The European Medicines Agency (EMEA) workshop on biosimilar monoclonal antibodies (mAbs), held July 2, 2009 at the EMEA headquarters on Canary Wharf in London, was a harbinger with potentially far-reaching implications for all groups interested in antibody therapeutics development. These groups include not only regulators and the innovator and generic biopharmaceutical industries, but also physicians, patients and payers. The workshop was led by Christian Schneider, chairman of EMEA's Similar Biological (Biosimilar) Medicinal Products Working Party (BMWP), with assistance by Falk Ehmann, Scientific Secretariat of the BMWP. Representatives of the Committee for Human Medicinal Products (CHMP), Biologics Working Party (BWP), Safety Working Party (SWP), Efficacy Working Party (EWP) and Scientific Advice Working Party (SAWP) also participated.
The objective of the workshop was to discuss and assess the feasibility of the development and authorization of mAbs using CHMP's biosimilar regulatory pathways. The workshop sequentially focused on questions relevant to three areas: (1) chemistry, manufacturing and controls (CMC); (2) non-clinical issues; and (3) clinical issues, including outcome measures. The CMC session was chaired by Jean-Hugues Trouvin (chairman of BWP), the non-clinical issues session was chaired by Beatriz Silva-Lima (chairwoman of SWP), and the clinical issues session was chaired by Dr. Schneider. Each session opened with presentations giving the perspectives of the innovator industry, the biosimilar industry and regulators. Discussion of various points then followed. Participation was by invitation only. Over 160 people attended, including representatives from regulatory agencies in the European Union (EU), United States (US) and Canada, and approximately 40 biopharmaceutical companies located worldwide. Presentations from the innovator industry were coordinated by the European Biopharmaceutical Enterprises (EBE) and the European Association for Bioindustries (EuropaBio), while the biosimilar industry presentations were coordinated by the European Generic Medicines Association (EGA).
It is important to note that the workshop itself follows on a long, complex history surrounding marketing approvals for biosimilar products that have occurred over the last decade. EMEA has been at the forefront of regulatory agency activities concerning approval of biosimilars, although the US Food and Drug Administration (FDA), Health Canada, Australia's Therapeutic Goods Administration and Japan's Ministry of Health, Labor and Wealth, as well as other regulatory agencies, have approved biosimilar therapeutics (Table 1). The products are referred to as biosimilars in the EU and other countries, but Health Canada and FDA use the terms ‘subsequent entry biologics’ and ‘follow-on protein products,’ respectively.1 The term biosimilars will be used herein.
Table 1.
Biosimilar therapeutic proteins approved in selected countries*
| Non-proprietary name |
Approximate MW |
Trade name |
Company |
Country (year) approved |
| Somatropin |
22 kDa |
Omnitrope |
Sandoz GmbH |
Australia (2004) |
|
|
|
|
EU (2006) |
|
|
|
|
US (2006) |
|
|
|
|
Canada (2009) |
|
|
|
|
Japan (2009) |
| Somatropin |
22 kDa |
Valtropin |
BioPartners GmbH |
EU (2006) |
| Epoetin alfa |
30–40 kDa |
Binocrit |
Sandoz GmbH |
EU (2007) |
| Epoetin alfa |
30–40 kDa |
Epoetin alfa Hexal |
Hexal AG |
EU (2007) |
| Epoetin alfa |
30–40 kDa |
Abseamed |
Medice Arzneimittel Putter GmbH |
EU (2007) |
| Epoetin zeta |
32–40 kDa |
Retacrit |
Hospira Enterprises B.V. |
EU (2007) |
| Epoetin zeta |
32–40 kDa |
Silapo |
STADA Arzneimittel AG |
EU (2007) |
| Filgrastim |
18.8 kDa |
TevaGrastim |
Teva Generics GmbH |
EU (2008) |
| Filgrastim |
18.8 kDa |
Biograstim |
CT Arzeimittel |
EU (2008) |
| Filgrastim |
18.8 kDa |
Ratiograstim |
ratiopharm GmbH |
EU (2008) |
| Filgrastim |
18.8 kDa |
Filgrastim ratiopharm |
ratiopharm GmbH |
EU (2008) |
| Filgrastim |
18.8 kDa |
Filgrastim Hexal |
Hexal AG |
EU (2009) |
| Filgrastim |
18.8 kDa |
Filgrastim Zarzio |
Sandoz GmbH |
EU (2009) |
| Glucagon |
3.5 kDa |
GlucaGen |
Novo Nordisk |
US (1998) |
| Hyaluronidase (bovine) |
55 kDa |
Amphadase |
Amphastar Pharm |
US (2004) |
| Hyaluronidase (ovine) |
55 kDa |
Vitrase |
ISTA Pharms |
US (2004) |
| Hyaluronidase (bovine) |
55 kDa |
Hydase |
PrimaPharm |
US (2005) |
| Hyaluronidase (human, rDNA) |
61 kDa |
Hylenex |
Halozyme Therapeutics |
US (2005) |
| Calcitonin (salmon, rDNA) |
3.5 kDa |
Fortical |
Unigene Laboratories |
US (2005) |
| Abciximab |
48 kDa |
Clotinab |
ISU Abxis Co. |
South Korea (2007) Chile (2009) |
