological techniques” [17]. The scope of the latter definition is restricted to proteins produced by recombinant DNA technology, a technique invented in the 1970s and commercialized in the 1980s [18]. Particularly this latter group has been of paramount importance in recent pharmaceutical innovation. Between 1982 and 2014, a total of 246 recombinant biologicals (166 with distinct active substances) have been approved in Europe and the US [19,20], ranging from recombinant proteins for very rare diseases (e.g. alglucosidase alfa for Pompe disease) to blockbuster monoclonal antibodies (e.g. adalim- 10 Chapter 1

umab for rheumatoid arthritis or rituximab for various indications). Nowadays, biologicals continue to being developed, with more than 40% of today’s non-proprietary name applications at the World Health Organization comprising biologicals [21]. Moreover, biologicals are progressively finding their way into clinical practice [22], and 8 out of 10 medicines with the highest expenditures in Europe now comprise biologicals [23]. With the increasing availability and use of biologicals, an increasing amount of pharmacovigilance data is being generated for these products. For example, the number of ADR reports for biologicals has gradually increased over time. Already in 2010, biologicals were reported as suspected drugs in a quarter of all spontaneous reports in EudraVigilance, the European database for collection of ADR reports (see Figure 1). In previous research, the challenges associated with the monitoring of this increasingly important group of medicine have been explored. It was, amongst others, demonstrated that biologicals are associated with different types of safety concerns [24-26], are more sensitive to changes in manufacturing conditions [27], and are associated with more uncertainties about the benefit-risk profile at market approval [28], as compared to small-molecule drugs. The dynamics in post-approval safety learning in relation to these specific challenges posed by biologicals have, however, not been explored in detail. In this thesis, several aspects of the post-approval safety learning are explored, including the dynamics in knowledge accrual on unexpected adverse effects and on identified uncertainties, as well as the challenges related to ensuring adequate exposure ascertainment to enable the detection of manufacturing source-specific risks. 0

0 300 Cases with 0 , 1 23.7% 25.0% suspected x r biological* a

e 250 24.4% y

r e p

s 200

t 12.9% r o p e r

150 R 11.5% D A

f 100 o

r 14.8% e b

m 50 u N

0 2005 2006 2007 2008 2009 2010 Year of reporting

Figure 1. Number of spontaneous ADR reports processed in EudraVigilance per year, and proportion of reports with biological as suspected drug. *Data obtained from European Medicines Agency (see chapter 3.1). Biologicals are defined as “pharmaceutical products consisting of (glyco)proteins”. “Backlog reports” were excluded from this analysis. General introduction 11

The detection of adverse effects of biologicals after approval Previous empirical work has shown that biologicals are associated with a different type of adverse effects, as compared to small-molecule drugs. Risks of biologicals are more frequently related to infections, neoplasms and immune reactions [24,25], and less frequently to psy- 1 chiatric or vascular disorders. These findings may partly be attributable to the differences in mechanism of action and approved indications. Biologicals relatively frequently comprise antineoplastic and immunomodulatory drugs, for which infections and neoplasms may be expected on basis of the pharmacology, and susceptibility of the treated patient population. Similar differences in ADR profile were, however, also observed between biologicals and small-molecule drugs approved within the same indication [26]. This finding suggests that, apart from the differences in indications, other factors also play a role in the observed differences in ADR profile, including potential differences in potency [29]. Monoclonal antibodies, for example, selectively and potently target specific pathophysiological processes, but overstimulation of certain pathways may tip the balance to the negative. This “problem” of potency is illustrated by the occurrence of progressive multifocal leukoencephalopathy (PML) with natalizumab. Although natalizumab effectively ameliorates the inflammatory processes involved in multiple sclerosis, it has also been associated with an increased risk of PML [30], an opportunistic brain infection until recently almost exclusively observed in severely immunocompromised patients with HIV/AIDS and patients with B-cell malignancies. The occurrence of such ADRs that have not been described as drug-induced conditions before presents challenges to pharmacovigilance, particularly with respect to recognition of these unexpected adverse effects, but also given the lack of diagnostic criteria and common case definitions. The increased propensity for immunogenicity is another characteristic of biologicals that sets them apart from small-molecule drugs [31]. The immunogenic potential is influenced by many factors, including the structural properties of the biological (amino acid sequence, glyco- sylation pattern), the presence of contaminations or process-related impurities, and the method of administration. Although the formation of antidrug antibodies is, in itself, no adverse outcome, it can ultimately trigger a wide variety of clinical conditions. These include attenuation or complete loss of treatment response due to formation of binding or neutralizing antibodies [32]; interference with normal physiological functioning due to formation of auto-antibodies (e.g. anaemia due to anti-erythropoietin antibodies[33]); occurrence of thrombotic events due to formation of immune complexes [34]; or the occurrence of infusion-related reactions [35]. Apart from these in vivo effects, immunogenicity may also result in interference with routine laboratory monitoring. Biologicals can, for example, induce the formation of heterophile antibodies, which have a weak affinity for an array of ill-defined antigens, possibly including antigens used in immunoassays, thereby leading to interference [36]. With the use of laboratory markers in pharmacovigilance [37], lab test interference could potentially result in incorrect diagnosis and the generation of false positive safety signals. 12 Chapter 1

Challenges in the detection of manufacturing source-specific risks Small-molecule drugs typically comprise relatively simple chemical compounds, which are manufactured through well-defined chemical processes. By contrast, biologicals are large and complex molecules produced in living cells, and their activity depends on the protein folding into complex conformational (secondary, tertiary, or even quaternary) structures, which may be (easily) disturbed [16]. As a result, the characteristics of biologicals are, more than small-molecule drugs, determined by the specifics of the production process (e.g. type of cell line, growth conditions, purification process), the product formulation, handling, and storage conditions. Since the production and formulation process of marketed biologicals may be frequently changed throughout the product life-cycle – e.g., as much as 37 times for Remicade, infliximab [38] – the quality characteristics may shift over time [39,40], which may potentially result in differences in safety and/or efficacy. Moreover, unintended changes in the production process, and/or the occurrence of process-related impurities, may lead to (batch-specific) drifts in the product quality, which may rarely affect the benefit-risk profile. Several recent examples illustrate the close link between the production and formulation process and the clinical characteristics of biologicals (see Table 1). Another aspect of this process determines the product-principle is that the development and regulation of generic versions of biologicals (“biosimilars”) poses new challenges. While the production process for small-molecule drugs can be reliably replicated by a third manufacturer to produce identical copies of the reference product, the production of similar biological substances by different manufacturers will by definition not result in identical copies, as a result of unavoidable differences in manufacturing conditions [31,41]. In fact, not even two package units from the same manufacturer can be considered identical, due to inherent variability of biological production systems. Separate approval pathways have therefore been established for biosimilars [15,42], which aim to confirm similarity in quality characteristics (i.e. confirm variability does not exceed variability of reference product), efficacy and safety. Subtle differences may however only be detected after approval and usage in clinical practice, also because biosimilars have their own life-cycles; there is no requirement to demonstrate similarity to the reference product for changes in manufacturing process issued after approval. Because manufacturing variability between or within products over time may result in product- or even batch-specific risks, detailed exposure information (i.e. batch-, formulation-, and product-specific) should be available in pharmacovigilance systems that are used for the monitoring the risks of biologicals. In view of the expected surge in availability of biosimilars due to patent expiries [43], adequate traceability will become increasingly important, and may be central to trust in the safety of available biosimilar therapies. At the same time, traceability will be increasingly challenged with the availability of multiple versions of the same substances from different manufacturing sources, and the possibility of patients switching between these (whether or not through pharmacist-led substitution), triggering a debate on the required naming conventions for biosimilars to ensure traceability General introduction 13

1 An increase in reporting TMA, a diverse group of microvascular microvascular of group a diverse TMA, in reporting increase An manufacturing after observed was Rebif disorders, for occlusive been had issued. changes An increased risk of inhibitor development was observed for observed was for development inhibitor of risk increased An (Kogenate, products VIII factor full-length second-generation products. VIII factor full-length other to compared as Helixate), confirm to required data more considered authorities Regulatory signal. In a clinical study investigating a new subcutaneous formulation formulation a new subcutaneous investigating a clinical study In PRCA. developed patients two biosimilar, HX575, a marketed for of the manufacturing in the tungsten of useto the related The risk aggregates. protein of formation the inducing syringes, pre-filled Variations in the manufacturing process for Octagam and Octagam and for process in the manufacturing Variations in (increase potential pro-coagulant increased Vivaglobin in TEE increase led to kallikrein), and XI and factor activated risk. Batches of OCS-contaminated heparin, issued from multiple multiple from issued heparin, OCS-contaminated of Batches linked to were products, used in multiple in China and sources in the US and reactions hypersensitivity fatal) (including severe animal extracted directly from a polysaccharide is Heparin EU. tissue. Formulation changes increased formation of aggregates (though (though aggregates of formation increased changes Formulation 17-fold estimated in an resulting still debated), mechanism anaemia. profound of condition a rare PRCA, of in risk increase Additional information Additional Unknown; change in change Unknown; process manufacturing Unknown; Unknown; immunogenicity between difference products New subcutaneous subcutaneous New (not formulation marketed) Change in Change resulted manufacturing shift in quality Contamination Contamination oversulfated with sulfate chondroitin (OCS) Replacement of human human of Replacement serum by album 80 polysorbate Root cause Thrombotic Thrombotic microangiopathy (TMA) Inhibitor Inhibitor development PRCA Thromboembolic Thromboembolic (TEE) events Severe hypersensitivity reactions Pure Red CellPure (PRCA) Aplasia Risk Case [55] reports Register [53,54] Register Clinical trial [52] Case [49- reports 51] Case [48] reports Case reports [27,33,47] Data source Data 2014 2013 2012 2008 - 2012 2007 - 2008 1998 - 2002 Year Rebif, interferon beta interferon Rebif, Kogenate/ Helixate, Helixate, Kogenate/ VIII factor HX575, epoetin alfa Octagam/ Vivaglobin, immunoglobulin Multiple, heparin Multiple, Eprex, epoetin alfa Eprex, Recent safety signals of manufacturing source-specific risks for biologicals. for risks source-specific manufacturing signals of safety 1. Recent Table Product 14 Chapter 1

[44-46]. The extent to which detailed exposure information for biologicals is captured in pharmacovigilance data sources, the extent to which existing supply chain standards assist in this, and how robust pharmacovigilance systems are to the potential for exposure misclassification is, however, unknown.

Addressing knowledge gaps through proactive pharmacovigilance Regulatory decisions to allow new drugs on the market by definition have to accept a level of uncertainty about the full benefit-risk balance [56]. Pre-approval studies typically provide information on a limited number of patients over a relative short follow-up period, and handle strict inclusion criteria with respect to patients’ comorbidities and use of concomitant medications [57,58]. Therefore, uncertainties exist regarding risks and in particular rare and long-term ADRs, as well as with regard to the benefit-risk profile in specific populations or context of use (e.g. in patients with renal impairment, or safety in home therapy). A previous study found that biologicals are on average associated with more identified uncertainties at approval, as compared to small-molecule drugs [28]. This can firstly be explained by the immunogenic nature and species-specific action of biologicals, which limits the predictive value of preclinical data to human pharmacodynamics [59,60]. This is exemplified by the catastrophic first-in-man study in which all volunteers developed life-threatening ADRs to a novel anti-CD28 antibody (TGN1412), which drug had proven safe in animal studies [61]. Secondly, the increased uncertainty for biologicals may also relate to their context of use, including use in orphan indications (with fewer clinical data available at approval [62]), and to complexity of the procedures involved in the preparation and administration of these medicines, and in the clinical monitoring of patients receiving them. An important aim of pharmacovigilance is to progressively increase knowledge and evidence on these uncertainties after approval. This may either involve “demonstrating safety” over time (i.e. to refute potential risks, or to confirm its safe use in specific populations), or confirming theoretical risks and taking appropriate measures to minimize the risk. In rare cases, the confirmation of risks may also result in the withdrawal of a product. For example, in 2004, the small-molecule drug rofecoxib (Vioxx) was withdrawn over an increased risk of cardiovascular death, a potential risk already described before approval [63]. To ensure that existing knowledge gaps are addressed in a more proactive and timely manner, European regulators since 2005 require a risk management plan (RMP) for all new drug applications [9]. The RMP contains a pharmacovigilance plan on the post-approval data acquisition for important identified uncertainties (“known unknowns”). It is, however, unknown to what extent these identified knowledge gaps for new drugs are filled over time along this proactive approach. This may be of particular importance for biologicals given the greater level of uncertainty at initial approval [28], and may become increasingly important in general, as novel approval pathways that allow a greater level of uncertainty [64,65] should rely on a robust pharmacovigilance net in which existing uncertainties are timely addressed. General introduction 15

Thesis aim

The overarching aim of this thesis is to explore the dynamics in the post-approval safety learning for biologicals. We thereby specifically focus on the dynamics in post-approval 1 knowledge accrual on unexpected adverse effects and on identified uncertainties, as well as on the challenges in ensuring adequate exposure ascertainment in view of the potential for manufacturing source-specific risks.

Thesis outline

This thesis includes seven studies divided over three chapters, followed by a general discussion about the implications of our findings, and recommendations for current practice and future research. Chapter 2 focuses on the detection of unexpected adverse effects after approval. In chapter 2.1 we compare the spontaneous reporting patterns for drug-induced progressive multifocal leukoencephalopathy (PML) between natalizumab and rituximab. PML has recently been identified as a serious drug-induced condition of several immunomodulatory biologicals, used under diverse clinical conditions. The difference in timing with respect to the recognition of the PML as a drug-induced condition of natalizumab (prior to re- approval) and rituximab (nine years after approval) has provided a unique opportunity to contrast the products with respect to their PML reporting patterns. Chapter 2.2 evaluates the potential for analytical interference by monoclonal antibodies within immunoassays. Interference in clinical laboratory testing may lead to incorrect diagnosing and faulty clinical decision-making, and could ultimately contribute to adverse patient outcomes and false positive safety signals. A unique characteristic of biologicals is that new risks, including previously unobserved adverse effects, may emerge at any point in the drug life-cycle as a result of manufacturing variability within or across products over time. Chapter 3 focuses on the challenges in ensuring adequate exposure ascertainment for biologicals, in view of this potential for manufacturing source-specific risks. We explore the current status of traceability of biologicals in the main European and US spontaneous reporting systems (chapter 3.1), and assess the robustness of such systems to the potential for exposure misclassification (chapter 3.2). In chapter 3.3 we explore the processes involved in ensuring traceability, and on-going advances herein, and provide recommendations for the traceability of biologicals that may contribute to more robust post-marketing surveillance of biologicals. Chapter 4 focuses on the post-approval knowledge accrual on identified uncertainties. For all new drug approvals in Europe, the main uncertainties in context of the benefit-risk balance of a drug are nowadays documented in a risk management plan (RMP) at approval. 16 Chapter 1

In chapter 4.1 we examine the evolution of safety concerns in the RMP after initial approval for a selected cohort of 31 small-molecule drugs and 17 biologicals, to provide insight into the knowledge gain over time. In chapter 4.2 we aim to provide an overview of potential cancer risks associated with new drugs approved in Europe, and examine the knowledge accrual on potential risks of cancer after approval and the contribution of pro-active pharmacovigilance activities. General introduction 17

References

1. World Health Organization 1969. Pharmacovigi- 12. Pirmohamed, M. et al. Adverse drug reactions lance internationale : rôle de l’ hôpital, rapport d’ as cause of admission to hospital: prospective une réunion de l’ OMS [tenue à Genève du 18 au analysis of 18 820 patients. BMJ 329, 15-19 1 23 novembre 1968]. Available from: http://apps. (2004). who.int/iris/handle/10665/37994 [Last accessed 13. Leendertse, A. J., Egberts, A. C., Stoker, L. J., 14 January 2015]. van den Bemt, P. M. & HARM Study Group. 2. World Health Organization 2006. The safety of Frequency of and risk factors for preventable medicines in public health programmes: Phar- medication-related hospital admissions in the macovigilance an essential tool. Available Netherlands. Arch. Intern. Med. 168, 1890-1896 from: http://www.who.int/medicines/areas/ (2008). quality_safety/safety_efficacy/pharmvigi/en/ 14. Bouvy, J. C., De Bruin, M. L. & Koopmanschap, [Last accessed 20 January 2015]. M. A. Epidemiology of Adverse Drug Reactions 3. Brill, E. in Hippocrates. Pseudepigraphic Writings. in Europe: A Review of Recent Observational Letters-Embassy-Speech from the Altar-Decree. Studies. Drug Saf. 38, 437-453 (2015). Edited and translated with an introduction by 15. European parliament and Council of 27 Wesley D. Smith (Brill Academic Pub, Leiden, January 2003. Directive 2001/83/EC of the 1990). European Parliament and of the Council of 4. Mcbride, W. G. Thalidomide and congenital 6 November 2001 on the community code abnormalities. The Lancet 278, 1358 (1961). relating to medicinal products for human use. 5. Lenz, W. A personal perspective on the thalido- Available from: http://ec.europa.eu/health/files/ mide tragedy. Teratology 46, 417-418 (1992). eudralex/vol-1/dir_2001_83_consol_2012/ 6. Meyboom, R. H., Egberts, A. C., Gribnau, F. W. & dir_2001_83_cons_2012_en.pdf [Last accessed Hekster, Y. A. Pharmacovigilance in perspective. 30 October 2014]. Drug Saf. 21, 429-447 (1999). 16. Crommelin, D. J. A. et al. Shifting paradigms: 7. van Puijenbroek, E. P. et al. A comparison of biopharmaceuticals versus low molecular weight measures of disproportionality for signal detec- drugs. Int. J. Pharm. 266, 3-16 (2003). tion in spontaneous reporting systems for adverse 17. Walsh, G. Biopharmaceuticals and biotechnology drug reactions. Pharmacoepidemiol. Drug Saf. medicines: an issue of nomenclature. Eur J Pharm 11, 3-10 (2002). Sci 15, 135-138 (2002). 8. Harmark, L. & van Grootheest, A. C. Pharma- 18. Wright, S. Recombinant DNA Technology and covigilance: methods, recent developments and Its Social Transformation, 1972-1982. Osiris 2, future perspectives. Eur. J. Clin. Pharmacol. 64, 303-360 (1986). 743-752 (2008). 19. Johnson, I. S. Human Insulin from Recombinant 9. Raine, J., Wise, L., Blackburn, S., Eichler, H. G. DNA Technology. Science 219, pp. 632-637 & Breckenridge, A. European perspective on risk (1983). management and drug safety. Clin. Pharmacol. 20. Walsh, G. Biopharmaceutical benchmarks 2014. Ther. 89, 650-654 (2011). Nat. Biotechnol. 32, 992-1000 (2014). 10. Arlett, P. et al. Proactively managing the risk 21. World Health Organization (WHO). July 2014. of marketed drugs: experience with the EMA Biological Qualifier an INN proposal. Available Pharmacovigilance Risk Assessment Committee. from: http://www.who.int/medicines/services/ Nat Rev Drug Discov 13, 395-397 (2014). inn/bq_innproposal201407.pdf [Last accessed: 11. Pouyanne, P., Haramburu, F., Imbs, J. L. & 26 February 2015]. Begaud, B. Admissions to hospital caused by 22. IMS Institute for Healthcare Informatics. Novem- adverse drug reactions: cross sectional incidence ber 2013. The Global Use of Medicines: Outlook study. French Pharmacovigilance Centres. BMJ through 2017. Available from: http://www. 320, 1036 (2000). imshealth.com/deployedfiles/imshealth/Global/ 18 Chapter 1

Content/Corporate/IMS%20Health%20Institute/ 34. Korswagen, L. A. et al. Venous and arterial Reports/Global_Use_of_Meds_Outlook_2017/ thromboembolic events in adalimumab-treated IIHI_Global_Use_of_Meds_Report_2013.pdf patients with antiadalimumab antibodies: a case [Last accessed: 26 February 2015]. series and cohort study. Arthritis Rheum. 63, 23. Generics and Biosimilars Initiative (GaBI). 27 877-883 (2011). June 2014. Biologicals dominate Europe’s best 35. Baert, F. et al. Influence of Immunogenicity on sellers. Available from: http://www.gabionline. the Long-Term Efficacy of Infliximab in Crohn’s net/Reports/Biologicals-dominate-Europe-s- Disease. N. Engl. J. Med. 348, 601-608 (2003). best-sellers [Last accessed: 16 January 2015]. 36. Levin, O., Morris, L. F., Wah, D. T., Butch, 24. Giezen, T. J. et al. Safety-related regulatory actions A. W. & Yeh, M. W. Falsely elevated plasma for biologicals approved in the United States and parathyroid hormone level mimicking tertiary the European Union. JAMA 300, 1887-1896 hyperparathyroidism. Endocr. Pract. 17, e1-4 (2008). (2011). 25. Giezen, T. J. et al. Mapping the safety profile of 37. Giezen, T. J. et al. Rituximab-induced thrombo- biologicals: a disproportionality analysis using cytopenia: a cohort study. Eur. J. Haematol. 89, the WHO adverse drug reaction database, 256-266 (2012). VigiBase. Drug Saf. 33, 865-878 (2010). 38. Schneider, C. K. Biosimilars in rheumatology: 26. Ebbers, H. C. et al. Differences between post- the wind of change. Ann Rheum Dis. 72, 315-318 authorization adverse drug reactions of biophar- (2013). maceuticals and small molecules. BioDrugs 27, 39. Schiestl, M. et al. Acceptable changes in quality 167-174 (2013). attributes of glycosylated biopharmaceuticals. 27. Ebbers, H. C., Mantel-Teeuwisse, A. K., Moors, Nat. Biotechnol. 29, 310-312 (2011). E. H., Schellekens, H. & Leufkens, H. G. Today’s 40. Ramanan, S. & Grampp, G. Drift, Evolution, Challenges in Pharmacovigilance: What can and Divergence in Biologics and Biosimilars we Learn from Epoetins? Drug Saf. 34, 273-287 Manufacturing. BioDrugs 28, 363-372 (2014). (2011). 41. Weise, M. et al. Biosimilars: what clinicians 28. Giezen, T. J. et al. Evaluation of post-authorization should know. Blood 120, 5111-5117 (2012). safety studies in the first cohort of EU Risk Man- 42. European Medicines Agency. Committee for agement Plans at time of regulatory approval. Medicinal Products for Human Use (CHMP). Drug Saf. 32, 1175-1187 (2009). Guideline on similar biological medicinal 29. Scott, C. T. The problem with potency.Nat Biotech products. CHMP/437/04 Rev 1. 23 October 2014. 23, 1037-1039 (2005). 43. Generics and Biosimilars Initiative (GaBI). 29 30. Bloomgren, G. et al. Risk of natalizumab-associ- June 2012. AUS$67 billion worth of biosimilar ated progressive multifocal leukoencephalopathy. patents expiring before 2020. Available from: N. Engl. J. Med. 366, 1870-1880 (2012). http://www.gabionline.net/Biosimilars/ 31. Schellekens, H. Bioequivalence and the im- General/US-67-billion-worth-of-biosimilar- munogenicity of biopharmaceuticals. Nat. Rev. patents-expiring-before-2020 [Last accessed: Drug Discov. 1, 457-462 (2002). 26 February 2015]. 32. Bartelds, G. M. et al. Development of Antidrug 44. The INN crowd.Nat. Biotechnol. 31, 1055-1055 Antibodies Against Adalimumab and Asso- (2013). ciation With Disease Activity and Treatment 45. Silverman, E. Biosimilars: what’s in a name? BMJ Failure During Long-term Follow-up. JAMA 348, g272 (2014). 305, 1460-1468 (2011). 46. Schellekens, H. A divided INN crowd. Nat. 33. Casadevall, N. et al. Pure red-cell aplasia and Biotechnol. 32, 132-132 (2014). antierythropoietin antibodies in patients treated 47. Boven, K. et al. The increased incidence of pure with recombinant erythropoietin. N. Engl. J. Med. red cell aplasia with an Eprex formulation in 346, 469-475 (2002). uncoated rubber stopper syringes. Kidney Int. 67, 2346-2353 (2005). General introduction 19

48. Kishimoto, T. K. et al. Contaminated Heparin 55. Hunt, D. et al. Thrombotic Microangiopathy Associated with Adverse Clinical Events and Associated with Interferon Beta. N. Engl. J. Med. Activation of the Contact System. N. Engl. J. 370, 1270-1271 (2014). Med. 358, 2457-2467 (2008). 56. Eichler, H. G., Pignatti, F., Flamion, B., Leufkens, 49. European Medicines Agency. 12 May 2011. H. & Breckenridge, A. Balancing early market 1 Assessment report for Octagam and associated access to new drugs with the need for benefit/ names (EMEA/H/A-31/1281). Available from: risk data: a mounting dilemma. Nat. Rev. Drug http://www.ema.europa.eu/docs/en_GB/docu- Discov. 7, 818-826 (2008). ment_library/Referrals_document/Octagam_31/ 57. Duijnhoven, R. G. et al. Number of patients WC500154855.pdf [Last Accessed: 10 February studied prior to approval of new medicines: 2015]. a database analysis. PLoS Med. 10, e1001407 50. European Medicines Agency. 7 May 2012. (2013). Referral assessment report for Vivaglobin and 58. Rothwell, P. M. External validity of randomised associated names (human normal immunoglo- controlled trials: “to whom do the results of this bin solution for injection - subcutaneous use) trial apply?”. Lancet 365, 82-93 (2005). (EMEA/H/A-36/1296). Available from: http:// 59. Dixit, R. & Boelsterli, U. A. Healthy animals and www.ema.europa.eu/docs/en_GB/document_li- animal models of human disease(s) in safety brary/Referrals_document/Vivaglobin_36/ assessment of human pharmaceuticals, including WC500154569.pdf [Last accessed: 10 February therapeutic antibodies. Drug Discov. Today 12, 2015]. 336-342 (2007). 51. Funk, M. B. et al. Thromboembolic events as- 60. van Meer, P. J. K. et al. Immunogenicity of mAbs sociated with immunoglobulin treatment. Vox in non-human primates during nonclinical safety Sang. 105, 54-64 (2013). assessment. mAbs 5, 810-816 (2013). 52. Haag-Weber, M. et al. Safety, immunogenicity 61. Suntharalingam, G. et al. Cytokine storm in and efficacy of subcutaneous biosimilar epoetin- a phase 1 trial of the anti-CD28 monoclonal alpha (HX575) in non-dialysis patients with renal antibody TGN1412. N. Engl. J. Med. 355, 1018- anemia: a multi-center, randomized, double- 1028 (2006). blind study. Clin. Nephrol. 77, 8-17 (2012). 62. Melnikova, I. Rare diseases and orphan drugs. 53. Gouw, S. C. et al. Factor VIII Products and Nat. Rev. Drug Discov. 11, 267-268 (2012). Inhibitor Development in Severe Hemophilia 63. Psaty, B. M. & Furberg, C. D. COX-2 inhibi- A. N. Engl. J. Med. 368, 231-239 (2013). tors—lessons in drug safety. N. Engl. J. Med. 352, 54. European Medicines Agency. 20 February 1133-1135 (2005). 2014. CHMP confirms PRAC recom- 64. Eichler, H. et al. Adaptive Licensing: Taking the mendations on Kogenate Bayer/Helixate Next Step in the Evolution of Drug Approval. NexGen (EMA/108793/20144). Available Clin. Pharmacol. Ther. 91, 426-437 (2012). from: http://www.ema.europa.eu/docs/ 65. Eichler, H. et al. From adaptive licensing to en_GB/document_library/Referrals_docu- adaptive pathways: Delivering a flexible life-span ment/Kogenate_Bayer_Helixate_NexGen_20/ approach to bring new drugs to patients. Clin. Opinion_provided_by_Committee_for_Medici- Pharmacol. Ther. 97, 234-246 (2015). nal_Products_for_Human_Use/WC500158680. pdf [Last accessed: 10 February 2015].

The detection of adverse effects of 2 biologicals after approval

Drug-induced Progressive 2.1 Multifocal Leukoencephalopathy (PML): Lessons learned from contrasting natalizumab and rituximab

Vermeer NS, Straus SMJM, Mantel-Teeuwisse AK, Hidalgo- Simon A, Egberts ACG, Leufkens HGM, De Bruin ML

Clinical Pharmacology & Therapeutics 2015 (in press)

Abstract

Background and objective Progressive Multifocal Leukoencephalopathy (PML) has been identified as a serious ADR of several immunomodulatory biologicals, used under diverse clinical conditions. In this study, we contrasted the ADR reporting patterns of PML for two biologicals for which the risk was identified at different points in their life-cycle: natalizumab (before re-approval) and rituximab (nine years post-approval). Methods Spontaneous reports on suspected drug-induced PML for natalizumab and rituximab were retrieved from EudraVigilance, the European database for collection of ADR reports. Information on case characteristics was extracted, and we explored temporal trends in reporting and assessed the completeness of the reports along the availability of data for eight predefined variables. Results A total of 375 and 287 reports on suspected drug-induced PML were retrieved for natalizumab and rituximab, respectively. We found that, apart from the differences in clinical characteristics (age, gender, indication, time to event, fatality), which reflect the diversity in context of use, PML reports for natalizumab were more complete (median data completeness 87.5% vs. 62.5%; p<0.001), and received sooner after occurrence (median 1 vs. 3 months; p<0.001). The completeness of the suspected PML reports for both products declined over time, with sequence of reporting. Conclusion This study serves as an important reminder that spontaneous reports should only be used with great caution to quantify and compare safety profiles across products over time. The observed variability in reporting patterns and heterogeneity of PML cases presents challenges to such comparisons. Lumping uncharacterized PML reports together without taking these differences into account may result in biased comparisons and flawed conclusions about differential safety. 24 Chapter 2.1

Introduction

Progressive multifocal leukoencephalopathy (PML) is a severe viral infection of the human brain associated with poor clinical outcomes, including disability and death [1]. The aetiol- ogy of PML involves reactivation of the latent polyomavirus JC in the presence of disorders associated with severe cellular immune deficiency. For a long time PML was a rare condition that was mostly observed among patients with haematological malignancies [2,3]. However, the incidence sharply increased with the onset of HIV/AIDS epidemic, and most cases have since then been diagnosed in HIV-infected patients [4,5]. Over the last decade, however, PML has increasingly been diagnosed in HIV-negative patients treated with immunomodulatory biologicals, particularly natalizumab and rituximab [6,7]. In February 2005, only three months after receiving accelerated marketing approval in the US, the marketing of natalizumab (Tysabri; Biogen Idec/Élan) was suspended after PML had been diagnosed in two patients participating in a clinical trial for multiple sclerosis [8-10]. A third case was subsequently discovered following a re-examination of a patient who had received natalizumab for inflammatory bowel disease, and who was initially falsely diagnosed with drug-induced fatal astrocytoma in 2003 [11]. After a reanalysis of previously natalizumab-treated patients confirmed that no additional PML cases had occurred, natalizumab was reintroduced in the US and first approved in Europe in June 2006. It was agreed that specific risk minimization and surveillance measures were to be undertaken to mitigate and further characterize the risk of PML, including a restricted distribution program in the US and a detailed risk management plan in Europe [12,13]. Since then, the number of Adverse Drug Reaction (ADR) reports for suspected drug-induced PML has vastly increased over time [14], and an increasing number of immunomodulatory drugs has been suspected or identified to increase the risk of PML [15,16]. Some of these products had been marketed for multiple years, suggesting that increased awareness may have contributed to the recognition of the risk of drug-induced PML. Rituximab (Rituxan/ Mabthera; Genentech/Hoffmann-La Roche) is an example of a biological that received increased scrutiny for drug-induced PML relatively late in its life-cycle [7,17]. Rituximab was first approved in 1997 for the treatment of relapsed and refractory follicular lymphoma. Yet, its use has considerably changed throughout its life-cycle, as new indications have been added (including rheumatoid arthritis), and the product became first- line therapy for several lymphoid malignancies. The risk of PML was first included in the US and European product label in 2007, more than nine years after initial approval, on the basis of data from spontaneous reports of suspected ADRs [18,19]. The difference in timing with respect to the recognition of PML as a drug-induced condition for natalizumab and rituximab is intriguing, and provides a unique opportunity to contrast the two products in relation to their respective ADR reporting patterns. Although PML has presented a common challenge in the pharmacovigilance of several immunomodu- Reporting patterns for drug-induced PML 25 latory biologicals, the differences in clinical context in which products are used, as well as the differences in regulatory history, including indication dynamics, of products, may impact on the ADR reporting patterns. This can present challenges to clinical and regulatory decision making. In this study, we contrasted the PML reporting patterns (e.g. case characteristics, temporality of reporting, completeness of the reports) between natalizumab and rituximab, given the distinct regulatory pathways over time, different indications and treatment populations of these two products.

Results 2 Case characteristics A total of 375 and 287 spontaneous reports on suspected drug-induced PML were retrieved for natalizumab and rituximab, respectively, from EudraVigilance, the European database for collection of ADR reports. As shown in Table 1, substantial differences were observed in patient and treatment characteristics between cases of natalizumab- and rituximab-associated PML. Compared to patients with rituximab-associated PML, patients with natalizumab-associated PML were on average younger (45 vs. 65 years; p<0.001), and were more frequently females (69.3% vs. 44.3%; p<0.001). The median time to onset from treatment initiation was 36 months (interquartile range [IQR]: 26-48 months) for natalizumab-associated PML, and 12 (IQR: 4-24) months for rituximab-associated PML (p<0.001). The outcome of the PML reaction was more frequently fatal for rituximab-associated PML (n=114, 39.7%), as compared to natalizumab-associated PML (n=40, 10.7%; p<0.001). For each case of natalizumab-associated PML, on average 3 (IQR: 2-5) reports had been received over time, involving either follow-up reports from the initial reporter, or multiple reports from different reporters on the same case. For rituximab-associated PML the average number of reports per case was 2 (IQR: 1-3). Medical doctors were most frequently involved in the reporting of PML cases for both natalizumab (n=348, 92.8%) and rituximab (n=251, 87.5%). Notably, patients were involved in the reporting of 89 (31.0%) cases for rituximab, and in 50 (13.3%) cases for natalizumab.

Temporal trends in reporting The first spontaneous PML report for natalizumab was received in July 2008, approximately two years after its initial marketing approval in Europe and being back on the market in the US (June 2006, see Figure 1). The first spontaneous report on rituximab-associated PML was received in December 2004, respectively 6.5 and 7 years after receiving initial marketing approval in Europe (June 1998) and the US (November 1997). The change-point analysis showed that the PML reporting rate was low (0.1 reports/month) for both natalizumab and rituximab in the year(s) following the first spontaneous report, but 26 Chapter 2.1

Table 1. Characteristics of spontaneous reports on suspected drug-induced PML for rituximab and natalizumab in EudraVigilance. Natalizumab (n=375) Rituximab (n=287) Patient and treatment information Age, median (IQR) 45 (37-52) 65 (57-73) Missing, n (%) 28 (7.5) 62 (21.6) Sex, n (%) Female 260 (69.3) 127 (44.3) Male 113 (30.1) 140 (48.8) Missing 2 (0.5) 20 (7.0) Indication, n (%)1 Multiple sclerosis 341 (90.9) - Lymphoid neoplasm - 218 (76.0) Rheumatoid arthritis/ other auto-immune disorder - 29 (10.1) Other condition 2 (0.5) 9 (3.1) Missing 33 (8.8) 34 (11.8) Time to onset, median months (IQR) 36 (26-48) 12 (4-24) Missing, n (%) 92 (24.5) 125 (43.6) Treatment duration, median months (IQR) 36 (25-48) 6 (3-20) Missing, n (%) 75 (20.0) 126 (43.9) Number of concomitant drugs, n (%) 0 concomitant drugs reported 221 (58.9) 76 (26.5) 1-5 concomitant drugs reported 109 (29.1) 139 (48.4) 6-10 concomitant drugs reported 27 (7.2) 51 (17.8) >10 concomitant drugs reported 18 (4.8) 21 (7.3) Outcome of PML reaction, n (%) Fatal 40 (10.7) 114 (39.7) Not recovered 205 (54.7) 82 (28.6) Recovered 29 (7.8) 6 (2.0) Recovering 4 (1.1) 7 (2.4) Missing 97 (25.9) 78 (27.2) Report information Reporter, n (%)1 Physician 348 (92.8) 251 (87.5) Pharmacist 9 (2.4) 20 (7.0) Other health professional 151 (40.3) 94 (33.8) Patient/ non-health professional 50 (13.3) 89 (31.0) Missing 1 (0.3) - Reporter region, n (%) Europe 233 (62.1) 161 (56.1) United States 119 (31.7) 79 (27.5) Other 23 (6.1) 47 (16.4) Reports per PML event, median (IQR) 3 (2-5) 2 (1-3) PML, progressive multifocal leukoencephalopathy; IQR, interquartile range. 1 Due to the possibility of multiple indications for rituximab/ natalizumab therapy per patient and the possibility of multiple reporters per event, these variables don’t add up to hundred per cent. Reporting patterns for drug-induced PML 27

20 Natalizumab (re)approved in US and EU EU DHPC on PML risk after long PML included in product label term use of natalizumab 15

Marketing of natalizumab voluntarily Number of spontaneous PML reports per month suspended over PML concerns EMA recommendation to suspend marketing * 10 of efalizumab over PML concerns

Natalizumab approved in US 5

0 1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7 8 9 10 11 12 1 2004 2005 2006 2007 2008 2009 2010 2011 2012 0

Rituximab approved in * US (1997) and EU (1998) 5

PML included in US product label 10

PML included in EU product label DHPC in US (Oct-09) and EU (Nov-09) on PML 15 cases in RA, including a fatal case in a patient DHPC in EU on PML cases in off-label 2 without other known risk factors indications of SLE and vasculitis 20 DHPC in US (Sep-08) and EU (Nov-08) on PML cases in patients with auto-immune disease, including RA, treated with rituximab 25

Figure 1. Temporal trends in the spontaneous reporting of cases suspected of drug-induced PML for natalizumab (above) and rituximab (below), according to the absolute number of reports per calendar-month (bars) and average number of reports per interval as found by the change-point analysis (lines). PML, progressive multifocal leukoencephalopathy; EMA, European Medicines Agency; DHPC, direct healthcare professional communication; SLE, system lupus erythematosus; RA, rheumatoid arthritis. *Data on Janu- ary 2013 is incomplete due to study end date (23 January 2013). increased over time. Overall, as shown in Figure 1, a total of respectively four and three change points in reporting rates were found for natalizumab and rituximab over time. For natalizumab, the reporting rate increased from 0.1 reports/month (until 9 September 2009) up to 24.2 reports/ month from 13 November 2012 onwards. For rituximab, the reporting rate was more constant over time, increasing from 0.1 reports/month (until 14 September 2006) up to 4.3 reports/month from 21 November 2008 onwards. However, a peak in reporting (11.3 reports/month) was noted between 30 September 2008 and 20 November 2008, with a higher proportion of reports originating from the US (n=12) relative to Europe (n=5) and other regions (n=3). This peak occurred shortly after a DHPC had been sent on rituximab-induced PML in the US (September 2008). Figure 2 shows the lag time between the occurrence of the PML reaction, and the receive date of the initial spontaneous report. Overall, the lag time was significantly shorter for cases of natalizumab-associated PML (median=1 month [IQR: 0-1]), as compared to cases of rituximab-associated PML (median=3 months [IQR: 1-8]; p<0.001). Notably, six cases of rituximab-associated PML in EudraVigilance had occurred prior to the receive date of the first spontaneous reports (December 2004), but were reported only thereafter. This was not observed for natalizumab. The peak in reporting for rituximab that was identified through change-point analysis, in particular involved cases with a long lag time (median=8 months [IQR: 1-51]) between reaction and reporting. As further shown in Figure 3, differences were 1-year lag | 3-year lag Rituximab PML reports (n= 187) 1-year lag | 3-year lag

28 Chapter 2.1 01-Jan-13 01-Jan-12 01-Jan-11 01-Jan-10 01-Jan-09 01-Jan-08 01-Jan-07 01-Jan-06 01-Jan-05 01-Jan-04 01-Jan-03 01-Jan-02 01-Jan-01 Date of PML reaction Date of PML 1-year lag Rituximab PML reports (n= 187) 1-year lag | 3-year lag | 3-year lag 01-Jan-13 01-Jan-12 01-Jan-11 01-Jan-10 01-Jan-09 01-Jan-08 01-Jan-07 01-Jan-06 01-Jan-05 01-Jan-04 01-Jan-03 01-Jan-02 01-Jan-01 Rituximab PML reports (n= 187) Date of PML reaction Date of PML Natalizumab Natalizumab PML (n=reports 317) Rituximab PML reports (n= 187) 1-year lag | 3-year lag

01-Jan-13 01-Jan-12 01-Jan-11 01-Jan-10 01-Jan-09 01-Jan-08 01-Jan-07 01-Jan-06 01-Jan-05 01-Jan-04 01-Jan-03 01-Jan-02 01-Jan-01 01-Jan-13 01-Jan-12 01-Jan-11 01-Jan-10 01-Jan-09 01-Jan-08 01-Jan-07 01-Jan-06 01-Jan-05 01-Jan-04 01-Jan-03 01-Jan-02 01-Jan-01

n o i t c a e r L M P n o t r o p e r t s r i f e t a d e v i e c e R 01-Jan-13 01-Jan-12 01-Jan-11 01-Jan-10 01-Jan-09 01-Jan-08 01-Jan-07 01-Jan-06 01-Jan-05 01-Jan-04 01-Jan-03 01-Jan-02 01-Jan-01 Date of PML reaction Date of PML Date of PML reaction Date of PML Natalizumab Natalizumab PML (n=reports 317) Rituximab PML reports (n= 187)

01-Jan-13 01-Jan-12 01-Jan-11 01-Jan-10 01-Jan-09 01-Jan-08 01-Jan-07 01-Jan-06 01-Jan-05 01-Jan-04 01-Jan-03 01-Jan-02 01-Jan-01

n o i t c a e r L M P n o t r o p e r t s r i f e t a d e v i e c e R 01-Jan-13 01-Jan-12 01-Jan-11 01-Jan-10 01-Jan-09 01-Jan-08 01-Jan-07 01-Jan-06 01-Jan-05 01-Jan-04 01-Jan-03 01-Jan-02 01-Jan-01 Natalizumab Natalizumab PML (n=reports 317) Date of PML reaction Date of PML Natalizumab Natalizumab PML (n=reports 317)

Lag time between (left)and natalizumab rituximab thefor occurrence suspectedof drug-induced PML reaction (X-axis) reporting and (Y-axis),

01-Jan-13 01-Jan-12 01-Jan-11 01-Jan-10 01-Jan-09 01-Jan-08 01-Jan-07 01-Jan-06 01-Jan-05 01-Jan-04 01-Jan-03 01-Jan-02 01-Jan-01

n o i t c a e r L M P n o t r o p e r t s r i f e t a d e v i e c e R

01-Jan-13 01-Jan-12 01-Jan-11 01-Jan-10 01-Jan-09 01-Jan-08 01-Jan-07 01-Jan-06 01-Jan-05 01-Jan-04 01-Jan-03 01-Jan-02 01-Jan-01

n o i t c a e r L M P n o t r o p e r t s r i f e t a d e v i e c e R Figure 2. Figure lag time. a zero indicates line diagonal The report. a spontaneous marks Each(right). in the bullet graphs leukoencephalopathy. multifocal PML, progressive Natalizumab Natalizumab PML (n=reports 317)

01-Jan-13 01-Jan-12 01-Jan-11 01-Jan-10 01-Jan-09 01-Jan-08 01-Jan-07 01-Jan-06 01-Jan-05 01-Jan-04 01-Jan-03 01-Jan-02 01-Jan-01

n o i t c a e r L M P n o t r o p e r t s r i f e t a d e v i e c e R Reporting patterns for drug-induced PML 29

Age + gender patient Age + gender patient 100% 100%

80% 80% Diagnostic test results Start date natalizumab Diagnostic test results Start date natalizumab 60% 60%

40% 40%

20% 20%

Outcome of PML 0% Duration of therapy Outcome of PML 0% Duration of therapy

PML reaction start date Indication of natalizumab PML reaction start date Indication of natalizumab

Previous drug history Previous drug history First 50 reports Natalizumab Natalizumab Last 50 reports

First 50 reports Rituximab Rituximab Last 50 reports Age + gender patient Age + gender patient 100% 100% 2 80% 80% Diagnostic test results Start date rituximab Diagnostic test results Start date rituximab 60% 60%

40% 40%

20% 20%

Outcome of PML 0% Duration of therapy Outcome of PML 0% Duration of therapy

PML reaction start date Indication of rituximab PML reaction start date Indication of rituximab

Previous drug history Previous drug history

Figure 3. Completeness of spontaneous reports on natalizumab-associated (above) and rituximab- associated (below) PML, stratified by first 50 case reports received (left) and latest 50 case reports (right). The inner plot reflects the first information received (initial report), the outer plot the most recent (“master” report) on the same cases. The radar chart shows the availability of data on the eight pre-defined variables in the reports. For example, 90% completeness of “start date natalizumab’” means that in 90% of the reports the start date of natalizumab had been provided by reporter. PML, progressive multifocal leukoencephalopathy. observed in the quality of the information on individual variables across the reports received over time. For rituximab, a particular decline in the reporting of information on diagnostic test results was observed over time: from 48% for the first fifty reports, to 14% for the latest fifty reports.

Trends in reporting completeness The completeness of PML reports was assessed along the availability of data for eight predefined variables on patient, treatment, and reaction details. Overall, taking into account any follow-up information received over time, the data completeness was 79% for natalizumab cases (median: 87.5% [IQR: 62.5-100%]), as compared to 60% for rituximab cases (median: 62.5% [IQR: 37.5-75%]; p<0.001). As shown in Table 2, the indication of therapy was frequently available for both natalizumab (91%) and rituximab (88%), but particular differences were observed with regard to the patient’s previous drug history (56% and 22%, respectively), and availability of diagnostic test results for the PML reaction (70% and 31%, respectively). 30 Chapter 2.1

For both natalizumab and rituximab, the completeness of the PML reports declined over time, with sequence of reporting (Figure 3). The decline was particularly pronounced for the initial reports, i.e. the first information received from the initial reporter. For natalizumab, the completeness of the initial reports declined from 71% (median: 75% [IQR: 50-87.5%]) for the first fifty reports, to 55% (median: 62.5% [IQR: 37.5-75%]; p=0.002) for the last reports. Similarly, for rituximab, the completeness declined from 53% (median: 62.5% [IQR: 37.5-75%]) to 40% (median: 37.5% [IQR: 12.5-62.5%]; p=0.013).

Table 2. Completeness of spontaneous reports on suspected natalizumab-induced and rituximab- induced PML. Natalizumab Rituximab (n=375) (n=287) n % n % Age and gender patient 345 92 223 78 Start date of therapy 316 84 190 66 End date and/or duration of therapy 300 80 161 56 Indication of therapy 342 91 253 88 Previous drug history 210 56 63 22 Start date of PML reaction 317 85 186 65 Outcome of PML reaction 278 74 209 73 Diagnostic tests results for PML 261 70 89 31 Total 79 60 PML, progressive multifocal leukoencephalopathy.

Discussion

This study highlights several differences in ADR reporting patterns for suspected cases of drug-induced PML between natalizumab and rituximab. We found that, apart from the differences in clinical characteristics (age, gender, indication of therapy, time to event, duration of use before event, and fatality rate subsequent to onset of PML), PML reports for natalizumab were more complete and were received sooner after occurrence, as compared to reports for rituximab. Furthermore, a time gap in reporting of suspected cases of rituximab-induced PML was observed within the first 7 years after approval. Several factors may account for these observed differences, including variability in patient and treating physician populations (health status, confounding by indication, clinical monitoring), as well as temporal issues related to the general awareness regarding the risk of drug-induced PML. The recognition of unexpected associations between drug exposure and clinical events is one of the main challenges in pharmacovigilance [20,21]. Rare but serious ADRs may therefore go unnoticed until first reported by a small number of observant clinicians. The Reporting patterns for drug-induced PML 31 here observed absence of spontaneous reports for rituximab-associated PML until 7 years after approval (December 2004), and low reporting rate until September 2006 (0.1 report/ month), also indicate that initial cases of drug-induced PML may have not have been recognized. As shown in Supplementary Figure 1, in which the PML reporting trends have been put into context of the use of natalizumab and rituximab, the PML reporting curves of both products are not proportional to the increase in patient exposure over time. With a median time to onset of 36 months, it is expected that the PML reporting of natalizumab lags behind a couple of years to the exposure curve. Though a shorter delay may be expected for rituximab, in view of the shorter time to onset (median 12 months), a larger delay was observed instead. Notably, by the time of the receipt of the first PML report for rituximab (December 2004), an estimated 540,000 patients had been exposed to the product in clinical 2 practice [22]. This suggests that particularly initial cases of rituximab-induced PML may not have been recognized, as further supported by our finding that six cases of suspected rituximab-induced PML in EudraVigilance occurred prior to 2004, but were only reported thereafter. The reason for the delay in recognition of PML as a potential drug-induced event of rituximab could be twofold: cases may either not have been diagnosed as PML, or may not have been attributed to rituximab but seen as a consequence of the disease (i.e. confounding by indication), or concomitant therapy. Rituximab was initially approved as third-line treatment for stage III-IV lymphoma. Diagnosing PML in this population can be particularly challenging because PML symptoms may be falsely interpreted as symptoms of central nervous system infiltration of lymphoma cells in relation to disease progression [23], or as symptoms of high-dose chemotherapy toxicity [24]. On the other hand, the elucidation of the causal involvement of rituximab presented a challenge, as lymphoma is an independent risk factor for PML [2], and rituximab was administered as third-line therapy in heavily pre-treated patients. By contrast, PML in multiple sclerosis patients is predominantly a phenomenon associated directly with natalizumab therapy. The recognition of PML as a drug-induced event of rituximab may thus relate to a shift in product usage; from third- to first-line treatment of lymphoma; and from use in the oncology setting to use in auto-immune diseases that have not traditionally been associated with PML, including rheumatoid arthritis [2]. On the other hand, safety learning between products could also have been important. Dissemination of information on the suspended marketing of natalizumab over PML concerns, may have contributed to the awareness among health professionals regarding the potential risk of rituximab-associated PML. Notably, regulatory authorities had received only one spontaneous report for rituximab-associated PML prior to the suspension of marketing of natalizumab over PML concerns, though, prior to this, observant clinicians had described a number of cases in medical literature [25-27]. The reporting of suspected rituximab-induced PML subsequently peaked after a safety alert had been issued, a phenomenon known as the “notoriety effect” [28]. Though not evaluated 32 Chapter 2.1

in this study, also media coverage may result in temporal increase in ADR reporting [29]. This may include social media coverage, particularly amongst multiple sclerosis patients, a patient group known to be very active on the internet. Apart from these effects, efforts to characterize and minimize the risk of drug-induced PML; including the establishment of a common case definition [30] and the formation of a global research agenda [14], may also have contributed to safety learning between products. Safety learning across and between products has repeatedly coined as a key activity for pharmacovigilance. Subsequent to the first documentation of QT-prolongation with quinidine therapy in 1964 [31], the risk has been identified for multiple other drugs, and eventually became the single most common reason for withdrawal and restriction of use of marketed drugs [32]. As a result, testing for drug-induced QT-prolongation is now a routine requirement for new drug approvals. More recently, subsequent to the identification of the risk of osteonecrosis of the jaw for the intravenous bisphosphonate zoledronic acid, and the hereto-related pharmacovigilance efforts [33], the whole class of bisphosphonates came under scrutiny [34], and a task force was formed to further characterize, prevent, and treat this drug-induced condition [35]. Over time, evolving regulatory pathways and learning within and between regulatory systems may result in earlier recognition and more timely regulatory actions for safety concerns like PML. Apart from the observed differences in temporal reporting patterns, we found that individual cases of suspected natalizumab-induced PML were generally received sooner after occurrence, when compared to cases of suspected rituximab-induced PML. This may be the result of routine MRI monitoring and evaluation of neurological status during natalizumab therapy as recommended by the risk management plan [12,13], and the potential for earlier diagnosis of PML in multiple sclerosis patients given the significant overlap in clinical signs and, therefore, earlier diagnostic workup. By contrast, as rituximab is primarily prescribed in oncology, in patients with more severe disease and the potential for disease progression and other side effects, physicians may be less likely to timely, if at all, investigate and report suspected cases of drug-induced PML. In view of the observed differences in reporting patterns, it is important to consider that spontaneous reports can only be used to quantify and compare the incidence of PML across products over time with great caution. Other pharmacovigilance databases should, therefore, be considered to further characterize and quantify the risk of PML, including patient registries, database of medical records and claims databases. Differences in patient and treating physician populations may also have been important in the observed differences in completeness of the PML reports. Quality and completeness of spontaneous reports is critical to efficient pharmacovigilance. Detailed information on individual cases may not only contribute to the timely identification but also further characterization of new risks, allowing the implementation of appropriate risk minimization strategies. For natalizumab, an algorithm was proposed to calculate an individual’s risk of developing drug-induced PML, on basis of information from spontaneous reports [36]. Pre- Reporting patterns for drug-induced PML 33 vious immunosuppressant use was one of the factors that increases the PML risk. The relative low availability and further decline of information on previous drug use in spontaneous reports for rituximab hampers the possibility to identify similar risk stratification factors if they would be present. The relative low availability of diagnostic test results is another aspect with potential undesirable consequences. Apart from clinical symptoms, PCR for JC virus DNA in cerebrospinal fluid, brain MRI, and brain biopsy/autopsy results form the basis to determine the diagnostic certainty of PML [30]. The Pharmacovigilance Risk Assessment Committee (PRAC) of the European Medicines Agency has recently proposed a labelling strategy for PML in which ascertainment of this diagnostic certainty plays a crucial role [37]. Paucity of diagnostic information hampers appropriate and consistent labelling resulting in suboptimal risk minimization. 2 In this study, we also observed substantial differences in clinical characteristics between cases of suspected natalizumab- and rituximab-induced PML. In accordance with previous case series [6,7], we found that patients with rituximab-associated PML had a significantly higher case fatality rate, as compared to patients with natalizumab-associated PML. The difference in mortality rate may firstly be explained by the difference in overall health status between the patient groups, including disease severity, prior treatments, and concomitant medication use. Apart from these patient-related factors, it has also been reported that PML associated with natalizumab is somewhat dissimilar from PML associated with HIV and hematologic disease [38], as it more frequently affects the frontal lobes, and is more commonly heralded by cognitive and behaviour disturbances. Furthermore, the potential for early diagnosis of PML in multiple sclerosis, as described above, may also contribute to a better prognosis. In conclusion, we have contrasted the occurrence of drug-induced PML for two biologicals with their own unique characteristics, temporal features and challenges. Despite all well-documented limitations, spontaneous reports remain critical to pharmacovigilance. This study serves as an important reminder that lumping uncharacterized PML reports together without taking into account variability in reporting patterns over time, differences in patient populations and treating physicians, may result in biased comparisons and flawed conclusions about differential safety.