Questions surrounding quality, non-clinical assessment and clinical evaluation of biosimilar therapeutics have been raised in the EU, US and other regions of the world. Interestingly, the questions apply to the current conundrum of requirements for biosimilar mAb approval, but also to the much older issue of assessing comparability of biological products following manufacturing process changes. The FDA issued guidance on this problem, in which “those steps that manufacturers may perform and which FDA may evaluate to allow manufacturers to make manufacturing changes without performing additional clinical studies to demonstrate safety and efficacy”2 were described, as early as April 1996. The problem is an ongoing concern—one role of BMWP, established in 2005 following the Biosimilar Task Force (2004–2007), is to provide recommendations to CHMP on the conduct of tests conducted to ensure the comparability of new and old versions of biologically similar products.3 CHMP also has a guideline, Guideline on comparability of biotechnology-derived medicinal products after a change in the manufacturing process: non-clinical and clinical issues, that came into effect in November 2007.
Innovator and biosimilar companies thus have some common general problems, but there are obvious differences in the specifics, i.e., whereas innovators are comparing versions of products produced using internally-vetted processes, biosimilar companies are comparing their products with externally-sourced material. The EMEA workshop focused on discussion of the points of commonality and difference.
Workshop Introduction
To open the proceedings, Dr. Schneider introduced ‘pro and con’ points of biosimilar mAbs marketing approvals. There are several key questions on this topic for regulators. One is how much do we need to know? Like an incomplete puzzle, the overall picture supplied by CMC, non-clinical and clinical data comparing only key elements of a biosimilar to a reference product might be sufficient to extrapolate the whole picture, or it might be missing key pieces. Another important question is how much ‘similarity’ do we need? With an average molecular weight of 150 kilo Dalton for full-size molecules, it would be quite difficult to verify that each atom of a biosimilar mAb mapped exactly to those in a reference product. In fact, for both reference and biosimilar mAbs, the product likely consists of more than one drug substance, with minor differences in glycosylation, aggregation or other characteristics occurring between batches and over time. The differences might be due to fluctuations in the manufacturing process, e.g., pH, temperature, culture media, or changes in the expression system. The small changes might be meaningless, or they might have a high impact. However, industry and regulators now have ample experience with mAbs as therapeutics, with 23 marketed in the US, and nearly as many marketed in the EU. Although complex,4 manufacturing processes have also become quite consistent through-out the industry.
For non-clinical testing, a central aspect is that mAbs are species-specific and so the animal species relevant for testing a therapeutic intended for humans is an important question. A relevant species, as defined in EMEA's Note for Guidance on preclinical safety evaluation of biotechnology derived pharmaceuticals (CPMP/ICH/302/95; ICH S6), is ‘one in which the test material is pharmacologically active due to the expression of the receptor or an epitope (in the case of monoclonal antibodies).’ Importantly for biosimilar mAb developers, the relevant species for approved mAbs have been described. Potency assays for various approved mAbs are also known and available, including an anti-proliferation bioassay used to evaluate bevacizumab (Avastin), inhibition of binding assays suitable for evaluation of basiliximab (Simulect) and omalizumab (Xolair), and an in vivo potency assay in cotton rats for evaluation of palivizumab (Synagis).
There is also extensive patient experience with mAbs. For example, the anti-TNF product infliximab (Remicade), which was first approved in 1998, is currently approved for seven indications; there is cumulative product safety data for approximately 576,000 patients totaling 1.34 million patient years. However, on the con side, there is increasing evidence that glycosylation differences can affect mAb function.5,6 The current methods for characterizing mAbs, including physicochemical characterization, antigen-antibody interaction and secondary structure detection are increasingly sensitive, but, if differences are observed, the questions becomes what, if anything, should be done about these differences. On the other hand, if no differences are observed, then the question might be: Were the methods sensitive enough to find them?