Methods

Setting. We used data from the EudraVigilance database of the European Medicines Agency (EMA). EudraVigilance contains reports of suspected (serious) Adverse Drug Reactions (ADRs) to medicines licensed in Europe, including reports from clinical studies, literature reports, and reports from post-marketing use by health professionals and patients. As required by EU law, EudraVigilance contains all serious ADR reports that occur in the EU, and all serious unexpected ADR reports occurring in the rest of the world. As of 31 December 34 Chapter 2.1

2013, more than 4.5 million unique case reports are stored in the EudraVigilance database from worldwide reporting sources [39].

Selection of PML cases. We included cases from post-marketing use (“spontaneous reports”) for natalizumab and rituximab in which the reaction term involved “progressive multifocal leukoencephalopathy”. Other types of reports, including cases emerging from clinical studies and literature, were excluded from the present study.

Data extraction. For each case, both the report from the initial reporter was retrieved, as well as the “master” report, which contains the most recent information on the same case, including any follow-up information received over time. Information on the following standardized data elements [40] was extracted from the reports: administrative details (source country, receive date, reporter qualification, number of reports per case), patient characteristics (date of birth, age, sex); reactions details (reaction, reaction start date, reaction outcome); results of tests and procedures relevant to investiga- tion of the patients, including autopsy results; and drug information (current and past drug use, indication for use, therapy start and stop dates, duration of use). The data extraction was carried out at the EMA. Although ADR reports from the Eudra- Vigilance database are published on http://www.adrreports.eu, these publicly available data contain too limited details to fulfil the aim of the present study. The data lock point (DLP) for data extraction was 28 January 2013. This date matched the cut-off date of a quality check of PML reports in EudraVigilance, in which duplicate reports were detected and handled according to a predefined algorithm [41], and followed by a manual deduplication step.

Data classification and analysis. Case characteristics. Information on patient, therapy and report characteristics was retrieved from the most recent available (“master”) reports. For cases in which the age was not reported, the age was calculated using the date of birth and the reaction start date (where available), or the receive date of the report. The treatment duration was calculated using the therapy start and stop dates, and the time to onset was calculated using the therapy and reaction start dates. When only the month and year had been provided, the fifteenth of the month was used by default. Indications for rituximab therapy were categorized into the following groups: lymphoid neoplasms, unspecified neoplasms, rheumatoid arthritis or other autoimmune disorders, and other conditions. For natalizumab, we categorized the indications into multiple sclerosis, and other conditions. The source country was categorized into Europe, the US and other.

Temporal trends in the reporting of suspected drug-induced PML. We used the receive date to calculate temporal trends in reporting. The receive date comprises the date on which the report was received by the initial stakeholder (i.e. regulatory agency or pharmaceutical Reporting patterns for drug-induced PML 35 company), thus before the report is actually transmitted to the EudraVigilance database. The reporting trend was calculated by the number of reports per calendar month, and graphi- cally presented over time. Changes in reporting rates for each product over time were identified by a previously described change-point analysis [42]. This method assumes that the number of reports follows a Poisson distribution with constant intensities between two subsequent change points. The procedure involves a multistep approach. In the first step, all data over the entire interval (0, T] are used to test the model of no change points (M0), against the model of one single change point (M1). If the data best fits model M0, the intensity is considered constant over the entire interval. If, however, model M1 is preferred over model M0, the procedure is repeated within the two newly formed intervals. The Bayesian information criterion (BIC) 2 approximation to the Bayes factor was used for model selection. An interval comprised a minimum of five reports and/or seven calendar days. The reporting rates in each interval between two change points was calculated by the average number of reports per month in the respective interval.

Cumulative patient exposure. Data on the estimated cumulative number of patient exposed to natalizumab and rituximab over time was extracted from publicly available sources, including literature articles, FDA safety communications, and from the websites of the marketing authorization holders, and was used to put the trends in reporting into context of the trends in patient exposure.

Lag time between reaction and reporting. The lag time between onset of PML and initial reporting was calculated using the reaction date and receive date. When only the month and year was provided for the reaction date, the fifteenth of the month was used by default, unless this was before the receive date. Cases with lacking information on the reaction date, or only referring to the year of onset, were excluded from this analysis.

Completeness of case reports. For each case, we assessed the completeness of both the initial and the “master” report. The completeness was assessed along the availability of data for the following eight predefined variables: age and gender of the patient; therapy start date; therapy end date and/or duration of therapy; indication of therapy; previous drug history; start date of PML reaction; outcome of PML reaction; and diagnostic test results. Regarding the latter, in line with a recently proposed case definition for PML [30], we assessed the availability of results for any of the following diagnostic tests: brain MRI characteristic of PML; positive test result for JC virus DNA in cerebrospinal fluid; or evidence from an autopsy.

We assessed the trends in data completeness by comparing the first 50 cases (both the initial and master report) to the last 50 cases (both the initial and master report). For this analysis 36 Chapter 2.1

we included only cases reported before 28 July 2012 (i.e. a half year before the DLP of the data extraction) to allow for sufficient time for collection of any follow-up information.

Statistical methods. Data were presented as means, medians, or proportions, as appropriate. Normality was tested using Shapiro-Wilk’s test. Independent sample t-tests were used to compare normally distributed variables, and nonparametric testing was performed using the Mann-Whitney test. Categorical data were tested using Pearson’s chi-square tests.

Acknowledgements

We wish to thank Dr. Svetlana Belitser (Utrecht University, the Netherlands) for her expert advice and assistance with statistical analysis. Reporting patterns for drug-induced PML 37

Literature

1. Tan, C. S. & Koralnik, I. J. Progressive multifocal ucm051761.htm [Last accessed 30 October leukoencephalopathy and other disorders caused 2014]. by JC virus: clinical features and pathogenesis. 11. Van Assche, G. et al. Progressive multifocal Lancet Neurol. 9, 425-437 (2010). leukoencephalopathy after natalizumab therapy 2. Brooks, B. R. & Walker, D. L. Progressive multifo- for Crohn’s disease. N. Engl. J. Med. 353, 362-368 cal leukoencephalopathy. Neurol. Clin. 2, 299-313 (2005). (1984). 12. FDA news release. FDA Approves Resumed 3. Astrom, K. E., Mancall, E. L. & Richardson, E. P. Marketing of Tysabri Under a Special Distribu- Progressive multifocal leuko-encephalopathy; a tion Program. 5 June 2006. http://www.fda. hitherto unrecognized complication of chronic gov/NewsEvents/Newsroom/PressAnnounce- lymphatic leukaemia and Hodgkin’s disease. ments/2006/ucm108662.htm. [Last accessed 30 2 Brain 81, 93-111 (1958). October 2014]. 4. Molloy, E. S. & Calabrese, L. H. Progressive mul- 13. European Public Assessment Report. Mabthera. tifocal leukoencephalopathy: A national estimate http://www.ema.europa.eu/docs/en_GB/docu- of frequency in systemic lupus erythematosus ment_library/EPAR_-_Scientific_Discussion/ and other rheumatic diseases. Arthritis Rheum. human/000603/WC500044690.pdf. [Last ac- 60, 3761-3765 (2009). cessed 30 October 2014]. 5. Engsig, F. N. et al. Incidence, Clinical Presenta- 14. Vinhas de Souza, M. et al. Drug-induced PML: tion, and Outcome of Progressive Multifocal a global agenda for a global challenge. Clin. Leukoencephalopathy in HIV-Infected Patients Pharmacol. Ther. 91, 747-750 (2012). during the Highly Active Antiretroviral Therapy 15. Piccinni, C. et al. Stronger association of drug- Era: A Nationwide Cohort Study. J. Infect. Dis. induced progressive multifocal leukoencepha- 199, 77-83 (2009). lopathy (PML) with biological immunomodulat- 6. Clifford, D. B. et al. Natalizumab-associated ing agents. Eur. J. Clin. Pharmacol. 66, 199-206 progressive multifocal leukoencephalopathy in (2010). patients with multiple sclerosis: lessons from 28 16. Schmedt, N., Andersohn, F. & Garbe, E. Signals cases. Lancet Neurol. 9, 438-446 (2010). of progressive multifocal leukoencephalopathy 7. Carson, K. R. et al. Progressive multifocal for immunosuppressants: a disproportional- leukoencephalopathy after rituximab therapy ity analysis of spontaneous reports within the in HIV-negative patients: a report of 57 cases US Adverse Event Reporting System (AERS). from the Research on Adverse Drug Events and Pharmacoepidemiol. Drug Saf. 21, 1216-1220 Reports project. Blood 113, 4834-4840 (2009). (2012). 8. Langer-Gould, A., Atlas, S. W., Green, A. J., 17. Allison, M. PML problems loom for Rituxan. Bollen, A. W. & Pelletier, D. Progressive multifo- Nat. Biotechnol. 28, 105-106 (2010). cal leukoencephalopathy in a patient treated 18. Rituxan label update. http://www.ac- with natalizumab. N. Engl. J. Med. 353, 375-381 cessdata.fda.gov/drugsatfda_docs/ (2005). label/2007/103705s5247Lbl.pdf. [Last accessed 9. Kleinschmidt-DeMasters, B. K. & Tyler, K. L. 30 October 2014]. Progressive multifocal leukoencephalopathy 19. Community register of medicinal products for complicating treatment with natalizumab and human use. Mabthera http://ec.europa.eu/health/ interferon beta-1a for multiple sclerosis. N. Engl. documents/community-register/html/h067.htm. J. Med. 353, 369-374 (2005). [Last accessed 30 October 2014]. 10. Public Health Advisory - Suspended Marketing of 20. Trontell, A. Expecting the Unexpected — Drug Tysabri (natalizumab). 28 February 2005. http:// Safety, Pharmacovigilance, and the Prepared www.fda.gov/Drugs/DrugSafety/Postmarket- Mind. N. Engl. J. Med. 351, 1385-1387 (2004). DrugSafetyInformationforPatientsandProviders/ 38 Chapter 2.1

21. Woosley, R. L. Discovering adverse reactions: 32. Roden, D. M. Drug-Induced Prolongation of why does it take so long? Clin. Pharmacol. Ther. the QT Interval. N. Engl. J. Med. 350, 1013-1022 76, 287-289 (2004). (2004). 22. Kimby E. Tolerability and safety of ritux- 33. Edwards, B. J. et al. Pharmacovigilance and imab (MabThera®). Cancer Treat Rev 2005 reporting oversight in US FDA fast-track process: 10;31(6):456-473. bisphosphonates and osteonecrosis of the jaw. 23. Ando, K. et al. Diffuse large B-cell lymphoma Lancet Oncol. 9, 1166-1172 (2008). showing CNS invasion by CD30-positive multi- 34. Bilezikian, J. P. Osteonecrosis of the Jaw — Do nuclear giant cells mimicking the clinical features Bisphosphonates Pose a Risk? N. Engl. J. Med. of progressive multifocal leukoencephalopathy. 355, 2278-2281 (2006). Rinsho Ketsueki 41, 507-512 (2000). 35. Khosla, S. et al. Bisphosphonate-Associated 24. Sikkema, T., Schuiling, W. J. & Hoogendoorn, Osteonecrosis of the Jaw: Report of a Task Force M. Progressive multifocal leukoencephalopathy of the American Society for Bone and Mineral during treatment with rituximab and CHOP Research. J. Bone Miner. Res. 22, 1479-1491 chemotherapy in a patient with a diffuse large (2007). B-cell lymphoma. BMJ Case Reports 2013 (2013). 36. Bloomgren, G. et al. Risk of natalizumab-associ- 25. Goldberg, S. L. et al. Unusual viral infections ated progressive multifocal leukoencephalopathy. (progressive multifocal leukoencephalopathy N. Engl. J. Med. 366, 1870-1880 (2012). and cytomegalovirus disease) after high-dose 37. Segec, et al. Strategy in regulatory decision mak- chemotherapy with autologous blood stem cell ing for management of PML. Clin. Pharmacol. rescue and peritransplantation rituximab. Blood Ther. (advanced online publication DOI: 10.1002/ 99, 1486-1488 (2002). cpt.199). 26. Matteucci, P. et al. Leukoencephalopathy and 38. Chalkley, J. J. & Berger, J. R. Progressive multifocal papovavirus infection after treatment with leukoencephalopathy in multiple sclerosis. Curr. chemotherapy and anti-CD20 monoclonal Neurol. Neurosci. Rep. 13, 408 (2013). antibody. Blood 100, 1104-1105 (2002). 39. European Medicines Agency. Inspections and 27. Steurer, M. et al. Progressive multifocal Human Medicines Pharmacovigilance. 2013 An- leukoencephalopathy after allogeneic stem nual Report on EudraVigilance for the European cell transplantation and posttransplantation Parliament, the Council and the Commission. rituximab. Transplantation 76, 435-436 (2003). Reporting period: 1 January to 31 December 28. Pariente, A., Gregoire, F., Fourrier-Reglat, A., 2013. 4 April 2014. Haramburu, F. & Moore, N. Impact of safety 40. International Conference on Harmonisation alerts on measures of disproportionality in spon- of Technical Requirements for Registration of taneous reporting databases: the notoriety bias. Pharmaceuticals for Human Use. ICH Expert Drug Saf. 30, 891-898 (2007). Working Group. Implementation Guide for 29. Faasse K, Gamble, G. Cundy, T. & Petrie, K.J. Electronic Transmission of Individual Case Impact of television coverage on the number Safety Reports (ICSRs). E2B(R3) Data Elements and type of symptoms reported during a health and Message Specification. Version 5.01, 12 April scare: a retrospective pre–post observational 2013. study. BMJ Open 2, 1-7(2013). 41. Committee for Medicinal Products for Human 30. Mentzer, D. et al. Case definition for progres- Use (CHMP. Guideline on Detection and sive multifocal leukoencephalopathy following Management of Duplicate Individual Cases and treatment with monoclonal antibodies. J. Neurol. Individual Case Safety Reports (ICSRs). 21 June Neurosurg. Psychiatry. 83, 927-933 (2012). 2012. 31. Selzer, A., H. W. Quinidine Syncope. Paroxysmal 42. Yang, T. Y. & Kuo, L. Bayesian Binary Segmen- Ventricular Fibrillation Occurring during Treat- tation Procedure for a Poisson Process with ment of Chronic Atrial Arrhythmias. Circulation Multiple Changepoints. J. Comp. Graph. Stat.10, 30, 17-26 (1964). 772-785 (2001). Reporting patterns for drug-induced PML 39

43. Yaldizli Ö, Putzki N. Natalizumab in the Treat- www.tapp.com.au/members/Tysabri_Safety_Up- ment of Multiple Sclerosis. Ther. Adv. Neurol. date_160812.pdf [Last accessed 28 March 2015]. Disord. 2(2), 115-128 (2009). 49. Kimby E. Tolerability and safety of rituximab 44. Food and Drug Administration. FDA Drug Safety (MabThera®). Cancer Treat Rev; 31(6), 456-73 Communication: FDA Drug Safety Communica- (2005). tion Information for Healthcare Professionals 50. Cohes S.B., et al. Rituximab for rheumatoid - Natalizumab Injection for Intravenous Use arthritis refractory to anti tumor necrosis factor (marketed as Tysabri.) August 2008. Available therapy: Results of a multicenter, randomized, from: http://www.fda.gov/Drugs/DrugSafety/ double-blind, placebo-controlled, phase III trial PostmarketDrugSafetyInformationforPatient- evaluating primary efficacy and safety at twenty- sandProviders/ucm126592.htm [Last accessed four weeks. Arthritis Rheum. 54(9), 2793-806 28 March 2015]. (2006). 45. Food and Drug Administration. FDA Drug 51. Genentech, Biogen Idec. New Release. Genentech 2 Safety Communication: FDA Drug Safety and Biogen Idec announce positive results from Communication: Risk of Progressive Multifo- a phase II trial of Rituxan in relapsing-remitting cal Leukoencephalopathy (PML) with the use multiple sclerosis. 28 August 2006. Available of Tysabri (natalizumab). 5 February 2010. from: http://www.roche.com/iru20060828.pdf Available from: http://www.fda.gov/Drugs/ [Last accessed 28 March 2015]. DrugSafety/PostmarketDrugSafetyInforma- 52. Roche. Media Release. US FDA approves Rituxan/ tionforPatientsandProviders/ucm199872.htm MabThera for the most common type of adult [Last accessed 28 March 2015]. leukemia. 19 February 2010. Available from: 46. Food and Drug Administration. FDA Drug Safety http://www.roche.com/media/store/releases/ Communication: Safety update on Progressive med-cor-2010-02-19.htm [Last accessed 28 Multifocal Leukoencephalopathy (PML) as- March 2015]. sociated with Tysabri (natalizumab). 22 April 53. Roche. Media Release. FDA approves Rituxan/ 2011. Available from: http://www.fda.gov/Drugs/ MabThera for first-line maintenance use in DrugSafety/ucm252045.htm [Last accessed 28 follicular lymphoma. 11 January 2011. Avail- March 2015]. able from: http://www.roche.com/media/store/ 47. Food and Drug Administration. FDA Drug releases/med-cor-2011-01-31.htm [Last accessed Safety Communication: New risk factor for 28 March 2015]. Progressive Multifocal Leukoencephalopathy 54. Roche. Media Release. Roche reports positive (PML) associated with Tysabri (natalizumab). studies of MabThera given by subcutaneous 20 January 2012. Available from: http://www.fda. injection. 8 December 2012. Available from: gov/Drugs/DrugSafety/ucm288186.htm [Last http://www.roche.com/media/store/releases/ accessed 28 March 2015]. med-cor-2012-12-08b.htm [Last accessed 28 48. Biogen Idec. Tysabri (natalizumab) Safety March 2015]. Update. 17 August 2012. Available from: http:// 40 Chapter 2.1

Cumulative number of PML reports Cumulative patient exposure

s t r ) o

350 s t p n e r e

100,000 i t L a M 300 p

f P

d r e e c

80,000 b u 250 d m n u i - n ( b

a e r

m 200 60,000 u u s z i o l p a x t e a

150 t n

n d

40,000 e i e t t c a e p

p 100 d s e u t s a

f 20,000 m o i

50 t r s e E b m u

N 0 0

Supplementary Figure 1-A. Cumulative number of suspected PML reports (left y-axis), and estimated cumulative patient exposure to natalizumab (right y-axis) over time.

Cumulative number of PML reports Cumulative patient exposure 3,000,000 s t r ) o s p t e n

r 250 e

i t L 2,500,000 a M p

P f

o d

r e e c 200 b u 2,000,000 d m n u i - n ( b

a e r m

i 150 u x

1,500,000 s u o t i p r

x d e

e t t n c 100 e e 1,000,000 i t p a s p u

s d

f e t o

a r 50 500,000 e m i b t s m E u N 0 0

Supplementary Figure 1-B. Cumulative number of suspected PML reports (left y-axis), and estimated cumulative patient exposure to rituximab (right y-axis) over time. *Data on estimated patient exposure for natalizumab and rituximab was extracted from publicly available sources [43-54].

Analytical interference by 2.2 monoclonal antibodies in frequently used immunoassays: an in vitro study

Vermeer NS, De Bruin ML, Egberts ACG, van Solinge WW, Lentjes EGWM

Submitted for publication

Abstract

Background and objective Drug interference in clinical laboratory testing may lead to misdiagnosis, faulty clinical decision- making, and could ultimately contribute to adverse patient outcomes. Immunoassays are particularly susceptible to interference by endogenously occurring heterophile antibodies, yet little is known about monoclonal antibody therapy as a potential exogenous source of interference. The aim of this study was to evaluate the potential for direct interference by monoclonal antibodies within immunoassays. Methods We explored the potential for analytical interference by monoclonal antibodies within an in vitro study. Patient serum samples were spiked with peak serum concentrations of monoclonal antibodies (rituximab, belimumab, tocilizumab, infliximab) and evaluated for interference within a number of frequently used immunoassays: free thyroxine, thyroid-stimulating hormone, cortisol, estradiol, and luteinizing hormone. Interference was tested by comparing the differences in observed analyte concentrations between spiked test samples and control samples, and was considered clinically relevant when exceeding predefined thresholds. Results We found statistically significant differences in mean measured analyte concentrations, as compared to the control samples, for 5 out of the 20 test samples that had been spiked with monoclonal antibodies. Overall, the observed difference in mean measured analyte concentrations between test and control samples was, however, modest. The largest, and clinically relevant, effect was observed for infliximab-spiked samples in the estradiol assay, which showed on average 19% lower estradiol levels as compared to the control sample. Conclusion This study shows that the susceptibility of immunoassays to monoclonal antibodies as direct exogenous source of interference is low, but in specific cases can lead to incorrect diagnoses and false positive drug safety signals. Therefore, when patients are on monoclonal antibody therapy unexpected results of laboratory immunoassay tests should be carefully evaluated. 44 Chapter 2.2

Introduction

Immunoassays are frequently used laboratory tests for quantifying endogenous and exogenous substances in patient plasma samples, and are important in diagnosis and in monitoring of medical conditions and the effects of their treatment [1,2]. The application of immunoassays has increased vastly over the past decades owing to its many advantages, including high sensitivity and, often, the absence of the need for prior extraction steps. Due to the lack of these prior extraction steps, immunoassay measurements are, however, vulnerable to interfering compounds that may potentially alter the correct value of the test result [3,4]. Im- munoassay interference could potentially – when not adequately recognized and anticipated upon – lead to an incorrect or missed diagnosis and faulty clinical decision-making, which may ultimately lead to adverse patient outcomes [5-7]. A case series, for example, described the incorrect diagnosing of malignant disease in twelve women, as a result of the occurrence of interference in an immunoassay for human chorionic gonadotropin, with a number of women consequently receiving needless chemotherapy and surgery [7]. Both endogenous and exogenous compounds may cause immunoassay interference. Among these, heterophile antibody interferences represent a commonly encountered form of interference in immunoassays [8,9]. Heterophile antibodies are endogenous polyspecific antibodies that have weak affinity for an array of ill-defined antigens, possibly including components of immunoassays. The major exogenous source of interference, on the other hand, comprises drugs or metabolites thereof, which may be present in patient samples. Much knowledge has been gained on the influence of frequently used small-molecule drugs on various immunoassays, which has been summarized in the standard reference by Young [10,11]. Relatively little is, however, known about the potential influence of monoclonal antibodies on immunoassays [12], despite their increasing deployment in clinical practice [13] as well as theoretical capability of interference. Monoclonal antibodies (mAbs) could through various, either direct or indirect, mechanisms cause immunoassay interference. Treatment with murine or partly murine (“chimeric”) mAbs could, for example, induce the formation of heterophile antibodies [14], and thereby indirectly lead to immunoassay interference as described in one report, where administration of muromonab-CD3 was linked to falsely elevated plasma parathyroid hormone levels mimicking tertiary hyperparathyroidism [6]. Direct mechanisms, on the other hand, may involve cross- reactivity of the mAb with any of the immunoassay components or the analyte; steric hindrance preventing the interaction of the ligand with the analyte; or altering of the characteristics of the matrix, which may potentially affect the reaction velocity or the fluorescence spectrophotometry. Recently, direct interference by daratumumab, a novel anti-CD 38 mAb for the treatment of multiple myeloma, was observed in an immunoassay used for blood group typing, causing difficulties in transfusion medicine decisions [15]. In view of this finding, we aimed to further evaluate the potential for direct interference by monoclonal antibodies within immunoassays in this study. Immunoassay interference by mAbs 45

Methods

Experimental design We evaluated the potential for analytical interference by monoclonal antibodies within an in vitro study. Aliquots of pooled patient serum samples were spiked with monoclonal antibodies, and evaluated for interference within a number of frequently used immunoassays, by comparing the assay results for the spiked samples with those of the control samples. Because monoclonal antibodies are in vivo degraded into amino acids, and do not form active metabolites that may act as potential interferents, the current in vitro approach was considered an adequate model for the approximation of the in vivo situation. 2 Materials

Monoclonal antibodies Four monoclonal antibodies were evaluated for their interference potential in this study: rituximab, belimumab, tocilizumab, and infliximab. These antibodies differ with respect to their pharmacological properties, indication, and antibody type (chimeric [25%/75% murine/ human protein], humanized [5%/95% murine/human], human [100% human]), but have in common that they all reach relatively high peak serum concentrations in vivo (ranging from 0.18 to 0.55 mg/ml) and may therefore be more likely to cause immunoassay interference. More details on the clinical, physiochemical and pharmaceutical characteristics of the four test drugs is provided in Table 1.

Immunoassays We selected five immunoassays that are used for measurement of hormones, or precursors thereof: free thyroxine (FT4), thyroid-stimulating hormone (TSH), cortisol, estradiol, and luteinizing hormone (LH). The choice of the immunoassays was based on their frequent deployment in clinical practice, and their recognized potential for interference by exogenous substances.

Serum base pool A serum base pool was created from whole blood of approximately forty donors, as per standardized procedures (mixing, centrifugation, aliquoting), with the following target analyte concentrations: cortisol (0.3 – 0.5 µmol/L); FT4 (13 – 20 pmol/L); TSH (1 – 5 mU/L); LH (8-20 U/L); and estradiol (200 – 500 pmol/L). 46 Chapter 2.2 1 1 10 mg/ml infliximab 10 mg/ml 20 mg/ml tocilizumab20 mg/ml 15 mM disodium phosphate sodium dihydrogen dodecahydrate/ dihydrate phosphate 80 polysorbate 0.5 mg/ml sucrose 50 mg/ml injection for Water 10 mg/ml rituximab 10 mg/ml sodium9.0 mg/ml chloride dehydrate sodium7.35 mg/ml citrate 80 polysorbate 0.7 mg/ml injection for Water 80 mg/ml belimumab 80 mg/ml 0.16 mg/ml citric acid citric 0.16 mg/ml 80 polysorbate 0.4 mg/ml sodium2.7 mg/ml citrate sucrose 80 mg/ml injection for Water sodium monobasic 0.22 mg/ml 0.61mg/ml monohydrate, phosphate dihydrate sodiumdibasic phosphate 80 polysorbate 0.05 mg/ml sucrose 50 mg/ml injection for Water Formulation, vial composition Formulation, 149,100 148,000 144,544 147,000 MW (Dalton) Chimeric Humanized Chimeric Human Physiochemical and pharmaceutical aspects pharmaceutical and Physiochemical Antibody typeAntibody = Maximum serum concentration; MW= molecular weight serum concentration; = Maximum max max 277 μg/ml 183 μg/ml 550 μg/ml 8) (after cycle 404 μg/ml 2) (after cycle 313 μg/ml Mean C Mean up up 2 3 to 5 mg/ 3 to kg, every 2-8 weeks 8 mg/kg (max (max 8 mg/kg every800mg), 4 weeks 375 mg/m to 4-8 cycles to 2x 1.000mg in 2x 1.000mg 14 days 10 mg/kg, 10 mg/kg, every 2-4 weeks Rheumatoid arthritis; Crohn’s arthritis; Crohn’s Rheumatoid colitis; disease; ulcerative spondylitis; ankylosing arthritis psoriatic psoriasis; Rheumatoid arthritis Rheumatoid Non-Hodgkin’s lymphoma; lymphoma; Non-Hodgkin’s lymphocytic chronic leukaemia; granulomatosis and polyangiitis with polyangiitis microscopic Rheumatoid arthritis Rheumatoid Systemic lupus erythematosus lupus Systemic Therapeutic indications and and posology indications Therapeutic TNF-alpha Interleukine-6 Interleukine-6 receptor (IL-6) CD-20 on CD-20 on human B-lymphocytes B lymphocyte B lymphocyte stimulator (BLyS) protein Clinical aspects Target Clinical, physicochemical and pharmaceutical aspects of monoclonal antibodies (data from product information and public assessment reports [16-23]). reports assessment public and information product from (data antibodies monoclonal of aspects pharmaceutical and physicochemical Clinical, Composition after reconstituted with water for injection, according to product information. C information. product to according injection, for water with reconstituted after Composition Infliximab Remicade Tocilizumab RoActemra Rituximab Mabthera Table 1. 1. Table 1 Belimumab Benlysta Immunoassay interference by mAbs 47

Experimental procedures

Sample preparation All samples were prepared from one serum pool. The test samples were spiked with literature- derived peak serum concentrations for the different mAbs [16-23], and a corresponding volume of water was added to the control samples (see Table 2).

Table 2. Preparation of test and control serum samples. Pipetting scheme (total volume = 500µl)

Vial content Mean Cmax, Test Cmax µl mAb µl Water µl serum (mg/ml) literature (mg/L) (mg/L) sample 2 Control - - - - 30 470 Test Rituximab 10 550 600 30 - 470 Belimumab 80 313 320 2* 28 470 Tocilizumab 20 183 200 5* 25 470 Infliximab 10 277 300 15* 15 470 *Belimumab, tocilizumab and infliximab were added after one prior dilution step with water, respectively 1:14; 1:5 and 1:1. Sample measurement

Sample measurement All samples were analysed in single within one analytical run, and measured using the Beckman DXI (Beckman All samples were analysed in single within one analytical run, and measured using the Beck- Coulter Inc., Brea, California) immunoassay system. man DXI (Beckman Coulter Inc., Brea, California) immunoassay system. The required numberThe of required samples number for the of different samples forimmunoassays the different to immunoassays ensure sufficient to ensure sufficient power to detect clinical power to detect clinical significance interference in the assay is was calculated using the significance interference in the assay is was calculated using the following equation [24], which takes into account following equation [24], which takes into account the significant level (α) and power (β) for hypothesis testing; the repeatabilitythe significant (or: levelwithin-run (α) and precision power ()β )of for the hypothesis measurement testing; procedure the repeatability (or: within-run precision) of the

(s); and the amount of interferencemeasurement that procedure is considered ( ); and to the be amountclinically of interferencerelevant (D maxthat): is considered to be clinically relevant (Dmax):

� � � �� The Westgard reference list [25] for biological variance was used to determine the Dmax for the individual The Westgard reference list [25] for biological variance was used to determine the Dmax for the individual immunoassays.immunoassays. The repeatability The repeatability (s) of ( the) of themeasurement measurement procedures procedures was was obtained from the internal validation obtained from the internal validation records, which have been performed according to records, which have been performeds according to standardized procedures [26]. We tested at two-sided 95% standardized procedures [26]. We tested at two-sided 95% confidence level, and 95% power. The required number of confidencesamples for level, each and of the95% immunoassays power. The required is described number in of Table 3 samples. for each of the immunoassays is described in Table 3.

Table 2. Preparation of test and control serum samples.

Pipetting scheme (total volume = 500µl)

Vial content Mean Cmax, Test Cmax µl mAb µl Water µl serum (mg/ml) literature (mg/L) (mg/L) sample Control - - - - 30 470 Test Rituximab 10 550 600 30 - 470 Belimumab 80 313 320 2* 28 470 Tocilizumab 20 183 200 5* 25 470 Infliximab 10 277 300 15* 15 470

*Belimumab, tocilizumab and infliximab were added after one prior dilution step with water, respectively 1:14; 1:5 and 1:1.

Table 3. Number of replications for different immunoassays.

Estimated analyte Biological variability Within-run precision Number of

1 2 3 concentration B% Dmax S% S replications Cortisol 0.4 µmol/L 12.8 0.0512 µmol/L 3.5 0.014 µmol/L 2 FT4 15 pmol/L 3.3 0.495 pmol/L 3 0.45 pmol/L 21

TSH 5 mU/L 7.8 0.39 mU/L 5 0.25 mU/L 11 Estradiol 200 pmol/L 8.3 16.6 pmol/L 8 16 pmol/L 24 LH 10 IU/L 8.9 0.89 IU/L 3.5 0.35 IU/L 4

1Mean of target analyte concentration within serum base pool (see above). 2B%: desirable specification for inaccuracy based on biological variation, obtained from [25]. 3Within-run precision, obtained from internal validation records.

48 48 Chapter 2.2 Immunoassay interference by mAbs 49

Table 3. Number of replications for different immunoassays. Estimated analyte Biological variability Within-run precision Number of 1 concentration 2 3 replications B% Dmax S% S Cortisol 0.4 µmol/L 12.8 0.0512 µmol/L 3.5 0.014 µmol/L 2 FT4 15 pmol/L 3.3 0.495 pmol/L 3 0.45 pmol/L 21 TSH 5 mU/L 7.8 0.39 mU/L 5 0.25 mU/L 11 Estradiol 200 pmol/L 8.3 16.6 pmol/L 8 16 pmol/L 24 LH 10 IU/L 8.9 0.89 IU/L 3.5 0.35 IU/L 4 1Mean of target analyte concentration within serum base pool (see above). 2B%: desirable specification for inaccuracy based on biological variation, obtained from [25]. 3Within-run precision, obtained from internal validation records. 2 Data analysis Means and standard deviations of the observed analyte concentrations were calculated for both the test and control samples, and independent sample Student’s T-tests were used to compare means at a two-sided alpha level of 0.05. The amount of interference was calculated as the mean difference in observed analyte concentrations between control sample and test samples, and was considered to be clinically relevant when the point estimate was signifi-

cantly different from the Dmax (i.e. the 95% confidence interval of the mean excluded Dmax).

Results

We found statistically significant differences in mean measured analyte concentrations, as compared to the control samples, for 5 out of the 20 test (i.e. monoclonal antibody-spiked) samples. As shown in Table 4, for the infliximab-spiked samples, significantly lower mean serum concentrations of estradiol (181.48 vs. 223.30 pmol/L) and cortisol (0.226 vs. 0.255 µmol/L), but higher serum concentrations of FT4 (17.86 vs. 16.80 pmol/L) were observed, as compared to the control samples. In addition, for the tocilizumab-spiked samples, significantly decreased mean cortisol levels (0.232 vs. 0.255 µmol/L), but increased mean TSH levels (1.88 vs. 1.78 mU/L) were observed. For the samples that had been spiked with either rituximab or belimumab, we found no statistically significant differences in mean measured analyte concentrations, as compared to the control samples. None of the tested mAbs interfered with the LH tests. Overall, the observed difference in mean measured analyte concentrations between test and control samples was only modest (see Table 5). However, for two spiked samples (both with infliximab) the difference with the control sample exceeded the pre-defined levels for

clinical relevance (Dmax). The mean measured FT4 concentration in the samples that were spiked with infliximab was 1.06 pmol/L (95% CI: 0.34 – 1.78 pmol/L), or 6%, higher than in

the control sample, and thereby exceeded the Dmax of 0.492 pmol/L. For the estradiol assay, 48 Chapter 2.2 Immunoassay interference by mAbs 49 -4% -3% +6.2 <1% % - - SD 0.447 0.548 0.483 Relative LH LH - - 11.0 11.8 11.4 12.0 11.3 Mean (IU/L) 95% CI 95% -0.50 – 0.70 -0.91 – 1.04 5 5 5 4 n 10 0.7 0.1 0.5 -0.3 Mean difference Mean (IU/L) SD -5% -7% -4% 26.46 29.25 41.53 35.61 24.70 -19% % Relative Estradiol Mean 207.56 233.16 212.96 223.30 181.48*

(pmol/L) 2 95% CI 95% Estradiol -37.1 – 5.6 -35.6 – 14.9 -18.8 – 38.6 -62.2 – -21.5 n 21 25 25 24 10 9.86 Mean difference Mean -10.34 -15.74 41.82* - (pmol/L) SD 0.116 0.083 0.144 0.170 0.103 <1% +6% +2% +2% Relative TSH 1.81 1.81 1.79 1.78 Mean 1.88* (mU/L) TSH n 12 12 12 11 10 95% CI 95% 0.02 – 0.18 -0.11 – 0.14 -0.07 – 1.27 -0.09 – 1.42 0.01 0.10 0.03 0.03 SD 0.65 0.77 0.61 0.65 1.32 Mean difference Mean (mU/L) <1% <1% +6% +3% FT4 16.86 17.23 16.86 16.80 Mean Relative 17.86* (pmol/L) n 21 22 22 21 10 FT4 95% CI 95% 0.09 – 2.02 -0.91 – 1.02 -0.91 – 1.04 -0.54 – 1.39 SD 0.011 0.008 0.019 0.019 0.022 0.06 0.06 1.06 0.43 Mean difference Mean (pmol/L) -9% <1% +5% 0.256 0.268 0.255 Cortisol -11% 0.226* 0.232* Mean (µmol/L) Relative 5 5 5 4 n 10 The 95%ofconsidered effect CI excludedfor estimate thetothe interference interference point Dmax; clinicallyamount predefinedbe rel - of the 95% CI 95% Cortisol -0.02 – 0.04 -0.02 – 0.03 -0.04 - -0.00 - -0.04 -0.05 – -0.01 Mean difference Mean 0.013 0.001 -0.023 -0.029 (µmol/L) Infliximab Tocilizumab Belimumab Rituximab n, number of replications; SD, standard deviation of mean; FT4, free thyroxine; TSH, thyroid-stimulating hormone; LH , luteinizing hormone. LH , luteinizing hormone; thyroid-stimulating TSH, FT4, free thyroxine; mean; of deviation standard SD, replications; of n, number sample (p<0.05). control from significantly differed *The measuredanalyte concentration Test Control Rituximab Tocilizumab Infliximab Belimumab Mean observed analyte concentrations for test and control samples. control and test for observed concentrations 4. Mean analyte Table samples. control and test between observed observed in concentrations analyte 5. Mean difference Table *Clinically relevant effect. LH (0.89 IU/L). and pmol/L); (16.6 (0.39 mU/L); estradiol TSH pmol/L); FT4 (0.492 cortisol (0.0512 µmol/L); immunoassays: the different for evant 50 Chapter 2.2

the infliximab-spiked samples resulted in a 41.82 pmol/L (95% CI: 62.2 – 21.5 pmol/L), or 19%, lower result as compared to the control samples, thereby exceeding the biological variability of 16.6 pmol/L. Only in the latter test, the 95% confidence interval excluded the

predefined Dmax, and therefore the interference between infliximab and the estradiol assay was regarded to be clinically relevant.

Discussion

This study shows that the susceptibility of the tested immunoassays to the tested monoclonal antibodies as direct exogenous source of interference is low. Overall, the observed interference effect of monoclonal antibodies in peak plasma concentrations was modest and mostly negligible in frequently used immunoassays. The largest effect was observed for infliximab- spiked samples in the estradiol assay, which showed on average 19% lower estradiol levels as compared to the control sample. Serum estradiol levels are an important laboratory marker in the diagnoses and monitoring of various clinical conditions. Estradiol measurements are, for example, used in the assessment of menstrual and reproductive function in women [27] and of gynecomastia in men [28], as well as in the monitoring of the efficacy of treatment with aromatase inhibitors in breast cancer [29]. Falsely elevated or decreased estradiol levels may therefore result in incorrect diagnosing and/or monitoring, including in the management of breast cancer [30]. Moreover, with the increasing use of automated laboratory data in drug safety research [31], erroneous result may result in false positive safety signals for drugs. The here observed interference effect of infliximab on the estradiol assay was, however, rather modest, and negligible when compared to, for example, the previously reported 26-fold increase in estradiol levels due to interference by efavirenz [32]. The potential effects of infliximab on the immunoassay are therefore not likely to be of clinical significance. It is important to note that, apart from the direct mechanisms that were evaluated in this study, indirect mechanisms may also be involved in analytical interference. Murine, or chimeric monoclonal antibodies like infliximab may for example induce the formation of heterophile antibodies, which could consequently cross-react with any of the components in the immunoassay and thereby result in erroneous results [6]. Because of the in vitro design of the current study we were not able to evaluate this potential. Furthermore, monoclonal antibodies may affect laboratory tests other than immunoassays. Interference with tests for activated partial thromboplastin time (aPTT) has, for example, been reported in patients with treated with certolizumab pegol, in whom an increase in aPTT was measured in the absence of any coagulation abnormalities [33]. In addition, unexpected interference with blood group serologic testing was recently observed in patients treated with a novel Immunoassay interference by mAbs 51 anti-CD38 mAb, which hampered the identification of irregular blood group antibodies in patients requiring blood transfusion [15]. Despite the low potential for interference observed in this study, health care professionals should therefore remain aware of the general risk of laboratory test interference by drugs, including monoclonal antibodies, in particular in view of the potential serious consequences. Falsely elevated blood glucose results in diabetic patients who were using medicinal products containing non-glucose sugars, for example, resulted in the death of thirteen patients due to insulin overdose in 2009 [34]. To increase the awareness among healthcare professionals, it is crucial that any available information regarding drug effects on laboratory tests are adequately communicated. A previous study, however, showed that only 15% of the known effects on laboratory markers are described in the product information of 2 the drug, leaving much room for improvement [11]. In conclusion, this study shows that the susceptibility of immunoassays to monoclonal antibodies as direct exogenous source of interference is low, but in specific cases can lead to incorrect diagnoses and false positive drug safety signals. 52 Chapter 2.2

References

1. Gosling JP. A decade of development in immuno- parison with 99mTc-labelled white blood cells. assay methodology. Clin Chem. 1990;36:1408-27. Eur J Nucl Med Mol Imaging. 2011;38(5):899- 2. Bock JL. The new era of automated immunoassay. 910. Am J Clin Pathol. 2000;113(5):628-46. 15. Oostendorp M, Lammerts van Bueren JJ, 3. Tate J, Ward G. Interferences in immunoassay. Doshi P, Khan I, Ahmadi T, Parren PWHI, et Clin Biochem Rev. 2004;25(2):105-20. al. When blood transfusion medicine becomes 4. Kroll MH, Elin RJ. Interference with clinical complicated due to interference by monoclonal laboratory analyses. Clin Chem. 1994;40:1996- antibody therapy. Transfusion. 2015;55:1555-62. 2005. 16. European medicines agency. mabthera product 5. Despres N, Grant AM. Antibody interference in information. march 26, 2015. available from: thyroid assays: A potential for clinical misinfor- http://www.ema.europa.eu/docs/en_GB/docu- mation. Clin Chem. 1998;44(3):440-54. ment_library/EPAR_-_Product_Information/ 6. Levin O, Morris LF, Wah DT, Butch AW, Yeh MW. human/000165/WC500025821.pdf [Last ac- Falsely elevated plasma parathyroid hormone cessed: May 25, 2015]. level mimicking tertiary hyperparathyroidism. 17. European medicines agency. mabthera public Endocr Pract. 2011;17(2):e1-4. assessment report. septemer 29, 2006. available 7. Rotmensch S, Cole LA. False diagnosis and from: http://www.ema.europa.eu/docs/en_GB/ needless therapy of presumed malignant document_library/EPAR_-_Scientific_Discus- disease in women with false-positive human sion/human/000165/WC500025817.pdf [Last chorionic gonadotropin concentrations. Lancet. accessed: May 25, 2015]. 2000; 355:712-5. 18. European medicines agency. benlysta product 8. Marks V. False-positive immunoassay results: A information. november 7, 2014. available from: multicenter survey of erroneous immunoassay http://www.ema.europa.eu/docs/en_GB/docu- results from assays of 74 analytes in 10 donors ment_library/EPAR_-_Product_Information/ from 66 laboratories in seven countries. Clinical human/002015/WC500110150.pdf [Last ac- Chemistry. 2002;48(11):2008-16. cessed: May 25, 2015]. 9. Emerson JF, Lai KKY. Endogenous antibody 19. European medicines agency. benlysta public interferences in immunoassays. Lab Medicine. assessment report. august 9, 2011. available 2013;44(1):69-73. from: http://www.ema.europa.eu/docs/en_GB/ 10. Young DS. Effects of drugs on clinical laboratory document_library/EPAR_-_Scientific_Discus- tests. Fifth Edition ed. Washington, DC: AACC sion/human/000165/WC500025817.pdf [Last Press; 2000. accessed: May 25, 2015]. 11. Geerts AF, De Koning FH, Egberts TC, De Smet 20. European medicines agency. RoActemra PA, Van Solinge WW. Information comparison product information. march 26, 2015. available of the effects of drugs on laboratory tests in drug from: http://www.ema.europa.eu/docs/en_GB/ labels and young’s book. Clin Chem Lab Med. document_library/EPAR_-_Product_Informa- 2012;50(10):1765-8. tion/human/000955/WC500054890.pdf [Last 12. Ostrov BE, Amsterdam D. The interference of accessed: May 25, 2015]. monoclonal antibodies with laboratory diagnosis: 21. European medicines agency. RoActemra public Clinical and diagnostic implications. Immunol assessment report. february 18, 2009. available Invest. 2013;42(8):673-90. from: http://www.ema.europa.eu/docs/en_GB/ 13. Walsh G. Biopharmaceutical benchmarks 2014. document_library/EPAR_-_Public_assess- Nat Biotechnol. 2014;32(10):992-1000. ment_report/human/000955/WC500054888. 14. Richter WS, Ivancevic V, Meller J, Lang O, Le pdf [Last accessed: May 25, 2015]. Guludec D, Szilvazi I, et al. 99mTc-besilesomab 22. European medicines agency. remicade product (scintimun) in peripheral osteomyelitis: Com- information. september 29, 2014. available Immunoassay interference by mAbs 53

from: http://www.ema.europa.eu/docs/en_GB/ 28. Braunstein GD. Gynecomastia. N Engl J Med. document_library/EPAR_-_Product_Informa- 2007;357(12):1229-37. tion/human/000240/WC500050888.pdf [Last 29. Smith IE, Dowsett M. Aromatase inhibitors in accessed: May 25, 2015]. breast cancer. N Engl J Med. 2003; 348(24):2431- 23. European medicines agency. remicade public 42. assessment report. december 7, 2005. Available 30. Dowsett M, Folkerd E. Deficits in plasma oestra- from: http://www.ema.europa.eu/docs/en_GB/ diol measurement in studies and management of document_library/EPAR_-_Scientific_Discus- breast cancer. Breast Cancer Res. 2005;7(1):1-4. sion/human/000240/WC500050885.pdf [Last 31. ten Berg MJ, Huisman A, van den Bemt PM, accessed: May 25, 2015]. Schobben AF, Egberts AC, van Solinge WW. 24. Clinical and laboratory standards institute Linking laboratory and medication data: New (CLSI). interference testing in clinical chemistry; opportunities for pharmacoepidemiological approved guideline - second edition. CLSI docu- research. Clin Chem Lab Med. 2007;45(1):13-9. 2 ment EP7-A2 (ISBN 1-56238-584-4) USA, 2005. 32. Sinicco A, Raiteri R, Rossati A, Savarino A, Di 25. Westgard QC. desirable biological variation Perri G. Efavirenz interference in estradiol ELISA database specifications - quality requirements. assay. Clin Chem. 2000;46(5):734-5. available from: http://www.westgard.com/ 33. Puli SR, Benage DD. Retinal vein thrombosis biodatabase1.htm [Last accessed June 2, 2015]. after infliximab (remicade) treatment for crohn’s 26. Clinical and laboratory standards institute (CLSI). disease. Am J Gastroenterol. 2003;98(4):939-40. user verification of precision and estimation of 34. FDA public health notification: Potentially fatal bias; approved guideline - third edition. CLSI errors with GDH-PQQ* glucose monitoring document EP15-A3 (ISBN 1-56238-966-1) USA, technology. august 13, 2009. Available from: 2014. http://www.fda.gov/MedicalDevices/safety/ 27. Nelson LM. Primary ovarian insufficiency. N AlertsandNotices/PublicHealthNotifications/ Engl J Med. 2009;360(6):606-14. ucm176992.htm [Last accessed: May 27, 2015].

Challenges in the detection of 3 manufacturing source-specific risks

Traceability of biopharmaceuticals in 3.1 spontaneous reporting systems: a cross- sectional study in the FDA Adverse Event Reporting System (FAERS) and EudraVigilance databases

Vermeer NS, Straus SMJM, Mantel-Teeuwisse AK, Domergue F, Egberts ACG, Leufkens HGM, De Bruin ML

Drug Safety 2013; 36: 617-25

Abstract

Background and objective Adverse drug reactions (ADRs) of biopharmaceuticals can be batch or product specific, resulting from small differences in the manufacturing process. Detailed exposure information should be readily available in systems for postmarketing safety surveillance of biopharmaceuticals, including spontaneous reporting systems (SRSs), in which reports of ADRs are collected. The aim of this study was to explore the current status of traceability of biopharmaceuticals in the US and the EU up to patient level in SRSs. Methods A cross-sectional study was conducted over the period 2004-2010, including ADR reports from two major SRSs: the FDA Adverse Event Reporting System (FAERS) in the US and EudraVigi- lance in the EU. The availability of batch numbers was determined for biopharmaceuticals, and compared with small-molecule drugs. For biopharmaceuticals for which a biosimilar has been approved for marketing in the EU, the identifiability of the product (i.e. the possibility of distinguishing the biosimilar from the reference biopharmaceutical) was determined. Results A total of 2,028,600 unique ADR reports were identified in FAERS, reporting a total of 591,380 biopharmaceuticals (of which 487,065 suspected). In EudraVigilance there were 2,108,742 unique ADR reports, reporting a total of 439,971 biopharmaceuticals (356,293 suspected). Overall, for 24.0% of the suspected biopharmaceuticals in FAERS and 7.4% of the small-molecule drugs (p<0.001) batch numbers were available. In EudraVigilance for 21.1% of the suspected biopharmaceuticals batch numbers were available, and for 3.6% of the small-molecule drugs (p<0.001). Among the 13,790 biopharmaceuticals (9,759 suspected) in EudraVigilance for which a biosimilar was approved, for 90.4% of these (and for 96.2% of the suspected) the product was identifiable. Conclusion This study underlines the need for improving traceability of biopharmaceuticals, in particular with respect to individual batches, allowing better identification and monitoring of postmarketing safety issues related to biopharmaceuticals. 58 Chapter 3.1

Background

Biopharmaceuticals, also known as “biologicals”, provide innovative and effective therapies for often severe and life-threatening diseases. Because of their specific characteristics, biopharmaceuticals have been associated with specific safety concerns, and challenges in pharmacovigilance and risk management [1,2]. One of the distinctive properties of biopharmaceuticals is that the safety profile may change over time, resulting from changes in the manufacturing or formulating process. The unexpected increased occurrence of pure red cell aplasia in patients treated with recombinant human erythropoietin outside the US is an example of a postmarketing safety issue associated with such a specific characteristic of biopharmaceuticals [3]. More than 200 cases of this rare and severe haematological disorder have been attributed to the altered immunogenicity of an epoetin alfa for which formulation changes had recently been issued [4]. Biopharmaceuticals are subject to frequent manufacturing changes once a marketing authorization has been granted. Although these changes are adopted to benefit public health, e.g. by improving product properties or product yield, alterations in the production process may adversely impact the product quality attributes [5,6]. Consequently, the clinical efficacy and safety, in particular immunogenicity, may be affected. Several examples, apart from epoetin alfa, are known whereby immunogenicity with potentially serious clinical consequences was associated with manufacturing and/or formulation changes, including thrombocytopenia with thrombopoietin and neutralizing antibodies with human growth hormone [6]. To ensure patient safety, regulatory authorities in the US and EU have adopted extensive guidance for evaluating, comparability of biopharmaceuticals pre- and post-manufacturing changes [7,8]. This comparability exercise represents a challenging task for manufacturers and regulatory authorities as there is no set of analytical techniques that can fully describe the structural properties of the biopharmaceutical [9]. Moreover, when substantial alterations in the structure products are found, regulatory authorities have the difficult task of deciding whether the identified changes are acceptable [10], i.e. that the changes don’t affect the clinical performance of the biopharmaceutical. Determining therapeutic equivalence is an even more difficult task when the whole manufacturing process of a biopharmaceutical is redeveloped by a second manufacturer, which is the case for biosimilars [11]. Manufacturers of biosimilars do not have access to the manufacturing process of the reference product since this is proprietary knowledge [12]. Consequently, the independent development of a new manufacturing process is likely to result in structural differences between biosimilars and their reference products, possibly affecting the product’s immunogenicity [13]. Recently, biosimilar development has been receiving increasing attention due to expiring patent protection of top-selling biopharmaceuticals [14] and the need for cutting healthcare spending in the Western world [15]. It Traceability of biologicals in ADR reports 59 is already estimated that in 2015 biosimilars will represent approximately 40% of the total worldwide biopharmaceutical market [16]. Ensuring the traceability of biopharmaceuticals up to batch and product level is essential in view of the risk of batch- or product-specific adverse drug reactions (ADRs). Detailed exposure information should be readily available in systems for postmarketing safety surveillance of biopharmaceuticals, including spontaneous reporting systems (SRSs), in which reports of ADRs are collected. It is known that SRS have played a pivotal role in detecting postmarketing safety issues for small-molecule drugs in the past. These systems could also play an essential role in detecting and monitoring any future batch- or product-specific safety issues of biopharmaceuticals, provided that this information is captured. The current study therefore aims to explore the current status of traceability of biopharmaceuticals in the US and the EU up to patient level in SRSs.

Methods

The traceability of biopharmaceuticals was studied in a cross-sectional study over the period 3 2004–2010, including spontaneous ADR reports from two major SRS: the FDA Adverse Event Reporting System (FAERS) in the US and EudraVigilance (EV) in the EU.

FDA Adverse Event Reporting System (FAERS) data content and structure The FAERS was established in 1969 to support the post-marketing safety surveillance of the Food and Drug Administration (FDA). The FAERS database encompasses individual case safety reports (ICSRs) for the majority of FDA approved medicinal products. An ICSR is defined as the information provided by a primary source to describe suspected ADRs related to the administration of one or more medicinal products to an individual patient at a particular point of time [17]. The FDA receives ICSRs directly from consumers or health care professionals or indirectly through manufacturers, when the ADR is initially brought to their attention. Manufacturers have an obligation to periodically report both serious and non-serious ADRs that occurred in domestic (US) clinical practice, and serious unexpected ADRs occurring in the US or a third country [18,19]. FAERS data is freely accessible under the Freedom of Information Act, and FAERS data from 2004 onwards is directly available from the website of the FDA (http://www.fda.gov/Drugs/InformationOnDrugs/ucm135151. htm). The FDA may receive multiple ICSRs, e.g. follow-up reports from the previous or a second reporter, referring to the same occurrence of ADRs in the same patient on the same time. All initial and follow-up ICSRs appear in FAERS. If correctly linked, follow-up reports on an initial ICSR are identifiable in FAERS upon identical “case numbers”. 60 Chapter 3.1

EudraVigilance data content and structure EudraVigilance (EV) was established in 2001 to collect ICSRs of (serious) ADRs to medicines licensed within the EU. ICSRs are received indirectly through EU national competent authorities and pharmaceutical companies. Pharmaceutical companies have a legal obligation to report all serious unexpected ADRs and any suspected transmission via a medicinal product of any infectious agents occurring outside the EU which are brought to their attention. National competent authorities are required to report any serious ADR occurring within the EU [20]. EV data was obtained through a request for Access to Data according to the EV Access Policy, as data from EV has only recently become accessible for research purposes. Recently, the EMA has begun publishing suspected ADR reports on their website to foster transparency (http://www.adrreports.eu). An ICSR in EV reflects the most recent and comprehensive information on an event of an ADR. Follow-up ICSRs by the same reporter will automatically replace the previous ICSR. Duplicate ICSRs from multiple reporters are merged into “master cases”, containing the most comprehensive information from the individual ICSRs[16].

Data extraction and handling All spontaneous ICSRs over the period 2004 – 2010 were selected from FAERS and EV. ICSRs originating from literature and clinical studies were not of interest, as the current study aims to describe batch and product traceability in clinical practice. All literature and study ICSRs were therefore omitted from FAERS (n=156,776). Only spontaneous ICSRs were requested from EV. All drug information from ICSRs in FAERS referring to the same case number were merged into a single cumulative ICSR, similar to the EV approach mentioned above. One cumulative ICSR contains all unique drugs reported over time within these ICSRs. Duplicate drugs within the same ICSR were merged in EV. These steps were undertaken to avoid duplication of data in our final unit of analysis: one medicinal product, subject of a suspected ADR report, administered to an individual patient at a particular point of time. Figure 1 illustrates how EV and FAERS data were processed to our unit of analysis. The following data was subsequently extracted from all cumulative ICSRs from EV and FAERS: information on name of the drug and/or active substances, batch number, name of marketing authorization holder (EV only), role code of the drug (see section “Classification of medicinal products” below), type of reporter and reporting date. In line with the ICH E2B guideline the following reporters are distinguished: physician, pharmacist, other health professional, lawyer and consumer. Since the reporter on a cumulative ICSR may be non- unique (due to duplicate reporting) a separate category was assigned for multiple reporters. Traceability of biologicals in ADR reports 61

Classification of medicinal products Biopharmaceuticals were defined as protein or nucleic based pharmaceuticals products used for therapeutic or in vivo diagnostic purposes [21]. Medicinal products were classified into two groups (i) biopharmaceuticals and (ii) small-molecule drugs using the WHO ATC classification system (http://www.whocc.no/atc_ddd_index/, see Supplementary Table 1). Two pharmacists confirmed the classification of the data. Medicinal products not classifiable to either group (e.g. verbatim data entered as “unspecified drug” or “radiation therapy”) were excluded. Moreover, vaccines (ATC class: J07) and whole blood or components of whole blood (ATC class: B05A) were excluded as they are subject to different reporting requirements. The role code of the medicinal product was recoded into suspected (classified as “primary suspect” or “secondary suspect” in FAERS or “suspect” in EV) and non-suspected (classified as “interacting” or “concomitant” in both SRS).

FAERS < EudraVigilance

Original dataset

ICSR 1, case nr i ICSR 2, case nr i ICSR 1 - Drug A - Drug A - Drug A 3 - Drug B - Drug B - Drug B - Drug B - Drug B - Drug C - Drug C

Dataset for analysis Cumulative ICSR 1 Cumulative ICSR 1 - Drug A - Drug A - Drug B - Drug B unit of analysis - Drug C - Drug C

Figure 1. Processing original FAERS and EudraVigilance ICSRs to cumulative ICSRs. FAERS, FDA Adverse Event Reporting System; ICSR, individual case safety report.

Classification of traceability Verbatim data entered in the designated field for batch number was recoded to a dichoto- mous variable describing the availability of the batch number. Any verbatim data entered into the designated field for batch number was considered to be a batch number. To validate whether the verbatim data did not contain any information referring to the unavailability of a batch number (e.g. “unknown”, or “discarded package”), data was aggregated and carefully reviewed. A second reviewer reviewed the determination of the availability of batch numbers. For biopharmaceuticals for which a biosimilar has been approved for marketing in the European Union (see Supplementary Table 2), the identifiability of the product in the cumulative ICSRs in EV was determined. For this analysis we included only ICSRs reported from the month following the approval of the first biosimilar within that product class: epoetin alfa, 62 Chapter 3.1

filgrastim, and somatropin. Product names were considered identifiable when the brand name or the international nonproprietary name (INN) plus name of marketing authorization holder were available. Products for which only the INN was available were considered non-identifiable, except for epoetin zeta for which product the INN differs from the innovator (epoetin alfa).

Data analysis Descriptive statistics were calculated as proportions to describe the traceability of batch numbers for biopharmaceuticals and small-molecule drugs in FAERS and EV. Results were further stratified by the role code of drug and the type of reporter to identify whether proportions differed over the different variables. For the top eight most frequently reported ATC classes (see Supplementary Table 1) the availability of batch numbers was calculated overall, and further stratified by type of reporter (medical doctor, pharmacist or consumer). The product identifiability, and subsequent batch traceability, for biopharmaceuticals for which a biosimilar has been approved in the EU were also calculated as proportions. Significance was tested using chi-square statistics. All analyses were performed using SPSS statistical software version 18 (Chicago, IL).

Results

A total of 2,028,600 and 2,108,742 cumulative ICSRs were respectively retrieved from FAERS and EV. Within these cumulative ICSRs a respective total of 591,380 (from the 6,603,489) and 439,971 (from the 6,431,175) medicinal products concerned biopharmaceuticals (see Figure 2). Overall, for 24.0% of the suspected biopharmaceuticals in FAERS and for 7.4% of the suspected small-molecule drugs (p<0.001) batch numbers were available. A similar pattern

Original dataset Dataset for analysis After exclusion of duplicate data: Classification of medicinal products:

FAERS 2,747,361 ICSRs 2,028,600 cumulative ICSRs 5,794,537 small molecule drugs 10,143,467 medicinal products 6,603,489 unique medicinal products: 591,380 biopharmaceuticals 217,572 (3.3%) excluded/ non-classifiable, including 4,773 blood products (B05A) and 3,253 vaccines (J07)

EudraVigilance 2,108,742 ICSRs 2,108,742 cumulative ICSRs 5,630,208 small molecule drugs 7,121,186 medicinal products 6,431,175 unique medicinal products: 439,971 biopharmaceuticals 360,996 (5.6%) excluded/ non-classifiable, including 7,397 blood products (B05A) and 189,978 vaccines (J07)

Figure 2. Number of ICSRs and medicinal products in original dataset, and dataset for analysis. EV, EudraVigilance; FAERS, FDA Adverse Event Reporting System; ICSR, individual case safety report. Traceability of biologicals in ADR reports 63 was seen in EV: for 21.1% of the suspected biopharmaceuticals batch numbers were available, compared to only for 3.6% of the small-molecule drugs (p<0.001). The traceability of individual batches for the overall group of drugs in FAERS and EV, including also concomitant and interacting drugs, was less well ensured (see Table 1).

Table 1. Availability of batch numbers reported biopharmaceuticals and small-molecule drugs in FAERS and EudraVigilance, stratified by role code. Drug type, role code FDA Adverse Event Reporting EudraVigilance system Total Drugs with batch Total Drugs with batch number of number available number of number available drugs [n, %] drugs [n (%)] Biopharma All 591,380 117,523 19.9% 439,971 75,713 17.2% ceutical Suspected 487,065 116,812 24.0% 356,293 75,272 21.1% Small-molecule All 5,794,537 164,755 2.8% 5,630,208 79,610 1.4% drug Suspected 2,221,229 163,670 7.4% 2,169,721 77,724 3.6% Total All 6,385,917 282,278 4.4% 6,070,179 155,323 2.6% Suspected 2,708,294 280,482 10.4% 2,526,014 152,996 6.1% 3

For the 487,065 suspected biopharmaceuticals in FAERS and 356,293 suspected biopharmaceuticals in EV the availability of batch numbers was calculated for different reporter types. Biopharmaceuticals reported by consumers most frequently contained a batch number (36.3% in FAERS and 40.7% in EV). Batch traceability of biopharmaceuticals reported by medical doctors in FAERS (13.3%) and EV (7.0%) was substantially lower, both p<0.001. Biopharmaceuticals reported by pharmacists or other health care professionals were also less likely to contain a batch number (see Table 2).

Table 2. Availability of batch numbers for reported suspected biopharmaceuticals, stratified by type of reporter. Reporter type FAERS (n=487,065) EudraVigilance (n=356,293) Total Drugs with batch Total number Drugs with batch number of number available of drugs1 number available [n, drugs1 [n, %] %] Medical doctor 112,770 15,026 13.3% 94,928 6,667 7.0% Pharmacist 12,971 2,984 23.0% 9,999 1,896 19.0% Other health 64,235 9,087 14.1% 46,765 5,366 11.5% professional Consumer 198,282 72,006 36.3% 117,411 47,800 40.7% Lawyer 1,489 10 0.7% 1,242 5 0.4% 1For a total of 97,318 biopharmaceuticals in FAERS and 85,948 in EudraVigilance the reporter type was not unique or unavailable; FAERS, FDA Adverse Event Reporting System. 64 Chapter 3.1

This pattern differed, however, between the eight most frequently reported pharmacological and therapeutic subgroups of biopharmaceuticals. Overall batch numbers were most frequently available for parathyroid hormone (H05AA) in FAERS (43.1%) and for immunoglobulins (J06) in EV (42.0%). Most notably, the overall availability of batch numbers for parathyroid hormone in FAERS was higher than in all three separate reporter categories (see Figure 3), representing the high availability of batch numbers in reports of which the reporter was not known (62.5%, not shown in figure). Respectively 43.1% and 44.2% of the TNF-alpha inhibitors, the most frequently reported ATC class, reported by consumers contained a batch number in FAERS and EV. Batch traceability of TNF-alpha inhibitors reported by medical doctors and pharmacists was lower in FAERS (respectively 16.2% and 13.5%), and particularly in EV (respectively 3.8% and 10.5%). Pharmacists did remarkably well in reporting batch numbers for immunoglobulins in FAERS (59.4%) and EV (55.2%). The traceability of antineoplastic monoclonal antibodies (ATC class L01XC) was overall poorly maintained (see Figure 3).

70% FAERS

60%

50%

40%

30%

20% Insulin (A10A) 10% Other antianemic preparations (B03X) Parathyroid hormone (H05AA) 0% Patient Medical doctor Pharmacist Overall Immunoglobulines (J06) Monoclonal antibodies, antineoplastic (L01XC) 60% EudraVigilance Colony stimulating factors (L03AA)

50% Interferons (L03AB) TNF alpha inhibitors (L04AB) 40%

30%

20%

10%

0% Patient Medical doctor Pharmacist Overall

Figure 3. Availability of batch numbers (%1) for eight groups of suspected biopharmaceuticals in FAERS and EudraVigilance, stratified by type of reporter. 1Calculated as number of biopharmaceuticals containing batch number/number of biopharmaceuticals reported by reporter within group. FAERS, FDA Adverse Event Reporting System.

There was a variability in traceability of suspected biopharmaceuticals over time for EV and FAERS (see Figure 4). Batch traceability of biopharmaceuticals in EV showed a sharp increase between 2007 (10.7% of reported biopharmaceuticals contained a batch numbers) and 2008 (22.8% did). Batch traceability of biopharmaceuticals in FAERS showed a peak Traceability of biologicals in ADR reports 65

FAERS EudraVigilance

40% ) % (

s 30% e h c t a b

o 20%

y t i l i b a

e 10% c a r T

0% 2004 2005 2006 2007 2008 2009 Year of reporting Figure 4. Availability of batch number for suspected biopharmaceuticals from 2004 until 2010 in FAERS and EudraVigilance. FAERS, FDA Adverse Event Reporting System.

in 2007, with 35.8% of the reported biopharmaceuticals containing a batch number, but declined thereafter. 3 A total of 13,790 biopharmaceuticals (of which 9,759 suspected) for which a biosimilar has been approved in the EU were extracted from cumulative ICSRs in EV. For 90.4% of these biopharmaceuticals, and for 96.2% of the suspected biopharmaceuticals the product was clearly identifiable. The batch traceability of biosimilars was interestingly poorly maintained (see Table 3).

Table 3. Identifiability of biosimilars1 in EudraVigilance. Biopharmaceutical, role code Total number of Drugs with identifiable Drugs with identifiable drugs product name [n, %] product name and traceable batch number [n, %] Epoetin alfa All 9,125 8,615 94.4% 320 3.5% Suspected 6,903 6,829 98.9% 318 4.6% Filgrastim All 2,227 1,702 76.4% 73 3.3% Suspected 706 600 85.0% 72 10.2% Somatropin All 2,438 2,148 88.1% 128 5.3% Suspected 2,150 1,963 91.3% 128 6.0% Total All 13,790 12,465 90.4% 521 3.8% Suspected 9,759 9,392 96.2% 518 5.7% 1Biopharmaceuticals for which a biosimilars has been approved in the EU 66 Chapter 3.1

Discussion

The present study showed that for 24.0% of the suspected biopharmaceuticals in FAERS and for 21.1% of the suspected biopharmaceuticals in EV a batch number was available. The current study in addition showed that for 96.2% of the suspected biopharmaceuticals for which a biosimilar was available in the EU the product name was clearly identifiable in EV. Ac- curate traceability of biopharmaceuticals up to batch and product level in these spontaneous reporting systems is essential for identifying and monitoring any batch or product specific safety issues, resulting from differences in the manufacturing process. Batch traceability may in addition help to distinguish between, and assess the safety profile over, different pharmaceutical forms and dosage strengths of biopharmaceuticals. Furthermore, accurate batch traceability is pivotal for relating any batch related problems of biopharmaceuticals, e.g. pathogen-contaminated batches [22] or other host cell impurities [23], to reported adverse drug reactions. Biopharmaceuticals might be at increased risk of batch related problems as the production process, which involves living expression systems, gives rise to a large number of host cell-, process-, and product-related impurities. The proportions of biopharmaceuticals containing a batch number in FAERS and EV were much higher than we found for suspected small-molecule drugs (respectively 7.4% and 3.6%), but lower than elsewhere reported for vaccines (54.4%) [24]. The lack of information on batch numbers for approximately 3 in 4 biopharmaceuticals in FAERS (approximately 4 in 5 in EV) could either be the result of incomplete recording of exposure information at dispensing, or of incomplete reporting of the available information to regulatory authorities and/or manufacturers. The reported differences in batch traceability between different pharmacological/therapeutic groups of biopharmaceuticals suggest in particular a role for incomplete recording of exposure information in clinical practice. Whereas consumers reported a batch number for 36.2-46.8% of frequently home-administered insulins, they did only for 1.3-7.5% of antineoplastic monoclonal antibodies, which are primarily administered in hospitals. For consumers it is relatively easy to obtain batch numbers if the medicine is applied at home, as is the case for insulins. As in the case of antineoplastic agents, the preparation, administration and reporting might very well be by different persons, and the patient or physician confronted with the ADR might not have access to the batch information. This indicates once the batch numbers are readily available consumers are likely to report this information. Another finding from the current study was that batch traceability was overall well maintained for immunoglobulins, in particular when reported by pharmacists. Batch numbers were overall available for 36.7-42.0% of the reported immunoglobulins, and even for 55.2-59.4% of the immunoglobulins reported by pharmacists. This might be explained by the fact that immunoglobulins historically were plasma-derived medicinal products, for which separate regulations are in place. The safety of blood-derived products has especially Traceability of biologicals in ADR reports 67 been in Europe under close scrutiny, following the HIV infected blood-products scandal that occurred in France in the 1990s [25], and concerns for transmission Creutzfeldt-Jakob Disease via whole blood and plasma-derived products in Europe [26]. Though the latter has only been a theoretical risk until 2009 [27]. The current study showed that patients play an important role in ensuring traceability of biopharmaceuticals. In respectively 36.3% and 40.7% of the consumer reports on biopharmaceuticals a batch number was available in FAERS and EV. These results underpin the importance of patient reporting of adverse events. Patients have been able to reports adverse to the FAERS, since its establishment in 1969. In most European countries patient reporting schemes have only recently been established, and in some countries patients are still not able to report adverse events directly to the competent authorities [28]. We showed on the other hand that physicians played only a minor role in ensuring traceability of biopharmaceuticals. Despite being a major contributor in the absolute number of reports, only in respectively 13.3% and 7.0% of the reports on biopharmaceuticals in FAERS and EV a batch number was available. The present study showed that product identifiability of biopharmaceuticals for which a biosimilars has been approved in the EU is reasonably well ensured in Europe, especially for 3 epoetin alfa. This is an important finding, taking into account that biosimilars are frequently given the same international nonproprietary (INN) name as the reference innovator. Of the seven currently approved biosimilars in Europe, six contain the same INN as the innovator (see Supplementary Table 2). It has therefore been recognized that the INN system, although playing an important role in global pharmacovigilance, cannot be relied upon for product identification of biosimilars [29]. As the number of biosimilars on the market is expected to increase in a vast pace, and a road for biosimilar registration is currently been paved in the US [30], traceability of biosimilars will only become increasingly important. The need of ensuring traceability is not unique to biopharmaceuticals, but also a well- known aspect in numerous other industries [31]. Especially the traceability of medical devices is receiving increased scrutiny following recent concerns in Europe on a possible association between frequently used breast implants (PIP implants) and cancer [32]. The European commission has already announced plans to enhance traceability of medical devices [33]. Similarly, several initiatives are currently ongoing to further promote the traceability of biopharmaceuticals [34]. These initiatives are not only fueled by increased interest in drug safety, but also by the need for improving supply chain efficacy and the need for taking measures against counterfeit medicines [31]. Two-dimensional barcodes that could include detailed product information like batch numbers is one of the presented solutions for promoting traceability of biopharmaceuticals [34]. When such information is automatically recorded in clinical practice, it is essential that patients and health care professionals are aware that this accurate exposure information is necessary to link the adverse event properly to the specific product. Patients and health care professional may be encouraged 68 Chapter 3.1

by regulatory authorities in the product information of biopharmaceuticals to reports such information. This may be one of increased efforts taken by the EU and its Member States to ensure traceability of biopharmaceuticals, demanded by recently adopted new European pharmacovigilance legislation [35]. It should be noted that spontaneous reporting systems may not be the sole point in the community where detailed exposure information on biopharmaceuticals is captured. In the current study we did not assess whether population-based databases or disease registries contain the necessary exposure information to monitor and ascertain the safety of biopharmaceuticals over different batches or products. A second limitation that should be addressed is that the databases we have used might contain a large number of duplicate reports. Especially for FAERS extensive duplication of reports has been reported [36]. To limit any influence of duplicate information on our study results, efforts were undertaken to reduce data duplication. As we were however not interested in the frequency of certain adverse event, but only in the reporting quality of the submitted reports, we feel any residual duplication might have little influence on our results. In summary, the present study was to our knowledge the first study to We have reported that the current system insufficiently ensures the traceability of individual batches of biopharmaceuticals, although the identifiability of biosimilars is reasonably well ensured. Stakeholders in pharmacovigilance should undertake efforts to improve the traceability of biopharmaceuticals. Traceability of biologicals in ADR reports 69

References

1. Giezen TJ, Mantel-Teeuwisse AK, Leufkens HG. 12. Mellstedt H, Niederwieser D, Ludwig H. Pharmacovigilance of biopharmaceuticals: chal- The challenge of biosimilars. Ann Oncol. lenges remain. Drug Saf. 2009;32(10):811–7. 2008;19(3):411–9. 2. Declerck PJ. Biotherapeutics in the era of bi- 13. Schellekens H. Bioequivalence and the im- osimilars: what really matters is patient safety. munogenicity of biopharmaceuticals. Nat Rev Drug Saf. 2007;30(12):1087–92. Drug Discov. 2002;1(6):457–62. 3. Casadevall N, Nataf J, Viron B, Kolta A, Kiladjian 14. Walsh G. Biopharmaceutical benchmarks 2010. JJ, Martin- Dupont P, et al. Pure red-cell aplasia Nat Biotechnol.2010;28(9):917–24. and antierythropoietin antibodies in patients 15. Wadman M. Bills target biosimilar drugs. Nature. treated with recombinant erythropoietin. N Engl 2009;458:394–5. J Med. 2002;346(7):469–75. 16. McCamish M, Woollett G. Worldwide experi- 4. Ebbers HC, Mantel-Teeuwisse AK, Moors EH, ence with biosimilar development. MAbs. Schellekens H, Leufkens HG. Today’s challenges 2011;3(2):209–17. in pharmacovigilance: what can we learn from 17. European Medicines Agency. Note for guid- epoetins? Drug Saf. 2011;34(4):273–87. ance—EudraVigilance human—processing of 5. Chirino AJ, Mire-Sluis A. Characterizing biologi- safety messages and individual case safety reports cal products and assessing comparability follow- (ICSRs). 15 October 2010. ing manufacturing changes. Nat Biotechnol. 18. US FDA. Postmarketing reporting of adverse 2004;22(11):1383–91. experiences. Code of Federal Regulations Title 3 6. Sharma B. Immunogenicity of therapeutic pro- 21 (21 CFR 600.80). Silver Spring: US FDA; 2012. teins. Part 3: impact of manufacturing changes. 19. US FDA. Postmarketing reporting of adverse Biotechnol Adv. 2007;25(3): 325–31. drug experiences, Code of Federal Regulation 7. European Medicines Agency. Note for guid- Title 21, (21 CFR 314.80). Silver Spring: US FDA; ance on biotechnological/biological products 2012. subject to changes in their manufacturing process 20. Regulation (EC) No 726/2004 of The European (CPMP/ICH/5271/03). June 2005. http://www. Parliament and of the Council of 31 March ema.europa.eu/docs/en_GB/document_library/ 2004, laying down Community procedures for Scientific_guideline/2009/09/WC500002805.pdf. the authorisation and supervision of medicinal [Last accessed May 7, 2013]. products for human and veterinary use and 8. US FDA. Demonstration of comparability of establishing a European Medicines Agency. human biological products, including therapeutic Available from: http://eur-lex.europa.eu/Lex- biotechnology-derived products. April 1996. UriServ/LexUriServ.do?uri=OJ:L:2004:136:00 http://www.fda.gov/Drugs/GuidanceCom- 01:0033:en:PDF. [Last accessed May 7, 2013]. plianceRegulatoryInformation/Guidances/ 21. Crommelin DJA, Storm G, Verrijk R, de Leede ucm122879.htm. [Last accessed May 7, 2013]. L, Jiskoot W, Hennink WE. Shifting paradigms: 9. Schellekens H. When biotech proteins go off- biopharmaceuticals versus low molecular weight patent. Trends Biotechnol. 2004;22(8):406–10. drugs. Int J Pharm. 2003;266(1–2):3–16. 10. Schiestl M, Stangler T, Torella C, Cepeljnik T, 22. Vonberg RP, Gastmeier P. Hospital-acquired Toll H, Grau R. Acceptable changes in quality infections related to contaminated substances. attributes of glycosylated biopharmaceuticals. J Hosp Infect. 2007;65(1):15-23. Nat Biotechnol. 2011;29(4):310–2. 23. Thorpe SJ, Fox BJ, Dolman CD, Lawrence J, 11. Buttel IC, Chamberlain P, Chowers Y, Ehmann Thorpe R. Batches of intravenous immunoglobu- F, Greinacher A, Jefferis R, et al. Taking im- lin associated with adverse reactions in recipients munogenicity assessment of therapeutic proteins contain atypically high anti-Rh D activity. Vox to the next level. Biologicals. 2011;39(2):100–9. Sang. 2003;85(2):80–4. 70 Chapter 3.1

24. Kurz X, Domergue F, Slattery J, Segec A, Szmi- 32. O’Dowd A. UK launches inquiry into safety of giel A, Hidalgo-Simon A. Safety monitoring PIP breast implants. BMJ. 2012;3(344):e11. of Influenza A/H1N1 pandemic vaccines in 33. Petitjean S. Commission to enhance traceability EudraVigilance. Vaccine. 2011;29(26):4378–87. of medical devices. http://www.europolitics. 25. de Bousingen DD. Former French ministers on info/commission-to-enhance-traceability-of- trial for infected blood-products scandal. Lancet. medical-devices-art324118.html. [Last accessed: 1999;353(9153):653. March 22,2012}. 26. Esmonde TF, Will RG, Slattery JM, Knight R, 34. European Federation of Pharmaceutical Harries-Jones R, de Silva R, et al. Creutzfeldt- Industries and Associations. EFPIA product Jakob disease and blood transfusion. Lancet. verification project. April 2011.http://www.efpia. 1993;341(8839):205–7. eu. [Last accessed January 12, 2012]. 27. Eaton L. Haemophilia patient had variant CJD 35. Directive 2010/84/EU of The European Par- agent in spleen. BMJ. 2009;18(338):b705. liament and of the Council of 15 December 28. van Hunsel F, Harmark L, Pal S, Olsson S, van 2010, amending, as regards pharmacovigilance, Grootheest K. Experiences with adverse drug Directive 2001/83/EC on the Community code reaction reporting by patients: an 11-country relating to medicinal products for human use. survey. Drug Saf. 2012;35(1):45–60. http://eudravigilance.ema.europa.eu/human/ 29. Zuniga L, Calvo B. Biosimilars: pharmacovigi- docs/Directives/dir_2010_84_en.pdf. [Last lance and risk management. Pharmacoepidemiol accessed May 7, 2013]. Drug Saf. 2010;19(7):661–9. 36. Hauben M, Reich L, DeMicco J, Kim K. 30. Fox JL. Debate over details of US biosimilar ‘Extreme duplication’ in the US FDA Adverse pathway continues to rage. Nat Biotechnol. Events Reporting System database. Drug Saf. 2012;30(7):577. 2007;30(6):551–4. 31. Lovis C. Traceability in healthcare: crossing boundaries. Yearb Med Inform. 2008;47(Suppl 1):105–13. Traceability of biologicals in ADR reports 71

Supplementary Table 1. Classification of biopharmaceuticals according to ATC codes. ATC code Chemical/pharmacological/therapeutic First level 2nd, 3rd, 4th or 5th level subgroups or substance name A, Alimentary tract and A10A Insulins and analogues metabolism A10BX04 Exenatide A10BX07 Liraglutide A16AB Enzymes B, Blood and blood forming B01AB02 Antithrombine III organs B01AC13 Abciximab B01AD Antithrombotic agents, enzymes B01AE01 Desirudin B01AE02 Lepirudin B02AB02 Alpha-1 antitrypsin B02AB03 C1-inhibitor B02BB Fibrinogen B02BD Blood coagulation factors B02BX04 Romiplostim B03X Other antianemic preparations 3 B05AA01 Albumin B06AA Enzymes C, Cardiovascular system C01DX19 Nesiritide D, Dermatologicals D03AX06 Becaplermine G, Genito uriny system and sex G03GA Gonadotropins hormones H, Systemic hormonal H01AB Thyrotropin preparations, excl sex hormones H01AC Somatropin and somatropin agonist H01AX Other anterior pituitary lobe hormones and analogues H04 Pancreatic Hormones H05AA Parathyroid Hormone H05BA Calcitonin preparations J, Anti-infectives for systemic J06 Immune sera and immunoglobulins use J06BB16 Palivizumab J06BC01 Nebacumab 72 Chapter 3.1

Supplementary Table 1. Classification of biopharmaceuticals according to ATC codes. (continued) ATC code Chemical/pharmacological/therapeutic First level 2nd, 3rd, 4th or 5th level subgroups or substance name L, Antineoplastic and L01XC Monoclonal antibodies immonomodulating agents L01XX02 Asparaginase L01XX29 Denileukin diftitoxin L03AA Colony stimulating factors L03AB Interferons L03AC Interleukins L03AX11 Tasonermin L04AA02 Muromonab-CD3 L04AA15 Alefacept L04AA21 Efalizumab L04AA23 Natalizumab L04AA24 Abatacept L04AA25 Eculizumab L04AA26 Belimumab L04AA28 Belatacept L04AB TNF-alpha inhibitors L04AC Interleukin inhibitors M, Musculo-skeletal system M03AX Other muscle relaxants M05BC Bone morphogenetic proteins M05BX04 Denosumab M09AX02 Autologous cultured chrondocytes R, Respiratory system R03DX05 Omalizumab S, Sensory Organs S01AD08 Fomivirsen S01LA04 Ranibizumab S01LA05 Alifercept V, Various V03AB24 Digoxin Immune Fab V03AF07 Rasburicase V03AF08 Palifermin V04CJ01 Thyrotropin V09IA04 Votumumab V09IA06 Arcitumomab V09IB02 Satumomab V09IB04 Capromab V09HA04 Sulesomab V10XA53 Tositumomab V10XX02 Ibritumomab Traceability of biologicals in ADR reports 73

Supplementary Table 2. Biosimilars approved for marketing within the EU. INN Trade name of biosimilar ATC code Reference product Date of approval Epoetin alfa Abseamed/ Binocrit/ Epoetine B03XA01 Eprex 28-08-2007 alfa Hexal Epoetin zeta Retacrit/ Silapo 18-12-2007 Filgrastim Biograstim/ Filgrastim L03AA02 Neupogen 15-09-2008 Ratiopharm1/ Ratiograstim/ Tevagrastim Zarzio/ Filgrastim Hexal 06-02-2009 Nivestim 08-06-2010 Somatropin Omnitrope H01AC01 Genotropin 12-04-2006 Valtropin Humatrope 24-04-2006 1Marketing authorization withdrawn on 20 April 2011

The effect of exposure 3.2 misclassification in spontaneous ADR reports on the time to detection of product-specific risks for biologicals: a simulation study

Vermeer NS, Ebbers HC, Straus SMJM, Leufkens HGM, Egberts ACG, De Bruin ML

Submitted for publication

Abstract

Background and objective The availability of accurate product-specific exposure information is essential in the pharmacovigilance of biologicals, because differences in the safety profile may emerge between products containing the same active substance. In spontaneous ADR reports drug exposure may, however, be misclassified, i.e. attributed to the incorrect product. The aim of this study was to explore the effect of exposure misclassification on the time to detection of product-specific risks in spontaneous reporting systems. Methods We used data simulations to explore the effect of exposure misclassification. We simulated an active substance-specific subset of a spontaneous reporting system, and used the proportional reporting ratio (PRR) for signal detection. The effect of exposure misclassification was evaluated in three test cases representing product-specific ADRs that may occur for biologicals, and studied in relative terms by varying the model parameters (market share, relative risk). Results We found that exposure misclassification results in the largest delay in identification of risks that have a weak association (RR< 2 or 3) with the product of interest, and in situations where the product associated with the unique risk has a large (>50%) market share. The absolute public health impact of exposure misclassification, in terms of cases/ time to detection, varied considerably across the test cases. Conclusion Exposure misclassification in ADR reports may result in a delayed detection of product-specific risks, particularly in the detection of weak drug-event associations. Our findings can help inform the future implementation and refinement of product- and batch-specific signal detection procedures. 76 Chapter 3.2

Introduction

The availability of accurate exposure information is essential in pharmacovigilance. Incor- rect information regarding an individual’s exposure status, including underascertainment and misclassification of exposure, may bias measures of association in drug safety research [1-3]. There are multiple facets to the characterization of exposure (product, dosage, treatment compliance, etc.), and the required level of detail depends on the drug-event combination under study. Adverse drug reactions (ADRs) can be evaluated at different levels (see Figure 1), most typically on the level of the active substance (e.g. PML associated with rituximab [4]), or therapeutic group (“class effects”, e.g. infections with TNF-alpha inhibitors [5]). Evaluations for manufacturing source-specific risk are uncommon for small-molecule drugs, notwithstanding exceptions like bowel perforation with Indosmos [6], but are routinely required for biologicals. The safety profile of biologicals is highly dependent on the manufacturing conditions and formulation process, and variability in these conditions within or across products, could potentially result in product- or batch-specific risks. An example hereof is the risk of epoetin-associated pure red cell aplasia, which was linked to a single manufacturing source of epoetin for which formulation changes had been issued (Eprex) [7].

Level 0: Erythropoiesis- Therapeutic class stimulating agents

Level 1: Epoetin alfa Darbepoetin alfa Active substance

Level 2: Abseamed Eprex Product

Level 3: Prefilled Powder for Manufacturing source- Formulation syringe injection specific exposure data

Level 4: Batch 1 Batch 2 Batch Figure 1. Different levels of exposure information.

Because subtle differences in manufacturing conditions for biologicals may give rise to previously unobserved adverse reactions, product- and batch-specific information is required for adequate safety evaluations of biologicals [8]. Previous studies have shown that batch-specific data is infrequently available in spontaneous ADR reports [9-12] although product-specific exposure information is commonly provided for biologicals [10-12]. The validity of the product and batch information is, however, unknown. Reporters may, for example, habitually provide a brand name, without verifying the actual product dispensed Impact of exposure misclassification 77

or used. The reported discrepancy between the high market share, but low number of ADR reports for generics [13,14], suggests that ADR reports could sometimes be incorrectly attributed to branded innovator drugs in clinical practice. Since the number of biosimilars is expected to expand over the coming years [15-17], product-specific safety evaluations of biologicals will become increasingly important. Spontaneous ADR reports are essential in product-specific pharmacovigilance. Many recent product-specific signals for biologicals that emerged after approval have been detected through case reports [7,18,19]. Regular screening of spontaneous reporting system databases for undiscovered product-specific safety signals may contribute to a timelier identification of product-specific risks, allowing for a timely implementation of risk mitigation strategies. Such screening involves the use of quantitative signal detection methods, in which the relative reporting of a given drug-event combination is compared to relative proportion of that event for all similar biological products [20,21]. For example, data for Eprex should be assessed in relation to the aggregated data on all other products containing the active substance epoetin alfa. Drug exposure misclassification (i.e. between similar products) may however distort such product-specific signal detection procedures, though the actual impact hereof is unknown. In thisMethods study, we therefore aimed to explore the effect of exposure 3 misclassification on the time to detection of product-specific risks in spontaneous reporting systems. We used data simulations to explore the effect of exposure misclassification. We simulated an active

Methods substance-specific subset of a spontaneous reporting system, in which the proportional reporting ratio (PRR) was used for signal detection. The effect of exposure misclassification was evaluated in absolute terms (cases/ time to We used data simulations to explore the effect of exposure misclassification. We simulated an active substance-specific subset detection)of a spontaneous in three reporting test system, cases in representing which the proportional product-specific ADRs that may occur for biologicals. Furthermore, the reporting ratio (PRR) was used for signal detection. The effect of exposure misclassification overall impact of exposure misclassification was studied by varying the model parameters, and assessing the effect of was evaluated in absolute terms (cases/ time to detection) in three test cases representing product-specific ADRs that maymisclassification occur for biologicals. in relative Furthermore, terms. the overall impact of exposure misclassification was studied by varying the model parameters, and assessing the effect of misclassification in relativeData terms. simulation procedure Data simulation procedure Number of ADR reports Number of ADR reports The expected number of spontaneousThe expected ADR reports number that ofis spontaneouscollected in a reporting ADR reports system that is collected in a reporting system database for a given drug-

database for a given drug-event combination (Ni,j) can be approximated by the following event combination ( ) can be approximated by the following equation [22-24]: equation [22-24]: ��

�� In this equation, describes the background incidence of event i in the treatment population; describes the patient

� � exposure to drug� j; describes the relative risk of event i for patients exposed to drug j;� and describes the

�� reporting probability ��for the drug-event combination. The reporting probability is determined by many �factors, including,

but not limited to, the seriousness and expectedness of the event, and the time since initial marketing of the drug.

Spontaneous reporting system database

As schematically represented in Figure 2, signal detection from a spontaneous reporting system database is based on

a cross-tabulation of all drug-event combinations ( ) that have been reported at least once. In this study, we only

�� simulated a substance-specific subset for similar biological� products of the database.

Signal detection method

Several measures of disproportionality are available to screen spontaneous reporting systems for previously unknown

associations drug exposures and events [20,21]. Most measures are –in essence– based on a two-by-two cross-

tabulation of the database (Figure 2), comparing the relative reporting for a given drug-event combination to the

relative reporting of that event for other drugs. For this study, we used the proportional reporting ratio (PRR) for signal

detection, which method is also used by the European Medicines Agency (EMA) [25]. Drug-event combinations for

which at least three reports have been received, and for which the lower bound of the 95% confidence interval (95%

CI) for the PRR is at least 1, are considered to represent signals of disproportionate reporting (i.e. safety signals):

� � � � ��

��

84 78 Chapter 3.2

In this equation, Ii describes the background incidence of event i in the treatment popula-

tion; Ej describes the patient exposure to drug j; RRi,j describes the relative risk of event i for

patients exposed to drug j; and pi,j describes the reporting probability for the drug-event combination. The reporting probability is determined by many factors, including, but not limited to, the seriousness and expectedness of the event, and the time since initial marketing of the drug.

Spontaneous reporting system database As schematically represented in Figure 2, signal detection from a spontaneous reporting

system database is based on a cross-tabulation of all drug-event combinations (Ni,j) that have been reported at least once. In this study, we only simulated a substance-specific subset for similar biological products of the database.

Spontaneous reporting system database Two-by-two cross-tabulation for signal detection

Drug1 Drug2 … Drug

j Drug1 Not drug1 Event1 …

Event Event2 … 1 A C

… … … … Not event1 B D

… Event …

Figure 2. Schematic representation of spontaneous reporting system database (left), and two-by- two cross-tabulation of spontaneous reporting system as used for signal detection (right).

Signal detection method Several measures of disproportionality are available to screen spontaneous reporting systems for previously unknown associations drug exposures and events [20,21]. Most measures are –in essence– based on a two-by-two cross-tabulation of the database (Figure 2), comparing the relative reporting for a given drug-event combination to the relative reporting of that event for other drugs. For this study, we used the proportional reporting ratio (PRR) for signal detection, which method is also used by the European Medicines Agency (EMA) [25]. Drug-event combinations for which at least three reports have been received, and for which the lower bound of the 95% confidence interval (95% CI) for the PRR is at least 1, are considered to represent signals of disproportionate reporting (i.e. safety signals):

Exposure misclassfication simulation study 15 Methods

We used data simulations to explore the effect of exposure misclassification. We simulated an active

substance-specific subset of a spontaneous reporting system, in which the proportional reporting ratio (PRR) was

used for signal detection. The effect of exposure misclassification was evaluated in absolute terms (cases/ time to

detection) in three test cases representing product-specific ADRs that may occur for biologicals. Furthermore, the

overall impact of exposure misclassification was studied by varying the model parameters, and assessing the effect of

misclassification in relative terms.

Data simulation procedure

Number of ADR reports

The expected number of spontaneous ADR reports that is collected in a reporting system database for a given drug-

event combination ( ) can be approximated by the following equation [22-24]:

��

�� In this equation, describes the background incidence of event i in the treatment population; describes the patient

� � exposure to drug� j; describes the relative risk of event i for patients exposed to drug j;� and describes the

�� reporting probability ��for the drug-event combination. The reporting probability is determined by many �factors, including,

but not limited to, the seriousness and expectedness of the event, and the time since initial marketing of the drug.

Spontaneous reporting system database

As schematically represented in Figure 2, signal detection from a spontaneous reporting system database is based on

a cross-tabulation of all drug-event combinations ( ) that have been reported at least once. In this study, we only

�� simulated a substance-specific subset for similar biological� products of the database.

Signal detection method

Several measures of disproportionality are available to screen spontaneous reporting systems for previously unknown

associations drug exposures and events [20,21]. Most measures are –in essence– based on a two-by-two cross-

tabulation of the database (Figure 2), comparing the relative reporting for a given drug-event combination to the

relative reporting of that event for other drugs. For this study, we used the proportional reporting ratio (PRR) for signal

detection, which method is also used by the European Medicines Agency (EMA) [25]. Drug-event combinations for

which at least three reports have been received, and for which the lower bound of the 95% confidence interval (95%

CI) for the PRR is at least 1, are considered toImpact represent of exposure signals misclassification of disproportionate79 reporting (i.e. safety signals):

� � � � ��

�� Test cases for data simulation Three test cases with product-specific ADRs were selected for this study, for which the characteristics of the drug-event combination were estimated from literature sources (see Box 1). Each of the test cases contained one unique ADR (event) that has a true association

(RRi,j > 1)84 with one product, but no association (RRi,j = 1) with any of the similar products. In two cases the background incidence of this unique event was low (< 10/100,000 patient-years) and the relative risk high (RR ≥ 5), whereas in the third case the background incidence was high (5,000/100,000 patient-years) and the relative risk low (RR = 1.5). The combined incidence of all other ADRs was equal for all products. The reporting probability was assumed non-differential, and 0.1 for any drug-event combination. The full list of model parameters assumptions for the test cases is provided in Table 1.

Box 1. Background information on test cases. 3 Case 1: Hypersensitivity reactions to infliximab Hypersensitivity reactions (HSR), including infusion-related reactions, are a common ADR of infliximab. In recent studies between 3% and 10% of the patients experienced HSR during infliximab treatment [35]. The risk was higher among patients who had developed antibodies to infliximab, and numerically (albeit not statistically significant) differed between similar infliximab-containing products, with a factor 1.25- to 3-fold across studies. Though the differences in incidence of HSR may have been a chance finding, potential differences in immunogenicity between products, which is known to be associated with an increased risk of infusion-related reactions [36], may have contributed to the observed differences. Case 2: Interferon beta-induced thrombotic microangiopathya Thrombotic microangiopathies (TMA) comprise a diverse group of severe microvascular occlusive disorders associated with haemolytic anaemia, thrombocytopenia, and organ injury [37]. A recent case series described the unexpected occurrence of TMA among multiple sclerosis (MS) patients treated with Rebif (interferon beta-1a) [19]. Because the increase in TMA cases (no data on relative risk available) coincided with the introduction of a new formulation of Rebif, the risk has been suspected to relate to changes in manufacturing conditions. The cases of Rebif-associated TMA predominantly comprised thrombotic thrombocytopenia purpura (TTP) and hemolytic-uremic syndrome (HUS). Both disorders are extremely rare, and the combined population incidence has been estimated to be between 2.2 and 11.3 per 1,000,000 person-years [38,39]. though the background incidence in MS patients is unknown Case 3: Epoetin alfa-induced pure red cell aplasia Pure red cell aplasia (PRCA) is a rare condition of profound anaemia, which may occur as a result of anti-erythropoietin antibodies, secondary to treatment with recombinant human erythropoietin (epoetin). As exemplified by previous incidents, the risk is highly dependent on the manufacturing and formulation process, and may accordingly vary between different formulations of epoetin [7,40]. Recent data from prospective registries showed that the background incidence of PRCA among patients exposed to epoetin alfa is between 14.0 and 35.8 per 100,000 patient-years [41] The incidence may however increase up to 17-fold, as estimated for the post-manufacturing change formulation for Eprex [42]. 80 Chapter 3.2

Evaluation of the impact of exposure misclassification

Direction of exposure misclassification In this study, we assumed misclassification of drug exposure to be non-differential of the outcome (i.e. irrespective of the type of ADR), and to occur in one direction only. That is, a varying proportion of all events for the product associated with the unique risks was misattributed to any of the similar products.

Effect of exposure misclassification in test cases For the three test cases, we calculated the number of cases and years to detection of the unique risk, along various levels (0 - 50%) of exposure misclassification. For this, we calculated the required patient exposure to generate a safety signal in the spontaneous reporting system for each test case in the situation of no misclassification, and compared this to the situation in which exposure information was misattributed. As described above, two conditions should be met to generate a safety signal: [I] the lower-bound of the 95-% CI of the PRR should be at least 1, and [II] at least 3 reports should be available for the drug-event combination. The required patient exposure to generate a safety signal was calculated by combining the two conditions with the equation for the expected number of spontaneous ADR reports for a

given drug-event combination (Ni,j), as further explained in the Supplementary material.

Overall impact assessment of exposure misclassification The overall impact of exposure misclassification was studied in general terms by varying the model parameters (relative risk, product market share) as described in Table 1, and assessing the effect of misclassification (0 – 99%) in relative terms. A number of model parameters (total patient exposure, background incidence unique event) were not included in this evaluation, because these parameters don’t impact on the relative effect.

Table 1. Parameter values for data simulations. Test cases TMA – interferon PRCA – epoetin Overall impact Parameter HSR – infliximab beta alfa assessment Background incidence unique 5,000 10 25 1:1001 event (per 100,000 patient-years) Relative risk for unique event 1.5 5 17 [1.5; 3; 5; 10; 15] Incidence all other events (per 10,000 10,000 10,000 100:11 100,000 patient-years) Reporting probability 0.1 0.1 0.1 n/a Total market (patients per year) 50,000 50,000 100,000 n/a Market share product with 50% 50% 50% [5%; 50%; 95%] unique risk PRCA, pure red cell aplasia; TMA, thrombotic microangiopathy; HSR, hypersensitivity reactions. 1The incidence rate ratio between the unique event and all other events is 1:100. Impact of exposure misclassification 81

Results

The analyses on the three test cases gave insight in the range of absolute effects of misclassification that can be expected in real-life signal detection, whereas the overall impact assessment shows which parameters are most affected by exposure misclassification. For the case of hypersensitivity reactions to infliximab, which assumes a high background incidence but weak association with the product of interest (RR=1.5, see Table 1), we observed a large impact of exposure misclassification. As shown inFigure 3 , the product- specific increased incidence is detectable through signal detection after the occurrence of 956 cases (of which 10% is assumed to be reported), which takes 0.5 years, in the situation where all ADR reports are attributed to the correct product (no misclassification scenario). In a scenario of 20% exposure misclassification, an additional 478 cases are required to detect the product-specific risk, taking an additional 0.25 years. This corresponds to a 49% increase in time and cases as compared to situation with no misclassification. A doubling in cases and time to detection of the product-specific risk was observed in the scenario of 34% misclassification. 3

3,000 1.6 n

o 1.4 i t 2,500 c e t

e 1.2 d

n o o t 2,000 i

s 1 c e e t s e a d c

1,500 0.8 f o t o

s r r e a

b 0.6 e

m 1,000 Y u n

l 0.4 a u

t 500 c 0% misclassification* 0.2 A

0 0 0% 10% 20% 30% 40% 50% Misclassification

Figure 3. Effect of exposure misclassification on the ability to detect a product-specific risk for hypersensitivity reactions to infliximab.1 1See Table 1 for model parameter assumptions. *In absence of exposure misclassification, it takes 956 cases/0.5 years (assuming 90% underreporting) to detect the product-specific risk.

As shown in Figure 4, a smaller impact of exposure misclassification was observed for the case of thrombotic microangiopathy with interferon beta, which assumed a low background incidence but relatively strong association (RR=5) with the product of interest. In the scenario of 20% misclassification, an additional 30 cases are required, taking an additional 2.4 years to detect the product-specific risk. This corresponds to a 34% increase in cases and 82 Chapter 3.2

250 20

n 18 o i t c

e 200 16 t e d

14 n o o t i

t s c

e 150 12 e t s e a d c

10 f o t o

s r r

e 100 8 a b e m Y

u 6 n

a 50 4 u t

c 0% misclassification*

A 2

0 0 0% 10% 20% 30% 40% 50% Misclassification

Figure 4. Effect of exposure misclassification on the ability to detect a product-specific risk for interferon beta-induced thrombotic microangiopathy.1 1See Table 1 for model parameter assumptions. *In absence of exposure misclassification, it takes 89 cases/7.1 years (assuming 90% underreporting) to detect the product-specific risk.

time as compared to situation with no misclassification. A doubling in cases and time to detect the product-specific risk was observed in the situation of 40% misclassification. For the case of epoetin alfa-induced PRCA, which assumes a low background incidence and very strong association (RR=17) with the product of interest, we observed only a modest impact of exposure misclassification. As shown in Figure 5, up to 22% misclassification will not result in a delayed identification of the safety signal, but, in contrast, in an earlier identification. This finding may be explained by the fact that low levels of exposure misclas-

0.8 160 n

o 0.7 i t

c 140 e t

e 0.6 d

120 n o o t i

s 0.5 c e 100 e t s e a d c

0.4 f o

80 t o

s r r e a

b 0.3

60 e m Y u n

l 40 0.2 a u t c 0% misclassification* 0.1

A 20

0 0 0% 10% 20% 30% 40% 50% Misclassification

Figure 5. Effect of exposure misclassification on the ability to detect a product-specific risk for epoetin alfa-induced pure red cell aplasia (PRCA).1 1See Table 1 for model parameter assumptions. *In absence of exposure misclassification, it takes 88 cases/0.4 years (assuming 90% underreporting) to detect the product-specific risk. Impact of exposure misclassification 83 sification ensure that sufficient data is available in the reference category (i.e. PRCA cases for other products) for the disproportionality measure to reach statistical significance. It should, however, be noted that the detection of product-specific risks is in similar scenarios, in which an event with a low background incidence has a (very) strong association with one product, is mainly determined by the required time to have at least one case in the reference

Figure 6A. 5% market share 10000% RR=1.5 RR=3 RR=5 RR=10 RR=15

s 1000% m r e t

e g a t e c r e p

100%

i 0% 10% 20% 30% 40% 50% 60% 70% 80% 90%

y a l e D

10%

1% Misclassification 3

Figure 6B. 50% market share 10000% RR=1.5 RR=3 RR=5 RR=10 RR=15

s 1000% m r e t

e g a t e c r e p

100%

i 0% 10% 20% 30% 40% 50% 60% 70% 80% 90%

y a l e D

10%

1% Misclassification

Figure 6C. 95% market share 10000% RR=1.5 RR=3 RR=5 RR=10 RR=15

s 1000% m r e t

e g a t e c r e p

100%

i 0% 10% 20% 30% 40% 50% 60% 70% 80% 90%

y a l e D

10%

1% Misclassification

Figure 6. Delay in identification of product-specific safety signals (in percentage terms) through exposure misclassification, stratified by market share and relative risk. 84 Chapter 3.2

category. This was, however, not a condition in our model in which continuous variables (i.e. also values smaller than 1) were allowed, and the actual impact of exposure misclassification may therefore even be smaller within similar scenarios. In relative terms, the effect of exposure misclassification was found to be mostly dependent on the market share of the product, and to a lesser extent of the relative risk of the unique risk. As shown in Figure 6, for products with a low market share, 50% misclassification will result in an approximate 100% delay (i.e. doubling in cases and time) to recognition of new safety signals, largely irrespective of the relative risk of the event (Figure 6A). By contrast, for products with a high market share, 5% misclassification will already result in a doubling in cases and time to detect new safety signals with a low relative risk (Figure 6B). As shown in the situation for products with a 50% market share (Figure 6C), the relative delay increases with lower relative risks.

Discussion

This study shows the effect of exposure misclassification (i.e. incorrect exposure attribution) in spontaneous ADR reports on the time to detection of product-specific risks in spontaneous reporting systems. More specifically, we focussed on situations where product-specific signal detection is performed, and an ADR associated with one particular product (e.g. a biosimilar) is incorrectly attributed to another product containing the same active substance (e.g. the innovator), or vice versa. We found that exposure misclassification results in the largest delay in the identification of risks that have a relative weak association with the product of interest, and in situations in which the product associated with the unique ADR has a large market share. By contrast, the detection of strong drug-event associations was found to be relatively robust to low levels of exposure misclassification. The absolute public health impact of exposure misclassification, in additional time and cases to detection of the product-specific risk, is highly dependent on the characteristics of the drug-event combination (patient exposure, background incidence event, etc.), and should therefore be assessed on a case-by-case basis. The finding that the identification of strong drug-event associations is particularly robust to the effects of exposure misclassification is important. Spontaneous reports are highly effective in detecting events that are strongly associated with a certain exposure, but have a low background incidence in the treatment population. Well-known examples hereof include the risk of progressive multifocal leukoencephalopathy (PML) with natalizumab [26], rhabdomyolysis with cerivastatin [27], and the occurrence of congenital anomalies with thalidomide [28]. Events that are, by contrast, only weakly associated with a certain exposure but occur with a relative high background rate are particularly difficult to detect. From the reporter’s perspective, it may be difficult to recognize commonly occurring events (e.g. car- Impact of exposure misclassification 85 diovascular disease) as drug-induced, particularly when the drug only imparts a small risk [29], Moreover, as previously demonstrated [23], only strong associations may sufficiently “compensate” for the variability in reporting to generate a signal in a spontaneous reporting system. For the detection of weak drug-event associations, which nevertheless may have a significant public health impact (e.g. cardiovascular events with rofecoxib [30]), one should therefore resort to other methods, and the observed effect of exposure misclassification in spontaneous reporting may be less relevant. A second finding from our study was that the product market share largely determines the relative impact of exposure misclassification. While a doubling in time and cases to detection was observed when 50% of the ADR for products with a low (5%) market share were misattributed, a similar effect was already observed when 5-10% of the ADR reports for products with a high (95%) market share were misattributed. The finding that products with a low market share are more robust to exposure misclassification is important, as it may be expected that the potential for misclassification will be higher for these products. For example, when a new product is gradually taken up in clinical practice, healthcare professionals and patients may initially not be familiar with the new product, and therefore more likely to incorrectly attribute the ADR to the innovator product. 3 In this study we focused on the occurrence of product-specific ADRs among biologicals, and explored the effect of exposure misclassification in product-specific signal detection procedures. Our findings may, however, also apply to evaluations for other manufacturer- specific risks. Previous studies have, for example, shown that quantitative signal detection methods could also be used to study potential formulation-specific (e.g. haemolytic events with liquid and lyophilized formulations of immunoglobulin [31]) and batch-specific ADRs (e.g. local reactions and fever with different batches of pandemic influenza vaccines [32]). Misclassification on the batch or formulation level will in such analysis resemble the effects observed in the present simulation study. It is important to note that quantitative signal detection methods are not the only strategy through which product-specific safety signals may be identified in spontaneous reporting systems. Historically, the detection of safety signals has been based on the manual review of every ADR report sent to a spontaneous reporting system. Such case-by-case evaluations are still routinely carried out, and play an important role in the identification of new ADRs [33]. It should also be noted that quantitative signal detection methods are only used to highlight potential signals for further (manual) review. For the actual confirmation of the safety signal a careful review of the individual case reports is required, including a clinical assessment of the strength and likelihood of the causal association. For such case-by-case evaluations of ADR reports it will be important to have reliable data on the product-specific exposure, and the impact of misattributed reports has not been evaluated. In this study we made several assumptions for the data simulation procedure, which should be taken into account when interpreting the results. First, we assumed the ADR 86 Chapter 3.2

reporting patterns to be similar across similar biological products, since no differences in reporting probabilities can be expected when reporters are unaware of the product-specific exposure status. The actual reporting pattern may in practice, however, very well differ between similar biological products, as shown by a recent analysis of the Italian pharmacovigilance database [12]. Patients and health professionals may particularly be triggered to report ADRs for novel biosimilars, as these products will be under increased scrutiny, which could lead to the generation of false positive product-specific signals. A second point to consider is that we did not include the underlying discrete probability distribution for the expected number of ADR reports in our model. As shown in a previous simulation study [23], the expected variance in reporting may result in the reporting of non-causal associations (i.e. false positive signals). Thirdly, we assumed the misclassification to occur exclusively in one direction. Exposure information may instead also be misclassified in two directions, or reported on active substance level rather than product level. The latter will resemble the effect of underreporting, for which the impact has been evaluated elsewhere [34]. In conclusion, the present study shows the direction and magnitude of the effect of incorrect exposure attribution in spontaneous ADR reports on the time to detection of product-specific safety signals. The largest effect was observed in the detection of weak drug- event associations, although the absolute public health impact of exposure misclassification was highly dependent on the characteristics of the drug-event combination, such as the patient exposure, and background incidence of the event. With the increasing availability and use of biologicals, including biosimilars, product-specific safety evaluations will become increasingly important in the near future. When product-specific signal detection will be implemented in the upcoming years and methods are refined for use in daily practice, it is important to keep in mind the challenges that we identified.

Acknowledgements

We wish to thank Dr. Marloes Bazelier (Utrecht University, the Netherlands) for her expert advice and help with the mathematical equations.

Supplementary material

Calculation of required patient exposure to generate a signal of disproportionate reporting:

Signal detection from a spontaneous reporting system database is based on a cross-tabulation of all drug-event combinations ( ) Supplementary material that have been reported at least once, as illustrated below: ��,� Supplementary material Impact of exposure misclassification 87 FigCalculation 2A. Spontaneous of required patientreporting exposure system to generatedatabase a signalFig of disproportionate 2B. Two-by-two reporting: cross-tabulation for signal detection

CalculationSignalSupplemen detection of required from patient ta aryspontaneous exposure mat reportinger to generateial system a signal database of disproportionate is based on a cross-tabulation reporting: of all drug-event combinations ( ) that have been reported at least once, as illustrated below: ��,� Signal detectionDrug1 from a Drug spontaneous2 …reporting systemDrug database is based on a cross-tabulation of all drug-event combinations ( ) that haveCalculation been reported of required at least once, patient as illustrated exposure below: toj generate a signal of disproportionate reporting: ��,� Fig 2A. Spontaneous reporting system database Fig 2B. Two-by-two cross-tabulation for signal detection Signal detection from a spontaneous reporting system database is based on a cross-tabulationDrug1 of all drug-eventNot drug1 Event1 …

Figcombinations 2A. Spontaneous (Ni,j) that reporting have been system reported database at least once, asFig illustrated 2B. Two-by-two below: cross-tabulation for signal detection �,� ��,� ��,� � Drug1 Drug2 … Drug Event1 A C Event2 Spontaneous reporting… system database Two-by-two cross-tabulation for signal detection j �,� �,� �,� Drug�,� 1 Drug�,� 2 … Drug�,� � �� … � � � � � � Drug1 Not drug1 Event1 … Not event1 B D … … … … j Drug1 Drug2 … Drug �,� �,� �,� Drug�,�1 �,� Not drug�,�1 �,� Event � � … �j �� … � � � 1 Event1 A �� … � � C� Event2 … … Drug1 Not drug1 Event Event1 …… �,� �,� �,� �,� �,� �,� � �,� � �,� � �,� � �� … � � � Event � � … � Event1 A C i 2 �,� �,� �,� Not event1 B D … … … … � Event1 A C Event� 2 � … �,� �,� �,� �,� �,� �,� � �� … � � � � � � ��,�� … � ��,�� ,� �,� Not event1 B �� … � � �D … … … … … B Event … … … …… Not event1 D �,� �,� �,� �,� In Figure 2, A, B, C and D are defined by the following equations: �� … � � � �� … � � � i �,� �,� … �,� Event Event � � …… �

i �,� �,� �,� � � � In Figure� � � 2 �, �A, � �� B, C � and� D are defined by the following equations:

In Figure In� Figure 2 �2,� A, � �B,,� CA, � and �B,� C D and are D defined are defined by the by following the following equations: equations: � � �� 3 � �� In which equations:�� E= patient exposure to drug1 In which� �equations:�� I= background incidence of event1 In which equations:�� E= patient exposure to drug1 RR=In which relative equations: risk for event1 among patient exposed to drug1 I= background incidence of event1 E= patient exposure to drug1 p= reportingE= patient probability exposure [note: to assumed drug1 non-differential in this study, and therefore not taken into account below] RR= relative risk for event1 among patient exposed to drug1 I= backgroundI= background incidence incidence of event1 of event1 y = proportionalRR= relative incidence risk for ofevent all other1 among events patient (≠event exposed1) to drug1 p= reporting probability [note: assumed non-differential in this study, and therefore not taken into account below] RR=p= relative reporting risk for eventprobability1 among [note: patient assumed exposed non-differential to drug1 in this study, and therefore not taken into ac- x = proportional patient exposure to all other drugs (≠drug1) p= reportingy = proportionalcount probability below] incidence [note: ofassumed all other non-differential events (≠event in1) this study, and therefore not taken into account below]

y = proportional incidence of all other events (≠event1) x = proportional patient exposure to all other drugs (≠drug1) y = proportional incidence of all other events (≠event1) The actual numberx = proportional of reports, considering patient exposure the extentto all other of misclassification drugs (≠drug1) is defined by:

x = proportional patient exposure to all other drugs (≠drug1) The actual number of reports, considering the extent of misclassification is defined by: The actual number of reports, considering the extent of misclassification φ, φ, is defined by: The actual number of reports, considering the extent of misclassification is φ, defined by: �� φ, �� The required�� patient exposure to generate a safety signal can be calculated by combining the above equations with the two TheThe �� required �� required patient patient exposure exposure to to generate generate a a safetysafety signal can can be be calculated calculated by bycombining combining the the above above equations equations with the two conditionswith for the the two a signal conditions of disproportionate for the a signal of reporting: disproportionate reporting: conditions for the a signal of disproportionate reporting: The required�� patient exposure to generate a safety signal can be calculated by combining the above equations with the two (I) The lower bound of the 95%-confidence interval (95% CI) of the PRR should be at least 1: conditions(I) for(I) the The a signal Thelower oflower disproportionatebound bound of the of 95%-confidencethe reporting:95%-confidence interval interval (95% (95% CI) CI)of theof the PRR PRR should should be at be least at least 1: 1:

(I) The lower bound of the 95%-confidence interval (95% CI) of the PRR should be at least 1:

�� (II) At least three reports should be available for the drug-event combination (A; drug1 - event1) (II) At least three reports�� should � � � be � available for the drug-event combination (A; drug1 - event1) �� (II)Exposure At misclassfication least three reports simulation�� should study � be � available for the drug-event combination (A; drug1 - event1) 15

ExposureExposure misclassfication misclassfication simulation simulation study study – supplementary– supplementary material material page 95/2page 95/2

Exposure misclassfication simulation study – supplementary material page 95/2 88 Chapter 3.2

(II) At least three reports should be available for the drug-event combination (A; drug1 - event1)

Condition I: Lower boundCondition of 95% CII: Lowerof PRR = 1 bound of 95% CI of PRR = 1 Condition I: Lower bound of 95% CI of PRR = 1

�� .�� .�� In this equation,�� we inserted�� the�� above�� equations for A” to�� D”,�� and�� rewrote�� it to an equation in which the re- In� �this � equation, we inserted theIn� above �this � equation,equations for we A” inserted to D”, and the rewrote above it to equationsan equation forin which A” to the D”, required and rewrote patient exposure it to an equation in which the required patient exposure quired patient exposure for drug1 (E) to generate a safety signal is described as a function of the above described parametersfor drug1 (E) includedto generate in aequations safetyfor signal drug A”-D”: is1 (E)described to generate as a function a safety of the signal above describedis described parameters as a function included inof equations the above A”-D”: described parameters included in equations A”-D”:

�� .�� .�� 1 � � ��

� � � 1 �� .�� ⟺ � � � � � � �� .�� ⟺ �.�� 1��1�� ⟺ � � ��1�� .�� 1��1�� 1��1��⟺ �.�� ⟺ ��1�� 1�� .�� 1��1��.�� ⟺ ⟺ � � � � ��1��

� � � � � � � � � � � � � � � � � � � � �.�� .�� ⟺ � � � � � �� ⟺ � � � � �� .�� ⟺ � � �.�� ⟺ � � � � � � � � � � � � � � � � � � � � �� 1 1 1 1 � � ⟺ �� 1.96� � � � � � � � � � � �� .�� ⟺ � � � � �� 1 1 1 1 ⟺ � � � � � 1.96 � �� 1 1 1 1 � � � � ⟺ �� 1.96� � � � � � �� 1 1 1 1 ⟺ � � � �� 1.96 �� 1 1 1 1 �� ⟺ �� 1 1 1 1 ⟺ � � �1.96 � � �� 1.96 � � � � � �1 � �� 1 � �� 1�� ⟺ � � � �� 1 1 1 1 1.96 ⟺ � � � � � 1 1 11.96 1 �� 1 � �� 1 � � ��1�� 1 1 1 1 � � �� ⟺ �� 1.96 � � � � � �1 � �� 1 � �� 1�� ⟺ � � � 1.96 � � � � �� 1 �� 1 � 1� �� 1 � � � � �� ⟺� � � � �1 � �� 1 � �� 1�� 1.96 1 1 1 1 � � � � � �. �� 1�� 1 � �� 1�� ⟺ � � � � � � � � � � � �� ⟺� � �� 1.96 1 1 1 1 Condition II: At least 3 ADR reports for the drug-event� combination� A � � �� 1 � �� 1 � �� 1��

� �This � ��equation can be rewritten as follows: �. �� ⟺ � � � � � � � � � � � �� Condition II: At least 3 ADR reports for the drug-event combination A

�This � ��equation can be rewritten as follows:

� � � � � � ��

Condition I: Lower bound of 95% CI of PRR = 1

�� .�� In� �this � equation, we inserted the above equations for A” to D”, and rewrote it to an equation in which the required patient exposure

for drug1 (E) to generate a safety signal is described as a function of the above described parameters included in equations A”-D”:

�� .�� 1 � �

� � � � �.�� ⟺ � � ��

� � � � �.�� 1��1�� ⟺ � � ��1��

� � � � �.�� ⟺ � � ��

� � � � � � � � � � � � � � � � � � � � �.�� ⟺ � � � � � � � � � � � � � � � � � � ��

� � � � � � �.�� ⟺ � � � � � � ��

� � � � �� 1 1 1 1 ⟺ �� 1.96� � � � � � ��

� � � � � � �� 1 1 1 1 ⟺ � � � � � 1.96 ��

� � � � � �� 1 1 1 1 ⟺ � � � � � 1.96 �� 1 1 1 1 ⟺ � � � � � 1.96 � � � � � �1 � �� 1 � �� 1�� ⟺ � 1.96 1 1 1 1 � � � � �� 1 � �� 1 � �� 1��

� � � � � �� ⟺ � � � 1.96 1 1 1 1 � � � � �� 1 � �� 1 � �� 1��

� �. �� . �� ⟺ � � � � � � � � � � �� Impact of exposure misclassification 89 � � � � ��

ConditionCondition II: II: At At least least 3 ADR 3 ADR reports reports for the drug-eventfor the drug-event combination combination A A

This�This � equation��equation can bebe rewrittenrewritten as as follows: follows:

� � � � � � ��

3 90 Chapter 3.2

References

1. Flegal, K. M., Keyl, P. M. & Nieto, F. J. Differential the Accuracy of Product Name Attribution of Misclassification Arising from Nondifferential US-Sourced Adverse Event Reports in MedWatch Errors in Exposure Measurement. Am J Epide- of Somatropins and Insulins. Ther Innov Regul miol 1991; 134: 1233-1246. Sci 2015 [e-pub ahead of print]. 2. van Staa, T., Abenhaim, L. & Leufkens, H. A study 12. Cutroneo, P. M. et al. Safety profile of biological of the effects of exposure misclassification due to medicines as compared with non-biologicals: the time-window design in pharmacoepidemio- an analysis of the italian spontaneous reporting logic studies. J Clin Epidemiol 1994; 47: 183-189. system database. Drug Saf 2014; 37: 961-970. 3. Ilkhanoff, L., Lewis, J. D., Hennessy, S., Berlin, 13. Bohn, J. et al. Patterns in spontaneous adverse J. A. & Kimmel, S. E. Potential limitations of event reporting among branded and generic electronic database studies of prescription non- antiepileptic drugs. Clin Pharmacol Ther 2015; aspirin non-steroidal anti-inflammatory drugs 97: 508-517. (NANSAIDs) and risk of myocardial infarction 14. Grampp, G. et al. Active and passive surveillance (MI). Pharmacoepidemiol Drug Saf 2005; 14: of enoxaparin generics: a case study relevant 513-522. to biosimilars. Expert Opin Drug Saf 2015; 14: 4. Carson, K. R. et al. Progressive multifocal 349-360. leukoencephalopathy after rituximab therapy 15. Kozlowski, S., Woodcock, J., Midthun, K. & in HIV-negative patients: a report of 57 cases Behrman Sherman, R. Developing the Nation’s from the Research on Adverse Drug Events and Biosimilars Program. N Engl J Med 2011; 365: Reports project. Blood 2009; 113: 4834-4840. 385-388. 5. Bongartz, T. et al. Anti-TNF antibody therapy 16. Schneider, C. K. et al. In support of the European in rheumatoid arthritis and the risk of serious Union biosimilar framework. Nat Biotechnol infections and malignancies: systematic review 2012; 30: 745-8; author reply 748-9. and meta-analysis of rare harmful effects in 17. Generics and Biosimilars Initiative (GaBI). 29 randomized controlled trials. JAMA 2006; 295: June 2012. AUS$67 billion worth of biosimilar 2275-2285. patents expiring before 2020. Available from: 6. Day, T. K. Intestinal perforation associated with http://www.gabionline.net/Biosimilars/ osmotic slow release indomethacin capsules. BMJ General/US-67-billion-worth-of-biosimilar- 1983; 287: 1671-1672. patents-expiring-before-2020 [Last accessed: 7. Casadevall, N. et al. Pure red-cell aplasia and 26 February 2015]. antierythropoietin antibodies in patients treated 18. Funk, M. B. et al. Thromboembolic events as- with recombinant erythropoietin. N Engl J Med sociated with immunoglobulin treatment. Vox 2002; 346: 469-475. Sang 2013; 105: 54-64. 8. Vermeer, N. S. et al. Traceability of biologicals: 19. Hunt, D. et al. Thrombotic Microangiopathy present challenges in pharmacovigilance. Expert Associated with Interferon Beta. N Engl J Med Opin Drug Saf 2015; 14: 63-72. 2014; 370: 1270-1271. 9. Getz, K. A., Stergiopoulos, S. & Kaitin, K. I. 20. van Puijenbroek, E. P. et al. A comparison of Evaluating the Completeness and Accuracy of measures of disproportionality for signal detec- MedWatch Data. Am J Ther 2012; 21: 442-446. tion in spontaneous reporting systems for adverse 10. Vermeer, N. S. et al. Traceability of biopharma- drug reactions. Pharmacoepidemiol. Drug Saf ceuticals in spontaneous reporting systems: a 2002; 11: 3-10. cross-sectional study in the FDA Adverse Event 21. Bate, A. & Evans, S. J. W. Quantitative signal Reporting System (FAERS) and EudraVigilance detection using spontaneous ADR reporting. databases. Drug Saf 2013; 36: 617-625. Pharmacoepidemiol. Drug Saf 2009; 18: 427-436. 11. Stergiopoulos, S., Brown, C. A., Grampp, G., Felix, T. & Getz, K. A. Identifying and Quantifying Impact of exposure misclassification 91

22. Finney, D. J. Statisical logic in the monitoring of 33. Hauben, M. & van Puijenbroek, E. P. Evaluation reactions to therapeutic drugs. Methods Inf Med of suspected adverse drug reactions. JAMA 2005; 1971; 10: 237-245. 293: 1324; author reply 1324-5. 23. Tubert, P., Bégaud, B., Péré, J., Haramburu, F. & 34. van der Heijden, P. G. M., van Puijenbroek, E. Lellouch, J. Power and weakness of spontane- P., van Buuren, S. & van der Hofstede, J. W. On ous reporting: A probabilistic approach. J Clin the assessment of adverse drug reactions from Epidemiol 1992; 45: 283-286. spontaneous reporting systems: the influence of 24. Roux, E., Thiessard, F., Fourrier, A., Begaud, under-reporting on odds ratios. Stat Med 2002; B. & Tubert-Bitter, P. Evaluation of statistical 21: 2027-2044. association measures for the automatic signal 35. European Medicines Agency. Assessment report generation in pharmacovigilance. Information Inflectra. 27 June 2013. Available from: http:// Technology in Biomedicine, IEEE 2005; 518-527. www.ema.europa.eu/docs/en_GB/document_li- 25. EudraVigilance Expert Working Group (EV- brary/EPAR_-_Public_assessment_report/hu- EWG). Guideline on the use of statistical signal man/002778/WC500151490.pdf [Last accessed: detection methods in the EudraVigilance data 2 March 2015]. analysis system. Available from: http://www. 36. Baert, F. et al. Influence of Immunogenicity on ema.europa.eu/docs/en_GB/document_library/ the Long-Term Efficacy of Infliximab in Crohn’s Regulatory_and_procedural_guideline/2009/11/ Disease. N Engl J Med 2003; 348: 601-608. WC500011437.pdf [Last accessed: 27 May 2015]. 37. George, J. N. & Nester, C. M. Syndromes of 26. Bloomgren, G. et al. Risk of natalizumab-associ- Thrombotic Microangiopathy. N Engl J Med ated progressive multifocal leukoencephalopathy. 2014; 371: 654-666. 3 N Engl J Med 2012; 366: 1870-1880. 38. Terrell, D. R. et al. The incidence of thrombotic 27. Staffa, J. A., Chang, J. & Green, L. Cerivastatin thrombocytopenic purpura-hemolytic uremic and Reports of Fatal Rhabdomyolysis. N Engl J syndrome: all patients, idiopathic patients, and Med 2002; 346: 539-540. patients with severe ADAMTS-13 deficiency. J 28. Mcbride, W. G. Thalidomide and congenital Thromb Haemost 2005; 3: 1432-1436. abnormalities. Lancet 1961; 278: 1358. 39. Miller, D. P. et al. Incidence of thrombotic 29. Frantz, S. Drug safety special: Chasing shadows. thrombocytopenic purpura/hemolytic uremic Nature 2005; 434: 557-558. syndrome. Epidemiology 2004; 15: 208-215. 30. Bresalier, R. S. et al. Cardiovascular events as- 40. Seidl, A. et al. Tungsten-Induced Denaturation sociated with rofecoxib in a colorectal adenoma and Aggregation of Epoetin Alfa During Primary chemoprevention trial. N Engl J Med 2005; 352: Packaging as a Cause of Immunogenicity. Pharm 1092-1102. Res 2012; 29: 1454-1467. 31. Berg, R., Jacob, D. & Fuellenhals, E. Hemolytic 41. Macdougall, I. C. et al. Incidence of erythropoi- events after the administration of lyophilized etin antibody-mediated pure red cell aplasia: versus liquid immune globulin: an analysis of a the Prospective Immunogenicity Surveillance single manufacturer’s safety database. Transfusion Registry (PRIMS). Nephrol Dial Transplantat 2015;55(8):1847-54. 2015; 30: 451-460. 32. van Puijenbroek, E. P., Broos, N. & van 42. Boven, K. et al. The increased incidence of pure Grootheest, K. Monitoring adverse events of red cell aplasia with an Eprex formulation in the vaccination campaign against influenza A uncoated rubber stopper syringes. Kidney Int (H1N1) in the Netherlands. Drug Saf 2010; 33: 2005; 67: 2346-2353. 1097-1108.

Traceability of biologicals: present 3.3 challenges in pharmacovigilance Vermeer NS, Spierings I, Mantel-Teeuwisse AK, Straus SMJM, Giezen TJ, Leufkens HGM, Egberts ACG, De Bruin ML

Expert Opinion on Drug Safety 2015; 14: 63-72

Abstract

Introduction Adequate recording of drug exposure information in clinical practice, and ensuring the correct transfer of exposure data to pharmacovigilance data sources is essential to pharmacovigilance. For biological medicinal products, batch- and product-specific exposure information (i.e. beyond the biological substance) should be recorded, as differences in the safety profile could emerge within one product (from batch-to-batch) over time, or between products containing the same active substance. With the expected expansion of the biosimilar market, there have been concerns about the ability to trace individual products within pharmacovigilance databases. Areas covered We discuss the present challenges in the traceability of biologicals in relation to pharmacovigilance, by exploring the processes involved in ensuring their traceability. We explore both the existing systems that are in place for the recording of exposure information in clinical practice, as well as the critical steps involved in the transfer of exposure data to various pharmacovigilance databases. Expert opinion The existing systems ensure the traceability of biologicals down to the manufacturer within pharmacy records, but do not support the routine recording of batch information. Expected changes in supply chain standards provide opportunities to systematically record detailed exposure information. Spontaneous reporting systems are the most vulnerable link in ensuring traceability, due to the manual nature of data transfer. Efforts to improve the traceability should, in the short term, be focused towards encouraging health professionals and patients to systematically record and report detailed exposure information, though long-term solutions lie in expanding the accessibility to, and increasing the electronic exchange of exposure data. 94 Chapter 3.3

Introduction

In 2002, a case series reported the unexpected occurrence of a rare form of anaemia, pure red-cell aplasia (PRCA), in thirteen patients treated with recombinant human erythropoietin (epoetin) [1]. Some of the patients had been receiving epoetin treatment for years, yet had only recently developed PRCA. At the time, the pharmacovigilance community had observed a similar unexpected surge in the reporting of epoetin-associated PRCA, and was up to the challenging task of addressing the emerging risk. In due course, it was found that most cases could be linked to a common manufacturing source of epoetin for which formulation changes had been issued in 1998 (Eprex, Ortho Biotech), upon which the risk could be successfully countered [2]. The sudden and dramatic rise in the risk of epoetin-associated PRCA has become exemplary for the potential for variability in the safety profile of biological products (see definition in Table 1). Due to the complex nature and manufacturing process of biologicals, small deviations in their manufacturing or formulation might impact the safety profile of the

Table 1. Definitions of biological medicinal products. Term Description Example, human growth hormone Biological “A substance that is produced by or extracted from a biological Recombinant human substance source and that needs for its characterisation and the growth hormone, determination of its quality a combination of physicochemical- somatropin biological testing, together with the production process and its control” [51]. Biological substances include substances derived from recombinant DNA technology or other biotechnological processes, substances derived from human blood or plasma, advanced therapy products (gene or somatic cell therapy), and immunological products (vaccines, toxins, serums or allergen products). Biological A medicinal product for which the active substance is a Genotropin (somatropin, medicinal biological substance [51]. Pfizer) product, or biological Similar A biological product containing the same biological Omnitrope (somatropin, biological substance as an already licensed biological product (i.e. the Sandoz), which product, or reference product), and which has demonstrated to be similar demonstrated similarity to biosimilar to the reference product in terms of quality characteristics, Genotropin biological activity, safety and efficacy [52,53]. Related A biological product containing the same biological Somatropin biological substance as an already licensed biological product, and Biopartners (somatropin, product which has been developed on the basis of an independent full Biopartners), a prolonged- application procedure, involving an independent benefit- release formulation of risk assessment, not necessarily with the intention to show somatropin similarity to any reference biological product. Challenges in traceability of biologicals 95 end product [3-5]. These differences in safety profile will most likely relate to immunological events. Also in the particular case of Eprex, antibody formation was found to underlie the PRCA [6]. Deviations in safety profile may not only emerge within products, but also between related and similar products (Table 1) containing the same active biological substance. A recent study, for example, reported unexpected differences in the risk of inhibitor development between second and third generation recombinant factor VIII products [7]. Also, the unexpected increase in cases of thrombotic microangiopathy among patients treated with a particular formulation of interferon beta for which manufacturing changes had been issued (Rebif, Merck) [8], may indicate that the safety profile of this formulation differs from the related product (Avonex, Biogen Idec). The above examples testify to the importance of having a pharmacovigilance system in place that can signal any potential variability between products, and within products, especially after a change in the production process or formulation, over time. For the conduct of product-specific pharmacovigilance, unique identifiers like the trade name need to be available in pharmacovigilance systems (see Figure 1). In addition, batch numbers are required to adequately assess the impact of manufacturing changes over time, or signal any 3 batch-specific issues.

Safety profile across different levels Level identifiers

a. INN 1. Active substance-specific Active substance, e.g. somatropin b. ATC code

Potential for between product variability a. Trade name 2. Product-specific* Manufacturer A Manufacturer B Manufacturer C b. INN** + manufacturer e.g. Genotropin e.g. Omnitrope e.g. Somatropin c. Trade number Biopartners

Potential for within product variability a. Batch number 3. Batch-specific Batch Batch Batch Batch Batch Batch b. Expiry date A1 A2 B1 B2 C1 C2

Figure 1. Potential for variability in the safety profile beyond the level of the active substance of biological products, and required unique identifiers. INN: international non-proprietary name; ATC: anatomical therapeutic chemical. *Products containing the same active biologicals substance comprise either similar biological products (established similarity in terms of quality, efficacy and safety), or related biologicals products (no implied similarity, see Table 1). The provided examples are for illustrative purposes only. ** The international nonproprietary name (INN), or generic name, is a globally recognized name that facilitates the unambiguous identification of active substances. However, since biosimilars and other related biologicals regularly share the same INN, other product identifiers like the trade name are required for the conduct of product-specific pharmacovigilance. 96 Chapter 3.3

With the arrival of the first biosimilars in European clinical practice, and the expected expansion of the global biosimilar market [9], traceability has become increasingly important and become subject of a widespread debate [10-13]. Although a previous study from our group found that over 96% of the biological products for which biosimilars had been available were traceable up to the specific manufacturer in a major European spontaneous reporting system [14], it is unknown to what extent product-traceability is ensured in other pharmacovigilance data sources. In addition, the same study found that individual batches were traceable in less than 25% of the biologicals products, leaving considerate room for improvement. The aim of the present article is to discuss the present challenges in the traceability of biologicals in relation to pharmacovigilance, by exploring the individual processes involved in ensuring their traceability. First, the existing systems that are in place for the recording of detailed exposure information in clinical practice are explored, as these form an important prerequisite to ensure traceability in pharmacovigilance databases. Secondly, the critical steps involved in the transfer of exposure data to various pharmacovigilance databases are explored. The primary focus will be on the clinical practice and pharmacovigilance systems in Europe, yet the principles and systems outlined here will in certain aspects be comparable to other countries and regions in the world.

Recording of exposure information in clinical practice

In the pharmaceutical supply chain from the manufacturer up to the patient, product- and batch-specific exposure information can be recorded during the pharmacy dispensing and/ or during the administration or intake of the drug. This step is essential to maintain a link between the administered biological and the patient, also after discarding of the outer packaging.

Pharmacy dispensing Biologicals are dispensed through a diversity of pharmacy distribution channels, including hospital pharmacies, community and outpatient pharmacies, and specialty pharmacies. During the pharmacy dispensing process, information about the biological is documented in the pharmacy system through scanning of the barcode, which is presented on the outer packaging. These barcodes typically comprise linear barcodes that hold information on the National Trade Item Number (NTIN) of the biological, like the “Pharmazentralnummer” in Germany, the National Drug Code in the USA, or the “Nordisk Varenummer” in Norway, Sweden and Finland [15,16]. Since the NTIN is unique to the manufacturer, dosage form and strength of a product, product-specific exposure information will automatically be recorded. Also in situations where no barcode technology is used to facilitate dispensing, Challenges in traceability of biologicals 97 the NTIN is likely to be recorded at the point of dispensing as this is required to support the back office (e.g. for drug procurement and reimbursement). A survey among hospital and outpatient pharmacies in the Netherlands (see methods in box 1) confirmed that biological products listed in the individual patient’s pharmacy records can always be traced down to the manufacturer. The product batch number should, however, be manually recorded when linear barcodes are used to facilitate dispensing, as these barcodes typically do not include variable product data due to limited capacity. Therefore, batch numbers of biologicals are expected to be infrequently captured in dispensing records. This was confirmed within the pharmacist survey: only 2 out of the 27 pharmacies (14%) stated batch numbers are routinely recorded for biological products. The most frequently cited reasons for not recording this information comprise the lack of specific tools to facilitate the registration (n=17; 63%), but also the perceived absence of any need (n=12; 44%). Interestingly, the survey showed that batch numbers are, on the other hand, routinely recorded for products that are prepared for administration in the hospital pharmacy, which probably relates to good preparation practice. Likewise, batch numbers are also expected to be routinely recorded for blood- and plasma-derived products, due to legal requirements [17]. 3 Several initiatives are currently ongoing, or have recently been implemented, to improve the supply chain traceability. As of 2011, France has been the first EU country to require a data matrix on all pharmaceutical packaging that not only encodes the NTIN, but also the batch number and expiry date [18]. Through barcode-scanning this variable product information will be automatically recorded in the patient’s pharmacy records. Also, three other EU countries (Belgium, Italy and Greece) have recently adopted a serialization requirement, allowing tracing of individual units of a medicinal product (i.e. each different package of the same presentation within the same batch) along a unique serial number [15]. The recently introduced EU legislation on falsified medicines will also ensure supply chain traceability down to the individual package level in other EU countries [19]. As of 2016, barcodes will be required on individual packages that will be checked into a database by the manufacturer, and checked out during the pharmacy dispensing process. However, it is still undecided whether these new barcodes (or radio-frequency tags) will also encode the product batch number and/or expiry date [20]. Similar initiatives to secure the pharmaceutical supply chain are ongoing in other countries and regions in the world, including in the United States and China [21,22].

Administration of biologicals Use of biologicals is often hospital-based, due to the multiple and complex procedures involved in the preparation and administration of these medicines, and in the clinical monitoring of the patients receiving them. The recording of exposure information will hereby depend on, among others, the type of medical records (electronic vs. paper-based), the existence 98 Chapter 3.3

of a linkage between pharmacy and medical records or full integration of both, the local procedures with regard to the recording of exposure information, and the type of biological. With respect to the latter, requirements may apply for specific products. For example, EU guidelines require that the product name and batch number for blood- and plasma-derived products are recorded at the point of administration [17]. Also, since the approval of the first infliximab biosimilar, the EU product information of infliximab-containing products recommends the recording of the product name and batch number in the patient file [23]. The patient alert card for infliximab has accordingly been updated with the recommendation for patients to also record the trade name and batch number. New developments that may result in the systematic recording of drug exposure information at the point of administration include the increasing use of barcode medication-verification technology at the site of administration [24,25]. A substantial proportion of biological therapies is nowadays self-administered by the patient, or at least administered in the home care setting. To facilitate the recording of detailed exposure information in home administration (i.e. product- and/or batch-specific information), specific aids have been developed like mobile phone applications [26,27].

Transferring exposure information to pharmacovigilance databases

Several data sources and methods are used to assess the benefit-risk balance of biologicals after initial licensure [28,29]. Spontaneous reporting systems, which rely on Adverse Drug Reaction (ADR) reporting from routine practice, have traditionally formed the mainstay of pharmacovigilance. Electronic health care databases, including databases of medical records, claims databases, and disease or drug registries, play however an increasingly important role in pharmacovigilance [29]. Due to the difference in data collection between spontaneous reporting systems and healthcare databases (ad hoc vs. systematic), differences may exist in the availability of detailed exposure information.

Spontaneous reporting systems Nowadays, virtually every country in the world has some sort of scheme in place to allow health professionals and/or patients to submit ADR reports to national or regional pharmacovigilance centers [30]. Although international standards have been developed on the data elements within these ADR reports [31], the actual amount of data provided is up to the discretion of the reporter. Namely, the individual data elements need to be manually filled into the electronic or paper-based ADR reporting form. The traceability in ADR reports will consequently both depend on the availability of exposure information to the reporter, as well as their willingness to report the information. Challenges in traceability of biologicals 99

Several factors come into play with respect to the availability of the exposure information to the reporter. For ADRs with a relative short time-to-onset, exposure information might be obtained from the still available packaging. However, in case the package has already been discarded, it will either depend on the reporter’s ability to recall exposure, or on the available information in the pharmacy and/or medical records, including the extent to which these records are accessible to the reporter. For example, the doctor confronted with the ADR might not have been the doctor prescribing the biological, and could consequently not have access to the required exposure information. Similarly, in case of patient reporting, it will be particularly challenging to retrieve the exposure information for hospital-administered biologicals, like monoclonal antibodies used in cancer treatments. The recently updated EU pharmacovigilance legislation has provided an important opportunity to improve the traceability of biologicals in ADR reports. Article 102(e) of the new directive states that Member States should ensure that “all appropriate measures are taken to identify clearly any biological medicinal product prescribed, dispensed, or sold in their territory which is the subject of a suspected adverse reaction report, with due regard to the name of the medicinal product, and the batch number” [32]. As a first deliverable, the guideline on good pharmacovigilance practices now requires follow-up on ADR reports pertaining to 3 biologicals in which information about the batches and products involved is lacking [33]. The further implementation of this legislation is however up to the individual Member States, for which specific legal obligations may be imposed on health professionals. A survey among European Member States (see methods in box 1) showed that 1 out of the 19 responding countries had used the opportunity to impose specific obligations related to traceability on health professionals. Specifically, the competent authority had developed an ADR reporting guideline that, after adoption into the national law, made it obligatory for health professionals to report brand names and batch numbers for biologicals. In addition, 9 out of the 19 (47%) responding countries had introduced or were intending to introduce new measures to improve biological traceability. As shown in Table 2, these measures include, among others, informing patients and health professionals about the need to provide detailed information of biological medicinal products when reporting ADRs (n=8), and the introduction of new functionalities to spontaneous reporting systems (n=4). One country was exploring the possibility of establishing a connection with hospital and pharmacy IT systems, allowing exposure information to be automatically transferred in ADR reports. 100 Chapter 3.3

Box 1. Description of research methods. Survey among Dutch pharmacists All 75 Hospital pharmacies and 20 outpatient pharmacies pertaining to the Utrecht Pharmacy Prac- tice Network for Education and Research (UPPER) [49] were approached to evaluate the measures taken in clinical practice to ensure the traceability of biological medicinal products. Two separate, standardized online questionnaires were developed. The questionnaires were in Dutch and consisted of five parts: general information about the pharmacy, prescription policies, registration of detailed product information, reporting of adverse drug reactions and final suggestions/remarks. Pharmacies were first approached on 20 December 2013, and a reminder was sent on 7 January 2014. Overall, 20 (27%) hospital pharmacies and 7 (35%) outpatient pharmacists completed the online questionnaire. Survey among national competent authorities in Europe A Non-Urgent Information (NUI) request was circulated among the national competent authorities of all 31 countries in the European Economic Area to evaluate the measures taken to ensure the traceability of biological medicinal products in adverse drug reaction (ADR) reports. A NUI is an established method for the exchange of non-urgent pharmacovigilance information between national competent authorities and the European Medicines Agency (EMA) [50]. The NUI request consisted of six questions relating to the procedures, systems and requirements in place to ensure traceability of biological products, and on the national implementation of the new pharmacovigilance directive. The NUI request was circulated on 16 December 2013, and a reminder was sent on 16 January 2014. Six weeks after the first mailing, the last responses were received. Overall, 19 (61%) countries responded to the NUI request.

Electronic healthcare databases Databases of electronic healthcare information have the advantage of ensuring the routine and systematic recording of clinical data. The availability of detailed exposure information consequently directly relates to the extent to which these data are systematically recorded in medical or pharmacy records, provided that this information is correctly and completely transferred. Claims databases are for example an important data source for studying the safety profile of biologicals in routine clinical care in the US [34,35]. Drug exposure in these databases is typically recorded along the NTIN (or reimbursement number) of the drug, which ensures that products are traceable up to the manufacturer, dosage form and strength. Therefore, these claims database can be a readily available tool for the conduct of product- specific pharmacovigilance, whereas batch-specific pharmacovigilance will not be possible. In Europe, on the other hand, academia-initiated registers play an important role in the pharmacovigilance of biologicals. A wide variety of registers are currently in place, ranging from small, single-center, drug registers, to large, multinational, disease registers that include multiple treatment arms and collect information on a range of clinical outcomes. Also, the procedures for the exchange of exposure data may vary according to the register. The majority of existing registers comprise specifically implemented long-term cohorts that collect information on treatment details and patient outcomes at-predefined intervals, e.g. Challenges in traceability of biologicals 101

Table 2. Planned or adopted measures by European Member States (n=19) to improve traceability of biological products in spontaneous reporting systems. Category of measures Member States, Specification of measures n (%) Provide extra 8 (42%) Set up educational programs, and/or start dialog with information stakeholders, to raise awareness and knowledge about the to healthcare traceability of biologicals. professionals and Provide information and guidelines on the website of the patients competent authority to inform patients, health care professional and marketing authorization holders about the need to provide detailed information of biological medicinal products when reporting ADRs. Functionalities to the 4 (21%) Develop or update electronic reporting systems, including mobile spontaneous reporting phone applications, to include a specific field for the registration system of batch numbers. Introduce new features to the electronic reporting system to facilitate brand traceability, e.g. by providing a list of brand names when a reporter only selects the INN for a biological product. Incorporate extra questions in the electronic reporting systems, to facilitate registration of extra information about previous used biological products and specification of these products. Follow-up on ADR 2 (11%) Follow-up on all ADR reports of biological products with missing 3 reports of biological brand name or batch number. products Perform quality checks on the ADR reports for biological products, by introducing a functionality that is able to identify ADR reports for biological products. Other 2 (11%) Develop legal tools and technical solutions which make tracing of biological products through the whole supply chain possible. Establish a connection with hospital and pharmacy IT systems. INN: international non-proprietary name; ADR, adverse drug reaction. the British rheumatology register [36,37] or the European PedNet Haemophilia register [38]. The availability of detailed exposure information will thus be determined by the agreed procedures for the reporting of exposure data. On the other hand, for the registers that are based upon the electronic exchange of routinely collected clinical data, e.g. the Swedish rheumatology register [36,37] or the Danish Multiple Sclerosis Registry [39], the exposure information will typically be equal to the data recorded in clinical practice. Though beyond of the scope of this review, it should be noted that good pharmacovigilance not only requires careful collection of exposure but also of outcome data. While, for example, some registers only capture routinely available clinical data, for other registers great efforts are made to collect information on the nature, seriousness and severity of adverse events, in much more detail than this would be possible based upon routine data. 102 Chapter 3.3

Conclusion

On basis of the review of the existing systems, it is concluded that product-traceability of biologicals is routinely ensured within the individual patient’s pharmacy records. Variable product information like the product batch number is, contrarily, expected to be infrequently captured in dispensing records at present, though this may differ according to national traceability regulations, local procedures, as well as the specific type of biological. Similarly, the extent to which product- and batch-specific exposure information is available to other health professionals and patients is assumed to be variable, and dependent on, among others, the existence of a linkage between pharmacy and medical records or full integration of both, and local procedures with regard to recording of exposure data. Once adequately available in clinical practice, it is essential that exposure data is correctly and completely transferred in spontaneous ADR reports, and databases of electronic healthcare information.

Expert opinion

Traceability of biologicals for pharmacovigilance purposes comprises a multifaceted challenge, involving both the presence of robust systems to ensure the traceability of individual products and batches throughout the pharmaceutical supply chain, as well as the correct and complete transfer of exposure information to pharmacovigilance data sources. Overcom- ing this challenge will, therefore, require a multifaceted approach, tackling both aspects of traceability. At the same time, traceability is not unique to pharmacovigilance, but actually at the interface of multiple needs. Repeated incidents of counterfeit biologicals entering the mainstream drug supply in Europe and the US have, for example, highlighted the need to further improve the supply chain integrity [40,41].

Ongoing challenges and potential solutions: Pharmaceutical supply chain Various countries around the world are currently implementing, or have recently implemented, enhanced pharmaceutical supply chain standards that will ensure the traceability of each individual unit of a medicinal product from the point of manufacturing up to the patient’s pharmacy record. Although these measures arose from the need to prevent falsified medicines from entering the legal pharmaceutical supply chain, they provide an important opportunity to systematically record detailed exposure information in the patient’s pharmacy records for pharmacovigilance purposes. This will mostly benefit the traceability of individual batches, as it was identified that existing systems already ensure the traceability of biologicals down to the manufacturer. The extent to which these dispensing data are consequently available to other health professionals and patients will depend on the existence of a linkage between pharmacy Challenges in traceability of biologicals 103 and medical records or full integration of both, and the possibility for patients to access their medical and/or pharmacy records. Long-term solutions therefore lie in integrating of pharmacy and medical data, and expanding patient access to medical records. It is, however, recognized that most countries are only in the beginning stages of implementing health information technologies, and data exchange across providers and with patients is not yet common practice [42,43]. In the short term, it will therefore remain important to adequately record detailed exposure information at the point of administration. A specific paragraph on the importance of traceability could, for example, be included in the healthcare label and/ or patient information leaflet. Also, patients may contribute to the traceability of hospital- administered products by recording the batch number and trade name, as for example recommended within the patient alert card for infliximab. As administration practices however remain a matter of clinical governance and good clinical practice, the actual implementation of these recommendations is uncertain.

Ongoing challenges and potential solutions: Pharmacovigilance data sources It is recognized that the data transfer to spontaneous reporting systems comprises a particularly vulnerable link in ensuring traceability. Spontaneous reporting systems rely on the 3 voluntary reporting of ADRs by health professionals and patients, and due to the ad hoc and manual nature of data transfer, exposure information might be incompletely or incorrectly attributed. Therefore, besides ensuring the recording of detailed exposure information in clinical practice, it is important to encourage reporters to provide product trade names, or other unique product identifiers, and batch numbers. As shown in the survey, several EU member states have plans to further raise awareness of the importance for traceability of biological products. Educational programs for ADR reporting that involve periodic mailings have previously, however, been shown to only have a temporal effect on the reporting behavior [44]. On the other hand, (time-consuming) educational outreach visits and directly accessible reminders did show prolonged effects on the reporting behavior [44,45]. Reminders regarding product traceability might for example be provided within the label information of the biological product, or in the (electronic) reporting form. An even more robust solution would be the establishment of direct links to the patient’s pharmacy or medical records, as proposed by one of the surveyed member states. Tools for integrating ADR reporting schemes into hospital information systems have previously been developed, and shown to facilitate and increase the rate of ADR reporting [46]. In particular, such integrated ADR reporting systems will eliminate the potential for errors associated with the manual transfer of data, as detailed exposure information can be automatically transferred to the reporting form. Though some regulatory authorities are already exploring the possibilities of integrating ADR reporting schemes into pharmacy/medical software [47], it is recognized that the actual implementation hereof could face many challenges, including legal barriers, concerns related to data protection and privacy, and technical dif- 104 Chapter 3.3

ficulties related to the diversity of healthcare systems. Therefore, while promising, this may constitute a rather long-term solution. As compared to the ad hoc data collection through spontaneous reports, electronic healthcare databases have the advantage of ensuring the routine and systematic recording of clinical data. With respect to traceability, it is therefore important to ensure that procedures are in place for the transfer or reporting of exposure information. In the ongoing debate about the International Non-proprietary Names (INN) policy for similar biological products, it has also been suggested that distinguishable INNs are needed to ensure adequate product-traceability of biological products [10-13]. According to a draft proposal, a four letter code, distinct from the INN, may be assigned to all biological substances [48]. This “Biological Qualifier code” uniquely identifies the manufacturer and manufacturing site of the biological. Reporters might erroneously attribute an ADR to the reference product instead of the biosimilar (or vice versa) , however there is currently no data that additional identifiers(on top of distinct brand names) will reduce the potential for misattribution. Regardless of the method used to facilitate the identification of individual products, adequate procedures need to be in place to ensure that either of these unique identifiers are recorded in clinical practice, and (where required) transferred to pharmacovigilance databases. Challenges in traceability of biologicals 105

References

1. Casadevall N, Nataf J, Viron B, et al. Pure red-cell 14. Vermeer NS, Straus SM, Mantel-Teeuwisse aplasia and antierythropoietin antibodies in pa- AK, et al. Traceability of biopharmaceuticals in tients treated with recombinant erythropoietin. spontaneous reporting systems: A cross-sectional N Engl J Med. 2002;346(7):469-75. study in the FDA adverse event reporting system 2. Ebbers HC, Mantel-Teeuwisse AK, Moors EH, (FAERS) and EudraVigilance databases. Drug et al. Today’s challenges in pharmacovigilance: Saf. 2013;6(8):617-25. What can we learn from epoetins? Drug Saf. 15. European Federation of Pharmaceutical 2011;34(4):273-87. Industries and Associations (EFPIA). A 3. Crommelin DJA, Storm G, Verrijk R, et al. vision for the coding and identification of Shifting paradigms: Biopharmaceuticals versus pharmaceutical products in Europe. Version low molecular weight drugs. Int J Pharm. 4.0, January 2008. Available from: http:// 2003;266(1-2):3-16. ec.europa.eu/health/files/counterf_par_trade/ 4. Sharma B. Immunogenicity of therapeutic pro- doc_publ_consult_200803/114_b_efpia_en.pdf teins. part 3: Impact of manufacturing changes. [Last accessed 25 September 2014]. Biotechnol Adv. 2007;25(3):325-331. 16. Code of Federal Regulations, Barcode label 5. Ramanan S, Grampp G. Drift, evolution, and requirements, title 21, sec.201.25 (21CFR201.25), divergence in biologics and biosimilars manu- revised as of April 1, 2013. Available from: http:// facturing. BioDrugs. 2014 Feb 25. www.accessdata.fda.gov/scripts/cdrh/cfdocs/ 6. Schellekens H, Jiskoot W. Erythropoietin- cfcfr/CFRSearch.cfm?fr=201.25 [Last accessed 3 associated PRCA: Still an unsolved mystery. J 30 July 2014] Immunotoxicol. 2006 Sep 1;3(3):123-30. 17. European Medicines Agency, Committee for 7. Gouw SC, van der Bom JG, Ljung R, et al. medicinal products for human use (CHMP). Factor VIII products and inhibitor develop- Guideline on plasma-derived medicinal ment in severe hemophilia A. N Engl J Med. products (EMA/CHMP/BWP/706271/2010); 2013;368(3):231-9. 21 july 2011. Available from: http://www.ema. 8. Hunt D, Kavanagh D, Drummond I, Weller B, europa.eu/docs/en_GB/document_library/ Bellamy C, Overell J, et al. Thrombotic micro- Scientific_guideline/2011/07/WC500109627. angiopathy associated with interferon beta. N pdf [Last accessed 30 July 2014] Engl J Med. 2014;370(13):1270-1. 18. Datamatrix adopted by French pharmaceutical 9. IMS Institute for Healthcare Informatics. The industry; 2011. Available from: http://eipg.eu/ global use of medicines: Outlook through 2017. wp-content/uploads/2013/11/EIP8-Feb11-P14. November. 2013. Available from: http://www. pdf [Last accessed 30 July 2014] imshealth.com/deployedfiles/imshealth/Global/ 19. European Parliament and Council of 1 July Content/Corporate/IMS%20Health%20Institute/ 2011. Directive 2011/62/EU of the European Reports/Global_Use_of_Meds_Outlook_2017/ Parliament and of the Council of 9 June 2011, IIHI_Global_Use_of_Meds_Report_2013.pdf amending directive 2001/83/EC on the com- [Last accessed 30 October 2014]. munity code relating to medicinal products for 10. Fox JL. Debate over details of US biosimilar human use, as regards the prevention of the entry pathway continues to rage. Nat Biotechnol. into the legal supply chain of falsified medicinal 2012;30(7):577. products. 11. The INN crowd. Nat Biotechnol. 2013;31(12):1055. 20. European Commission, health systems and 12. Silverman E. Biosimilars: What’s in a name? BMJ. products, pharmaceuticals. Delegated act on 2014;348:g272. the detailed rules for a unique identifier for 13. Schellekens H. A divided INN crowd. Nat medicinal products for human use, and its Biotechnol. 2014;32(2):132. verification. Concept paper submitted for public consultation. Brussels, 18 November 2011. 106 Chapter 3.3

Available from: http://ec.europa.eu/health/files/ 31. International Conference on Harmonisation counterf_par_trade/safety_2011-11.pdf [Last of technical requirements for registration of accessed 30 July 2014] pharmaceuticals for human use. ICH expert 21. Drug Supply Chain Security Act ; 2013. Available working group. Implementation guide for from: http://www.fda.gov/Drugs/DrugSafety/ electronic transmission of Individual Case Safety DrugIntegrityandSupplyChainSecurity/Drug- Reports (ICSRs). E2B(R3) Data elements and SupplyChainSecurityAct/ [Last accessed 30 July message specification. Version 5.01, 12 April 2014] 2013. 22. Infosys. Pharmaceutical serialization track & 32. European Parliament and Council of 31 trace; 2014. Available from: http://www.infosys. December 2010. Directive 2010/84/EU of The com/SAP/collaterals/Documents/pharmaceuti- European Parliament and of the Council of cal-serialization-guide-to-country-wise.pdf [Last 15 December 2010, amending, as regards to accessed 30 July 2014] pharmacovigilance, Directive 2001/83/EC on the 23. Summary of Product Characteristics Inflectra, Community Code relating to medicinal products Remicade, and Remsima. Available from: http:// for human use. www.ema.europa.eu/docs/en_GB/document_li- 33. European Medicines Agency. Heads of Medicines brary/EPAR_-_Product_Information/human/ Agencies. Guideline on good pharmacovigilance [Last accessed 30 July 2014] practices (GVP). module VI – Management and 24. Poon EG, Keohane CA, Yoon CS, et al. Ef- reporting of adverse reactions to medicinal prod- fect of bar-code technology on the safety of ucts. 22 June 2012. Available from: http://www. medication administration. N Engl J Med. ema.europa.eu/docs/en_GB/document_library/ 2010;362(18):1698-707. Scientific_guideline/2012/06/WC500129135.pdf 25. Pedersen CA, Schneider PJ, Scheckelhoff DJ. [Last accessed 30 July 2014] ASHP national survey of pharmacy practice in 34. Solomon DH, Massarotti E, Garg R, et al. hospital settings: Monitoring and patient educa- Association between disease-modifying anti- tion—2012. American Journal of Health-System rheumatic drugs and diabetes risk in patients Pharmacy. 2013;70(9):787-803. with rheumatoid arthritis and psoriasis. JAMA 26. Bar coding on pharmaceutical packaging cuts 2011;305(24):2525-31. costs and improves patient safety. Available from: 35. Bernatsky S, Lix L, O’Donnell S, Lacaille D, http://www.gs1ie.org/Download_Files/Health- CANRAD Network. Consensus statements for care_Files/Haemophilia_Treatment_Trace- the use of administrative health data in rheumatic ability_Case_Studyv2.pdf [Last accessed 30 July disease research and surveillance. J Rheumatol. 2014] 2013;40(1):66-73. 27. Chen WC, Chang P, Chen LF. Developing a 36. Zink A, Askling J, Dixon WG, et al. European camera-phone-based drug barcode reader biologicals registers: Methodology, selected and support system. AMIA Annu Symp Proc. results and perspectives. Annals of the Rheumatic 2006:882. Diseases. 2009 August 01;68(8):1240-6. 28. Harmark L, van Grootheest AC. Pharmaco- 37. Strangfeld A, Hyrich K, Askling J, et al. Detection vigilance: Methods, recent developments and and evaluation of a drug safety signal concerning future perspectives. Eur J Clin Pharmacol. 2008 pancreatic cancer: Lessons from a joint approach Aug;64(8):743-52. of three European biologics registers. Rheumatol- 29. Wise L, Parkinson J, Raine J, Breckenridge A. New ogy. 2011;50(1):146-51. approaches to drug safety: A pharmacovigilance 38. Fischer K, Ljung R, Platokouki H, et al. Prospec- tool kit. Nat Rev Drug Discov. 2009;8(10):779-82. tive observational cohort studies for studying rare 30. Pal SN, Duncombe C, Falzon D, Olsson S. WHO diseases: The European PedNet Haemophilia strategy for collecting safety data in public health Registry. Haemophilia. 2014;20(4):e280-6. programmes: Complementing spontaneous reporting systems. Drug Saf. 2013;36(2):75-81. Challenges in traceability of biologicals 107

39. Koch-Henriksen N. The Danish Multiple Scle- 48. World Health Organization (WHO). Biological rosis Registry: A 50-year follow-up. Multiple Qualifier, an INN proposal, INN Working Doc Sclerosis. 1999;5(4):293-6. 14.342; Revised draft July 2014. Available from: 40. European Medicines Agency alerts EU healthcare http://www.who.int/medicines/services/inn/ professionals after vials of falsified Herceptin bq_innproposal201407.pdf [Last accessed 25 identified; 16 April 2014. Available from: http:// September 2014] www.ema.europa.eu/docs/en_GB/document_li- 49. Koster E, Blom L, Philbert D, et al. The Utrecht brary/Press_release/2014/04/WC500165501.pdf pharmacy practice network for education and [Last accessed 30 July 2014] research: A network of community and hospital 41. Counterfeit version of Avastin in U.S. distribu- pharmacies in the Netherlands. Int J Clin Pharm. tion; 10 July 2012. Available from: http://www. 2014;36(4):669-74. fda.gov/drugs/drugsafety/ucm291960.htm [Last 50. Rules Governing Medicinal Products in the accessed 30 July 2014] European Union, Volume 9A,Guidelines on 42. Jha AK, Doolan D, Grandt D, et al. The use of Pharmacovigilance for Medicinal Products health information technology in seven nations. for Human Use; 2008. Available from: http:// Int J Med Inf. 2008;77(12):848-54. Ec.Europa.Eu/Health/Files/Eudralex/Vol-9/Pdf/ 43. Pagliari C, Detmer D, Singleton P. Potential vol9a_09-2008_en.pdf [Last accessed 30 July of electronic personal health records. BMJ. 2014] 2007;335:330-3. 51. European parliament and Council of 27 44. Castel J, Figueras A, Pedrós C, et al. Stimulating January 2003. Directive 2001/83/EC of the adverse drug reaction reporting. Drug Safety. European Parliament and of the Council of 3 2003;26(14):1049-55. 6 November 2001 on the community code 45. Figueiras A, Herdeiro MT, Polónia J, Gestal- relating to medicinal products for human use. Otero J. An educational intervention to improve Available from: http://ec.europa.eu/health/files/ physician reporting of adverse drug reactions: eudralex/vol-1/dir_2001_83_consol_2012/ A cluster-randomized controlled trial. JAMA. dir_2001_83_cons_2012_en.pdf [Last accessed 2006;296(9):1086-93. 30 October 2014] 46. Ortega A, Aguinagalde A, Lacasa C, et al. Ef- 52. European Medicines Agency, Committee for ficacy of an adverse drug reaction electronic medicinal products for human use (CHMP). reporting system integrated into a hospital in- Guideline on similar biological medicinal formation system. Annals of Pharmacotherapy. products; 30 October 2005. Available from: 2008;42(10):1491-6. http://www.ema.europa.eu/docs/en_GB/ 47. Medicines and Healthcare products Regula- document_library/Scientific_guideline/2009/09/ tory Agency (MHRA). New study launched for WC500003517.pdf [Last accessed 30 July 2014] pharmacists to report adverse drug reactions 53. European Medicines Agency. Questions and electronically; 21 September 2010. Available answers on biosimilars medicines (similar from: http://www.ema.europa.eu/docs/en_GB/ biological medicinal products); 27 September document_library/Press_release/2014/04/ 2012. Available from: http://www.ema.europa. WC500165501.pdf [Last accessed 25 September eu/docs/en_GB/document_library/Medicine_ 2014] QA/2009/12/WC500020062.pdf [Last accessed 30 July 2014]

Dynamics in post-approval 4 knowledge accrual on identified uncertainties

Risk Management Plans as a Tool 4.1 for Proactive Pharmacovigilance: A Cohort Study of Newly Approved Drugs in Europe

Vermeer NS, Ebbers HC, Straus SMJM, Leufkens HGM, Egberts ACG, De Bruin ML

Clinical Pharmacology & Therapeutics 2014; 96: 723-31

Abstract

Background and objective Risk management plans (RMPs) were introduced in Europe in 2005 to support a proactive approach in gaining knowledge on safety concerns through early planning of pharmacovigilance activities. The rate at which uncertainties in the safety profile are resolved along this proactive approach is currently unknown. Methods We examined the evolution of safety concerns in the RMP after initial approval for a selected cohort of 31 small-molecule drugs and 17 biologicals (approved 2006 - 2009), to provide insight into the knowledge gain over time. Information on the safety concerns (identified risks, potential risks, or missing information) was extracted from the initial RMP at approval, and were followed up in subsequent post-approval updates up to December 2012. Uncertainties (potential risks and missing information) were considered resolved when changed to identified risks, or when removed from the RMP. Results A total of 640 safety concerns were described in the baseline RMPs for the 48 products, among which 489 (76.4%) comprised uncertainties (247 potential risks and 242 concerns pertaining to missing information). The RMPs for biologicals contained more uncertainties at baseline (median=11; IQR: 7–16) than those for small-molecule drugs (median=9; IQR: 8-12; p=0.15). Overall, respectively 9.8% and 20.7% of the uncertainties were resolved at 3 and 5 years after approval. The rate at which uncertainties were resolved did not differ between small-molecule drugs and biologicals. Conclusion The relatively modest accrual of knowledge, as demonstrated in this study through resolution of uncertainties, suggests that opportunities for optimization exist while ensuring feasible and risk-proportionate pharmacovigilance planning. 112 Chapter 4.1

Introduction

In 2005, the European Medicines Agency introduced the risk management plan (RMP) as the instrument for the planning of pharmacovigilance activities and risk minimization for new drugs [1,2]. This introduction marked a shift toward a more proactive approach in gaining knowledge on safety concerns for newly approved drugs [3]. Earlier, regulatory authorities had mainly relied on spontaneous reports and industry or investigator-initiated studies, using a more reactive approach [4]. This change resulted from long-acknowledged limitations of pre-approval studies in predicting the risks of drugs in everyday practice [5,6]. In addition, regulators had been confronted with high-profile drug withdrawals and increasingly complex drugs and treatment regimens [7-9]. Biologicals, for example, have now become an important class [10,11], but they pose novel challenges in pharmacovigilance [12,13]. Together, these developments paved the way for renewed methods for the assessment of a drug’s risk-benefit profile after approval, demanding a planned approach starting ahead of licensing [14].

Box 1. Definitions in proactive pharmacovigilance, adapted from [2,15].

Safety concern: An important identified risk, important potential risk or missing information.

Important identified risk: “An untoward occurrence for which there is adequate evidence of an association with the medicinal product of interest”. For example, an adverse reaction adequately demonstrated in non-clinical studies and confirmed by clinical data. “Important” indicates that the identified risk could have an impact on the risk-benefit balance of the product or have implications for public health.

Important potential risk: “An untoward occurrence for which there is some basis for suspicion of an association with the medicinal product of interest but where this association has not been confirmed”. For example, toxicological findings seen in non-clinical safety studies which have not been observed or resolved in clinical studies. “Important” indicates that the potential risk could have an impact on the risk-benefit balance of the product or have implications for public health.

Missing information: Information about the safety of a medicinal product which is not available at the time of submission and which represents a limitation of the safety data with respect to predicting the safety of the product in the marketplace. For example, information about use in specific subpopulations like patients with renal or hepatic impairment.

Routine pharmacovigilance: A set of activities required to fulfil the legal requirement for pharmacovigilance, and which should therefore be conducted for all medicinal products. For example, collecting and reviewing suspected adverse drug reaction reports.

Additional pharmacovigilance activities: Post-marketing studies proposed in the RMP on top of routine pharmacovigilance. A safety concern may have no, or a number of, additional pharmacovigilance activities associated with it depending upon its nature, the degree to which it has already been characterized, and the feasibility of studying it. Examples of such additional measures include drug registries and clinical trials. Evolution of uncertainties in the RMP 113

Core to the RMP is the safety specification that describes all safety concerns in detail. These safety concerns are classified as either “important identified risks” or “important potential risks” and “missing information”, reflecting the uncertainties at approval (see definitions in Box 1 [2,15]). This safety specification forms the basis for a pharmacovigilance plan tailored to study the safety concerns post-approval. Depending on the need to further characterize identified risks or resolve existing uncertainties, additional activities might be proposed apart from routine pharmacovigilance. For example, post-approval studies could be required to study risks suspected on basis of the new drug’s pharmacology. After initial approval, the RMP is updated regularly over time. As the knowledge about the safety profile expands, safety concerns may change among the three categories (e.g., important potential risks becoming important identified risks), or be removed from the RMP when there is “demonstration of safety” over time; on the other hand, new safety concerns may be added (see Figure 1). It is expected that uncertainties in the safety profile at approval (important potential risks and missing information) will be resolved during a drug’s life- cycle. However, the extent and the rate at which uncertainties are resolved is unknown. Some previous studies focused on the content and quality of the RMP at the moment of approval [16-18], and the adherence to committed activities over time [19], but its dynamics in the post-approval phase has not yet been explored.

Safety concerns in RMP at approval (t=0)

Certainties in safety profile

Important Characterized concern 4 identified risk

t = x t = x Introduced t = x t = x Removed in RMP from RMP

Important Missing New concern potential risk t = x information

Uncertainties in safety profile Figure 1. Theoretical directions in which safety concerns existing in the RMP at approval (t=0) could change, or be newly added within the post-approval phase (t=x). Larger arrowheads indicate an increased assumed likelihood

In this study, we examined the development of safety concerns in the RMP after approval, to provide insight into the knowledge gain over time. We quantified changes in safety concerns by reviewing subsequent RMP updates for a cohort of recently approved drugs, and 114 Chapter 4.1

studied factors associated with change. Two groups of products were selected for which it was anticipated that there might be a particular need to further study existing uncertainties after approval: biologicals and small-molecule drugs intended for chronic use. Biologicals were selected, because these have been associated with specific risks and more uncertainties at the moment of approval [16] and drugs intended for chronic use were chosen because exposure time is relatively limited at the time of approval [6].

Results

Product characteristics A total of 48 products (17 biologicals and 31 small-molecule drugs intended for chronic use, hereafter referred to as “small-molecule drugs”), with a centralized approval between 31 January 2006 and 30 November 2009, were included (Supplementary Table 1 for a full list). The majority of the biologicals encompassed antineoplastic and immunomodulating agents (n=10; 59%, Table 1). For small-molecule drugs, the largest groups consisted of drugs for alimentary tract and metabolic disorders (n=7; 23%), drugs for cardiovascular disease (n=6;

Table 1. Product and approval details, stratified by product type. Small-molecules for Biologicals Total chronic use (n=17) (n=48) (n=31) Therapeutic group, n (%) Alimentary tract and metabolism 7 (23) 3 (18) 10 (21) Blood and blood forming organs 1 (3) 2 (12) 3 (6) Cardiovascular system 6 (19) 0 6 (13) Antiinfectives for systemic use 6 (19) 0 6 (13) Antineoplastic and immunomodulating 0 10 (59) 10 (21) Nervous system 5 (16) 0 5 (10) Other group 6 (19) 2 (12) 8 (17) Indication for long-term use, n (%) 31 (100) 13 (77) 44 (92) First product within class, n (%) 16 (52) 10 (59) 26 (54) Approval details, n (%) Orphan indication 5 (16) 5 (29) 10 (21) Exceptional circumstances 0 3 (18) 3 (6) Conditional approval 1 (3) 1 (6) 2 (4) No specific conditions 25 (81) 10 (59) 35 (73) First approval worldwide, n (%) Europe 13 (42) 4 (24) 17 (35) United States 13 (42) 8 (42) 21 (44) Rest of the world 5 (16) 5 (29) 10 (21) Evolution of uncertainties in the RMP 115

20%), and anti-infectives for systemic use (n=6; 20%). Biologicals were more often licensed for orphan diseases than were small-molecule drugs (29 vs. 16%, respectively; p=0.28), and biologicals less frequently comprised the first worldwide approval than did small-molecule drugs (24 vs. 42%, respectively; p=0.20).

Safety concerns listed in the RMP at time of approval A total of 640 safety concerns were described in the baseline RMPs (i.e., the adopted version at approval) for the 48 products, corresponding to a median of 13 (interquartile range (IQR): 10-16) safety concerns per product (Table 2). Overall, more safety concerns per product

Table 2. Characteristics of safety concerns listed in the risk management plan at time of drug approval, stratified by product type. Small-molecules Biologicals Total for chronic use (n=251) (n=640) (n=389) Safety concerns by type, n (%) Important identified risks 102 (26.2) 49 (19.5) 151 (23.6) Important potential risks 148 (38.0) 99 (39.4) 247 (38.6) Missing information 139 (35.7) 103 (41.0) 242 (37.8) Nature of safety concern by risk1,2, n (%) 232 (59.6) 154 (61.4) 386 (60.3) Investigations 21 (5.4) 21 (8.4) 42 (6.6) Infections and infestations 5 (1.3) 27 (10.8) 32 (5.0) Nervous system disorders 23 (5.9) 6 (2.4) 29 (4.5) Cardiac disorders 20 (5.1) 7 (2.8) 27 (4.2) Metabolism and nutrition disorders 20 (5.1) 5 (2.0) 25 (3.9) 4 Immune system disorders 8 (2.1) 16 (6.4) 24 (3.8) Gastrointestinal disorders 16 (4.1) 7 (2.8) 23 (3.6) Neoplasms 7 (1.8) 14 (5.6) 21 (3.3) Other3 112 (28.8) 51 (20.3) 163 (25.4) Nature of safety concern by context of use2, n (%) 178 (45.8) 124 (49.4) 302 (47.2) Patients with comorbidities 41 (10.5) 33 (13.1) 74 (11.6) Specific age groups 39 (10.0) 18 (7.2) 57 (8.9) Concomitant medication/ interaction 33 (8.5) 21 (8.4) 54 (8.4) Pregnancy, lactation 28 (7.2) 16 (6.4) 44 (6.9) Off-label indications 8 (2.1) 7 (2.9) 15 (2.3) Abuse, misuse, medication error 3 (0.8) 11 (4.4) 14 (2.2) Long-term use 6 (1.5) 3 (1.2) 9 (1.4) Unstudied ethnicity 5 (1.3) 4 (1.6) 9 (1.4) Other3 15 (3.9) 11 (4.4) 26 (4.1) 1 Adverse drug reactions categorized according to Medical Dictionary for Regulatory Activities (MedDRA). 2 For 51 safety concerns a combination of both risk and context was specified (e.g. infections + interactions). For 3 safety concerns (“limited size of safety database”) neither was specified. 3 Not pertaining to the eight most frequently occurring categories. 116 Chapter 4.1

were described for biologicals (median=15; IQR: 11-19) than for small-molecule drugs (median=12; IQR: 10-16; p=0.14). Among the 640 safety concerns, a total of 489 (76.4%) reflected uncertainties (important potential risks or missing information). The RMPs for biologicals contained more uncertainties at baseline (median=11; IQR: 7–16) than those for small-molecule drugs (median=9; IQR: 8-12; p=0.15). The number of uncertainties listed in the RMP at approval increased over time, from 5 uncertainties (IQR: 3.5-9.5) for products approved in 2006, to 9 (IQR: 7.25-11.75) for those approved in 2007, to 8.5 (IQR: 8-10.75) for those approved in 2008, and to 13 (IQR: 9.5-17) for those approved in 2009 (p=0.001). In 386 (60.3%) of the 640 safety concerns, reference was made to specific adverse drug reactions (ADRs), most commonly nervous system disorders (n=23/389; 5.9%) for small- molecule drugs, and infections and infestations (n=27/251; 10.8%) for biologicals. For 302 (42.7%) of the safety concerns, a “context of use” was defined: a specific aspect of use that had either not been studied during clinical development or that was known to relate to (specific) risks. This context of use most commonly involved use in patients with specific comorbidities (n=74; 11.6%), such as use in patients with renal or hepatic impairment. Ad- ditional pharmacovigilance activities were significantly more often proposed for biologicals (n=186/251; 74.1%) than for small-molecule drugs (n=182/389; 46.8%; p<0.001).

Outcome 1: Post-approval changes in baseline safety concerns Baseline safety concerns (i.e. safety concerns included in the RMP at approval) the RMP at approval) were followed up for changes within subsequent post-approval RMP updates until 31 December 2012. Overall, the median follow-up from approval was 56 (IQR: 44-68.75) months, with a median of 6.5 (IQR: 5-8) RMP update submissions (1.35 updates/year on average). Two products (rimonabant [Acomplia] and sitaxentan [Thelin]) were withdrawn from the European market before the end of the study (on 21 October 2008 and 2 March 2011, respectively). For 539 of the 640 (84.2%) baseline safety concerns, no change was recorded during follow-up. Overall, a very similar pattern was seen for small-molecule drugs and biologicals (Table 3). The most common change (n=47) involved the change from potential to identified risks. A total of 44 (6.9%) of the baseline safety concerns were removed from the RMP during follow-up, including 8 safety concerns that were considered identified risks at approval. All removed identified risks referred to a specific ADR; of these, most were mild in nature (e.g., “headache” or “dizziness”). For two of the three identified risks that changed to potential risks during follow-up, it was stated that based on clinical trial and postmarketing data, the risks should be considered potential risks, whereas for the remaining risk no reasoning was provided. Evolution of uncertainties in the RMP 117

Table 3. Status of baseline safety concerns after last change or end of follow-up stratified by product type and type concern. After last change or end of follow-up, n (%) Identified Potential Missing Concern Safety concerns at approval, n (%) risk risk information removed No change Small-molecules for chronic use Identified risk 102 (100) NA1 3 (2.9) 0 5 (4.9) 94 (92.2) Potential risk 148 (100) 28 (18.9) NA 0 4 (2.7) 116 (78.4) Missing information 139 (100) 2 (1.4) 1 (0.7) NA 18 (12.9) 118 (84.9) All concerns 389 (100) 27 (6.9) 328 (84.3) Biologicals Identified risks 49 (100) NA 0 0 3 (6.1) 46 (93.9) Potential risks 99 (100) 19 (19.2) NA 1 (1.0) 9 (9.1) 70 (70.7) Missing information 103 (100) 0 3 (2.9) NA 5 (4.9) 95 (92.2) All concerns 251 (100) 17 (6.8) 211 (84.1) All products Identified risks 151 (100) NA 3 (2.0) 0 8 (5.3) 140 (92.7) Potential risks 247 (100) 47 (19.0) NA 1 (0.4) 13 (5.3) 186 (75.3) Missing information 242 (100) 2 (0.8) 4 (1.7) NA 23 (9.5) 213 (88.0) All concerns 640 (100) 44 (6.9) 539 (84.2) 1NA, not applicable, not involving a change.

Outcome 2: Resolved baseline uncertainties Of the 489 baseline important potential risks and concerns categorized as missing information, 85 advanced to identified risks or were removed from the RMP during follow-up. 4 For the purpose of this study, these uncertainties were therefore considered resolved, as knowledge gain led to a change. The overall cumulative incidence of resolved uncertainties was 9.8% at 3 years after approval and 20.7% at 5 years after approval. As shown in Table 4, no difference was observed between the rates at which uncertainties for small-molecule drugs and biologicals (both 0.04/year (i.e., 4 of 100 per year); p=0.95) were resolved. Uncertainties for which additional pharmacovigilance activities (i.e., additional postmarketing studies) had been agreed at time of drug approval were resolved at a similar rate as those without additional activities (“routine pharmacovigilance”) (both 0.04/year; p=0.97). However, the rate at which uncertainties were resolved differed according to the type and nature of the uncertainty. Overall, important potential risks were resolved at a higher rate (0.06/year) as compared with missing information (0.02/year; hazard ratio: 2.61; 95% confidence interval (95% CI): 1.65-4.17). Similarly, as shown in Figure 2, uncertainties that referred to a specific ADR were resolved faster (0.06/year) than uncertainties that related to a specific context of use (0.02/year; hazard ratio: 2.51; 95% CI: 1.58-3.98). The similar direction of these two outcome determinants is related to the correlation between the type 118 Chapter 4.1

Table 4. Univariate determinants for resolving uncertainties. Number of Number of Number Follow- Rate per HR (95% CI) products uncertainties resolved up time year (cluster (months) level) Product type Small-molecule 31 287 52 14,951 0.041 [ref] Biological 17 202 33 9,651 0.042 0.99 (0.64-1.53) Pharmacovigilance activities Routine pharmacovigilance - 226 40 11,558 0.042 [ref] Additional studies - 263 45 13,044 0.041 1.01 (0.66-1.54) Type of uncertainty Potential risk - 247 60 11,852 0.061 2.61 (1.64-4.17) Missing information - 242 25 12,750 0.024 [ref] Nature of uncertainty by risk No specific ADR defined - 246 26 12,886 0.024 [ref] Specific ADR defined - 243 59 11,716 0.060 2.51 (1.58-3.98) Nature of uncertainty by context No context of use defined - 204 50 9,871 0.061 [ref] Context of use defined - 285 35 14,731 0.029 0.47 (0.30-0.72) ADR, adverse drug reaction; HR, hazard ratio; 95% CI, 95% confidence interval.

Risk defined 1.0 Context defined

0.8

0.6

0.4 Cumulative likelihood

0.2

0 1 2 3 4 5 6 Years since approval 243 236 222 207 101 72 16 Risk defined 285 282 273 267 137 87 31 Context defined Number at risk Figure 2. Cumulative likelihood of uncertainties existing at approval being resolved after initial approval, stratified by the nature of the uncertainty (risk-defined vs. context-defined) Evolution of uncertainties in the RMP 119 and nature of the uncertainty: for 87% (214/247) of the potential risks, a risk category was defined, in contrast to 12% (29/242) of safety concerns categorized as missing information. On the other hand, for 95% (229/242) of safety concerns categorized as missing information, a specific context of use was defined, against 23% (56/247) of the potential risks. Among the uncertainties that referred to a specific ADR, those related to immune system disorders or gastrointestinal disorders were resolved at the fastest rate (0.25/year and 0.13/ year, respectively), whereas uncertainties related to neoplasms or nervous system disorders were resolved at the lowest rate (0.01/year and 0.03/year, respectively). Among the uncertainties for which a specific context of use was defined, concerns related to long-term use or to abuse, misuse, and medication errors were resolved at the fastest rate (both 0.05/year), whereas uncertainties related to pregnancy and lactation or non-studied age groups were resolved at the lowest rate (0.00/year and 0.02/year, respectively).

Outcome 3: Newly added safety concerns A total of 157 new safety concerns were added in the RMP post-approval, corresponding to a median of 3 (range: 0-14; IQR: 1-4.75) per product. The numbers were similar for biologicals (median=3; IQR: 1-5) and small-molecule drugs (median 3; IQR: 1.5-4). For 8 of the 48 products, no new safety concerns were added. The newly added safety concerns comprised 45 identified risks and 112 uncertainties (69 potential risks and 43 concerns as having missing information). Among the new concerns, 71 (45.2%) resulted from a change in indication or formulation (n=25), or from an issue not studied during the clinical development program (n=46). The newly added uncertainties (n=112) were different in nature as compared with those present at approval (n=489). As compared with the uncertainties at approval, new uncertainties more frequently made reference to a specific ADR (66.1 (n=74) vs. 49.7% (n=243); p=0.002) 4 but less frequently related to a specific context of use (46.4 [n=52] vs. 58.3% [n=285]; p=0.023). The newly added uncertainties (n=112) were followed up in subsequent RMP updates, and 21 (18.8%) were resolved. This indicates a higher rate (0.10/year) than that for uncertainties added at the time of initial approval (0.04/year; p<0.001).

Safety concerns at end of follow-up Overall, the number of important identified risks increased by 65% from approval (n=151) to the end of this study (n=249). This increase is explained by baseline uncertainties changing to identified risks, the inclusion of new identified risks, and newly added uncertainties changing to identified risks. The number of uncertainties remained approximately equal throughout the study period because the number of resolved uncertainties (both baseline and newly added uncertainties, n=106) was equal to the number of newly added uncertainties (n=112), see Figure 3. At the product level, the median difference in uncertainties between the time at approval and at the end of follow-up was similar for small-molecule drugs (0; IQR: −1 to 2), and biologicals (0; IQR: −1 to 1). 120 Chapter 4.1

Identified risks Uncertainties Resolved uncertainty New uncertainty

750

700 Total number of safety concerns Cumulative number of resolved or new uncertainties

650

600 Identified risks 100 75 550 50 25 500 0 25 450 50 75 400 100

350

300

250 Uncertainties

200

150

100

0 0 1 2 3 4 Years since approval

48 48 48 47 29 (100%) (100%) (100%) (98%) (60%)

Number of products in follow-up Figure 3. Evolution of the RMP: Number of listed safety concerns in the RMP for the cohort of 48 drugs from approval onwards (left axis), and cumulative number of resolved and newly included uncertainties (right axis). Follow-up censored at the median follow-up of 4.67 years (56 months); RMP, risk management plan

Discussion

This study provides important insights into the dynamic evolution of RMPs and the post- approval knowledge gain for newly approved drugs. In the first 5 years after approval, 20.7% of the uncertainties that had been identified at approval were resolved. The rate at which uncertainties were resolved did not differ between small-molecule drugs and biologicals, and was unrelated to the required pharmacovigilance activities. Uncertainties about specific ADRs (mostly potential risks) were resolved faster than uncertainties related to a specific context of use (mostly missing information). On average, the number of uncertainties per drug remained equal throughout the first 5 years after approval, as the number of resolved Evolution of uncertainties in the RMP 121 uncertainties was counterbalanced by an equal number of newly added uncertainties. This dynamic nature of safety specifications is in line with long-established principles that the knowledge about safety will evolve during a product’s lifetime and although some uncertainties will be resolved, new questions or concerns may arise from everyday medical practice or use in a new indication. The finding from this study that one-fifth of the uncertainties at approval had been resolved 5 years after approval is important. Over the years, it has been suggested that a more proactive and tailored pharmacovigilance approach, as operationalized in Europe through the RMP, would enable a timelier characterization of uncertainties after approval [7,20,21]. It has even been suggested that this might ultimately lead to a reduced amount of data needed before approval [22]. Because the RMP actually enabled us to quantify the knowledge gain on uncertainties over time, it is not feasible to compare the present findings with the situation before the RMP requirement. Although we found that uncertainties progressively became resolved, their number is relatively low. Several factors might explain this low number of resolved uncertainties. First, the available follow-up time may be rather short. For those uncertainties for which additional studies had been required, there might not have been sufficient time during this study follow-up to complete the requested studies and for regulatory assessment of the results. A previous study indicated that the median planned duration of European Medicines Agency requested post-approval studies was 2.5 years, starting on average 1 year after initial approval [19]. However, the actual duration is unknown, and the completion of a planned study may face many challenges. In addition, several uncertainties might, independently of the chosen pharmacovigilance methodology, intrinsically be difficult to characterize. Resolving, for example, an uncertainty related to the carcinogenic potential of a drug requires long-term 4 follow-up [23], as illustrated in this study by the low rate at which uncertainties regarding neoplasms became resolved. Similarly, we showed that uncertainties related to a specific aspect of use (e.g., use during pregnancy) also resolved at a relatively low rate. Another explanation for the low number of resolved uncertainties could be that the planned activities were not sufficiently robust. Several data sources and methods are available for pharmacovigilance, and selecting the most appropriate strategy for evaluating a particular safety concern requires a multifaceted approach [24,25]. For example, a potential risk with a high background incidence in the exposed population cannot normally be resolved by routine activities only, due to the difficulty of distinguishing between the background incidence and causal involvement of the drug in spontaneous reports. Within Europe, the development of detailed methodological guidance [26], as well as research on better methods [27], specifically aims to strengthen study design through capacity building and new methods. Relatively little is known about the efficacy of pharmacovigilance systems and regulatory actions in terms of ensuring safe use of drugs in clinical practice and improving patient 122 Chapter 4.1

outcomes. Evaluations of such have been limited to date [28,29]. Assessing the effect of regulatory activities is challenging due to the multitude of determinants that may impact use or outcomes apart from the single regulatory action under study. Similarly, it not known whether accrual of knowledge, as identified in our study, will contribute to improved clinical outcomes. Further research toward better understanding the performance of pharmacovigilance systems and regulatory systems is needed and may help to further strengthen drug regulatory systems [9,30,31]. As part of the 2010 pharmacovigilance legislation [32], operational since July 2012, there have been substantial improvements in the European system of pharmacovigilance. A new independent committee (Pharmacovigilance Risk Assessment Committee) has been in- stalled, with a broad remit covering all aspects of pharmacovigilance, including RMP review and assessment. In addition, the new legislation foresees the possibility of imposing safety and efficacy studies. Furthermore, a new set of guidelines has been developed, the so-called good pharmacovigilance practices. These efforts will likely have a positive impact that is not yet captured in our study, which covered the period from 2006 to 2012 [33]. In this study, we included two groups of products, for which a number of differences were noted at approval. In accordance with previous findings, more uncertainties had been identified at initial approval for biologicals, as compared with small-molecule drugs [16]. Additional pharmacovigilance activities had more often been required for biologicals, as compared with small-molecule drugs. These findings indicate more perceived uncertainty for biologicals, which concurs with the perception that the risk–benefit profile of biologicals might be more complex to characterize (e.g., due to the reduced predictive value of preclinical animal studies [34] and the potential for immunogenicity [35]). In line with previous findings, we found that safety concerns for biologicals were more often related to the risk of infections, neoplasm, and immune system disorders [12]. Although our findings are likely attributable to the difference in approved indications, a previous study showed a similar difference in ADR profile between small-molecule drugs and biologicals within the same therapeutic class of antineoplastic agents [36]. Notably, despite all differences observed at approval, the overall rate at which uncertainties were resolved was similar for small-molecule drugs and biologicals. We also explored the addition of new safety concerns after approval. Our study showed that on average three new safety concerns were added within the first 5 years post-approval. Approximately half of these were related either to a change in the indication or formulation, or to a specific aspect of use that had not been studied within the clinical development program (e.g., concomitant medications or patients with comorbidities). This exemplifies how new use of drugs in other patient populations than those initially studied is often associated with new safety concerns and underpins the need for continuous risk management planning. Evolution of uncertainties in the RMP 123

In this study, we used changes in safety concerns as a proxy for knowledge gain regarding new drugs. Because safety concerns in the RMP are, by definition, important [15], changes within the RMP over time reflect relevant knowledge gain. This proxy, however, did not allow us to study subtle changes in the listed safety concerns, which might have resulted in some underestimation of actual knowledge gain. A second limitation related to our outcome measure is that some changes in safety concerns might be due to regulatory requirements and/or dynamics between regulatory authorities and industry, rather than resulting from knowledge gain. A minority of the changes from important potential risks to identified risks were not based on new information and thus did not involve any actual knowledge gain. In addition, discussion of the contents of the RMP could lead to a situation in which safety concerns move in and out of the RMP. To overcome this, safety concerns that were removed from the RMP and reintroduced later upon request of the regulatory authority were not considered. Finally, our cohort of small-molecule drugs was limited to drugs intended for chronic use. The extent to which our findings are generalizable to other small-molecule drugs is unknown and an area for future research. In conclusion, this study shows that one-fifth of the uncertainties at approval were resolved 5 years after approval. Because new uncertainties were included in the RMP at a similar rate, the overall number remained approximately equal, underpinning the life-cycle nature of risk management. RMPs are established tools within Europe and have delivered greater proactivity and planning of pharmacovigilance and risk minimization. The relatively modest accrual of knowledge, as demonstrated in this study through resolution of uncertainties, suggests that opportunities for optimization exist while ensuring feasible and risk- proportionate pharmacovigilance planning. 4

Methods

Study population. From the European Commission’s Community Register (http://ec.euro pa.eu/health/documents/community-register/), medicinal products were identified that had been approved between 1 November 2005 and 31 December 2009. Only products approved on the basis of a full application dossier for new active substances were included in the study. The drugs were classified into small-molecules or biologicals, according to the Eu- ropean legal definition of biologicals [37]. Only small-molecules intended for chronic use were included, which were identified according to a previously described definition [6]. We excluded vaccines and cellular therapeutic products, as these have specific characteristics not representative of all biologicals. From the 56 eligible products, 48 were included in the current study. For the excluded products, we were either unable to retrieve the required RMPs (i.e., the baseline RMP and at least one post-approval RMP update) (n=6) or the 124 Chapter 4.1

safety concerns in the RMPs were not classified according to the formal format (important identified risk, important potential risk, and missing information) (n=2).

Data source. RMPs—both the baseline RMP (i.e., the adopted version at approval) and all post-approval RMP updates—were retrieved from the European Medicines Agency’s databases. Data were collected from RMP updates until 31 December 2012, thereby allowing for a minimum follow-up of 3 years for the included products. An RMP update is warranted when new safety information becomes available post-approval or when there is a significant change in the indication of the product. Because these factors may vary between products, the interval between two RMP updates and the number of RMP updates per product may differ.

Data extraction and classification. Data were extracted from the baseline RMP and subsequent updates by two researchers (N.S.V. and R.G.D.) and were crosschecked for consistency by a third (M.L.D.B.) using a prespecified standardized data-extraction format. Data were checked for discrepancies, which were resolved by discussion. Further information on approval details (e.g., exceptional circumstances), and on the status after initial approval (whether withdrawn), was collected from the European Medicines Agency’s website.

Products. Medicinal products were classified into related therapeutic groups, on basis of the 1st level of the Anatomical Therapeutic Chemical (ATC) classification system (http://www. whocc.no/atc_ddd_index/). To determine whether products were first in class, all products were further categorized into subgroups consisting of products with the same indication and target receptor similar to the method used by Stefansdottir et al [38].

Safety concerns. In the RMP, safety concerns are classified as “important identified risks”, “important potential risks”, or “missing information” (see definitions in Box 1). Information on the status of each safety concern was extracted from the baseline RMPs, and subsequent post-approval updates. To enable longitudinal follow-up, the status of the safety concerns was recorded for every RMP update. For each change the data source cited was recorded. To provide further information on the nature, each safety concern was classified according to the adverse drug reaction listed as a risk, and to the context of use the risk was referring to. Risks were classified into System Organ Classes (SOCs), according to the Medical Diction- ary for Regulatory Activities (MedDRA), version 16. For the context of use we used the following predefined categories: use in specific (unstudied) age groups, use during pregnancy/ lactation, use in patients with specific (unstudied) co-morbidities, use with concomitant medications (including interactions), use in unstudied indications, abuse/ misuse/ medication errors, treatment discontinuation, long-term use, and use in unstudied ethnicities. A Evolution of uncertainties in the RMP 125 separate category was assigned for risks and context of use that referred to multiple categories, or that were not classifiable to any category.

Pharmacovigilance activities. Information on proposed pharmacovigilance activities (routine or additional, see Box 1) was extracted from baseline RMPs, as well as for each newly added safety concerns in the updates.

Study outcome. The outcomes of interest in this study were (i) change in baseline safety concerns, specifically focusing on (ii) resolved baseline uncertainties (combined end point of “potential risks” or “missing information” changing to “identified risks” or being removed from the RMP) and (iii) newly added safety concerns post0-approval. Figure 1 summarizes all possible directions in which safety concerns could change, or be newly added, during the product’s life-cycle. With respect to our first outcome, safety concerns described in the baseline RMP (t = 0) could change in multiple directions after initial approval, depending on the status of the concern at baseline. First of all, any type of safety concern could be removed from the RMP. We hypothesized that there are two reasons for a safety concern to be no longer monitored: (i) the safety concern has been sufficiently characterized and/or additional monitoring or risk minimization is not considered necessary, or (ii) there may be a “demonstration of safety” over time, i.e., no adequate evidence is found for an association between the safety concern and the medicinal product based on robust research. Second, “missing information” in the baseline RMP can change to the category “potential risks” or “identified risks,” when there is evidence (or at least suspicion) of an actual risk. Finally, “potential risks,” indicating a possible association of the safety concern with the medicinal product, could advance to 4 “identified risks” when enough evidence supports a true causal relation. For the calculation of the number of resolved uncertainties, concerns categorized as either “potential risks” or “missing information” were combined in the group “uncertainties,” as these reflect lack of sufficient information at time of approval. As the knowledge about the safety profile increases over time, “potential risks” and “missing information” may become “identified risks” or may be removed from the RMP (as described above) and may therefore be considered as uncertainties that are resolved. Uncertainties that were, however, removed from the RMP and reintroduced in the subsequent update were not considered resolved. With respect to our third outcome, safety concerns could be newly added, once new risks or uncertainties are observed post-approval.

Outcome determinants. The following determinants were considered for possible association with the number of resolved uncertainties: type of product (small-molecule drug intended for chronic use vs. biological), proposed pharmacovigilance activities (routine vs. 126 Chapter 4.1

additional), type of concern (potential risk vs. missing information), and the nature of the concern (whether related to a specific risk or to a specific context of use).

Data analysis. Descriptive statistics for the included medicinal products and related safety concerns were analyzed as medians (IQR), proportions, or count data, as appropriate. Dif- ferences in proportions and medians were tested using c2 square tests, and Mann-Whitney U-tests or Kruskal-Wallis tests, respectively. Kaplan-Meier survival curves were presented to identify time patterns in resolving uncertainties, as well as to estimate cumulative incidence rates. Univariate Cox regression analysis was used to examine the association of the above determinants on the outcome of resolved uncertainties. All data were analyzed using SPSS version 20.

Acknowledgements

We wish to thank for many helpful discussions and for their valuable advice Dr. Stella Black- burn (European Medicines Agency, United Kingdom) Dr. Lisette Hoogendoorn (Medicines Evaluation Board, the Netherlands), and Dr. Almath Spooner (Irish Medicines Board, Ireland). Evolution of uncertainties in the RMP 127

References

1. European Parliament and Council of 30 April the European Union. JAMA 300, 1887-1896 2004. Directive 2004/27/EC of The European (2008). Parliament and of the Council of 31 March 13. Vermeer, N. S. et al. Traceability of biopharma- 2004, amending Directive 2001/83/EC on ceuticals in spontaneous reporting systems: a the Community Code relating to medicinal cross-sectional study in the FDA Adverse Event products for human use. (2004). 14. International Conference on Harmonisation 2. Rules governing Medicinal Products in of Technical Requirements for Registration of the European Union, Volume 9A , http:// Pharmaceuticals for Human Use (ICH). Final ec.europa.eu/health/files/eudralex/vol-9/pdf/ Concept Paper. E2E: Pharmacovigilance Plan- vol9a_09-2008_en.pdf (2008). ning. 11 September 2002. http://www.ich.org/ 3. Raine, J., Wise, L., Blackburn, S., Eichler, H. G. fileadmin/Public_Web_ Site/ICH_Products/ & Breckenridge, A. European perspective on risk Guidelines/Efficacy/E2E/Concept_papers/ management and drug safety. Clin. Pharmacol. E2E_Concept_Paper.pdf. Ther. 89, 650-654 (2011). 15. European Medicines Agency. Heads of Medicines 4. Friedman, M. A. et al. The safety of newly ap- Agencies. Guideline on good pharmacovigilance proved medicines: do recent market removals practices (GVP). Module V – Risk management mean there is a problem? JAMA 281, 1728-1734 systems. 2 July 2012. (1999). 16. Giezen, T. J. et al. Evaluation of post-authorization 5. Rothwell, P. M. External validity of randomised safety studies in the first cohort of EU Risk Man- controlled trials: “to whom do the results of this agement Plans at time of regulatory approval. trial apply?”. Lancet 365, 82-93 (2005). Drug Saf. 32, 1175-1187 (2009). 6. Duijnhoven, R. G. et al. Number of patients 17. Frau, S., Font Pous, M., Luppino, M. R. & studied prior to approval of new medicines: Conforti, A. Risk Management Plans: are they a database analysis. PLoS Med. 10, e1001407 a tool for improving drug safety? Eur. J. Clin. (2013). Pharmacol. 66, 785-790 (2010). 4 7. Psaty, B. M. & Furberg, C. D. COX-2 inhibi- 18. Zomerdijk, I. M. et al. Risk minimization activi- tors—lessons in drug safety. N. Engl. J. Med. ties of centrally authorized products in the EU: a 352, 1133-1135 (2005). descriptive study. Drug Saf. 35, 299-314 (2012). 8. Avorn, J. Evaluating Drug Effects in the Post- 19. Blake, K. V. et al. European Medicines Agency Vioxx World: There Must Be a Better Way. review of post-authorisation studies with impli- Circulation 113, 2173-2176 (2006). cations for the European Network of Centres for 9. Lumpkin, M. M. et al. Advancing the science of Pharmacoepidemiology and Pharmacovigilance. medicines regulation: the role of the 21st-century Pharmacoepidemiol. Drug Saf. 20, 1021-1029 medicines regulator. Clin. Pharmacol. Ther. 92, (2011). 486-493 (2012). 20. Eichler, H. G., Pignatti, F., Flamion, B., Leufkens, 10. Walsh, G. Biopharmaceutical benchmarks 2010. H. & Breckenridge, A. Balancing early market Nat. Biotechnol. 28, 917-924 (2010). access to new drugs with the need for benefit/ 11. Evens, R. P. & Kaitin, K. I. The Biotechnology risk data: a mounting dilemma. Nat. Rev. Drug Innovation Machine: A Source of Intelligent Discov. 7, 818-826 (2008). Biopharmaceuticals for the Pharma Industrym- 21. Fontanarosa, P. B., Rennie, D. & DeAngelis, C. dash]Mapping Biotechnology/’s Success. Clin. D. Postmarketing surveillance—lack of vigilance, Pharmacol. Ther. 95, 528-532 (2014). lack of trust. JAMA 292, 2647-2650 (2004). 12. Giezen, T. J. et al. Safety-related regulatory actions for biologicals approved in the United States and 128 Chapter 4.1

22. Eichler, H. G. et al. The risks of risk aversion 31. Gispen-de Wied, C. C. & Leufkens, H. G. M. in drug regulation. Nat. Rev. Drug Discov. 12, From molecule to market access: Drug regula- 907-916 (2013). tory science as an upcoming discipline. Eur. J. 23. Smith, U. & Gale, E. A. Does diabetes therapy Pharmacol. 719, 9-15 (2013). influence the risk of cancer? Diabetologia 52, 32. European Parliament and Council of 31 1699-1708 (2009). December 2010. Directive 2010/84/EU of The 24. Meyboom, R. B. et al. Principles of Signal European Parliament and of the Council of Detection in Pharmacovigilance. Drug Safety 15 December 2010, amending, as regards to 16, 355-365 (1997). pharmacovigilance, Directive 2001/83/EC on the 25. Wise, L., Parkinson, J., Raine, J. & Breckenridge, Community Code relating to medicinal products A. New approaches to drug safety: a pharma- for human use. covigilance tool kit. Nat. Rev. Drug Discov. 8, 33. Arlett, P. et al. Proactively managing the risk 779-782 (2009). of marketed drugs: experience with the EMA 26. The European Network of Centres for Phar- Pharmacovigilance Risk Assessment Committee. macoepidemiology and Pharmacovigilance Nat Rev. Drug Discov. 13, 395-397 (2014). (ENCePP), Guide on Methodological Stan- 34. Dixit, R. & Boelsterli, U. A. Healthy animals and dards in Pharmacoepidemiology (Revision animal models of human disease(s) in safety 2), EMA/95098/2010, http://www.encepp.eu/ assessment of human pharmaceuticals, including standards_and_guidances (2013). therapeutic antibodies. Drug Discov. Today 12, 27. Pharmacoepidemiological Research on Outcomes 336-342 (2007). of Therapeutics by a European Consortium 35. Schellekens, H. Bioequivalence and the im- (PROTECT), Protect periodic report number munogenicity of biopharmaceuticals. Nat. Rev. 4, http://www.imi-protect.eu/documents/PRO- Drug Discov. 1, 457-462 (2002). TECT4thPeriodicReport_Sep2012-Aug2013.pdf 36. Ebbers, H. C. et al. Differences between post- (2013). authorization adverse drug reactions of biophar- 28. Bührlen, B., Reiß, T., Beckmann, C., Gassner, maceuticals and small molecules. BioDrugs 27, U.M. & Gleiter, C.H. Assessment of the European 167-174 (2013). Community System of Pharmacovigilance. Final 37. Directive 2001/83/EC of the European Parlia- Report - Final Version, http://publica.fraunhofer. ment and of the Council of 6 November 2001 de/documents/N-49939.html (2006). on the Community code relating to medicinal 29. Briesacher, B. A. et al. A critical review of products for human use (Consolidated version methods to evaluate the impact of FDA regula- : 16/11/2012). (2012). 30. Meagher, E. A. & Fitzgerald, G. A. Efficient drug 38. Stefansdottir, G. et al. Safety learning from drugs approval and monitoring must rely on sound of the same class: room for improvement. Clin. regulatory science. Nat. Med. 17, 1535-1535 Pharmacol. Ther. 91, 872-880 (2012). (2011). Evolution of uncertainties in the RMP 129

Supplementary Table 1. List of medicinal products included in cohort (n=48). Product name Generic name Indication Approval date Small-molecules for long-term use (n=31) Macugen Pegaptanib Wet Macular Degeneration 31-Jan-06 Neupro Rotigotine Parkinson Disease 15-Feb-06 Acomplia Rimonabant Obesity 21-Jun-06 Baraclude Entecavir Hepatitis B, chronic 26-Jun-06 Thelin Sitaxentan Hypertension, Pulmonary 11-Aug-06 Exjade Deferasirox Beta-Thalassemia, Iron Overload 28-Aug-06 Byetta Exenatide Diabetes Mellitus Type 2 20-Nov-06 Inovelon Rufinamide Epilepsy 16-Jan-07 Prezista Darunavir HIV Infections 12-Feb-07 Xelevia Sitagliptin Diabetes Mellitus Type 2 21-Mar-07 Toviaz Fesoterodine Overactive Urinary Bladder 20-Apr-07 Sebivo Telbivudine Hepatitis B, Chronic 24-Apr-07 Rasilez Aliskiren Hypertension 22-Aug-07 Celsentri Maraviroc HIV Infections 18-Sep-07 Galvus Vildagliptin Diabetes Mellitus Type 2 26-Sep-07 Isentress Raltegravir HIV Infections 20-Dec-07 Adenuric Febuxostat Gout 21-Apr-08 Volibris Ambrisentan Hypertension, Pulmonary 21-Apr-08 Ranexa Ranolazine Angina Pectoris 9-Jul-08 Intelence Etravirine HIV Infections 28-Aug-08 Vimpat Lacosamide Epilepsy 29-Aug-08 Kuvan Sapropterin Phenylketonurias 2-Dec-08 4 Thymanax Agomelatine Major Depressive Disorder 19-Feb-09 Efient Prasugrel Myocardial Infarction 23-Feb-09 Conbriza Bazedoxifene Postmenopausal Osteoporosis 17-Apr-09 Zebinix Eslicarbazepine Epilepsy 21-Apr-09 Samsca Tolvaptan Inappropriate ADH Syndrome 3-Aug-09 Onglyza Saxagliptin Diabetes Mellitus Type 2 1-Oct-09 Resolor Prucalopride Constipation 15-Oct-09 Multaq Dronedarone Atrial Fibrillation 26-Nov-09 Oslif Breezhaler Indacaterol COPD 30-Nov-09 Biologicals (n=17) Tysabri Natalizumab Multiple sclerosis 27-Jun-06 Myozyme Alglucosidase alfa Glycogen storage disease Type II 29-Mar-06 Elaprase Idursulfase Hunter Sydnrome 8-Jan-07 Orencia Abatacept Arhtritis, Rheumatoid 21-mei-07 Increlex Mecasermin Laron Syndrome 3-Aug-07 130 Chapter 4.1

Supplementary Table 1. List of medicinal products included in cohort (n=48). (continued) Product name Generic name Indication Approval date Lucentis Ranibizumab Age-related macular degeneration 22-Jan-07 Mircera Epoetin Beta Symptomatic anaemia 20-Jul-07 Vectibix Panitumumab Colorectal cancer 3-Dec-07 Soliris Eculizumab Paroxysmal hemoglobinuria 20-Jun-07 RoActemra Tocilizumab Arthritis, Rheumatoid 16-Jan-09 Stelara Ustekinumab Psoriasis 16-Jan-09 Nplate Romiplostim Immune thrombocytopenia purpura 4-Feb-09 Removab Catumaxomab Malignant ascites 20-Apr-09 Victoza Liraglutide Diabetes Mellitus Type 2 30-Jun-09 Simponi Golimumab Arthritis, Rheumatoid/ psoriatic 1-Oct-09 Cimzia Certolizumab pegol Arthritis, Rheumatoid 1-Oct-09 Ilaris Canakinumab Carcinoma of the colon or rectum 23-Oct-09

Cancer risks of medicines 4.2 approved in the European Union: what is known at licensing and what is solved post-marketing?

Duijnhoven RG, Vermeer NS, Straus SMJM, Hoes AW, de Boer AT, De Bruin ML

Abstract

Background and objective Signals of potential cancer risks may be identified years after use in clinical practice, but could also be observed early in the development of a new medicine. Studies may be agreed in a risk management plan (RMP) to assess potential risk after licensing. It is unknown how often potential cancer risks are known before approval, and how much evidence is yielded by the agreed studies after approval. The aim of this study was to provide an overview of potential cancer risks associated with medicines newly approved in the EU, and examine the contribution of the agreed studies to the post-approval knowledge accrual. Methods New and innovative biologicals and small-molecular medicines for chronic use approved in the EU between 1 November 2005 and 31 December 2009 were included. The public data on cancer risks at time of licensing and the agreed studies to investigate these risks were extracted. Products with potential cancer risks were subsequently followed until 5 May 2015 for new data on the suspected cancer risks from agreed studies, or other sources as reported by regulatory authorities. Results In total, 48 medicines were included, of which 18 (38%) were associated with potential elevated cancer risks at licensing, more often biologicals (11/17, 65%) than small-molecular medicine (7/31, 23%). Additional studies to investigate potential cancer risks were agreed for 16 of the 18 medicines. At the end of follow-up, the results of the agreed studies led to changes in the product information of three medicines. Other sources of information than the agreed studies led to the update of the product information of another four medicines. The product information of the remaining nine medicines was not updated on cancer risks. Conclusion Concerns about possible increased risks for cancer at time of licensing are common, particularly for biologicals. Though additional studies were frequently agreed to investigate these concerns, their added value appears to be modest. 134 Chapter 4.2

Introduction

During the past decades, a number of medicines have been associated with an increased risk for cancer. An example of medicines with concerns of increased cancer risks is anti-TNF therapy, which has been associated with lymphomas [1-4]. While genotoxicity and increased risks of secondary malignancies may be perceived as inevitable in for instance chemotherapy, it is generally difficult to accept cancer risks for most other medicines. Signals of potential cancer risks are sometimes observed before licensing of a new medicine, and the uncertainty about these risks needs to be weighed against the positive effects of the medicine. At time of approval, however, knowledge of a new medicine’s safety profile is based on information from a relatively standardised set of tests, experiments and clinical studies. Preclinical testing most often involves a series ofin vitro and in vivo carcinogenicity studies, which predict cancer risk in patients only to a certain extent [5]. Clinical studies will often be too limited in size, and have a too short duration of follow-up to identify potential cancer risks [6]. Therefore, uncertainties exist when a product is approved. In order to facilitate the monitoring and assessment of medicines’ risks after approval, risk management plans (RMPs) were introduced in the EU in 2005 [7]. The RMP summarizes important safety concerns that may have a substantial impact on the benefit-risk balance of the product. These safety concerns are categorised as either important “identified risk”, important “potential risk” or “missing information”. In addition to an overview of these safety issues, the RMP sets out the agreed activities that need to be conducted after approval by the pharmaceutical company. This may be routine collection of spontaneously reported adverse effects, but can also include additional studies. Furthermore, the plan may include additional risk minimisation measures to mitigate known risks [8]. After approval, the RMP is repeatedly updated during the whole life-cycle of the medicine to incorporate the latest information about the benefits and risks of the product. Although the accrual of knowledge on potential risks after approval has been studied previously, it is unknown to what extent the agreed studies may have contributed to the knowledge gain [9]. In this study we therefore aimed to provide an overview of potential cancer risks associated with medicines newly approved in the EU, and to examine the contribution of the agreed studies in the post-approval knowledge accrual.

Methods

Data sources and selection of products A selection of innovative medicines approved between 1 November 2005 and 31 December 2009, including biologicals and small-molecular medicines for chronic use, were included as previously described [9]. Biologicals were included as these products have previously Cancer risks of new medicines 135 been shown to have a relatively high number of safety-related regulatory actions related to neoplasms after approval [10], while small-molecular medicines for chronic use may be associated with potential cancer risks because of their specific pattern of use. Detailed information on cancer risks for the included medicines at time of approval was extracted from summaries of RMPs that are included in the Scientific Discussion of European Public Assessment Reports (EPARs). These overviews include updates to the product information, but also other procedures where the benefit-risk balance of a medicine was assessed such as license renewals, as well as procedures which involve submissions of studies to the competent authorities, irrespective whether or not the product information should be updated based on the new information from such a study. EPARs are publicly available from the EMA website and updated periodically to reflect the latest information on medicines with an appropriate level of detail [11].

Evidence at time of approval From the summary information of RMPs at time of licencing, all safety concerns relating to cancer, or neoplasms were selected and the following characteristics were extracted: type of safety concern (“important identified risks”, “important potential risks” or “missing information”), studies agreed to further investigate the concern, including the type of study (observational, e.g. cohort, case-control studies or registries; or experimental, i.e. trials) and additional risk minimisation measures (e.g. educational material) to mitigate the risks. Study protocol numbers were extracted and used to identify study results later during follow-up. The source of initial evidence leading to the potential cancer concern included in the RMP was categorised as preclinical (e.g. from in vitro experiments) or clinical data (e.g. differences in the occurrence of cancer between treatment groups in the clinical development studies), or pharmacological class effects (e.g. for immunosuppressive medicine) based on the information provided in the Scientific Discussion. 4

New evidence post-approval We assessed the overviews of procedural steps and scientific information of medicines associated with cancer concerns in the period after approval up to 5 May 2015 to examine the knowledge accrual over time. The overviews were screened for: 1.) updates related to the cancer concern of interest [either based on evidence from the agreed studies or from other sources], 2.) updates related to the agreed studies [either confirmative, inconclusive, or contradictive of an increased risk, or merely reporting on other concerns than the cancer risk]. If the product licence of a medicine was revoked after approval the date and main safety issue for revocation were noted. The data supporting updates is described in the overview published online, and can include studies sponsored by the company, information identified in the literature, or evidence from case-reports. In the overview of updates the study protocol numbers are generally reported along with the results and the implementation of these results in the product information, allowing the results to be linked to the original studies from the RMP. 136 Chapter 4.2

Results

Forty-eight innovative medicines licensed in the period 1 November 2005 - 31 December 2009 were included in the study. Products were indicated for diverse indications; cardiovascular diseases or diabetes (n=8), rheumatoid arthritis (n=6), infectious diseases (n=7), endocrine and metabolic diseases (n=6), and other conditions each of which was less common (n=21) (see Supplementary Table 1, chapter 4.1). Thirty-one of the products were small-molecular medicines and 17 were biologicals. Ten products were orphan medicines, of which five small-molecular medicines and five biologicals.

Concerns at time of approval Eighteen (38%) of the medicines included were associated with potential cancer risks, more often biologicals (11/17, 65%) than small-molecular medicines (7/31, 23%). None of the medicines had a confirmed increased risk for cancer at time of licensing. Of all products with concerns for cancer, four products were orphan medicines; one small-molecular medicine and three biologicals. Two of the eighteen products associated with a potential cancer risks at time of approval were licenced under exceptional circumstances. Concerns over cancer risks were included in the RMPs most often based on data from preclinical studies in the case of small-molecular medicines (5 out of 7), while for biologicals the concern was primarily (9 out of 11) based on their pharmacological effects (Table 1). Additional observational or interventional studies were agreed for 16 of the 18 products (89%) potentially associated with cancer risks. For most of these medicines (n=15, 83%) observational studies were agreed, involving registries for ten of these medicines. Post- approval interventional studies, in which the incidence of cancer was monitored, were conducted for 10 products (56%), and generally involved long-term extensions of pre-approval studies (including open-label studies). For 9 (50%) products both observational studies and interventional studies were agreed to investigate the potential cancer risks. Additional risk minimisation measures to warn prescribers, patients or both for a potentially increased risk for cancer (educational materials), were implemented for 4 medicines.

Post-approval knowledge accrual All 18 products associated with a potential increased risk of cancer were followed-up after approval to examine the knowledge accrual on the potential risk of cancer over time. The median follow-up time from the date of approval to the censoring date on 5 May 2015 was six years and eleven months (interquartile range: 71 - 97 months). Over the period since licensing up until the end of follow-up in May 2015, there were 27 updates related to the cancer concern of interest and/or related to the agreed studies, and these pertained to 11 of the 18 products (Table 1, see Supplementary Table 1 for a complete Cancer risks of new medicines 137 yes yes concern on cancer cancer on contributed contributed Agreed study study Agreed to knowledge knowledge to 2 1 3 2 2 0 1 0 0 5 3 related related Updates Updates to cancer cancer to concern or or concern agreed studies agreed 44 43 35 39 34 37 28 26 29 48 43 Total Total updates product product information information Trials Yes No No Yes Yes No No No No Yes No Obser vational studies No No Yes* Yes* Yes Yes* Yes Yes* No Yes Yes* Risk Minimisation Minimisation Risk (RMM) Measures Routine RMM Routine Routine RMM Routine Educational program Educational Routine RMM Routine Routine RMM Routine Physicians Educational Educational Physicians Program Education material: Patient Patient material: Education neoplasia for card Routine RMM Routine Routine RMM Routine Routine RMM Routine Physcian Education Physcian 4 Source concern Source Preclinical study Preclinical Clinical study Expected pharmacology Preclinical study Preclinical Preclinical study Preclinical Expected pharmacology Preclinical study Preclinical Preclinical study Preclinical Preclinical study Preclinical Clinical study Combi: Clincal Combi: expected study/ pharmacology Cancer concerns at approval at concerns Cancer Malignancies Malignant Neoplasms Malignant Malignancies, including including Malignancies, lymphoma Malignancy (both AIDS and (both AIDS and Malignancy related) non-AIDS Renal carcinoma and adenoma and Renal carcinoma Malignancy Neoplasia Potential for haematological haematological for Potential blood dyscrasias Prolactin-induced mammary Prolactin-induced carcinogenesis Malignancies Progression of existing existing of Progression malignancies hematological MDS or Type SM SM B SM SM B B SM SM SM B New New medicines between approved 2005 November and December 2009 where cancer concerns were listed as in well the as RMP, updates of the product information until May 2015. May until information product Table 1. Table Product name Baraclude Byetta Cimzia Celsentri Conbriza Ilaris Increlex Inovelon Multaq Isentress Nplate 138 Chapter 4.2 yes concern on cancer cancer on contributed contributed Agreed study study Agreed to knowledge knowledge to 2 0 0 4 2 0 0 related related Updates Updates to cancer cancer to concern or or concern agreed studies agreed 75 42 65 55 42 68 28 Total Total updates product product information information Trials Yes Yes No Yes Yes Yes Yes Obser vational studies Yes* Yes* Yes* Yes* Yes* Yes Yes Risk Minimisation Minimisation Risk (RMM) Measures Routine RMM Routine Routine RMM Routine Routine RMM Routine Routine RMM Routine STELARA Education STELARA Education TB including Program appropriate screening, selection, key patient efficacysafety and information Routine RMM Routine Routine RMM Routine Source concern Source Combi: Preclinical Preclinical Combi: Expectedstudy/ pharmacology Expected pharmacology Expected pharmacology Combi: Clincal Combi: expected study/ pharmacology Expected pharmacology Expected pharmacology Clinical study Cancer concerns at approval at concerns Cancer Malignancies: Lymphoma, Lymphoma, Malignancies: Breast cancer, NMSC, Lung cancer Malignancies Malignancies/ hematologic hematologic Malignancies/ abnormalities Malignancy Malignancy (potential risk) & risk) (potential Malignancy with in patients Use a malignancy or concurrent malignancy (missing history of information) Malignancy Medullary Thyroid cancer & cancer Medullary Thyroid neoplasms Type B B B B B B B New New medicines between approved 2005 November and December 2009 where cancer concerns were listed in as the well as RMP, updates of the Table 1. Table 2015. (continued) May until information product Product name Orencia RoActemra Soliris Simponi Stelara Tysabri Victoza - MDS, myelo tuberculosis; TB, plan; risk management RMP, be performed in a registry; measures; RMM, risk minimisation biological; small-molecular B, medicine; SM, *To skin cancer. NMSC, non-melanoma syndrome; dysplastic Cancer risks of new medicines 139 list). For the remaining seven products there was no new information added on either the planned studies or the cancer concerns. Of the 16 products for which studies were agreed in the RMP, for three medicines the studies generated evidence that was used to update the product information on the cancer concern. Results from ongoing trials for Isentress (raltegravir) and Celsentri (maraviroc) showed that there was no increased cancer incidence in the long-term for both medicines. By contrast, the results from the three-year extension of a trial for Simponi (golimumab) indicated an increased risk for malignancies. A subsequent cumulative review of data from all interventional and observational studies (including those in registries) conducted for Simponi, also provided additional evidence of an increased risk for skin cancers. For 4 of the 16 products for which additional studies had been agreed, new information on cancer risk became available from other sources than the studies agreed in the RMP. For two products (Baraclude [entecavir], Orencia, [abatacept]) the new evidence refuted the potential cancer risks, which lead to an update of the product information for Orencia. For the other two products (Cimzia [certolizumab pegol], Nplate [romiplostim]), the new data confirmed the potential cancer risks, leading to updates of the product information. For the remaining 9 products there was no new information at the end of follow-up on the potential cancer risks identified at time of licensing. In addition, there were two medicines with cancer concerns without agreed studies to address this. For one of these (Byetta [exenatide]), new evidence after approval showed a potential increased risk for pancreatic cancer and thyroid neoplasms, for which a new non-interventional study was requested during the routine regulatory assessment that was performed as part of the renewal of the license. 4 Discussion

In this study we describe the cancer risks associated with newly approved, innovative medicines in the EU in the period 2005-2009, with a follow-up until May 2015. At the time of licensing there are frequently concerns about potential cancer risks (18 of 48 products, 38%). Our results show that additional observational or interventional studies are agreed at time of approval for 89% of these products. For only three of these medicines the results of the agreed studies led to updates where new information on the cancer concern was incorporated in the product information. More often evidence on cancer risks of medicines were established by studies outside the RMP (four products). The high percentage of cancer concerns for biologicals is likely to be partly explained by the mechanisms of action of these medicines, which include growth hormones and antineoplastic and immunomodulatory therapies. Several of these are, therefore, expected to have some sort of effect on cell proliferation or on the ability of the immune system to respond to malignant cells. Consequently, a potential cancer risk may be included in risk management 140 Chapter 4.2

plans even though there is actually no direct evidence of harm. Eight of the medicines associated with potential cancer risks at approval had immunomodulatory action; several of which were monoclonal antibodies intended for rheumatoid arthritis. There was a high percentage of agreed additional studies in the RMP for medicines with potential elevated cancer risks at approval (16 out of 18 product; 89%). The contribution of these studies to the knowledge gained on the cancer concerns was, however, modest. For only three medicines the agreed study was the main source of information that could be used to update the product information on the specific cancer concern. The product information of another four was updated to incorporate evidence on the cancer concern, but the data came from other sources than the studies agreed in the RMP. These results cast doubts on the usefulness and effectiveness of RMPs, and the possibilities to achieve timely results from agreed studies, at least for cancer risks. Our results indicate that the initially agreed studies are often surpassed by other sources of information. This could be studies independently conducted and published in the literature or sponsored studies that are not included in the RMP or case reports that are evaluated constantly by companies and regulators or any other another source of information. All of these sources can contribute to the knowledge accrual of a certain medicine and potential adverse effects. These data sources are all valuable, but the RMP, which is intended to facilitate the post- approval management of risks, is only a very small contributor. While cancer concerns for newly approved medicines, to our knowledge, have not been studied before, the progress of post-authorisation safety studies has. A study conducted by the EMA reported that compliance of pharmaceutical companies with the request to conduct a study was “very good” [12]. Studies in the RMP, as described in our study can be regarded as such studies, but our results contradict quite clearly with the observations of Blake et al. This can be explained by the definition used in the aforementioned study: study conduct was the primary interest, rather than completion, or useful results from the agreed study. In practice this means that any delay in the conduct of the study for example due to problems of inclusion, are unobserved in the study by Blake. Another explanation for the discordant findings can be that for cancer risks, it is particularly challenging to address the uncertainties with post-approval studies, or that our median follow-up of almost 7 years was merely not sufficient to show the added value of the planned studies for these long-term risks. For this study we selected only new and innovative medicines that were approved through European centralised procedure. As a result, the studied products do not necessarily represent all products that are available on the market. However, the medicines selected for our study were regarded to have the highest chance for cancer related safety concerns and thus are expected to receive optimal scrutiny. In conclusion, concerns about possible increased risks for cancer at time of licensing are common, particularly for biologicals. Additional studies were frequently agreed to investigate these concerns, but their added value appears to be modest as more often other sources Cancer risks of new medicines 141 of information provided evidence on the cancer risks before new evidence was generated by the initially agreed studies.

4 142 Chapter 4.2 Cancer risks of new medicines 143

References

1. Askling J, Fored CM, Baecklund E, et al. Haema- 7. Raine J, Wise L, Blackburn S, et al. European topoietic malignancies in rheumatoid arthritis: Perspective on Risk Management and Drug lymphoma risk and characteristics after exposure Safety. Clin Pharmacol Ther 2011;89(5):650-54. to tumour necrosis factor antagonists. Annals of 8. Zomerdijk IM, Sayed-Tabatabaei FA, Trifirò G, the rheumatic diseases 2005;64:1414-20. et al. Risk Minimization Activities of Centrally 2. Bongartz T, Sutton AJ, Sweeting MJ, et al. Anti- Authorized Products in the EU. Drug Safety TNF antibody therapy in rheumatoid arthritis 2012;35:299-314. and the risk of serious infections and malignan- 9. Vermeer NS, Duijnhoven RG, Straus SMJM, et al. cies: systematic review and meta-analysis of rare Risk Management Plans as a Tool for Proactive harmful effects in randomized controlled trials. Pharmacovigilance: A Cohort Study of Newly JAMA 2006;295(19):2275-85. Approved Drugs in Europe. Clin Pharmacol 3. Raaschou P, Simard JF, Holmqvist M, et al. Ther 2014;96(6):723-731. Rheumatoid arthritis, anti-tumour necrosis 10. Giezen TJ, Mantel-Teeuwisse AK, Straus factor therapy, and risk of malignant melanoma: SMJM, et al. Safety-related regulatory actions nationwide population based prospective cohort for biologicals approved in the United States study from Sweden. BMJ (Clinical research ed) and the European Union. JAMA 2008; 300 (16): 2013;346:f1939. 1887-1896. 4. Thompson AE, Rieder SW, Pope JE. Tumor 11. European Medicines Agency. European necrosis factor therapy and the risk of serious Public Assesment Reports: Background and infection and malignancy in patients with early context. Available from: http://www.ema. rheumatoid arthritis: a meta-analysis of random- europa.eu/ema/index.jsp?curl=pages/ ized controlled trials. Arthritis and rheumatism medicines/general/general_content_000433. 2011;63:1479-85. jsp&midWC0b01ac058067fa25 [Last accessed: 5. International Conference on Harmonisation July 17, 2015] of Technical Requirements for the Registration 12. Blake K, Prilla S. Medicines Agency review of of Pharmaceuticals. Need for Carcinogenicity post‐authorisation studies with implications Studies of Pharmaceuticals S1A, 2004. for the European Network of Centres for Phar- 4 6. Duijnhoven RG, Straus SMJM, Raine JM, et al. macoepidemiology and Pharmacovigilance. Number of Patients Studied Prior to Approval Pharmacoepidemiology and Drug Safety of New Medicines: A Database Analysis. PLoS 2011;9:1021-29. Med 2013;10(3):e1001407. 142 Chapter 4.2 Cancer risks of new medicines 143 no Info on on Info study yes yes Info on on Info of concern interest no Two-hundred Two-hundred . and continued continued and

AI463901 AI463901 [HCC]; and “disease progression” defined by defined progression” “disease and [HCC]; hepatocellular carcinoma hepatocellular Submission of the final study report, 5 years follow-up of Study AI463048. This submission submission This AI463048. Study of follow-up years 5 report, study final the of Submission high-level a provide analyses final AI463048 study The measure. post-authorisation a fulfils with observed be can that outcomes clinical the of understanding limited and descriptive HBV. chronic to due function liver decompensated with patients of treatment long-term (death; population decompensated this in interest of events outcome clinical key Three of diagnosis the over assessed were decompensation) of [AEs] events adverse specified of occurrence the identified. been have concerns specific of signals No period. observation study entire the these of view the in needed is Baraclude for information product the of update no Therefore, data. Description Update of section 5.1 with long-term data on the effects of entecavir on liver liver on entecavir of effects the on data long-term with 5.1 section of Update HBeAg- nucleos(t)ide-naive of subset a in evaluated as fibrosis and necroinflammation Study from subjects HBV chronic HBeAg-negative and positive AI463022 studies pivotal the from patients negative HBeAg and positive HBeAg three and study to roll-over to eligible were respectively AI463027, and also initially and daily once mg 1 of dose increased an at but (ETV) entecavir with treatment biopsies, long-term and baseline evaluable with patients seven fifty the Of lamivudine. with the Of score. fibrosis Ishak in decrease 1-point a had 50 and improvement histologic had 55 fibrosis of Regression decrease. 2-point a had 58% 2, of score fibrosis Ishak baseline a with 43 57 all biopsy, the of time the At cirrhosis. or fibrosis advanced with patients 10 all in seen was 57 All ULN. X 1 ALT serum had (86%) 49/57 and copies/ml DNA < 300 HBV had patients long-term that demonstrated study this of results The HBsAg. for positive remained patients B hepatitis chronic with patients nucleoside-naïve in years 6 approximately for therapy ETV long- of benefit clinical The fibrosis. and necroinflammation liver of improvement in resulted data. histology long-term with confirmed been now has HBV of suppression viral term 4 Date of of Date opinion 23-04-09 25-04-14 Variation ID Variation II/0020 II/0043 ) and/or study proposed in the RMP the in proposed study and/or ( underlined) interest of concern safety to related information product of updates Post-approval RMP studies ** studies RMP AI463080 - Long- - AI463080 study, outcomes term malignancies. including and 050 AI463-901, longterm includes - 049 surveillance cancer in years 5 during 1500 approximately patients Safety concern concern Safety ** Malignancies Supplementary Table 1. Table 1. Supplementary )*. ( italic Product Baraclude 144 Chapter 4.2 yes yes no no Info on on Info study yes unclear Info on on Info of concern interest yes yes leukaemia leukaemia required in the the in required conducted to to conducted rhabdomyolysis and other other and rhabdomyolysis , malignancies and investigating also other events. These These events. other also investigating and

. The latter will be further investigated in a new new a in investigated further be will latter The . post authorisation safety studies safety authorisation post long term data from open label extensions of pivotal phase phase pivotal of extensions label open from data term long associated with the use of TNF-antagonists. From the data data the From TNF-antagonists. of use the with associated . Data from open label observational extension phase (treatment duration up to to up duration (treatment phase extension observational label open from Data . . serious infectious events with maraviroc treatment was consistent with the incidence seen at at seen incidence the with consistent was treatment maraviroc with events infectious serious studies. the in time-points earlier evaluate the risk of non-AIDS malignancies non-AIDS of risk the evaluate activity pharmacovigilance additional an as conducted been have studies RMP Description III studies III hepatic events, AIDS-defining death, of incidence that showed studies pivotal of years) 5 ischaemia, infarction/cardiac myocardial failure, Renewal of the marketing authorisation. Based upon the data that have become available available become have that data the upon Based authorisation. marketing the of Renewal the that considers CHMP the Authorisation, Marketing initial the of granting the since safety its that considers but positive, remains (exenatide) Byetta of balance benefit-risk have issues safety of number A reasons: following the for monitored closely be to is profile and exenatide between association potential the particular in Byetta, for identified been neoplasms thyroid and cancer pancreatic tacrolimus and exenatide between interaction drug possible the Also study. epidemiological a on agreed CHMP the addition, In evaluation. further needs lamotrigine and exenatide and following formulation) release prolonged (exenatide Bydureon and Byetta for PSUR common 726/2004. Regulation per required as Bydureon of frequency PSUR for timelines the MAH the that concluded CHMP the Byetta, of profile safety the upon based Therefore, time. years’ 5 in application renewal additional one submit should with 4.8 section SmPC of Update observational of results of Submission of cases to related information include to 4.8 and 4.4 section SmPC of Update and paediatric malignancies paediatric and are there that agreed is it malignancies, paediatric and leukaemias on MAH the by presented anti-TNF between association causal a is there whether regarding uncertainties of number a mechanism the given Nevertheless, not. or malignancies paediatric or leukaemias and agents with malignancies such of development the for risks possible the agents, these of action of product the in risks these addressing By excluded. be cannot class this of agents of use Cimzia for balance benefit/risk the activities, up follow ongoing the with and information, positive. remains 23-01-14 24-06-10 24-10-13 Date of of Date opinion 21-07-11 II/0037 II/0006 Variation ID Variation II/0033 R/0028 ) and/or study proposed in the RMP the in proposed study and/or ( underlined) interest of concern safety to related information product of updates Post-approval Active surveillance surveillance Active registry through Maraviroc Safety Safety Maraviroc (A4001067). Registry TE/ in trials Ongoing subjects TN RMP studies ** studies RMP none Malignancies, Malignancies, including lymphoma Malignancy Malignancy AIDS (both non-AIDS and related) Malignant Malignant Neoplasms Safety concern concern Safety ** Cimzia Celsentri Supplementary Table 1. Table 1. Supplementary Product Byetta )*. (continued) ( italic Cancer risks of new medicines 145 no Info on on Info study Info on on Info of concern interest yes (MCC) carcinoma cell Merkel Description Update of sections 4.4 and 4.8 of the SmPC in order to add add to order in SmPC the of 4.8 and 4.4 sections of Update as a new adverse event with unknown frequency. This update is based on a review of post- of review a on based is update This frequency. unknown with event adverse new a as The analysis. disproportionality and search literature a cases, trials clinical and marketing pegol certolizumab with associated MCC of cases of review cumulative a presented MAH MCC of cases Two analyses. disproportionality and review literature a by supported (CZP), MCC No cases. marketing post were reports Both review. cumulative the from retrieved were attribution causal of evidence lacked cases described Both trials. clinical observed were cases medications; prior by confounded was and onset to time short a had cases the of One CZP. to CZP with associated MCC of cases No disregarded. completely be not could case other the suspicion the supports strongly presented data the however literature the in identified were disproportionality presented The MCC. of development on TNF-blockers of effect class a of whether clear not is it although Overall, effect. class a of suspicion the supports also analysis such factors of number a to due be might CZP receiving patients in MCC of appearance the the age, patient’s the diseases, autoimmune underlying the therapy, inhibitor TNF other as of contribution possible the therapy, immunosuppressant non-biologic other to exposure ‘Undesirable 4.8 section to added is MCC Therefore excluded. be cannot risk the to use CZP seriousness and severity The known”. “not of category frequency a with SmPC, the of effects’ precautions and warnings ‘Special 4.4 section to addition its justify also MCC of event the of patients in reported been have MCC of cases that physicians prescribing the warn to use’ for examination, skin periodic recommend to and CZP including TNF-antagonists with treated cancer. skin for factors risk with patients for particularly 4 Date of of Date opinion 21-03-13 Variation ID Variation II/0028 ) and/or study proposed in the RMP the in proposed study and/or ( underlined) interest of concern safety to related information product of updates Post-approval RMP studies ** studies RMP Safety concern concern Safety ** Supplementary Table 1. Table 1. Supplementary Product )*. (continued) ( italic 146 Chapter 4.2 Cancer risks of new medicines 147 yes no Info on on Info study unclear Info on on Info of concern interest yes for for

. hepatosplenic T cell cell T hepatosplenic to add safety and efficacy information after long-term treatment long-term after information efficacy and safety add to extension of the pivotal Study 301-WW Study pivotal the of extension (HSTCL). A post-marketing case of HSTCL concomitant with certolizumab certolizumab with concomitant HSTCL of case post-marketing A (HSTCL). up to 5-years from an an from 5-years to up Description lymphoma a and short was event the of onset to time the Although reviewed. was use (CZP) pegol was agents) anti-TNF other (including treatment immunosuppressive extended previous in Therefore, excluded. totally be cannot CZP of contribution the factor, confounding a added is HSTCL concerning warning a agents, anti-TNF other for wording the with line HSTCL of development the for risk a that professionals healthcare inform to 4.4 section to RA the for review safety cumulative A excluded. be cannot Cimzia with treated patients in consistent are presented AEs of pattern the and incidence the that showed studies population with treated subjects RA for expected those with and CZP with experience previous with number the concerning updated therefore are 5.1 and 4.8 Sections agent. anti-TNFα an discontinued who patients of proportion the trials, clinical RA in exposed patients of and malignancies infections, for rates incidence the and reaction drug adverse an to due CZP. to antibodies and reactions site injections disorder, lymphoproliferative Update of section 4.4 of the SmPC in order to add the risk of of risk the add to order in SmPC the of 4.4 section of Update discussion scientific the to refer Please Characteristics. Product of Summary of Update 4.4, 4.2, sections of update an concerns variation II type This “Conbriza/H/C/000913/II/10”. SPC the of 5.1 and 4.8 4 20-01-11 Date of of Date opinion 25-04-13 II/0010 Variation ID Variation II/0031 ) and/or study proposed in the RMP the in proposed study and/or ( underlined) interest of concern safety to related information product of updates Post-approval 4 year extension extension year 4 301-WW. study of carcinoma cell Renal PASS of part as pharmacoepidemiology studies RMP studies ** studies RMP Renal carcinoma carcinoma Renal adenoma and Safety concern concern Safety ** Conbriza Supplementary Table 1. Table 1. Supplementary )*. (continued) ( italic Product 146 Chapter 4.2 Cancer risks of new medicines 147 Info on on Info study yes Info on on Info of concern interest no . In addition, upon request by the CHMP following following CHMP the by request upon addition, In . Description the assessment of PSUR 4, the MAH proposes the addition of wording in section 5.1 of the the of 5.1 section in wording of addition the proposes MAH the 4, PSUR of assessment the Study in observed cancer ovarian and cancer thyroid of incidences the include to SmPC information include to SmPC the of 4.8 and 4.7 sections of update the and 3068A1-301-WW, on events these of effects the regarding advice and reported events adverse ocular the on authorisation marketing the supporting data safety and efficacy The drive. to ability the and placebo randomized, double-blind, multicentre, the on based were bazedoxifene for the for women postmenopausal in GL 3068A1-300 study 3 Phase controlled, raloxifene randomized, double-blind, multicentre, month, 36 the and osteoporosis of prevention postmenopausal older, in WW 3068A1-301 study 3 Phase controlled, raloxifene and placebo 39808) (CSR WW 301 of study Core year 3 The osteoporosis. of treatment the for women I (Extensions extensions study placebo-controlled double-blind, year 2 two by extended was fracture in reduction the in bazedoxifene of safety and efficacy long-term the assess to II) and subjects 1,732 of total A (…) osteoporosis. with women postmenopausal in incidence bazedoxifene n=560, mg: 20 (bazedoxifene extension 2-year second the into continued after uterus, the on effects to reference (…)With n=590). placebo: and n=582, mg: 40/20 not did group mg 20 bazedoxifene the in thickness endometrial the treatment, of years 7 the in cancer endometrial of cases no were there placebo; to similar remained and change the In (p<0.008). group placebo the in cases 7 to compared group mg 20 bazedoxifene among years), 66 age, (mean women postmenopausal 7,492 in study treatment osteoporosis (0.69 cancer thyroid of cases 5 were there mg), (20 bazedoxifene with treated subjects 1,886 cancer thyroid of case 1 was there placebo, with treated subjects 1,885 among and 1,000) per mg 40 the in cancer thyroid of cases no were There treatment. of years 7 after 1,000) per (0.14 1,000) per (0.69 cancer ovarian of cases 5 were there Further, years. 5 to up group treatment 7 after cancer ovarian of cases 0 were there placebo, with treated subjects 1,885 among and to up group treatment mg 40 the in cancer ovarian of case one was There treatment. of years bazedoxifene the in cancer breast of cases 13 were there treatment, of years 7 After years. 5 1,000 per (1.50 group placebo the in cases 11 and women-years) 1,000 per (1.78 group mg 20 (…) women-years). This type II variation concerns an update of SmPC sections 4.4, 4.8 and 5.1 with relevant relevant with 5.1 and 4.8 4.4, sections SmPC of update an concerns variation II type This in bazedoxifene with treatment of years 7 following obtained data efficacy and safety 3068A1-301-WW Study to II extension 4 Date of of Date opinion 19-04-12 Variation ID Variation II/0021 ) and/or study proposed in the RMP the in proposed study and/or ( underlined) interest of concern safety to related information product of updates Post-approval RMP studies ** studies RMP Safety concern concern Safety ** Supplementary Table 1. Table 1. Supplementary )*. (continued) ( italic Product 148 Chapter 4.2 Info on on Info study yes Info on on Info of concern interest no registry registry that aims to enrol 100 patients who have reached near-adult height and who will will who and height near-adult reached have who patients 100 enrol to aims that Description Renewal of the Marketing Authorisation. Based on the CHMP review of data on quality, quality, on data of review CHMP the on Based Authorisation. Marketing the of Renewal authorisation marketing the since introduced variations all including efficacy, and safety treatment the in Increlex of balance benefit-risk the that considers CHMP the granted, was growth insulin-like primary severe with adolescents and children in failure growth of recommends therefore and favourable remains IGFD) (primary deficiency (IGF-1) factor-1 CHMP The circumstances. exceptional under authorisation marketing the of renewal the the to subject Increlex, for Authorisation Marketing the of renewal the recommended the that considers CHMP The Opinion. the to II Annex in down laid as conditions be to is profile safety its that considers but positive, remains Increlex of balance benefit-risk exceptional under MA a granted was Increlex reasons: following the for monitored closely severe with subjects 76 to limited was data safety available the time the at and circumstances since period the during gathered been has data safety clinical more Although IGFD. primary available are data long-term whom for patients of number the MA, initial the of granting the the into patients any enrol to possible yet not was it example, For limited. still is substudy on effect potential a including Increlex, of safety long-term the Thus, years. 5 for followed be the that decided CHMP The established. fully been yet not has neoplasia, of incidence the CHMP. the by agreed otherwise until cycle yearly the follow will product the for cycle PSUR renewal additional one submit should MAH the that concluded CHMP the Therefore, application, renewal the of part as submitted data new of view In time. years’ 5 in application do changes These IIIB. and IIIA II, I, Annexes the to amendments recommends CHMP the positive. remains which product, the of balance benefit-risk the affect not Date of of Date opinion 24-05-12 Variation ID Variation R/0019 ) and/or study proposed in the RMP the in proposed study and/or ( underlined) interest of concern safety to related information product of updates Post-approval RMP studies ** studies RMP Surveillance study Surveillance incidence, of Review duration and severity months; 6 every sub-study Surveillance safety; long-term for follow-up, neoplasia biennial Neoplasia Safety concern concern Safety ** Supplementary Table 1. Table 1. Supplementary Product Increlex )*. (continued) ( italic Cancer risks of new medicines 149 yes Info on on Info study unclear Info on on Info of concern interest , however the the however , 019 Protocol and 018 Protocol confirmed the previously observed safety profile safety observed previously the confirmed longer term data warranted inclusions of further adverse events (rash-related events), in the the in events), (rash-related events adverse further of inclusions warranted data term longer rate < 1% a at occurred that AEs non-severe included not previously also Furthermore, SPC. current the with line in changes other with together Information Product the in listed were guidance. Description Update of sections 4.1, 4.5, 4.8 and 5.1 of the SPC based on 48 week safety and efficacy data data efficacy and safety week 48 on based SPC the of 5.1 and 4.8 4.5, 4.1, sections of Update II Phase ongoing from data safety long-term the as well as studies, clinical pivotal two from the with CHMP the provide to committed MAH the (SOB), Obligation Specific a As studies. III Phase ongoing the from data efficacy and safety 48-week for review to further support the benefit/risk assessment. The MAH has provided this data data this provided has MAH The assessment. benefit/risk the support further to review for and pharmacokinetics as well as SOB, above the fulfil to order in variation this of part a as presented data pharmacokinetics The studies. II Phase ongoing from data safety long-term picture clearer a gave medications concomitant to relation in particular in variation, this in not did raltegravir that include to updated was SPC the of 4.5 Section interactions. the of contraceptives hormonal of pharmacokinetics the on effect meaningful clinically a have data pharmacokinetics submitted The data. pharmacokinetics related the include to and as such inhibitors UGT1A1 potent less of effect the of mentioning the warranted further existing the of update an supported further data The SPC. the in saquinavir and indinavir the efficacy, of terms In omeprazole. and tenofovir for data interaction pharmacokinetic Authorisation, Marketing initial the for assessed effects the confirmed overall data 48-weeks 24-weeks the replace to 48 week at outcomes efficacy the with updated was SPC the however and responses virologic of terms in information further gave analysis the addition, In data. submitted data safety The SPC. the in included also was which 48, week to up rebound viral overall variation this for 4 Date of of Date opinion 20-11-08 Variation ID Variation II/0001 ) and/or study proposed in the RMP the in proposed study and/or ( underlined) interest of concern safety to related information product of updates Post-approval RMP studies ** studies RMP Monitor reports of of reports Monitor from malignancies trials clinical ongoing 021, 019, 018, 005, (004, 033) 032, 023, 022, post Observational safety authorization study Malignancies Safety concern concern Safety ** Supplementary Table 1. Table 1. Supplementary Product Isentress )*. (continued) ( italic 150 Chapter 4.2 yes Info on on Info study no yes unclear Info on on Info of concern interest yes yes does does is an ongoing ongoing an is . Study 021 Study or efficacy of raltegravir. raltegravir. of efficacy or confirm the safety the confirm long-term (2-year) carcinogenicity studies of of studies carcinogenicity (2-year) long-term ) in treatment naïve patients. patients. naïve treatment in ) in treatment experienced and naïve patients, no new safety safety new no patients, naïve and experienced treatment in Protocol 021 Protocol in fulfilment of a post-authorisation commitment. The study in study The commitment. post-authorisation a of fulfilment in

studies 018, 019 and 021 and 019 018, studies Sections 4.8 and 5.1 of the SmPC with were updated with this new information. new this with updated were with SmPC the of 5.1 and 4.8 Sections double-blind, randomised, active-controlled, non-inferiority trial evaluating raltegravir 400 400 raltegravir evaluating trial non-inferiority active-controlled, randomised, double-blind, (tenofovir Truvada with combination in administered each efavirenz, versus day, a twice mg achieving patients of proportion the was endpoint efficacy primary The emtricitabine). + The 96. Week at endpoint efficacy secondary a with 48 Week at copies/mL RNA <50 HIV Week The 240. Week through observations blinded continued for allows extension study application present the with submitted data 156 Description not appear to be any specific cancer risk associated with raltegravir with associated risk cancer specific any be to appear not The additional safety and efficacy data (week 96) generally confirm the previous conclusions conclusions previous the confirm generally 96) (week data efficacy and safety additional The data 96 Week The patients. naïve and antiretroviral-experienced in use raltegravir regarding maintained 48 week and/or 24 Week at responded had who patients of majority the that show non- and placebo over raltegravir for benefit a and 96 Week to response virological their 3 the of 96 to 48 Weeks including follow-up this In confirmed. was efavirenz to inferiority pivotal there that assurance further provide data clinical updated The identified. were concerns the from data efficacy/safety weeks 156 the with SmPC the of 5.1 and 4.8 sections of Update ( study III Phase ongoing Update of section 5.3 of the SPC based on on based SPC the of 5.3 section of Update rodents in raltegravir rats, In tested. levels dose raltegravir any at carcinogenicity of evidence no showed mice high in identified were nasopharynx / nose the of carcinoma) cell (squamous tumours secondary be to considered were neoplasms These animals. group dose intermediate and nose, and nasopharynx the within present also inflammation, and irritation chronic to a in negative was Raltegravir drug. study the of aspiration reflux of consequence a and of risk relevant a indicate not did data The studies. clastogenicity and genotoxicity of series humans. for carcinogenicity Date of of Date opinion 23-04-09 18-02-10 19-01-12 Variation ID Variation II/0009 II/0017 II/0025 ) and/or study proposed in the RMP the in proposed study and/or ( underlined) interest of concern safety to related information product of updates Post-approval RMP studies ** studies RMP Safety concern concern Safety ** Supplementary Table 1. Table 1. Supplementary Product )*. (continued) ( italic Cancer risks of new medicines 151 yes Info on on Info study unclear Info on on Info of concern interest (section 4.8) 4.8) (section 021 study product information was updated with these long-term safety long-term these with updated was information product and efficacy data (section 5.1) on the use of ISENTRESS in treatment-naïve HIV-infected HIV-infected treatment-naïve in ISENTRESS of use the on 5.1) (section data efficacy and patients. Description (multi-centre, randomised, double-blind, active-control trial) evaluates the safety and anti- and safety the evaluates trial) active-control double-blind, randomised, (multi-centre, a in bedtime, at mg 600 SUSTIVA vs. daily twice mg 400 ISENTRESS of activity retroviral efficacy Primary patients. HIV-infected treatment-naïve in TRUVADA, with combination II/25). II/17, II/10, procedures (variations CHMP the by reviewed previously were results through data the submit to CHMP by made request the with complied presently MAH The The 240. Week Update of sections 4.8 and 5.2 of the SmPC with the 240 weeks efficacy/safety data from from data efficacy/safety weeks 240 the with SmPC the of 5.2 and 4.8 sections of Update /_ STARTMRK patients. naïve treatment in 021) (Protocol study III Phase the 4 Date of of Date opinion 17-01-13 Variation ID Variation II/0037 ) and/or study proposed in the RMP the in proposed study and/or ( underlined) interest of concern safety to related information product of updates Post-approval RMP studies ** studies RMP Safety concern concern Safety ** Supplementary Table 1. Table 1. Supplementary Product )*. (continued) ( italic 152 Chapter 4.2 unclear no no Info on on Info study Info on on Info of concern interest yes yes yes . use in in use FU2 024.2 FU2 based on the final report report final the on based (MDS) ) further to the assessment of of assessment the to further ) (MDS , further to the assessment of data submitted as procedure OTH28. In this this In OTH28. procedure as submitted data of assessment the to further , progression of existing Myelodysplastic Syndromes Myelodysplastic existing of progression Description To amend section 4.4 of the SmPC to reflect updated information regarding the the regarding information updated reflect to SmPC the of 4.4 section amend To syndromes myelodysplastic with patients with patients for recommendations update to SmPC the of 4.8 and 4.4 sections in Changes myelogenous acute to progression of risk the on (MDS) syndrome myelodysplastic (AML) leukaemia a from data available with SmPC the of 4.8 and 4.4 sections updated has MAH the variation myelodysplastic with associated thrombocytopenia with patients of study clinical randomised leukaemia myelogenous acute to progression of risk increased an which in (MDS) syndrome addition, In placebo. to compared romiplostim with treated patients in observed was (AML) that highlight to reinforced been have leaflet package and SmPC the in messages key the of treatment the for established been only has romiplostim for benefit/risk positive a the for used be not must romiplostim that and ITP chronic to associated thrombocytopenia other thrombocytopenia of cause other any or MDS to due thrombocytopenia of treatment trials. clinical outside ITP than related information safety the update to order in SmPC the of 4.8 and 4.4 sections of Update to and 4.4 sections variation this In 029. FUM of assessment the following 20060198 study from to related information safety latest the reflect to order in updated been have SmPC the of 4.8 from report final the on based (MDS) Syndromes Myelodysplastic existing of progression a to due stopped prematurely was romiplostim with treatment study this In 20060198. study greater blasts circulating in increase an and AML to progression disease of excess numerical was survival Overall placebo. to compared romiplostim receiving patients in 10% than placebo. to similar 21-07-11 27-06-13 Date of of Date opinion 18-10-10 Variation ID Variation II/0017 II/0029 IB/0013 ) and/or study proposed in the RMP the in proposed study and/or ( underlined) interest of concern safety to related information product of updates Post-approval RMP studies ** studies RMP Romiplostim Patient Patient Romiplostim and (US Registry Safety Canada) registry Prospective of safety assess to adult in romiplostrim in ITP with patients Countries Scandinavian 20070797) (Study Progression Progression existing of hematological or malignancies MDS Safety concern concern Safety ** Supplementary Table 1. Table 1. Supplementary Product Nplate )*. (continued) ( italic Cancer risks of new medicines 153 no no Info on on Info study Info on on Info of concern interest yes yes with abatacept in humans. in abatacept with ” under the Malignancy section. Section 4.5: inclusion inclusion 4.5: Section section. Malignancy the under ” increased risk of malignancy of risk increased Non-melanoma skin cancers skin Non-melanoma Description Update of the section 4.4 and 4.8 of the summary of product characteristics with safety safety with characteristics product of summary the of 4.8 and 4.4 section the of Update in submitted as 22.06.08) - 23.12.07 covered: (Period PSUR last the in included information the of some that reflect to amended been has SPC Orencia’s the of 4.4 Section 2008. August fatal, been have abatacept with reported pneumonia and sepsis including infections, serious 49 identified 2008 June 22 on database safety post-marketing the of search cumulative a as SPC the of 4.8 Section infection. to due were deaths 19 these of outcome; fatal a with reports reactions hypersensitivity drug and anaphylaxis, hypersensitivity, that reflect to amended was open-label and controlled during abatacept with treated patients in reported rarely were data exposure additional on based that, reflect to amended also was 4.8 Section trial. clinical patient-years), 10,365 during patients abatacept-treated (4149 period time current the in an be to appear not does there Following 4.8. and 4.5 4.4, section update to variations of group a for application an was This 22 to 2010 December 23 from period 1-year the (covering 7 PSUR Orencia’s of assessment update SmPC an for request a including conclusions adopted CHMP the 2011), December “Infections” –section under text of Change 4.4: Section information: following the with cases of experience post-marketing the on sentence a of inclusion tuberculosis; regarding “ with Section “Vaccinations”. under study vaccination pneumococcal the from information of include to effects”-section “undesirable 4.8 the under table existing the of replacement 4.8: spontaneous post-marketing and studies safety post-authorisation trials, clinical from data Orencia the in CHMP the by requested changes the implemented has MAH The reports. fatal regarding information product the in information the updated also has MAH The PI. in Orencia of discontinuation the on recommendations further include to and anaphylaxis CHMP. the by agreed was This reactions. anaphylactic of case 4 25-04-13 Date of of Date opinion 23-04-09 Variation ID Variation II/0062/G II/0023 ) and/or study proposed in the RMP the in proposed study and/or ( underlined) interest of concern safety to related information product of updates Post-approval RMP studies ** studies RMP Open extensions of of extensions Open up trials clinical pivotal years 5 to studies: Epidemiology (IM101- Registry US and RALLY) 045B, registries EU three ARTIS; (IM101-125, BSRBR; IM101-126, RABBIT) IM101-127, malignancies assess will lymphoma, including and cancer, lung NMSC, cancer breast and reports (annual vs abatacept risk relative DMARDs non-biologic pre-defined at milestones) Malignancies: Malignancies: Lymphoma, Lung NMSC, Breast cancer, cancer Safety concern concern Safety ** Supplementary Table 1. Table 1. Supplementary Product Orencia )*. (continued) ( italic 154 Chapter 4.2 no no Info on on Info study Info on on Info of concern interest yes yes , and psoriasis- and , and inclusion of of inclusion and

leukaemia occurrence of malignancies in paediatric patients, leukaemia patients, paediatric in malignancies of occurrence Description Update of the SmPC in sections 4.4 and 4.8 with reagrd to malignancies and and malignancies to reagrd with 4.8 and 4.4 sections in SmPC the of Update TNF-antagonists of administration concurrent as well as disorders lymphoproliferative of analysis safety previous A DMARDS. biological between switching on and anakinra and signal safety potential a revealed agents TNF-blocking of data post-marketing and clinical the involving of frequency the regarding SmPC the of 4.8 section of Update product the updated has MAH The request. CHMP a to response in latter the sarcoidosis, leukemia of report a following ADR an as leukemia of frequency the regarding information based CHMP from request a following Furthermore, subject. trial clinical golimumab a from (including disease granulomatous non-necrotizing of review cumulative the on report the on has “sarcoidosis” agents, anti-TNF of class entire the as well as golimumab with sarcoidosis) ADR. disorder system Immune rare a as added been like lesions associated with this class of products. In order to better reflect these safety safety these reflect better to order In products. of class this with associated lesions like was review addditional an Simponi, for information product the in appropriately concepts data the From MAH. the to available cases individual the assess to MAH the by undertaken are there that agreed is it malignancies, paediatric and leukaemias on MAH the by presented anti-TNF between association causal a is there whether regarding uncertainties of number a mechanism the given Nevertheless, not. or malignancies paediatric or leukaemias and agents use with malignancies such of development the for risks possible the agents, these of action of section in warning the to amendments the Thus, excluded. be cannot class this of agents of paediatric of development the for risk potential the addressing introduced been have 4.4 also MAH The TNF-antagonists. with treated patients in leukaemia and malignancies such regions major in labelling product TNF of evaluation class independent an conducted added has MAH the result, a As review. literature medical TNF a as well as EU, and US the as of use concomitant the regarding information product Simponi the to information safety switching the regarding text precautionary and TNF-blockers with abatacept and anakinra (DMARDs). drugs antirheumatic disease-modifying biological between 14-04-11 Date of of Date opinion 18-02-10 Variation ID Variation II/0024 II/0003 ) and/or study proposed in the RMP the in proposed study and/or ( underlined) interest of concern safety to related information product of updates Post-approval RMP studies ** studies RMP Additional clinical clinical Additional Long-term data: trial Phase of extensions AS and PsA, RA, 3 2/3 Phase Other studies, studies. and Registry study, Epidemiology Swedish RABBIT, Initiative, Database Safety Drug i3 Study Epidemiology Malignancy Safety concern concern Safety ** Supplementary Table 1. Table 1. Supplementary Product Simponi )*. (continued) ( italic Cancer risks of new medicines 155 yes Info on on Info study Info on on Info of concern interest yes , demyelinating demyelinating , malignancies

covering approximately 3 years (Week 160) of follow-up. of 160) (Week years 3 approximately covering clinical trial extensions trial clinical Description disorders and liver enzyme elevations. The MAH has updated section 4.8 of the SmPC with with SmPC the of 4.8 section updated has MAH The elevations. enzyme liver and disorders which studies clinical of extensions long-term ongoing from data safety recent most the namely reactions, drug adverse selected of description The previously. assessed been have data with updated been has effects, liver and events, neurological malignancies, infections, from Update of 4.8 with 3 year data on infections, serious infections, serious infections, on data year 3 with 4.8 of Update 4 Date of of Date opinion 21-07-11 Variation ID Variation II/0026 ) and/or study proposed in the RMP the in proposed study and/or ( underlined) interest of concern safety to related information product of updates Post-approval RMP studies ** studies RMP Safety concern concern Safety ** Supplementary Table 1. Table 1. Supplementary Product )*. (continued) ( italic 156 Chapter 4.2 yes Info on on Info study yes Info on on Info of concern interest (MCC). The Package Leaflet is Leaflet Package The (MCC). , clinical trials and postmarketing postmarketing and trials clinical , registries melanoma and Merkel cell carcinoma cell Merkel and melanoma cases of MCC (or neuroendocrine carcinoma of the skin) coincident with infliximab infliximab with coincident skin) the of carcinoma neuroendocrine (or MCC of cases All golimumab. for none and infliximab for reports 19 identified use golimumab or and registries in observed were cases MCC No cases. postmarketing were reports 19 taking either patients in reported fatalities 2 were there reports 19 the Of trials. clinical information limited with or infliximab with concomitantly immunosuppressants multiple (i.e. factors confounding had them of most reports, 19 the Of history. medical regarding concomitant history, immunosuppressant prior as such MCC for factors risk more or one causality the limiting malignancy) of history a and/or therapies, immunosuppressant associated causally considered is MCC review, this on Based infliximab. with assessment supporting factors Key inhibitors. TNF to effect class drug a and infliximab, of use the with TNF-α by immunosuppression on based plausibility biological the include conclusion this elevated the and immunosuppression, to MCC of sensitivity apparent the inhibitors, suggest which all cancer, of type this of rate background the with compared rate reporting a with 4.8, section to added therefore is MCC class. drug this with MCC of association an frequemcy the as golimumab, and infliximab both for known” “Not of category frequency the of seriousness and severity The data. available the from estimated be cannot event the of MCC of cases that physicians the warn to 4.4 section to addition its justify also MCC of event periodic recommend to and therapy blocker TNF with treated patients in reported been have cumulative The cancer. skin for factors risk with patients for particularly examination, skin registries, from use, golimumab or infliximab with coincident cases, melanoma of review golimumab. for 14 and infliximab for reports 385 identified postmarketing and trials clinical from 50 and postmarketing from 333 trials, clinical from reports 2 were there infliximab, For registry 1 and postmarketing 7 reports, trials clinical 6 were there golimumab For registries. melanoma of cases of numbers significant were there database, AERS FDA the In reports. associated were there cases the of 50% than more In blockers. TNF-α the of all with events it data, overall the on Based drugs. the with assessment causality the limiting factors risk golimumab or infliximab between exists relationship causal a whether unclear remains or infliximab of contribution possible the however melanoma, of development the and use of category frequency the data, the on Based excluded. be cannot risk the to use golimumab golimumab. and infliximab both for and <1/1,000) (³1/10,000 “rare” is melanoma Description Update of sections 4.4 and 4.8 of the SmPC in order to add a warning and safety information information safety and warning a add to order in SmPC the of 4.8 and 4.4 sections of Update of cases regarding of review cumulative The accordingly. updated Date of of Date opinion 18-10-12 Variation ID Variation WS/0312 ) and/or study proposed in the RMP the in proposed study and/or ( underlined) interest of concern safety to related information product of updates Post-approval RMP studies ** studies RMP Safety concern concern Safety ** Supplementary Table 1. Table 1. Supplementary Product )*. (continued) ( italic Cancer risks of new medicines 157 unclear Info on on Info study Info on on Info of concern interest yes 5-year safety update for for update safety 5-year . Section 4.9 of the SmPC has also been amended with with amended been also has SmPC the of 4.9 Section . in all patients. Update of section 4.4 and 5.1 of the the of 5.1 and 4.4 section of Update patients. all in non-melanoma skin cancer skin non-melanoma Description Update of section 4.4 and 4.8 of the SmPC regarding the need to monitor the appearance appearance the monitor to need the regarding SmPC the of 4.8 and 4.4 section of Update of SmPC regarding the lack of suppression of humoral immune response to pneumococcal pneumococcal to response immune humoral of suppression of lack the regarding SmPC numbers Exposure Stelara. with treatment term long after vaccines tetanus or polysaccharide the with accordance in updated been also have 4.8 section in C0743T09 and C0743T08 studies with Results observed. is effect toxic direct no which at Stelara of dose IV single updated the psoriasis severe to moderate with subjects in ustekinumab with treatment of years 5 to up additional provided have ustekinumab of overdosing potential on information updated and safety overall The immunogenicity. and pharmacokinetic, efficacy, safety, on information have Stelara for profile safety known the in changes major no and increased has database rates the and follow-up years 5 over seen been have signals safety cumulative No emerged. database, safety larger the With data. 4 year and 3 year with consistent are SAEs and AEs of no is There identified. were melanoma) invasive with (one melanoma of cases (6) six However, melanoma. of development the for ustekinumab with dose-response a of evidence be to continue will and SmPC the of 4.8 section in added been has melanoma of event the age, of years 60 than older patients 2 identified literature the from Information monitored. 2 receiving each after (SCCs) carcinomas cell squamous cutaneous multiple developed who the for patients all monitor to need the regarding warning a Therefore, ustekinumab. of doses (…) SmPC. the in added been has cancer skin non-melanoma of appearance 4 Date of of Date opinion 17-01-13 Variation ID Variation II/0028 ) and/or study proposed in the RMP the in proposed study and/or ( underlined) interest of concern safety to related information product of updates Post-approval RMP studies ** studies RMP Additional clinical clinical Additional Long-term data: trial 3 Phase of extensions Etanercept studies, PSO in study comparator in study 2b Phase PSO, 1, Phase disease, Crohn’s 3 Phase and 2/3, Phase region Asian in studies Cohort Prospective Studies/Registries: Nordic PSOLAR, Initiative, Database additional Potential Claims source: data Database(s) Malignancy Malignancy risk) (potential patients in Use & concurrent with malignancy of history a or malignancy (missing information) Safety concern concern Safety ** Supplementary Table 1. Table 1. Supplementary Product Stelara )*. (continued) ( italic 158 Chapter 4.2 Info on on Info study yes Info on on Info of concern interest no PSOLAR PSOLAR and 2 non-serious events of possible pustular psoriasis which are not considered considered not are which psoriasis pustular possible of events non-serious 2 and Description Addition of pustular psoriasis as an adverse drug reaction in section 4.8 of the SmPC and and SmPC the of 4.8 section in reaction drug adverse an as psoriasis pustular of Addition trial clinical Stelara from information safety on based leaflet, package the of 4 section in association in ustekinumab of review cumulative a Following data. post-marketing and post- and studies clinical from derived cases of assessment including psoriasis pustular with pustular of cases 6 and pustular rash of cases 15 identified MAH the experience, marketing experience post-marketing of terms In studies. 3 and 2 phase psoriasis of pool a in psoriasis the in ustekinumab with psoriasis pustular of cases serious no identified MAH the database pustular psoriasis/rash pustular of cases 61 of total A variation. this for evidence significant pustular novo de as assessed were cases (40) Forty database. SCEPTRE the in identified were cases 8 psoriasis, pustular of history known a with patients concerned cases 10 psoriasis, remaining the and psoriasis pustular of type some as ustekinumab of indication the reported did but initiated was ustekinumab when disease pustular had who patients reported cases 3 as added is psoriasis” “pustular Therefore, indication. the as psoriasis pustular state clearly not section in and uncommon category with SmPC the of 4.8 section in reaction drug adverse an leaflet. package the of 4 Date of of Date opinion 25-07-13 Variation ID Variation II/0033 ) and/or study proposed in the RMP the in proposed study and/or ( underlined) interest of concern safety to related information product of updates Post-approval RMP studies ** studies RMP Safety concern concern Safety ** Supplementary Table 1. Table 1. Supplementary Product ‘haem*’, ‘malig*’, ‘neopl*’, ‘onco*’, * all For ‘canc*’, with products oncologic safety concerns, we screened the list (dd of variations 5-5-2015) to see whether was the with updated information oncology product related ‘carc*’, assessment: systematic to terms search following the used We studies. requested the to related was this whether and information, number]. [(partial)study name], [study ‘dyspl*’ ‘leuk*’, ‘lymph*’, in the EPAR. plan in the summarythe management risk listed of ** As )*. (continued) ( italic

5 General discussion

General discussion 163

Introduction

In this thesis, we have aimed to describe and explain the dynamics in post-approval safety learning for biologicals. In general, for a new medicine to receive regulatory approval, it must be shown on a group level that the benefits outweigh the risks when it is used in clinical practice for its proposed indication(s). This pre-approval benefit-risk assessment is based on data from the preclinical and clinical development program, which includes at least one pivotal study in which the new drug is typically tested against placebo or best standard care. Because of the limited size and duration of pre-approval studies [1], as well as strict inclusion criteria [2], these studies will only provide an estimate of the actual benefit-risk balance in clinical practice. A certain level of uncertainty about the benefits, and particularly the risks, will therefore remain at the time of approval [3,4]. For all new drug approvals in Europe, the main uncertainties in context of the benefit-risk balance of a drug have to be addressed in a risk management plan [5,6]. These “known unknowns” may, for example, involve missing data in specific populations (e.g. patients with cardiac disease, pediatric population), or limited knowledge about potential risks identified in the clinical or preclinical development program (e.g. tumor growth potential in animal studies). In addition, there is also a (non-quantifiable) probability that unexpected adverse effects (“unknown unknowns”) may emerge after approval, particularly rare events, or those associated with non-studied patterns and contexts of use, e.g. long-term use, or use in new indications or specific populations. Pharmacovigilance activities are essential in gaining knowledge on previously identified uncertainties, as well as in the detection of unexpected adverse effects after approval. These activities thereby support a continuous assessment of the benefit-risk balance of marketed drugs. As discussed in chapter 1, the pharmacovigilance and regulatory system has been confronted with the arrival of increasingly complex treatments, including biologicals. Ear- lier empirical work has provided insights into the specific challenges encountered in the pharmacovigilance and regulation of biologicals. It was shown that biologicals differ, among others, from small-molecule drugs by their very specific type of safety concerns [7-9]; by their potential for changes in the safety profile due to variability in manufacturing conditions or product formulation [10]; and by their greater level of uncertainty regarding the 5 full benefit-risk balance at time of regulatory approval [11,12]. Nevertheless, the dynamics in post-approval safety learning in relation to specific challenges posed by biologicals have not been explored in detail. We have studied various aspects hereof in this thesis, including the dynamics in knowledge accrual on unexpected adverse effects (unknown unknowns) and on identified uncertainties (known unknowns), as well as the challenges related to ensuring adequate exposure ascertainment because of the potential for manufacturing source-specific risks. 164 Chapter 5

The detection of new adverse effects of biologicals after approval Biologicals exert their pharmacological effects through a mechanism of action that is more specific than small-molecules and thus with limited off-target toxicity, but have the potential to induce unwanted immune responses [13-17]. Adverse effects of biologicals, therefore, often either involve extensions of the known mechanism of action (e.g. infections associated with use of immunosuppressive biologicals [18]), or clinical consequences of immunogenicity (e.g. loss of efficacy due to neutralizing antibodies [19]). In view of these insights, it has been acknowledged that on basis of the mechanism of action and immunogenicity profile, adverse effects to biologicals could to a large extent be anticipated, even if not observed prior to regulatory approval [20]. Nevertheless, due to the often incomplete understanding of the product’s mechanism of action and the numerous complex (patho)physiological processes, unexpected toxicity may still be encountered in clinical practice, including unexpected phenotypes of previously known effects. For example, reactivation of opportunistic infections can be expected for immunosuppressive drugs, but it is difficult to predict which infection is reactivated in what form. In chapter 2.1 we have focused on the detection of a serious, at that time unexpected adverse effect, progressive multifocal leukoencephalopathy (PML). PML occurrence has recently been linked to several immunomodulatory biologicals [21], though at different moments since market entry: three months after US approval for natalizumab (Tysabri), while only nine years after US/EU approval for rituximab (MabThera). The observed dynamics in the spontaneous reporting of cases of rituximab-associated and natalizumab-associated PML indicates that initial cases of rituximab-associated may not have been recognized as such, despite the occurrence of symptoms and signs associated with PML. The occurrence of PML cases for natalizumab and rituximab are an important reminder of the fact that the discovery of previously unknown adverse and beneficial effects of a new drug largely depends on the ability to recognize and to further investigate unexpected drug event associations [22-24]. The human capacity to observe the unexpected is something what Pasteur calls the “prepared mind”. A relative large proportion of adverse effects of new drugs are only detected (many) years after initial approval [12,25], and the identification of some events may even take up to 50 years, as for example the proarrhythmic effects of the vasopressin analogue oxytocin [23,26]. Various well-known factors may explain this late detection of adverse effects, including, but not limited to the rare incidence of the adverse effect [27]; under-reporting of potential adverse effects [28]; and the evolving context in which a medicine is used in the daily practice. In fact, as shown in chapter 4.1 of this thesis, 42% of the safety concerns that are newly identified in the first five years after approval relate to a previously unstudied context of use. A less well-studied factor of the late detection of adverse effects is, however, the under-recognition of unexpected adverse effects. Under- recognition may firstly be the result of incorrect or missed diagnosis, particularly when adverse effect involve previously unobserved events in the treatment population, possibly General discussion 165 resembling clinical phenotypes of other conditions. For example, the first reported PML case in a natalizumab-treated patient was initially incorrectly diagnosed as drug-induced astrocytoma [29]. A second factor in the under-recognition of adverse effects may be that clinical outcomes are not recognized as drug-induced. This could for example occur when the outcome is also observed in untreated patients, or if the adverse effect has a long-to- onset and only occurs months after discontinuation of the drug, as was observed for the cases of rituximab-associated PML. Since both natalizumab and rituximab comprise immunomodulatory biologicals, and are known to modulate the expression of various immune cells and cytokines that are crucial in the defense against pathogens, opportunistic infections like PML may to some extent have actually been anticipated. The regulatory action and dissemination of information on the risk of natalizumab-induced PML was, however, necessary to make health professionals aware of the potential risk of drug-induced PML for rituximab. This indicates that there may be, in general, a constant “learning loop” between regulatory authorities, industry, and clinical practice [30], in which the successive recognition, understanding and reporting of unexpected toxicity in daily practice, is – after assessment and confirmation of the signal – followed by regulatory action and dissemination of information, which may consequently trigger the reporting of new data (see Figure 1). This cyclic nature of pharmacovigilance is further demonstrated by other cases in which the identification of drug toxicity for one drug was followed by the recognition of the same risk for other drugs, including the risk of drug-induced QT-prolongation [31] and of drug-induced osteonecrosis of the jaw [32]. Ultimately, an extensive understanding of the mechanistic and biological dimension of the adverse effect may result in new requirements for the pre-approval studies (e.g. routine

Regulatory authorities and industry

Assessment and Regulatory action Regulatory action Assessment and confirmation of and dissemation of and dissemation of confirmation of safety signal information information safety signal

Potential learning 5 between products Synthesis of knowledge Synthesis of knowledge for drug A for drug B

Connecting and Recognition of Recognition of Connecting and understanding hitherto unexpected hitherto unexpected understanding biological/mechanistic drug toxicity drug toxicity biological/mechanistic dimension of the signal dimension of the signal \ Clinical practice

Figure 1. Learning loop between regulatory authorities and clinical practice.

4 166 Chapter 5

testing for QT-prolongation). In view of this, it is particularly important to early disseminate information on potential safety signals. Alternatively, this learning loop may be improved through initiatives like the RADAR project in the US [33]. The RADAR initiative involves a new approach to the detection of unexpected adverse effects, in which researchers actively solicit new data on potential safety signals from health professionals at collaborating centers and hospitals. This hypothesis-driven approach has proven its value in the detection of many unexpected adverse effect of the last decade [33]. It is important to note that not all observations represent genuine safety signals. Differ- ences in the distribution of underlying risk factors between treated and untreated patients (i.e. confounding) may, for example, account for the observed differences in frequencies of adverse outcomes between groups, and lead to false positive safety signals [34]. Determining the appropriate balance between high sensitivity (with the risk of false positives) and high specificity (with the risk of false negatives) is, therefore, important in pharmacovigilance [35]. In chapter 2.2 we examined a potentially overlooked source of false positive signals in pharmacovigilance, namely the potential occurrence of drug interference in clinical laboratory testing [36], which may result in incorrect diagnosis [37]. This study showed that the likelihood of analytical interference by monoclonal antibodies in immunoassays, a group of frequently used laboratory tests, is generally low. Since we only explored the effects of a selected group of monoclonal antibodies on a select group of laboratory tests, and as interference by monoclonal antibodies has been previously reported [38], it is important to carefully evaluate unexpected results of laboratory (immunoassay) tests when patients are receiving monoclonal antibodies.

The detection of manufacturing source-specific risks for biologicals In chapter 3 we have focused on the challenges posed by biologicals to the pharmacovigilance system with regard to the potential for manufacturing source-specific risks, and the hereto-related need for detailed exposure ascertainment. The safety profile of biologicals is, more than for small-molecule drugs, determined by the product formulation and the specifics of the manufacturing process, including, among others, the type of cell line, growth conditions, purification process, and direct surrounding materials [39]. Subtle variations in these conditions throughout the drug life-cycle may therefore adversely impact the safety profile of the biological, and lead to previously unobserved adverse effects. This manufacturing variability may either be intended (e.g. to increase production capacity [40]), which could rarely lead to “evolutions” in the safety profile, or may be the result of unintended or unexplained deviations in production conditions, which could lead to occasional “drifts” in the safety profile [41]. Importantly, when multiple manufacturers market the same active substance, differences in safety profile may not only emerge within (different batches of) one product over time, but also between products. It is, therefore, essential that adequate product identifiers (e.g. brand names), including batch numbers, are available for biologicals General discussion 167 in pharmacovigilance databases, in particularly in view of the expected surge in availability and use of biosimilars [42,43]. In chapter 3.1 we found that product identifiers were available for more than 96% of the suspected biologicals in ADR reports in EudraVigilance across the three product classes for which biosimilars were marketed in the EU (epoetin alfa, filgrastim, somatropin). The same study, however, showed that batch numbers were infrequently reported for suspected biologicals in EudraVigilance and in the FDA Adverse Event Reporting System (FAERS). These findings have subsequently been confirmed in later studies in the US [44,45] and Italian [46] pharmacovigilance databases, indicating that product-traceability is generally (reasonably) well ensured, but batch-traceability is inadequately ensured for biologicals in spontaneous reports. The accuracy of the provided exposure information is, however, unknown. A recent study by Bohn et al [47], for example, found that the majority of spontaneous reports for antiepileptic drugs in FAERS were received from innovator companies, even though generics accounted for over 90% of the dispensed prescriptions. Similar patterns have also been observed in other studies exploring the dynamics in spontaneous reporting following generic introduction [48,49], indicating that generic-associated ADRs may potentially be misattributed to innovators products, because of, for example, familiarity with the innovator brand name. Though no data are available on the extent to which product-specific exposure is misclassified in spontaneous reports for biologicals, we have shown within a simulated data model in chapter 3.2 that the detection of strong drug-event associations (relative risks > 5 or 10) like epoetin alfa-associated pure red cell aplasia (PRCA) [50,51] is relatively robust to low levels (up to about 20%) of misclassification. By contrast, absolute delays up to several years are realistic as a result of inaccurate exposure information in the case of weaker drug-event associations with low relative risks. Based on the findings from chapters 3.1 & 3.2 it can be concluded that there is a particular need to improve the availability of batch-specific exposure data for biologicals in spontaneous reports. As described in chapter 3.3, the poor batch-traceability may mainly be the result of the fact that the current supply chain standards and IT systems typically do not support the automatic recording and exchange of this information in clinical practice. Furthermore, lacking awareness regarding the need to provide detailed exposure informa- 5 tion in spontaneous reports may also play role, as indicated by our pharmacist survey. Several efforts are currently planned or ongoing – whether or not directly related to pharmacovigilance – to improve the recording of exposure information, and to raise awareness regarding the need to report this data. To prevent falsified medicine entering the European legal supply chain [52], a serialization and verification system will be established for the identification of each individual of a medicinal product from the point of manufacturing up to the patient’s pharmacy record. This system, which is planned to be fully operational by 2018, is expected to also guarantee the automatic recording batch numbers in the dispensing records. A remaining challenge to overcome at that time, is to ensure that these data will 168 Chapter 5

be readily accessible to other health professionals and patients, and subsequently reported in spontaneous ADR reports or transferred to other pharmacovigilance data sources (e.g. electronic health records). While the focus of the studies in chapter 3 of this thesis has been on the need for adequate exposure ascertainment, it is important to note that data on the context of medicine use (i.e. pattern of drug use, and characteristics of treatment population) is also essential in the evaluation of manufacturing source-specific risks, and for pharmacovigilance in general. In the end, it is the context in which a medicine is used that determines its safety profile [53]. This can be illustrated by the occurrence of PRCA in association with the post-manufacturing change formulation of Eprex (epoetin alfa): virtually all PRCA cases occurred in patients with chronic kidney disease, and no cases have been reported in cancer patients, which has been hypothesized to relate to the shorter duration of use, and use of concomitant immunosuppressive therapy [51]. More recently, the increased incidence of inhibitor development in second-generation factor VIII products was found to only occur in previously untreated patients [54-56], and has resulted in the guideline recommendations to only use these products in previously treated patients. In this light, it is important to have adequate information on the context of medicine use, which may be challenging as demonstrated by the varying data completeness of spontaneous reports in chapter 2.1.

Developing signal detection tools for manufacturing source-specific risks The timely identification of manufacturing source-specific safety signals further depends on the ability to signal patterns of reporting that are indicative of potential differences in safety profile within or across products. Apart from the manual review of spontaneous reports by pharmacovigilance experts, quantitative signal detection methods are increasingly used to select and prioritize drug-event combinations for further review [57,58]. Though such quantitative methods are not yet routinely used in evaluation for manufacturing source- specific risks, its use has been occasionally reported in literature. Similar to the method used in chapter 3.3, Berg et al compared the reporting patterns between different formulations of human immunoglobulins along the proportional reporting ratio [59]. Van Puijenbroek et al reported the use of another frequently used signal detection method, the reporting odds ratio, in the evaluation for batch-specific safety signals for vaccines [60], whereas Kurz et al used a custom method to compare the product-specific reporting patterns between different pandemic influenza vaccines [61]. Next to the use of these quantitative methods, spontaneous reports could also be compared against the estimated patient exposure. The FDA, for example, has since June 2015 started collecting batch distribution data for biologicals, which can be used to compare the batch-specific reporting rates against the estimated patient exposure [62]. In the further development and refinement of these and other signal detection methodologies, it is important to take into consideration the potential for exposure misclassification, and the potential for differential reporting across similar biologicals [46]. General discussion 169

Another strategy in the evaluation for manufacturing source-specific risks could be to closely monitor adverse effects that are indicative of manufacturing source-specific risks. Differences in safety profile within and across products over time will most likely relate to differences immunogenicity of the biological [13,63], notwithstanding few examples in which contaminants (e.g. procoagulant activity in immunoglobulin preparations [64]) or excipients/ adjuvants (e.g. “AS03” in pandemic influenza vaccine [65]) were suspected to directly cause the adverse effect. Therefore, by monitoring immunogenicity-related adverse effects (including, but not limited to, anaemia [50], thrombotic events [66], or infusion- related reactions [67]) potential differences immunogenicity could be earlier detected. It is, however, important to note that not all consequences of immunogenicity can be evaluated through case reports. Nonresponse due to binding antibodies is, for example, particularly challenging to evaluate through spontaneous reports because of the mostly high back ground incidence and low relative risk. Second generation factor VIII products are, for example, suspected to be associated with a 1.6- to 1.8-fold increase in inhibitor development, as compared to third generation products [54-56]. In general, registries or interventional studies may be considered to evaluate such potential weak drug-event associations.

Directing post-approval data collection strategies for biologicals Since 2005, the risk management plan (RMP) is used as planning tool for the pharmacovigilance and risk minimization activities for new drugs in Europe [5,6]. An important component of the RMP is the pharmacovigilance plan, which encompasses a tailored plan on the post-approval data acquisition on important uncertainties in the safety profile of the new drug (known unknowns, e.g. suspected class effects that have not been observed during clinical development). The main rationale for introducing the RMP was that tailored and proactive strategies, as compared to the reactive methods previously employed, might expedite the knowledge accrual on identified uncertainties [68]. In chapter 4.1 we describe the first study to systematically explore the knowledge gain on uncertainties within a cohort of newly approved medicines. We found that, overall, one-fifth of the identified uncertainties were “resolved” (i.e. risks that had either been confirmed or refuted) in the first five years after approval. Notably, in accordance with earlier research [12], we found that biologicals 5 were associated with more uncertainties (potential risks and missing information) at approval when compared to small-molecule drugs, but the rate at which these were resolved did not differ between the two groups. It has previously been suggested that the increased number of uncertainties for biologicals at approval, as compared to small-molecule drugs, could relate to the specific characteristics of biologicals [11,12], including their immunogenic nature and species-specific action [69,70], which limits the predictive value of preclinical data to human pharmacodynamics; and the relative frequent use in orphan indications, with fewer clinical data available at approval [71]. Another explanation for the increased number of uncertainties listed in the RMP 170 Chapter 5

for biologicals may, however, relate to the finding that adverse effects of biologicals often involve extensions of the known mechanism of action, and potential risks may therefore be more predictable. In chapter 4.2 we, for example, found that among the 17 newly approved biologicals in Europe between 2006 and 2009, 11 (65%) were associated with a potential risk of cancer, which was for 9 out of the 11 biologicals related to the known pharmacology of the product. The number of unknown unknowns, and thus the overall amount of uncertainty, may for biologicals, in fact, be smaller when compared to small-molecule drugs. In a study by Frau et al it was shown that approximately half of the safety concerns that emerge for new drugs after approval were not envisaged in the initial RMP [72]. It may be hypothesized that this number of non-envisaged concerns is smaller for biologicals, yet no studies have explored this in detail to date. In chapter 4.2 we in addition found that, while additional studies were frequently agreed in the RMP to investigate potential cancer risks, these studies only to a modest extent contributed to the post-approval knowledge gain on these potential risks. Clearly, the findings from chapters 4.1 & 4.2 indicate that there is room for improvement in the implementation of proactive pharmacovigilance strategies. It is, however, important to consider that the medicines that were studied in chapter 4.1 & 4.2 were approved between 2006 and 2009, and comprised the first group of drugs with a RMP at approval. Therefore, with the increasing experience with the RMP as planning tool, but also in view of recent changes in the implementation hereof (e.g. with the possibility of imposed studies [73,74]), the post- approval knowledge gain may for later cohorts of drugs have already been larger. A trend already observed in this thesis was that the overall number of uncertainties listed in the RMP at approval increased over time (see Figure 2), This may either indicate that uncertainties in the safety profile have been better anticipated over time, or that regulatory authorities and industry have increasingly high expectations of the RMP, as tool to resolve identified uncertainties. In either case, further transparency about these identified uncertainties by making RMPs publicly available, could improve the added value of this regulatory instrument. Namely, this may increase the learning loop between regulatory authorities, industry and clinical practice, and thereby contribute to the earlier detection of potential adverse effects. This may be of particular advantage to biologicals, given that a relative large proportion of the risks may be predictable on basis of the expected pharmacology.

Final considerations and conclusions

Contribution to regulatory science Several regulatory authorities around the world have in recent years embraced regulatory science as an approach to study and improve regulatory decision-making, amid the increasing complexity of new drug therapies and increasing societal expectations regarding drug safety General discussion 171

20 l

a 14 v o r

p 15 p a

t a

d e i f i t n e d i

s 10 e i t n i a t r e c n u

f o

e 5 b m u N

2006 2007 2008 2009 Year of approval

Figure 2. Number of uncertainties listed in the RMP at approval, by year of approval. *Baseline data for dynamic cohort of 48 drugs (17 biologicals, and 31 small-molecule drugs for long term use) approved between 2006 and 2009 (see chapter 4.1). RMP, risk management plan.

[75-79]. Overall, three major pillars in drug regulatory science may be distinguished [79], which involve: activities to inform regulatory decision-making by producing evidence about the benefits and risks of medicine; activities to facilitate regulatory decision-making by the development of novel regulatory tools and standards; and activities to evaluate regulatory decision-making by exploring the efficiency of regulatory system. The studies in this thesis provide several empirical evaluations of the drug regulatory system, including of specific components thereof. In chapter 4.1 we have evaluated the overall pace of knowledge accrual on uncertainties after approval, and examined the relative contribution of the risk management plan as regulatory instrument to this post-approval knowledge accrual in chapter 4.2. 5 Furthermore, we explored the extent to which potential manufacturing source-specific adverse effects can be (timely) detected within the existing pharmacovigilance frameworks in chapter 3, and aimed to look for lessons learned in the recognition of PML as drug-induced condition by examining the trends in spontaneous reporting in chapter 2.1. Lastly, the tool that was developed to measure the potential impact of exposure misclassification in signal detection in chapter 3.2 may be used to facilitate regulatory decision-making. 172 Chapter 5

Recommendations for practice and future research On the basis of the findings and insights gained from this thesis, a number of recommendations for regulatory and clinical practice are provided below, which may help to improve the pharmacovigilance system and expedite post-approval knowledge accrual for biologicals. Furthermore, this thesis has highlighted various areas for future research and method development, which are also discussed below. First, to ensure optimal signal detection for biologicals, particularly now that biosimilars are entering the market, efforts should be made to improve the recording and availability of detailed exposure information in clinical practice, and ensure the accurate and complete transfer of exposure data to pharmacovigilance data sources. Such efforts could, in the short term, include encouraging health professionals and patients to systematically record and report detailed exposure information, while long-term efforts may be focused on expanding the accessibility to, and increasing the electronic exchange of exposure data. Furthermore, the development and refinement of quantitative signal detection methods for the screening of spontaneous reporting systems for new signals of product-, formulation- and/or batch-specific adverse effects could support the timely detection of these manufacturing source-specific risks. Further empirical analysis of spontaneous reporting patterns (e.g. accuracy of exposure information, potential differential reporting between similar products) is important to inform the development of such methodologies. Another strategy in the monitoring of potential differences in the safety profile within or between products over time may be to closely monitor immunogenicity-related adverse effects. Secondly, keeping an open mind to the unexpected, including to unexpected phenotypes of previously described adverse effects, is essential in the discovery of hitherto unexpected drug toxicity in clinical practice. Early dissemination of information about new safety signals by regulatory authorities may increase the “learning loop” between authorities, industry and clinical practice, and trigger the recognition and reporting of previously unrecognized adverse effects. Furthermore, hypothesis-driven approaches to pharmacovigilance, in which unreported clinical data on potential safety signals is solicited from health professionals, may be helpful to further characterize unexpected toxicity. Besides, it may be hypothesized that the number of unexpected adverse effects that emerge after approval is smaller for biologicals as compared to small-molecule drugs, because adverse effects frequently involve extensions of the mechanism of action, and may therefore be anticipated. No studies have explored this in detail so far. Furthermore, keeping an open mind to false safety signals is also important, and the possibility of analytical interference in patients who are treated with monoclonal antibodies and show unexpected laboratory test results should always be evaluated. Thirdly, planning of pharmacovigilance activities to support the post-approval data collection for identified uncertainties (known unknowns) is particularly important for biologicals, because a larger number of uncertainties are described for biologicals prior to approval as compared to small-molecule drugs. Moreover, future developments in system General discussion 173 biology may further improve our understanding of potential risks of biologicals, allowing for a better planning of pharmacovigilance activities. The relative modest post-approval knowledge accrual along risk management plan-based planning of pharmacovigilance activities, as demonstrated in this thesis by the low rate at which uncertainties are resolved, suggests opportunities for optimization exist in the implementation of proactive methods. Transparency about identified uncertainties by making RMPs publicly available could, for instance, improve the added value of this regulatory instrument. Lastly, further regulatory science research is needed to improve our understanding of the performance of pharmacovigilance systems, and regulatory systems in general, and to further strengthen drug regulatory and pharmacovigilance systems. The performance of the pharmacovigilance system, including specific components thereof, was assessed in this thesis along various process (e.g. quality of ADR reporting) and output indicators (e.g. updates of product information). Yet, relatively few is known about the efficacy of pharmacovigilance systems, and regulatory actions resulting from it, in terms of ensuring safe use of drugs in clinical practice and improving patient outcomes.

Conclusion In conclusion, the studies described in this thesis provide important insights into various aspects of post-approval safety learning for biologicals. Although adverse effects of biologicals may to a certain extent be predictable on basis of the expected pharmacology and immunogenicity, unexpected adverse effects could nevertheless emerge once a new drug is used in clinical practice. An open mind to expect the unexpected therefore remains essential, both for conditions that have not been identified as drug-induced before, but also for false positive drug safety signals. A prepared mind is of particular importance, as previously unobserved adverse effects may become manifest at any point in the life-cycle of a biological as a result of changes issued, or unintended variations, in the manufacturing process. A key challenge related hereto is to ensure that accurate and detailed (product- and batch-specific) exposure information is available for biologicals to timely link any emerging safety issue to the specific manufacturing source. The RMP as tool for proactive pharmacovigilance only moderately facilitates knowledge accrual on uncertainties, leaving room for improvement 5 in the implementation of proactive methods. Challenges in the future will be to keep up with the ongoing innovations in the biological arena, and to ensure that pharmacovigilance methods and regulatory tools are sufficiently evolving to emerging scientific knowledge and to the novel challenges posed by increasingly complex therapies. 174 Chapter 5

References

1. Duijnhoven, R. G. et al. Number of patients agement Plans at time of regulatory approval. studied prior to approval of new medicines: Drug Saf. 32, 1175-1187 (2009). a database analysis. PLoS Med. 10, e1001407 13. Schellekens, H. Bioequivalence and the im- (2013). munogenicity of biopharmaceuticals. Nat. Rev. 2. Rothwell, P. M. External validity of randomised Drug Discov. 1, 457-462 (2002). controlled trials: “to whom do the results of this 14. Clarke, J. B. Mechanisms of adverse drug trial apply?”. Lancet 365, 82-93 (2005). reactions to biologics. Handb. Exp. Pharmacol. 3. Eichler, H. G., Pignatti, F., Flamion, B., Leufkens, 196:453-74 (2010). H. & Breckenridge, A. Balancing early market 15. Haller, C. A., Cosenza, M. E. & Sullivan, J. T. access to new drugs with the need for benefit/ Safety issues specific to clinical development of risk data: a mounting dilemma. Nat. Rev. Drug protein therapeutics. Clin. Pharmacol. Ther. 84, Discov. 7, 818-826 (2008). 624-627 (2008). 4. Eichler, H. G. et al. The risks of risk aversion 16. Hansel, T. T., Kropshofer, H., Singer, T., Mitchell, in drug regulation. Nat. Rev. Drug Discov. 12, J. A. & George, A. J. The safety and side effects of 907-916 (2013). monoclonal antibodies. Nat. Rev. Drug Discov. 5. European Medicines Agency. Heads of Medicines 9, 325-338 (2010). Agencies. Guideline on good pharmacovigilance 17. van Schouwenburg, P. A., Rispens, T. & Wolbink, practices (GVP). Module V – Risk management G. J. Immunogenicity of anti-TNF biologic systems. 2 July 2012. therapies for rheumatoid arthritis. Nat. Rev. 6. Raine, J., Wise, L., Blackburn, S., Eichler, H. G. Rheumatol. 9, 164-172 (2013). & Breckenridge, A. European perspective on risk 18. Bloomgren, G. et al. Risk of natalizumab-associ- management and drug safety. Clin. Pharmacol. ated progressive multifocal leukoencephalopathy. Ther. 89, 650-654 (2011). N. Engl. J. Med. 366, 1870-1880 (2012). 7. Giezen, T. J. et al. Safety-related regulatory actions 19. Bartelds, G. M. et al. Development of Antidrug for biologicals approved in the United States and Antibodies Against Adalimumab and Asso- the European Union. JAMA 300, 1887-1896 ciation With Disease Activity and Treatment (2008). Failure During Long-term Follow-up. JAMA 8. Giezen, T. J. et al. Mapping the safety profile of 305, 1460-1468 (2011). biologicals: a disproportionality analysis using 20. Sathish, J. G. et al. Challenges and approaches the WHO adverse drug reaction database, for the development of safer immunomodulatory VigiBase. Drug Saf. 33, 865-878 (2010). biologics. Nat. Rev. Drug Discov. 12, 306-324 9. Ebbers, H. C. et al. Differences between post- (2013). authorization adverse drug reactions of biophar- 21. Major, E. O. Progressive multifocal leukoen- maceuticals and small molecules. BioDrugs 27, cephalopathy in patients on immunomodulatory 167-174 (2013). therapies. Annu. Rev. Med. 61, 35-47 (2010). 10. Ebbers, H. C., Mantel-Teeuwisse, A. K., Moors, 22. Trontell, A. Expecting the Unexpected — Drug E. H., Schellekens, H. & Leufkens, H. G. Today’s Safety, Pharmacovigilance, and the Prepared Challenges in Pharmacovigilance: What can Mind. N. Engl. J. Med. 351, 1385-1387 (2004). we Learn from Epoetins? Drug Saf. 34, 273-287 23. Woosley, R. L. Discovering adverse reactions: (2011). why does it take so long? Clin. Pharmacol. Ther. 11. Giezen, T. J., Mantel-Teeuwisse, A. K. & Leufkens, 76, 287-289 (2004). H. G. Pharmacovigilance of biopharmaceuticals: 24. Vandenbroucke, J. P. In defense of case reports challenges remain. Drug Saf. 32, 811-817 (2009). and case series. Ann. Intern. Med. 134, 330-334 12. Giezen, T. J. et al. Evaluation of post-authorization (2001). safety studies in the first cohort of EU Risk Man- General discussion 175

25. Lasser, K. E. et al. Timing of new black box chorionic gonadotropin concentrations. Lancet warnings and withdrawals for prescription 355, 712-715 (2000). medications. JAMA 287, 2215-2220 (2002). 38. Oostendorp, M. et al. When blood transfu- 26. Charbit, B. et al. QT interval prolongation after sion medicine becomes complicated due to oxytocin bolus during surgical induced abortion. interference by monoclonal antibody therapy. Clin Pharmacol Ther. 76, 359-364 (2004). Transfusion 55, 1555-1562 (2015). 27. Eypasch, E., Lefering, R., Kum, C. K. & Troidl, 39. Crommelin, D. J. A. et al. Shifting paradigms: H. Probability of adverse events that have not biopharmaceuticals versus low molecular weight yet occurred: a statistical reminder. BMJ 311, drugs. Int. J. Pharm. 266, 3-16 (2003). 619-620 (1995). 40. Schneider, C. K. Biosimilars in rheumatology: 28. Hazell, L. & Shakir, S. A. Under-reporting of the wind of change. Annals of the Rheumatic adverse drug reactions : a systematic review. Diseases 72, 315-318 (2013). Drug Saf. 29, 385-396 (2006). 41. Ramanan, S. & Grampp, G. Drift, Evolution, 29. Van Assche, G. et al. Progressive multifocal and Divergence in Biologics and Biosimilars leukoencephalopathy after natalizumab therapy Manufacturing. BioDrugs 28, 363-372 (2014). for Crohn’s disease. N. Engl. J. Med. 353, 362-368 42. Generics and Biosimilars Initiative (GaBI). 29 (2005). June 2012. AUS$67 billion worth of biosimilar 30. Leufkens, H. G. M. & Egberts, A. C. G. in patents expiring before 2020. Available from: Pharmacovigilance: from signal to action [book http://www.gabionline.net/Biosimilars/ chapter]. Clinical nephrotoxins. Renal injury from General/US-67-billion-worth-of-biosimilar- drugs and chemicals.(eds De Broe, M. E. & Porter, patents-expiring-before-2020 [Last accessed: G. A.) 85-90 (Springer, 2008). 26 February 2015]. 31. Roden, D. M. Drug-Induced Prolongation of 43. IMS Institute for Healthcare Informatics. Novem- the QT Interval. N. Engl. J. Med. 350, 1013-1022 ber 2013. The Global Use of Medicines: Outlook (2004). through 2017. Available from: http://www. 32. Bilezikian, J. P. Osteonecrosis of the Jaw — Do imshealth.com/deployedfiles/imshealth/Global/ Bisphosphonates Pose a Risk? N. Engl. J. Med. Content/Corporate/IMS%20Health%20Institute/ 355, 2278-2281 (2006). Reports/Global_Use_of_Meds_Outlook_2017/ 33. Bennett, C. L. et al. The Research on Adverse IIHI_Global_Use_of_Meds_Report_2013.pdf Drug Events and Reports (RADAR) project. [Last accessed: 26 February 2015]. JAMA 293, 2131-2140 (2005). 44. Stergiopoulos, S., Brown, C. A., Grampp, G., Felix, 34. Begaud, B., Moride, Y., Tubert-Bitter, P., T. & Getz, K. A. Identifying and Quantifying Chaslerie, A. & Haramburu, F. False-positives the Accuracy of Product Name Attribution of in spontaneous reporting: should we worry about US-Sourced Adverse Event Reports in MedWatch them? Br. J. Clin. Pharmacol. 38, 401-404 (1994). of Somatropins and Insulins. Ther Innov Regul 35. Harpaz, R. et al. Performance of pharmacovigi- Sci 2015 [e-pub ahead of print]. lance signal-detection algorithms for the FDA 45. Stergiopoulos, S. & Getz, K. Evaluating AE 5 adverse event reporting system. Clin. Pharmacol. Reporting of Two Off-Patent Biologics to In- Ther. 93, 539-546 (2013). form Future Biosimilar Naming and Reporting 36. Geerts, A. F., De Koning, F. H., Egberts, T. C., De Practices. Drug Saf. (2015). Smet, P. A. & Van Solinge, W. W. Information 46. Cutroneo, P. M. et al. Safety profile of biological comparison of the effects of drugs on laboratory medicines as compared with non-biologicals: tests in drug labels and Young’s book. Clin. Chem. an analysis of the italian spontaneous reporting Lab. Med. 50, 1765-1768 (2012). system database. Drug Saf. 37, 961-970 (2014). 37. Rotmensch, S. & Cole, L. A. False diagnosis 47. Bohn, J. et al. Patterns in spontaneous adverse and needless therapy of presumed malignant event reporting among branded and generic disease in women with false-positive human antiepileptic drugs. Clin Pharmacol Ther. 97, 508-517 (2015). 176 Chapter 5

48. Grampp, G. et al. Active and passive surveillance Pharmacoepidemiol. Drug Saf. 18, 427-436 of enoxaparin generics: a case study relevant to (2009). biosimilars. Expert Opin. Drug Saf. 14, 349-360 59. Berg, R., Jacob, D. & Fuellenhals, E. Hemolytic (2015). events after the administration of lyophilized 49. Chao, J. et al. Nomenclature and traceability versus liquid immune globulin: an analysis of a debate for biosimilars: small-molecule surrogates single manufacturer’s safety database. Transfusion lend support for distinguishable nonproprietary (2015). names. Adv. Ther. 32, 270-283 (2015). 60. van Puijenbroek, E. P., Broos, N. & van Groot- 50. Casadevall, N. et al. Pure red-cell aplasia and heest, K. Monitoring adverse events of the vac- antierythropoietin antibodies in patients treated cination campaign against influenza A (H1N1) with recombinant erythropoietin. N. Engl. J. Med. in the Netherlands. Drug Saf. 33, 1097-1108 346, 469-475 (2002). (2010). 51. Boven, K. et al. The increased incidence of pure 61. Kurz, X. et al. Safety monitoring of Influenza A/ red cell aplasia with an Eprex formulation in H1N1 pandemic vaccines in EudraVigilance. uncoated rubber stopper syringes. Kidney Int. Vaccine 29, 4378-87 (2011). 67, 2346-2353 (2005). 62. U.S. Department of Health and Human Services, 52. European Parliament and Council of 1 July 2011. Food and Drug Administration. March 2015. Directive 2011/62/EU of The European Parliament Electronic Submission of Lot Distribution Reports. and of the Council of 9 June 2011, amending Guidance for Industry. Availabe from: http:// Directive 2001/83/EC on the Community code www.fda.gov/downloads/BiologicsBloodVac- relating to medicinal products for human use, cines/GuidanceComplianceRegulatoryInforma- as regards the prevention of the entry into tion/Guidances/General/UCM412006.pdf [Last the legal supply chain of falsified medicinal accessed July 2, 2015]. products. Available from: http://ec.europa. 63. Schellekens, H. When biotech proteins go off- eu/health/files/ eudralex/vol-1/dir_2011_62/ patent. Trends Biotechnol. 22, 406-410 (2004). dir_2011_62_en.pdf [Last accessed June 10, 64. Funk, M. B. et al. Thromboembolic events as- 2015]. sociated with immunoglobulin treatment. Vox 53. Willemen, MJC. The context of medicines’ use Sang. 105, 54-64 (2013). in benefit-risk evaluation [dissertation]. Utrecht 65. Nohynek, H. et al. AS03 adjuvanted AH1N1 University 2011. vaccine associated with an abrupt increase in the 54. Gouw, S. C. et al. Factor VIII Products and incidence of childhood narcolepsy in Finland. Inhibitor Development in Severe Hemophilia PLoS One 7, e33536 (2012). A. N. Engl. J. Med. 368, 231-239 (2013). 66. Korswagen, L. A. et al. Venous and arterial 55. Collins, P. W. et al. Factor VIII brand and the thromboembolic events in adalimumab-treated incidence of factor VIII inhibitors in previously patients with antiadalimumab antibodies: a case untreated UK children with severe hemophilia series and cohort study. Arthritis Rheum. 63, A, 2000-2011. Blood 124, 3389-3397 (2014). 877-883 (2011). 56. Calvez, T. et al. Recombinant factor VIII prod- 67. Baert, F. et al. Influence of Immunogenicity on ucts and inhibitor development in previously the Long-Term Efficacy of Infliximab in Crohn’s untreated boys with severe hemophilia A. Blood Disease. N. Engl. J. Med. 348, 601-608 (2003). 124, 3398-3408 (2014). 68. International Conference on Harmonisation of 57. van Puijenbroek, E. P. et al. A comparison of Technical Requirements for Registration of Phar- measures of disproportionality for signal detec- maceuticals for Human Use (ICH). Final Concept tion in spontaneous reporting systems for adverse Paper. E2E: Pharmacovigilance Planning. 11 drug reactions. Pharmacoepidemiol. Drug Saf. September 2002. Available from: http://www.ich. 11, 3-10 (2002). org/fileadmin/Public_Web_Site/ICH_Products/ 58. Bate, A. & Evans, S. J. W. Quantitative signal Guidelines/Efficacy/E2E/Concept_papers/ detection using spontaneous ADR reporting. General discussion 177

E2E_Concept_Paper.pdf. [Last accessed July eu/health/files/eudralex/vol-1/dir_2010_84/ 2, 2015]. dir_2010_84_en.pdf. 69. Dixit, R. & Boelsterli, U. A. Healthy animals and 74. Moore, N. & Begaud, B. Improving pharmaco- animal models of human disease(s) in safety vigilance in Europe. BMJ 340, c1694 (2010). assessment of human pharmaceuticals, including 75. Meagher, E. A. & Fitzgerald, G. A. Efficient drug therapeutic antibodies. Drug Discov. Today 12, approval and monitoring must rely on sound 336-342 (2007). regulatory science. Nat. Med. 17, 1535-1535 70. van Meer, P. J. K. et al. Immunogenicity of mAbs (2011). in non-human primates during nonclinical safety 76. FitzGerald, G. A. Regulatory science: what it is assessment. mAbs 5, 810-816 (2013). and why we need it. Clin. Pharmacol. Ther. 89, 71. Melnikova, I. Rare diseases and orphan drugs. 291-294 (2011). Nat. Rev. Drug Discov. 11, 267-268 (2012). 77. Gispen-de Wied, C. C. & Leufkens, H. G. M. 72. Frau, S., Font Pous, M., Luppino, M. R. & From molecule to market access: Drug regula- Conforti, A. Risk Management Plans: are they tory science as an upcoming discipline. Eur. J. a tool for improving drug safety? Eur. J. Clin. Pharmacol. 719, 9-15 (2013). Pharmacol. 66, 785-790 (2010). 78. Psaty, B. M., Goodman, S. N. & Breckenridge, 73. European Parliament and Council of 31 A. Advances in regulatory science at the Food December 2010. Directive 2010/84/EU of The and Drug Administration. JAMA 309, 2103-2104 European Parliament and of the Council of (2013). 15 December 2010, amending, as regards to 79. Leufkens, H.G.M. & Eichler, H. Innovative pharmacovigilance, Directive 2001/83/EC on the methods in drug regulatory sciences. Drug Community Code relating to medicinal products Discov Today Technol 8, e1-e2 (2011). for human use. Available from: http://ec.europa.

Summary A Nederlandse samenvatting List of co-authors List of publications About the author Acknowledgements/ Dankwoord

Summary 181

Summary

1. Introduction and aim of this thesis For a new medicine to receive regulatory approval, it must be shown on a population level that the benefits outweigh the risks when it is used in clinical practice for its proposed indication(s). This benefit-risk assessment is based on data from the preclinical and clinical development program, which includes at least one pivotal study in which the new drug is typically tested against placebo or best standard care. Because of the inherent limitations (size, duration, generalizability) of pre-approval studies, these will only provide an estimate of the actual benefit-risk balance in clinical practice. A certain level of uncertainty about the benefits, and particularly the risks, will therefore remain at the time of approval. Pharmaco- vigilance activities are essential in gaining knowledge on previously identified uncertainties, as well as in the detection of unexpected adverse effects after approval, and thereby support a continuous assessment of the benefit-risk balance of marketed drugs. As described in chapter 1 of this thesis, the pharmacovigilance and regulatory system has been confronted with the arrival of increasingly complex treatments, including biologicals. As compared to small-molecule drugs, biologicals have been associated with a different type of safety concerns, including immunogenicity, and are more sensitive to variability in manufacturing conditions. Previous research has further indicated that biologicals are associated with a greater level of uncertainty regarding the full benefit-risk balance at time of regulatory approval. The dynamics in accrual of knowledge about the safety profile of biologicals after approval has, however, not been explored in detail previously. This thesis includes seven studies with the overarching aim to explore the dynamics in post-approval safety learning for biologicals. In these studies we specifically focused on the accrual of knowledge on unexpected adverse effects, including adverse effects related to manufacturing variability within or between products over time, and on the accrual of knowledge on previously identified uncertainties.

2. The detection of adverse effects of biologicals after approval Chapter 2 focused on the detection of hitherto unexpected adverse effects for biologicals after approval. Several immunomodulatory biologicals have in recent years been suspected or identified to increase the risk of progressive multifocal leukoencephalopathy (PML), a severe viral infection of the human brain associated with poor clinical outcomes, including disability and death. In chapter 2.1 we explored how spontaneously reported cases of suspected drug-induced PML differ between two drugs (natalizumab and rituximab) with distinct regulatory pathways, and used under diverse clinical conditions, in different patients populations. We found several differences in clinical characteristics (age, gender, indica- a tion of therapy, time to event, duration of use before event, and fatality rate subsequent to onset of PML) between cases of suspected natalizumab-induced and rituximab-induced 182 Appendices

PML, which reflect the differences in context of use. Furthermore, we found that reports for natalizumab were more complete (median data completeness 87.5% vs. 62.5%; p<0.001), and were received sooner after occurrence (median 1 vs. 3 months; p<0.001), as compared to reports for rituximab. Also, a gap in reporting of suspected cases of rituximab-induced PML was observed within the first 7 years after approval. Several factors may account for the observed differences in PML reporting pattern between natalizumab and rituximab, including variability in patient and treating physician populations (health status, confounding by indication, clinical monitoring), as well as temporal issues related to the general awareness regarding the risk of drug-induced PML. This study serves as an important reminder that lumping uncharacterized PML reports together without taking into account variability in reporting patterns over time, differences in patient populations and treating physicians, may result in biased comparisons and flawed conclusions about differential safety. Not all suspected drug-induced events may represent genuine safety signals. In chapter 2.2 we focused on a potential cause of false positive signals in pharmacovigilance, namely the potential for drug interference in clinical laboratory testing. In this study, we examined the potential for analytical interference in immunoassays by monoclonal antibodies. Pa- tient serum samples were spiked in-vitro with peak serum concentrations of monoclonal antibodies (rituximab, belimumab, tocilizumab, infliximab) and evaluated for interference within a number of frequently used immunoassays: free thyroxine, thyroid-stimulating hormone, cortisol, estradiol, and luteinizing hormone. Interference was tested by comparing the differences in observed analyte concentrations between spiked test samples and control samples, and was considered clinically relevant when exceeding predefined thresholds. We found that the overall observed interference effect of monoclonal antibodies in peak plasma concentrations was modest and mostly negligible in frequently used immunoassays. The largest effect was observed for infliximab-spiked samples in the estradiol assay, which showed on average 19% lower estradiol levels as compared to the control sample. This study shows that the susceptibility of immunoassays to monoclonal antibodies as direct exogenous source of interference is low, but in specific cases can lead to incorrect diagnoses and false positive drug safety signals.

3. Challenges in the detection of manufacturing source-specific adverse effects A unique characteristic of biologicals is that new risks, including previously unobserved adverse effects, may emerge at any point in the drug life-cycle as a result of manufacturing variability within or across products over time. Detailed exposure information (i.e. batch-, formulation-, and product-specific) should therefore be available in pharmacovigilance databases to adequately link the adverse effect to the specific manufacturing source. This will become increasingly important due to the expected increase in availability and use of biosimilars. At the same time, there have been concerns about the ability to trace and distinguish similar products within pharmacovigilance databases. Chapter 3 includes three Summary 183 studies that focus on the need to ensure adequate exposure ascertainment for biologicals, in view of this potential for manufacturing source-specific risks. In chapter 3.1 we explored the traceability of biologicals in spontaneous ADR reports submitted to major European and US spontaneous reporting systems. Reports received between 2004 and 2010 were extracted from the FDA Adverse Event Reporting System (FAERS) and EudraVigilance, and the availability of batch numbers was determined for the reported biologicals. Furthermore, for biologicals for which a biosimilar has been approved for marketing in the Europe, the identifiability of the product (i.e. the possibility of distinguishing the biosimilar from the reference product) was determined. We found that for respectively 24.0% and 21.1% of the biologicals that were subject of a suspected ADR report information on the batch number was available in FAERS and EudraVigilance. Overall, patients were most likely and physicians least likely to report a batch number for biologicals. Among the groups for which a biosimilar was available, for 96.2% of the biologicals that were subject of a suspected ADR report the product name was identifiable in EudraVigilance. We concluded that this study underlines the need for improving traceability of biologicals, in particular with respect to individual batches, allowing better identification and monitoring of postmarketing safety issues related to biologicals. In chapter 3.2 we evaluated the effect of exposure misclassification on the time to detection of product-specific risks in spontaneous reporting systems. Exposure information may sometimes be misclassified (i.e. attributed to the incorrect product), and previous studies have suggested that reporters may particularly tend to misattribute generic-associated ADRs to innovator products. In this study, we examined the potential (public health) impact hereof within a simulated data model. We simulated an active substance-specific subset of a spontaneous reporting system, and used the proportional reporting ratio (PRR) for signal detection. The effect of exposure misclassification was evaluated in three test cases representing product-specific ADRs that may occur for biologicals, and studied in relative terms by varying the model parameters (market share, relative risk). We found that exposure misclassification results in the largest delay in identification of risks that have a weak association with the product of interest (RR< 2 or 3), and in situations where the product associated with the unique risk has a large (>50%) market share. The overall public health impact, in terms of additional time and cases to detection of the product-specific risk, was found to be highly dependent on the characteristics of the drug-event combination, including patient exposure, background incidence event, etc. Findings from this study can help inform the future implementation and refinement of product- and batch-specific signal detection procedures. In chapter 3.3 we reviewed the existing systems that are in place for the recording of detailed exposure information in (European) clinical practice, and explored the critical steps a involved in the transfer of exposure data to various pharmacovigilance databases. Apart from literature data, we presented results of a survey among 95 hospital and outpatient 184 Appendices

pharmacists in the Netherlands (completion rate 28%) to evaluate the recording practices in clinical practice, and the results of a survey among 31 countries within the EEA (completion rate 61%) to evaluate the measures taken to ensure the traceability of biological medicinal products in adverse drug reaction (ADR) reports. We found that existing systems ensure the traceability of biologicals down to the manufacturer within pharmacy records, but do not support the routine recording of batch information. This may, however, differ according to national traceability regulations, local procedures, as well as the specific type of biological. Expected changes in supply chain standards provide opportunities to systematically record detailed exposure information. Once adequately available in clinical practice, it is essential that exposure data is correctly and completely transferred in spontaneous ADR reports, and databases of electronic healthcare information. Spontaneous reporting systems are the most vulnerable link in ensuring traceability, due to the manual nature of data transfer. We concluded that efforts to improve the traceability should, in the short term, be focused towards encouraging health professionals and patients to systematically record and report detailed exposure information, though long-term solutions lie in expanding the accessibility to, and increasing the electronic exchange of exposure data.

4. Accrual of knowledge on previously identified uncertainties Chapter 4 focused on the accrual of knowledge on previously identified uncertainties after approval. For all new drug approvals in Europe, the main “uncertainties” (potential risks and missing information) in context of the benefit-risk balance of a drug are nowadays documented in a risk management plan (RMP). The RMP was introduced in 2005 to support a proactive approach in gaining knowledge on safety concerns through early planning of pharmacovigilance activities. Yet, the rate at which uncertainties in the safety profile are resolved along this approach was unknown. In chapter 4.1 we explored the evolution of safety concerns in the RMP after initial approval for a selected cohort of 31 small-molecule drugs and 17 biologicals approved between 2006 and 2009, to provide insight into the knowledge gain over time. Data on safety concerns was extracted from the baseline RMP (i.e., the adopted version at product approval), and from subsequent post-approval updates until 31 December 2012. We found that baseline RMPs for biologicals contained more uncertainties (median = 11; IQR: 7-16) than those for small-molecule drugs (median = 9; IQR: 8-12; p=0.15). These baseline uncertainties were followed-up in subsequent post-approval RMP updates, and overall 9.8% and 20.7% of the uncertainties were resolved at respectively 3 and 5 years after approval. The rate at which uncertainties were resolved did not differ between small-molecule drugs and biologicals. Uncertainties about specific ADRs (mostly potential risks) were resolved faster than uncertainties related to a specific context of use (mostly missing information). On average, the number of uncertainties per drug remained equal throughout the first 5 years after approval, as the number of resolved uncertainties was Summary 185 counterbalanced by an equal number of newly added uncertainties. We concluded that the relatively modest accrual of knowledge, as demonstrated in this study through resolution of uncertainties, suggests that opportunities for optimization exist while ensuring feasible and risk-proportionate pharmacovigilance planning. In chapter 4.2 we provided an overview of potential cancer risks associated with medicines newly approved in the EU, and aimed to examine the contribution of the agreed studies in the RMP to the post-approval knowledge accrual. For this study, we included the same cohort of 48 drugs that were studied in chapter 4.1. Public data on cancer risks at time of approval and the agreed studies to investigate these risks were extracted, and the products with potential cancer risks were subsequently followed until 5 May 2015 for new data on the suspected cancer risks from agreed studies, or other sources as reported by regulatory authorities. We found that 18 of the 48 drugs (38%) were associated with potential elevated cancer risks at approval, more often biologicals (11/17, 65%) than small-molecule drugs (7/31, 23%). Concerns over cancer risks were included in the RMPs most often based on data from preclinical studies in the case of small-molecular medicines (5 out of 7), while for biologicals the concern was primarily (9 out of 11) based on their pharmacological effects(e.g. immunosuppression). None of the drugs had a confirmed increased risk for cancer at time of approval. All 18 products associated with a potential increased risk of cancer were followed- up after approval to examine the knowledge accrual on the potential risk of cancer over time. Of the 16 products for which studies were agreed in the RMP, for three drugs the studies generated evidence that was used to update the product information on the cancer concern, while for four drugs new information on cancer risk became available from other sources than the studies agreed in the RMP. We concluded that concerns about possible increased risks for cancer at time of licensing are common, particularly for biologicals. Additional studies were frequently agreed to investigate these concerns, but their added value appears to be modest as more often other sources of information provided evidence on the cancer risks before new evidence was generated by the initially agreed studies.

5. General discussion and conclusion Chapter 5 contains a general discussion about the implications of the findings of this thesis, and provides recommendations for current regulatory and clinical practice and for future research. We conclude that the studies described in this thesis provide important insights into various aspects of post-approval safety learning for biologicals. Although adverse effects of biologicals may to a certain extent be predictable on basis of the expected pharmacology and immunogenicity, unexpected adverse effects could nevertheless emerge once a new drug is used in clinical practice. An open mind to expect the unexpected therefore remains essential, both for conditions that have not been identified as drug-induced before, but also for a false positive drug safety signals. A prepared mind is of particular importance, as previously unobserved adverse effects may become manifest at any point in the life-cycle of a biological 186 Appendices

as a result of changes issued, or unintended variations, in the manufacturing process. A key challenge related hereto is to ensure that accurate and detailed (product- and batch-specific) exposure information is available for biologicals to timely link any emerging safety issue to the specific manufacturing source. The RMP as tool for proactive pharmacovigilance only moderately facilitates knowledge accrual on uncertainties, leaving room for improvement in the implementation of proactive methods. Challenges in the future will be to keep up with the ongoing innovations in the biological arena, and to ensure that pharmacovigilance methods and regulatory tools are sufficiently evolving to emerging scientific knowledge and to the novel challenges posed by increasingly complex therapies.

Nederlandse samenvatting 189

Nederlandse samenvatting

1. Introductie en aanleiding voor dit proefschrift Geen enkel geneesmiddel is zonder risico. Voordat een nieuw geneesmiddel op de markt komt, wordt het uitvoerig onderzocht om aan te tonen dat de verwachte effectiviteit van het geneesmiddel bij gebruik in de dagelijkse praktijk opweegt tegen de verwachte risico’s van het geneesmiddel. Dit onderzoek omvat ten minste een centrale klinische studie waarin het nieuwe geneesmiddel wordt vergeleken met een placebogeneesmiddel of de huidige behandelstandaard binnen de beoogde indicatie van het nieuwe geneesmiddel. Enkel voor geneesmiddelen waarvoor geconcludeerd kan worden dat de verwachte baten-risico balans op populatieniveau positief is, wordt uiteindelijk een handelsvergunning verstrekt voor de Nederlandse en/of Europese markt. Op dit moment van eerste marktoelating is de kennis over de effectiviteit en met name bijwerkingen van het nieuwe geneesmiddel nog relatief beperkt. De centrale klinische studie omvat over het algemeen slechts een beperkt aantal patiënten die over een relatief korte periode worden gevolgd, waardoor zeldzamere bijwerkingen alsmede langtermijnbijwerkingen doorgaans worden gemist. Daarnaast is op het moment van toelating over het algemeen weinig bekend over de baten-risico balans van het nieuwe geneesmiddel in specifieke patiëntenpopulaties, bijv. in patiënten met bijkomende aandoeningen zoals hart- en vaatziekten, of patiënten die gelijktijdig meerdere geneesmiddelen gebruiken. Het is daarom belangrijk om verdere kennis te verwerven over een geneesmiddel nadat het op de markt is toegelaten en wordt gebruikt door een grote en heterogene groep patiënten. Dit geldt voor de effectiviteit, en met name voor de bijwerkingen. Farmacovigilantie houdt zich bezig met het identificeren en karakteriseren van bijwerkingen en het bewaken van de baten-risico balans van tot de markt toegelaten geneesmiddelen. Spontane meldingen van vermoede bijwerkingen door zorgverleners en patiënten spelen hierbij traditioneel een belangrijke rol. In de loop der jaren zijn er voorts tal van nieuwe methoden en regulatoire instrumenten ontwikkeld om bijwerkingen sneller te kunnen detecteren, waaronder analysemethoden voor farmacovigilantie databases en instrumenten om farmacovigilantie-activiteiten voor markttoelating te plannen. Tegelijkertijd worden geneesmiddelen steeds complexer, wat specifieke uitdagingen met zich meebrengt binnen de geneesmiddelregulering en farmacovigilantie. Biologicals vormen hiervan een belangrijk voorbeeld. Biologicals zijn geneesmiddelen waarvan het actieve bestanddeel geproduceerd of geëxtraheerd is uit een biologische bron. Hierbij onderscheiden biologicals zich van chemisch geproduceerde geneesmiddelen zoals aspirine (acetylsalicylzuur), ook wel bekend als “klein-moleculaire geneesmiddelen”. Sinds de commercialisering van recombinant-DNA-technologie in de jaren ’80 heeft de ontwikkeling en het gebruik van biologicals een enorme vlucht genomen. Tegenwoordig nemen biologicals een belangrijke a plaats in in de behandeling van tal van aandoeningen, waaronder kanker, reuma, multiple sclerosis, alsmede vele zeldzame aandoeningen. Vanwege de specifieke eigenschappen zijn 190 Appendices

biologicals echter geassocieerd met bepaalde risico’s (waaronder immunologische reacties) en uitdagingen in de farmacovigilantie, zoals gevonden in eerder onderzoek. De dynamiek in kennisverwerving omtrent de bijwerkingen van biologicals na toelating op de markt is echter onbekend. Dit proefschrift omvat zeven studies waarin verschillende aspecten zijn onderzocht van de kennisverwerving omtrent de bijwerkingen van biologicals na toelating op de markt. We hebben daarbij specifiek gekeken naar de dynamiek in kennisverwerving omtrent onbekende bijwerkingen, waaronder bijwerkingen die specifiek zijn voor een bepaald merk of productiebatch, en voorts gekeken naar de kennisverwerving over onzekerheden die op het moment van marktoelating waren omschreven. De verschillende studies zijn uitgevoerd binnen het samenwerkingsverband tussen het College ter Beoordeling van Geneesmiddelen (CBG) en de Universiteit Utrecht. Het CBG is medeverantwoordelijk voor de toelating en bewaking van geneesmiddelen in Europa, en eindverantwoordelijk voor de Nederlandse markt. Op het gebied van farmacovigilantie werkt het CBG hierbij nauw samen met het Nederlands Bijwerkingencentrum, Lareb.

2. Het ontdekken van nieuwe bijwerkingen na toelating op de markt Hoofdstuk 2 omvat twee studies waarin we verschillende aspecten hebben onderzocht met betrekking tot de ontdekking van tot nog toe onbekende bijwerkingen. In de afgelopen jaren zijn verscheidene biologicals in verband gebracht met het risico van progressieve multifocale leukoencephalopathie (PML), een levensbedreigende virale infectie van de hersenen die tot voorkort voornamelijk werd gezien bij patiënten met een sterk verminderde weerstand als gevolg van HIV/AIDS en kanker van witte bloedcellen. Het voorkomen van deze bijwerking vormt een vergelijkbare uitdaging binnen de farmacovigilantie van de verschillende biologicals. Echter kan het patroon van spontane meldingen van vermoede gevallen van PML verschillen tussen de producten. In hoofdstuk 2.1 hebben we onderzocht hoe spontane PML meldingen verschillen tussen twee biologicals die worden gebruikt in een uiteenlopende klinische setting en uiteenlopende patiëntenpopulaties, en waarvan daarnaast het risico van PML op verschillende momenten na marktoelating werd ontdekt: natalizumab en rituximab. Natalizumab wordt hoofdzakelijk toegepast bij multiple sclerosis en is vanaf toelating in verband gebracht met het optreden van PML. Rituximab wordt daarentegen binnen de oncologie en tal van auto-immuunziekten toegepast en is pas 9 jaar na toelating in verband gebracht met PML. In deze studie gebruikten we gegevens uit EudraVigilance, de Europese databank voor spontane meldingen. We vonden dat de PML meldingen voor natalizumab en rituximab op tal van punten verschilden (o.a. wat betreft leeftijd, geslacht, indicatie, tijd tot aanvang van PML reactie, en mortaliteit na aanvang van PML reactie), wat verband houdt met het verschil in patiëntenpopulaties. Daarnaast vonden we dat de meldingen voor natalizumab over het algemeen meer compleet waren (gemiddeld 87.5% vs. 62.5%) en eerder na aanvang van de PML reactie werden ontvangen (gemiddeld 1 en 3 maanden). Nederlandse samenvatting 191

Verschillende factoren kunnen de gevonden verschillen in meldingspatronen verklaren, waaronder verschil in patiëntenpopulaties in behandelend artsenpopulaties, alsmede tijdsef- fecten met betrekking tot het bewustzijn van het risico van geneesmiddelgeïnduceerd PML. Deze studie laat zien dat spontane meldingen van vermoede bijwerkingen alleen met grote bedachtzaamheid, en met inachtneming van de mogelijke variabiliteit in meldingspatronen, gebruikt kunnen worden om bijwerkingen tussen producten en over de tijd te vergelijken. Niet alle signalen van mogelijke bijwerkingen betreffen daadwerkelijk nieuwe bijwerkingen. In de studie beschreven in hoofdstuk 2.2 hebben we ons gericht op een potentiële oorzaak van vals positieve signalen, namelijk de mogelijkheid van geneesmiddelinterferentie in klinisch laboratoriumonderzoek. Sommige substanties, waaronder geneesmiddelen, kunnen interfereren in bepalingen van stoffen (hormonen, zouten, eiwitten etc.) in bloed en andere lichaamsvochten. Dit kan leiden tot incorrecte diagnoses, en uiteindelijke tot verkeerde klinische beslissingen en behandeling van patiënten en/of vals positieve signalen in farmacovigilantie. In deze studie, onderzochten we, naar aanleiding van een recent geval, of een specifieke klasse biologicals (monoklonale antilichamen) kunnen interfereren in immunoassays, een groep laboratoriumbepalingen die gebruikmaakt van antigeen-antistof reacties. We voegden daarom in-vitro maximale serumconcentraties van monoklonale antilichamen (rituximab, belimumab, tocilizumab, infliximab) toe aan serummonsters en onderzochten of deze interfereren in immunoassays bepalingen voor vrij thyroxine, thyroïd stimulerend hormoon, cortisol, oestradiol, en luteïniserend hormoon. We vonden een gering interferentie effect van de onderzochte biologicals op de betreffende immunoassays. Het grootste, en enige klinische relevante, effect werd gezien voor rituximab op de oestradiolbe- paling, waarbij gemiddeld een 19% te hoge concentratie werd gemeten. We concludeerden dat het interferentie effect van monoklonale antilichamen in immunoassays laag is, maar in specifieke gevallen kan leiden tot incorrect diagnoses en vals positieve signalen in farmacovigilantie.

3. Het signaleren van merken batch-specifieke bijwerkingen van biologicals Een unieke eigenschap van biologicals is dat het veiligheidsprofiel van een product kan veranderen over de tijd als gevolg van wijzigingen – bedoeld of onbedoeld – in het productieproces. Hierbij kan het risico op een reeds bekende bijwerking groter worden, of kunnen nieuwe bijwerkingen ontstaan die niet eerder met het product waren geassocieerd. Een voorbeeld hiervan is het optreden van pure rode cel aplasie (PRCA), een zeldzame en levensbedreigende aanmaakstoornis van rode bloedcellen, bij patiënten die werden behandeld met het epoëtine alfa, een eiwit voor behandeling van bloedarmoede. In totaal zijn er tussen 1998 en 2002 meer dan 200 gevallen van PRCA beschreven bij met epoëtine alfa-behandelde patiënten. Na uitvoerig onderzoek werd uiteindelijk ontdekt dat deze bijwerking gerelateerd a kon worden aan een specifiek merk van epoëtine alfa waarvoor in 1998 wijzigingen in het productieproces hadden plaatsgevonden, namelijk Eprex. Ook meer recent zijn er geval- 192 Appendices

len geweest waarbij wijzigingen in het productieproces binnen verschillende batches van een product, maar ook verschillen in de productie tussen producten (d.w.z. verschillende merken) met hetzelfde actieve biologische bestanddeel, gerelateerd konden worden aan specifieke (nieuwe) bijwerkingen. Voor het signaleren van dergelijke bijwerkingen die specifiek terug te voeren zijn tot een bepaalde productiewijze van een biological, is het van belangrijk dat adequate (batch- en product-specifieke) blootstellingsgegevens beschikbaar zijn in databases die gebruikt worden voor de farmacovigilantie van biologicals. Dit is met name van belang omdat de beschikbaarheid en het gebruik van biosimilars in de komende jaren naar verwachting sterk zal stijgen. Biosimilars zijn biologicals die hetzelfde actieve biologische bestanddeel bevatten als een reeds verhandeld product, en waarvan op basis van uitgebreid onderzoek is vastgesteld dat beide producten therapeutisch gelijkwaardig zijn. Als gevolg van wijzigingen in het productieproces bij een van beide producten na initiële marktoelating, kan het veiligheidsprofiel van de producten echter uiteen gaan lopen, en is het van belang dat de respectievelijke producten duidelijk traceerbaar zijn in farmacovigilantie databases. Hoofdstuk 3 omvat drie studies waarin de specifieke uitdagingen worden onderzocht met betrekking tot het verkrijgen van adequate blootstellingsgegevens voor biologicals. In de studie beschreven in hoofdstuk 3.1 hebben we de traceerbaarheid van biologicals onderzocht in spontane meldingen van vermoede bijwerkingen. Spontane meldingen door zorgverleners (artsen, apothekers, enz.) of patiënten spelen een grote rol bij het signaleren van onbekende bijwerkingen van geneesmiddelen in de dagelijkse praktijk. Het was echter onbekend of de traceerbaarheid van biologicals voldoende was gewaarborgd in deze spontane meldingen om mogelijke producten/of batch-specifieke bijwerkingen op te sporen. In deze studie hebben we gevonden dat in grote spontanerapportagesystemen in Europa (EudraVigilance) en de VS (FDA Adverse Event Reporting System, FAERS) voor respectievelijk 21% en 24% van de gerapporteerde biologicals tussen 2004 en 2010 gegevens over de specifieke productiebatch beschikbaar waren. Patiënten waren over het algemeen het meest en artsen het minst geneigd om deze batchgegevens te melden. Daarnaast hebben we gevonden dat voor de groepen biologicals waarvoor een biosimilar beschikbaar was ten tijde van de melding (d.w.z. dus meerdere producten/merken met hetzelfde actieve biologische bestanddeel) in meer dan 96% van de spontane meldingen het specifieke product identificeerbaar was. We concludeerden dat de traceerbaarheid van biologicals, met name van individuele batches, in spontane meldingen verbeterd kan worden, om de risico’s van biologicals beter te kunnen volgen. Hoewel de studie in hoofdstuk 3.1 belangrijke inzichten biedt in de beschikbaarheid van product-specifieke blootstellingsgegevens voor biologicals is het onbekend of de verstrekte gegevens in spontane meldingen correct zijn. Eerdere studies hebben namelijk gesuggereerd dat bijwerkingen soms ten onrechte aan het specialité worden toegeschreven, terwijl de patiënt een generieke versie gebruikte. In hoofdstuk 3.2 hebben we in een gesimuleerd Nederlandse samenvatting 193 data model onderzocht wat het effect is van deze misclassificatie op de tijd tot detectie van product-specifieke bijwerkingen. We gebruikten daarvoor drie test scenario’s van product- specifieke bijwerkingen die kunnen voorkomen bij biologicals, en varieerden de hoeveelheid misclassificatie in spontane meldingen van bijwerkingen voor de respectievelijke producten in ons simulatiemodel. Voor het opsporen van de bijwerkingen gebruikt we zogenaamde signaaldetectie, een veelgebruikte analysemethode om signalen voor onbekende bijwerkingen te detecteren. We vonden dat misclassificatie zorgde voor de grootste vertraging in het opsporen van de product-specifieke bijwerkingen wanneer de bijwerking relatief zwak geassocieerd was met het specifieke product (twee tot driemaal verhoogd risico t.o.v. basaal), en wanneer het product een relatief groot (>50%) marktaandeel had. Het effect op de volks- gezondheid, in termen van tijd en aantal gevallen van de bijwerking tot detectie, varieerde sterk afhankelijk van onder andere het absolute risico op de bijwerking en het gebruik van het geneesmiddel. De simulatie biedt belangrijke inzichten voor de verdere ontwikkeling en implementatie van product-specifieke signaaldetectieprocedures. In hoofdstuk 3.3 hebben we de verschillende systemen in de klinische praktijk voor het vastleggen van blootstellingsgegevens van biologicals onder de loep genomen, alsmede de kritische stappen onderzocht in de overdracht van deze gegevens naar farmacovigilantiedatabases. Naast literatuurgegevens, presenteerden we resultaten van een enquête onder 95 ziekenhuis- en poliklinisch apothekers naar de procedures in de klinische praktijk, en de resultaten van een enquête onder 31 landen binnen de EEG naar de genomen maatregelen om de traceerbaarheid van biologicals in spontane meldingen te waarborgen. We vonden dat de huidige systemen voldoende de traceerbaarheid van biologicals in de klinische praktijk waarborgen tot op productniveau, maar het routinematig vastleggen van batchgegevens niet ondersteunen. Dit varieert echter afhankelijk van de nationale regelgeving, lokale procedure, en het type biological. Aanpassingen in de regelgeving voor de geneesmiddelketen zullen er in de nabije toekomst voor zorgen dat ook batchgegevens routinematig worden vastgesteld. We concludeerden dat inspanningen om de traceerbaarheid van biologicals te verbeteren op de korte termijn gericht moeten zijn op het bewustmaken van patiënten en zorgverleners systematisch blootstellingsgegevens vast te leggen en te rapporteren, terwijl langetermijnop- lossingen liggen in het vergroten van de toegankelijkheid tot, en elektronische uitwisseling, van deze gegevens.

4. Kennisverwerving over geïdentificeerde onzekerheden Voor alle nieuwe geneesmiddelen in Europa worden tegenwoordig voor marktoelating de belangrijkste geïdentificeerde risico’s en onzekerheden vastgelegd in een risicomanagementplan (RMP). Deze “onzekerheden” kunnen zowel potentiële risico’s omvatten (bijv. risico’s gezien in dierstudies waarvan de relevantie voor de mens onbekend, zoals effecten op hart- en a bloedvaten of mogelijk verhoogd kankerrisico), alsmede missende informatie (bijv. risico’s van het geneesmiddel bij gebruik in niet-bestudeerde patiëntenpopulaties, zoals patiënten 194 Appendices

met verminderde nierziekte of hartaandoeningen). In het RMP worden daarnaast de met de geneesmiddelfirma gemaakte afspraken vastgelegd over hoe deze onzekerheden na toelating verder zullen worden bestudeerd. Dit kunnen bijvoorbeeld nieuwe studies zijn naar een specifieke bijwerking, of naar risico’s van een geneesmiddel binnen een specifieke patiënten- populatie. Door middel van deze proactieve aanpak wordt beoogd om de kennisverwerving over deze onzekerheden te versnellen. Of en hoe snel deze onzekerheden na moment van registratie worden opgelost was echter tot op heden niet bestudeerd en onbekend. Hoofdstuk 4 omvat twee studies waarin we de dynamiek in kennistoename over deze onzekerheden vanaf toelating hebben bestudeerd. In de studie beschreven in hoofdstuk 4.1 hebben we voor een groep van 48 geneesmiddelen (waaronder 17 biologicals) die zijn toegelaten op de Europese markt tussen 2006 en 2009, de toename in kennis over onzekerheden onderzocht. We maakten daarvoor gebruik van het initiële RMP dat was vastgesteld ten tijde van marktoelating, en de daaropvolgende RMP updates na toelating tot en met december 2012. De belangrijkste uitkomst in deze studie was het oplossen van de onzekerheden, inhoudende [i] het verwijderen van de onzekerheid uit het RMP (wat zou impliceren dat deze voldoende is bestudeerd, en geen risico omvat), of [ii] het veranderen van de onzekerheid in een geïdentificeerd risico. We vonden in deze studie dat het initiële RMP voor biologicals gemiddeld meer onzekerheden beschreef dan het RMP voor kleine-moleculaire geneesmiddelen (respectievelijk 11 versus 9 onzekerheden). De onzekerheden uit het initiële RMP werden opgevolgd over de tijd, en 3 en 5 jaar na marktoelating was respectievelijk 9,8% en 20,7% van deze onzekerheden opgelost. We vonden geen verschil in de snelheid waarmee onzekerheden werden opgelost tussen biologicals en klein- moleculaire geneesmiddelen. De totale hoeveelheid onzekerheid bleef gelijk in de eerste 5 jaar na toelating omdat de hoeveelheid opgeloste onzekerheden gelijk was aan het aantal nieuw toegevoegde onzekerheden. We concludeerden dat de bescheiden kennistoename, zoals gemeten in deze studie aan de hand van het oplossen van onzekerheden, laat zien dat er mogelijkheden voor verbetering zijn in de implementatie van proactieve farmacovigilantie. In hoofdstuk 4.2 hebben we specifiek gekeken naar onzekerheden omtrent een potentieel verhoogd kankerrisico. We maakten daarvoor gebruik van hetzelfde cohort van 48 geneesmiddelen als in hoofdstuk 4.1 en onderzochten hoe vaak onzekerheden omtrent kankerrisico’s voorkwamen bij toelating, en hoe de kennis hieromtrent zich ontwikkelde over de tijd (t/m mei 2015). We vonden in deze studie dat 38% van de nieuwe geneesmiddelen (18 van de 48) waren geassocieerd met een potentieel verhoogd kankerrisico bij marktoelating, waarbij biologicals vaker (65%; 11 van de 17) dan klein-moleculaire geneesmiddelen (23%; 7 van de 31). Bij klein-moleculaire geneesmiddelen kwamen deze onzekerheden het vaakst voort (7 van de 9) uit preklinisch onderzoek (bijv. dierstudies of onderzoek met menselijk weefsel), terwijl bij biologicals onzekerheden omtrent mogelijk verhoogde kankerrisico’s het vaakst (9 van de 11) waren gebaseerd op het beredeneerde/verwachte farmacologische effect (zoals onderdrukking van het immuunsysteem). Voor geen van de nieuwe geneesmiddelen was er Nederlandse samenvatting 195 sprake van een bevestigd risico. Voor 16 van de 18 producten met een potentieel verhoogd kankerrisico’s waren studies afgesproken om dit potentiële risico verder te onderzoeken na markttoelating. We vonden dat bij 3 van deze 16 geneesmiddelen de resultaten van de voorgestelde studie aanleiding gaven tot een aanpassing van de bijsluiter na toelating, terwijl bij 4 andere middelen de productinformatie werd aangepasten op basis van andere studies en/of bronnen. We concludeerden dat zorgen omtrent mogelijk verhoogde kankerrisico’s frequent voorkomen bij nieuwe geneesmiddelen, met name bij biologicals. Hoewel studies vaak worden voorgesteld om deze onzekerheden te onderzoeken leiden deze tot een relatief bescheiden kennistoename, omdat vaker kennis wordt gegenereerd door andere studies/ bronnen.

5. Algemene discussie en conclusie Hoofdstuk 5 omvat een algemene discussie over de implicaties van onze bevindingen en bevat aanbevelingen voor de klinische en regulatoire praktijk en voor verder onderzoek. We concluderen dat dit proefschrift belangrijke inzichten biedt in verschillende aspecten van de kennisverwerving omtrent de bijwerkingen van biologicals na toelating op de markt. Hoewel bijwerkingen van biologicals tot op zekere hoogte voorspelbaar zijn op basis van het verwachte farmacologische en immunologische effect, kunnen onverwachte bijwerkingen desalniettemin ontdekt worden gedurende gebruik in de klinische praktijk. Bedachtzaam- heid op dergelijke bijwerkingen is daarom cruciaal, alsmede bedachtzaamheid op vals positieve veiligheidssignalen. Dit is met name van belang omdat veranderingen in het productieproces binnen een product over de tijd aanleiding kunnen geven tot nieuwe risico’s, inclusief bijwerkingen die tot dan niet met het product waren geassocieerd. Een belangrijke uitdaging hieraan gerelateerd is dat gedetailleerde en juiste producten batch-specifieke blootstellingsgegevens voor biologicals beschikbaar is in farmacovigilantiedatabases. Het risicomanagementplan blijkt maar matig effectief als instrument om verdere kennis te verwerven over geïdentificeerde onzekerheden, en laat hiermee ruimte voor verbetering in de implementatie proactieve farmacovigilantie. Uitdagingen voor de toekomst omvatten het bijblijven met de voortdurende innovaties in het biotechnologie veld, en om ervoor te zorgen dat de bestaande methoden en systemen voldoende evolueren naar de nieuwe wetenschappelijke inzichten en nieuwe uitdagingen van toenemend complexe geneesmiddelen.

Acknowledgements/ Dankwoord 197

Dankwoord

In de afgelopen jaren hebben velen bijgedragen aan de totstandkoming van dit proefschrift. Op deze laatste pagina’s wil ik graag een aantal personen in het bijzonder bedanken voor de wetenschappelijke en sociale ondersteuning die ik van hen heb ontvangen. Allereerst, grote dank aan mijn promotoren, prof. dr. HGM Leufkens & prof. dr. ACG Egberts, alsmede aan mijn copromotoren, dr. ML De Bruin & dr. SMJM Straus. Hierbij wil ik graag ook dr. AK Mantel-Teeuwisse betrekken, die de afgelopen jaren nauw bij dit project betrokken is geweest. Een groot team van begeleiders betekent veel verschillende meningen en invalshoeken, maar ook vele interessante discussies en een grote hoeveelheid aan expertise, en wetenschappelijke en sociale ondersteuning waar ik uit heb kunnen putten. Bert, ik wil je van harte bedanken voor je motiverende en inspirerende begeleiding, je voortdurende enthousiasme, en voor de onmisbare hulp en het vertrouwen bij de laatste eindspurt voor de afronding van dit proefschrift. Toine, ik heb jouw grote betrokkenheid en persoonlijke benadering altijd als zeer prettig ervaren. Op een aantal cruciale momenten heb jij ingegre- pen om projecten niet te laten verzandden in details en/of randzaken, zodat deze tijdig met een mooie publicatie konden worden afgerond. Veel dank hiervoor. Marieke, ik ben je veel dank verschuldigd voor alle hulp en ondersteuning. Onze wekelijkse overleggen zijn erg belangrijk geweest om voldoende vaart en richting in dit promotietraject te houden. Ook gedurende de week was je altijd bereikbaar voor een kort overleg, wat scherpzinnige input, of een aantal bemoedigende woorden. Sabine, ik wil je van harte bedanken voor je kritische inbreng door de jaren heen, vaak met oog voor politieke gevoeligheden en voor regulatoire relevantie. Aukje, wegens jouw betrokkenheid bij twee voorgaande promotietrajecten op het gebied van farmacovigilantie van biologicals ben jij vanaf het begin aangeschoven bij de overleggen. Omdat dit goed beviel, ben jij vervolgens gedurende het hele traject de overleggen bijgewoond. Beste Bert, Toine, Marieke, Sabine & Aukje heel veel dank voor de prettige samenwerking in de afgelopen jaren. De leden van de leescommissie, dr. TJ Giezen, prof. dr. JMW Hazes, prof. dr. OH Klun- gel, prof. dr. EP van Puijenbroek & prof. dr. H Schellekens, wil ik graag bedanken voor hun snelle beoordeling van het manuscript van dit proefschrift gedurende de vakantieperiode. Ik wil daarnaast graag de verschillende coauteurs bedanken waar ik de afgelopen jaren mee heb samengewerkt. A special thanks to the colleagues from the European Medicines Agency, Ana Hidalgo-Simon, Francois Domergue & Peter Arlett, for their contribution to respectively chapters, 2.1, 3.1 & 4.1 of this thesis. Svetlana Belitser, bedankt voor je bijdrage aan de statistische analyses voor hoofdstuk 2.1, en voor het feit dat ik altijd bij je langs kon lopen met uiteenlopende vragen over statistiek. Eef Lentjes & Wouter van Solinge, hartelijk dank voor jullie hulp en expertise bij de totstandkoming van hoofdstuk 2.2, en voor de a geboden mogelijkheid om onderzoek te doen in het klinisch chemisch lab. Marloes, bedankt voor je onmisbare hulp bij het oplossen van de vergelijkingen voor hoofdstuk 3.2. Hans, 198 Appendices

dank voor je bijdrage aan hoofdstuk 3.2, en bovenal voor de vele korte overlegmomenten in de afgelopen jaren. Kevin Klein & Pieter Stolk, bedankt voor de prettige samenwerking binnen het Escher project “Pharmacovigilance of biologics”. Irina, veel dank voor je hulp bij de opzet en dataverzameling van hoofdstuk 3.3, en super dat we dit hebben kunnen omzetten in een mooie publicatie. Thijs, bedankt voor je hulp bij de totstandkoming van hoofdstuk 3.3, en voor de verscheidene keren dat we in de afgelopen hebben kunnen overleggen. Ik ben verheugd dat we nu ook tijdens mijn promotie van gedachten kunnen wisselen. Ruben, veel dank voor de samenwerking voor hoofdstukken 4.1 & 4.2 van dit proefschrift, en daarnaast voor de gezelligheid, vele discussies, borrels en fietstochten in de afgelopen jaren. Het was erg prettig om tegelijk met jou dit promotietraject af te leggen bij de universiteit en het CBG. Mijn collega’s van de divisie farmaco-epidemiologie & klinische farmacotherapie wil ik graag bedanken voor alle gezelligheid, leerzame discussies en het delen van de onder- zoekservaringen, waaronder Corrine, Francisco, Geert, Joris, Jacoline, Jarno, Lydia, Marcel, Michelle, Niloufar, Yaser, Renate, Rik, Sanni, Soulmaz en Susanne. In het bijzonder, Maarten, Paul, Rianne, Ruben en Sander, bedankt voor de lange lunchdiscussies, en de gezelligheid. Ook dank aan de dames van het secretariaat, Anja Ineke en Suzanne, voor hun hulp door de jaren. Tijdens mijn promotie ben ik tevens werkzaam geweest bij het College ter Beoordeling van Geneesmiddelen. Ik wil mijn collega’s van de afdeling farmacovigilantie bedanken voor hun belangstelling in mijn onderzoek en hun support. In het bijzonder dank aan mijn di- recte collega’s bij FT3, Daniël, Maarten, Sara, Sophia, Stéphany, Ursula, voor hun flexibiliteit. Daarnaast wil ik ook mijn mede-promovendi en ex-promovendi bij het CBG, Alexandra, Ineke, Kartini, Menno, Nico, Quirine, bedanken voor het delen van de onderzoekservarin- gen en -frustraties. Anja, fijn dat ik altijd bij je terecht kon met vragen over EudraVigilance en over wet- en regelgeving gedurende mijn promotie. Inge, bedankt voor alle support en sportieve afleiding in de afgelopen jaren. Zonder de belangstelling en aanmoediging van mijn ouders, broer & zus, schoonouders, en vrienden van farmacie was dit proefschrift er nooit gekomen. Lieve ouders, Dimitri, Iris en Niki, Gabby & Goof, Evert, Jos, Arti, Johan, Maarten, Martijn en Roland, dank voor alle steun en afleiding in de afgelopen jaren. Tenslotte, allerliefste Julia, dankjewel, x.

List of co-authors 201

List of co-authors

Peter R Arlett European Medicines Agency, London, United Kingdom

Anthonius de Boer Utrecht Institute for Pharmaceutical Sciences (UIPS), Division of Pharmacoepidemiology and Clinical Pharmacology, Utrecht University, Utrecht, the Netherlands

Marie L De Bruin Utrecht Institute for Pharmaceutical Sciences (UIPS), Division of Pharmacoepidemiology and Clinical Pharmacology, Utrecht University, Utrecht, the Netherlands Medicines Evaluation Board, the Netherlands

Francois Domergue European Medicines Agency, London, United Kingdom

Ruben G Duijnhoven Utrecht Institute for Pharmaceutical Sciences (UIPS), Division of Pharmacoepidemiology and Clinical Pharmacology, Utrecht University, Utrecht, the Netherlands Medicines Evaluation Board, the Netherlands

Hans C Ebbers Utrecht Institute of Pharmaceutical Sciences (UIPS), Division of Pharmaceutics, Utrecht University, Utrecht, the Netherlands

Toine CG Egberts Utrecht Institute for Pharmaceutical Sciences (UIPS), Division of Pharmacoepidemiology and Clinical Pharmacology, Utrecht University, Utrecht, the Netherlands Department of Clinical Pharmacy, University Medical Center Utrecht, Utrecht, the Neth- erlands

Thijs J Giezen Pharmacy Foundation of Haarlem Hospitals , Haarlem, The Netherlands

Ana Hidalgo-Simon European Medicines Agency, London, United Kingdom a 202 Appendices

Arno W Hoes Julius Center for Health Sciences and Primary Care, University Medical Center Utrecht, the Netherlands

Eef G Lentjes Department of Clinical Chemistry and Haematology, University Medical Center Utrecht, Utrecht, the Netherlands

Hubert GM Leufkens Utrecht Institute for Pharmaceutical Sciences (UIPS), Division of Pharmacoepidemiology and Clinical Pharmacology, Utrecht University, Utrecht, the Netherlands Medicines Evaluation Board, the Netherlands

Aukje K Mantel-Teeuwisse Utrecht Institute for Pharmaceutical Sciences (UIPS), Division of Pharmacoepidemiology and Clinical Pharmacology, Utrecht University, Utrecht, the Netherlands

Wouter W van Solinge Department of Clinical Chemistry and Haematology, University Medical Center Utrecht, Utrecht, the Netherlands

Irina Spierings Utrecht Institute for Pharmaceutical Sciences (UIPS), Division of Pharmacoepidemiology and Clinical Pharmacology, Utrecht University, Utrecht, the Netherlands

Sabine MJM Straus Medicines Evaluation Board, Utrecht, the Netherlands Department of Medical Informatics, Erasmus Medical Center, Rotterdam, the Netherlands

List of publications 205

List of publications

Related to this thesis Vermeer NS, Straus SMJM, Mantel-Teeuwisse AK, Hidalgo-Simon A, Egberts ACG, Leufkens HCG, De Bruin ML. Drug-induced Progressive Multifocal Leukoencephalopathy (PML): lessons learned from contrasting natalizumab and rituximab. Clin Pharmacol Ther 2015 (in press)

Vermeer NS, Spierings I, Mantel-Teeuwisse AK, Straus SMJM, Giezen TJ, Leufkens HCG, Egberts ACG, De Bruin ML. Traceability of biologicals: Present challenges in pharmacovigilance. Expert Opinion on Drug Safety 2015; 14: 63-72

Vermeer NS, Duijnhoven RG, Straus SMJM, Mantel-Teeuwisse AK, Arlett PR, Egberts ACG, Leufkens HCG, De Bruin ML. Risk Management Plans as tool for proactive pharmacovigilance: A cohort study of newly approved drugs in Europe. Clin Pharmacol Ther 2014; 96: 723-31.

Vermeer NS, Straus SMJM, Mantel-Teeuwisse AK, Domergue F, Egberts ACG, Leufkens HCG, De Bruin ML. Traceability of biopharmaceuticals in spontaneous reporting systems: a cross-section study in the FDA Adverse Event Reporting Systems (FAERS) and EudraVigilance. Drug Safety 2013; 36: 617-25.

Not included in this thesis Segec A, Keller-Stanislawski B, Vermeer NS, Macchiarulo C, Straus SMJM, Hidalgo-Simon A, De Bruin ML. Strategy in regulatory decision making for management of PML. Clin Phar- macol Ther 2015 (in press)

Lalmohamed A, Vermeer NS, de Vries F. Harmful effects of proton pump inhibitors: discrepancies between observational studies and randomized clinical trials. JAMA Intern Med 2013; 173: 1559.

Vermeer NS, De Bruin ML. Research into the risk of malignancy from biologicals: utility of meta-analysis hampered. Ned Tijdschr Geneeskd. 2013; 157: A5884.

Vermeer NS, Bajorek BV. Utilization of evidence-based therapy for the secondary prevention of acute coronary syndromes in Australian practice. J Clin Pharm Ther 2008; 33: 591-601. a

About the author 207

About the author

Niels Sebastiaan Vermeer was born in Den Helder on July 25, 1985. Niels grew up in a small village in the north of the Netherlands and attended secondary school at RSG Wiringherlant in Wieringerwerf, where he completed his A-levels in 2003. In the same year, he moved to Utrecht and started his pharmacy studies at Utrecht University. During the last three and a half years of his studies, Niels worked as an undergraduate at the intensive care unit pharmacy of the Academic Medical Centre in Amsterdam. For his master thesis, he spent seven months as a visiting research student at the faculty of pharmacy at the University of Sydney and Royal North Shore Hospital, where he studied the utilization of evidence-based therapy for the secondary prevention of acute coronary syndromes. Niels received his Bachelor’s degree with distinction in pharmacy in 2006, and Pharmacist degree (MSc, PharmD) with distinction in 2009. Following graduation, Niels worked for one and a half years as a pharmacist in primary care health centers in Almere. In March 2011, he subsequently started his PhD project at the Division of Pharmacoepidemiology and Clinical Pharmacology of the Utrecht Institute for Pharmaceutical Sciences (UIPS), Utrecht University. He combined this with a position as pharmacovigilance assessor at the Dutch Medicines Evaluation Board (MEB). During this period, Niels was enrolled in the postgraduate Masters program in Epidemiology at the Julius Centre of University Medical Center Utrecht, where he obtained his Masters in Health Sciences with a GPA of 4.0 in September 2014. In 2015, a publication of Niels was nominated for the UIPS PhD high potential article award. Niels currently works as pharmacovigilance assessor at the MEB, and lives with Julia in Amsterdam. In his spare time, he enjoys outdoor activities such as surfing, running and speed-skating, travelling (to former Soviet states) and politics. As of January 2016, Niels will join the pharmacovigilance department of the European Medicines Agency (London, UK) as a seconded national expert.