Chemistry, Manufacturing and Controls Session
The stage was thus set for a point-counterpoint exchange on CMC by the innovator and biosimilar industries. EMEA had established key questions regarding CMC to be addressed, including: Are mAbs considered to be ‘well-characterized’ biologicals? Is available guidance for quality characterization sufficient for biosimilar mAbs? How well do current methods detect physicochemical differences between mAbs? To what extent do biological and functional assays substitute for a gap in sensitivity? What role should the biological assays play in comparisons of biosimilar mAbs? Can quality data substitute for gaps in knowledge in functional assays? How similar does the glycosylation need to be? Does a biosimilar mAb need to have the same distribution of antibody variants compared to the innovator product? What differences should be considered acceptable? What role should ICH Q8 and Q9 (quality risk analysis and risk management) play?
The innovator industry presentation was given by Georg-Burkhard Kresse (Hoffmann-LaRoche), who emphasized that the available guidance for quality characterization is applicable for biosimilar mAbs. These guidelines are quality characterization of mAbs (CHMP/BWP/157653/2007) and quality issues of biosimilar products (CHMP/BWP/49348/2005). However, he cautioned that it is not possible to characterize the quality attributes of mAbs completely by physicochemical analysis alone, or fully predict the impact of differences on clinical efficacy and safety. Similarity thus has to be shown in terms of quality, efficacy and safety in head-to-head comparative studies. A key point was that the ‘biosimilarity’ scenario differs from the ‘comparability after manufacturing changes’ scenario regulated by the International Conference on Harmonization (ICH) document Q5E.
Dr. Kresse noted that antibody modes of action are complex and may involve contributions from multiple mechanisms, and that the in vivo net contribution of different modes of action described for one mAb is often incompletely understood and may also be different in different indications. As a consequence of this multi-functionality, mAb characterization should include both Fab and Fc mediated functions unless there is justification to omit these studies. He emphasized that biosimilars must have the same amino acid sequence as the reference product, and that both reference and biosimilar mAb products will be micro-heterogeneous mixtures of a large number of post-translationally modified molecular species. The relevance of major variants on clinical efficacy and safety thus has to be established, and, since the exact composition of the mixture cannot be reproduced using a different manufacturing process, comparative non-clinical and clinical data will always be necessary for biosimilar mAbs.
The key question of glycosylation differences was then raised by Dr. Kresse. The fact that glycosylation can be critical for the biological function of mAbs has been established.7,8 In theory, IgGs can contain up to approximately 500 different glycoforms due to Fc glycosylation, and these differences may influence solubility, stability, clearance, immunogenicity and immune effector functions of the molecules. Even small differences in glycosylation, e.g., deletion of fucose residues,9 can have significant effects. Up to 30% of human IgGs contain N-linked oligosaccharides in the Fab region, and the functional significance of these has not been fully evaluated, e.g., impact of Fab galactosylation on hypersensitivity reaction. The experience of the innovator industry is that the pattern of glycosylation will vary between products because it depends on the manufacturing process. As a consequence of the critical nature of this attribute, the impact of glycosylation differences on clinical properties, which may be different for mAbs using different modes of action, should be proven or disproven.
In discussing functional assays, Dr. Kresse noted that quality data cannot substitute for gaps in knowledge. In the experience of the innovator industry, it may be difficult to understand critical quality attributes and predict the impact of differences on clinical efficacy and safety. For example, XOMA and Genentech separately produced batches of efalizumab that were found to have minor physicochemical differences, but gave notably different clinical results. Hence, only those differences known or proven to have no impact on clinical efficacy and safety should be acceptable without additional justification. Gaps in functional knowledge present at an early stage in the development process will lead to the requirement for additional non-clinical and clinical data, the specifics of which should be based on knowledge of the mode of action of the particular mAb. Quality (i.e., CMC), non-clinical and clinical aspects, are linked, and so a ‘holistic’ approach is needed for the evaluation of mAb-based drugs to connect analytical data with clinical safety and efficacy results.
Dr. Kresse's final point addressed the role of ICH Q8 (Pharmaceutical Development) and ICH Q9 (Quality Risk Management); he stated that these are applicable for biosimilar manufacturers for their own development processes in the same way as for the originators. However, he noted that the ‘design space’ concept depends on a particular manufacturing process connected to clinical studies results, and cannot be ‘borrowed’ from an innovator and used to demonstrate similarity of a biosimilar product to a reference product made by a different process. A biosimilar company has no access to the proprietary data of the innovator company needed to assess product quality attributes and batch-to-batch variability, or to understand batch difference relevance or impact on clinical safety and efficacy. As a consequence, the design space of the reference product cannot be utilized by a biosimilar manufacturer, and this manufacturer needs to establish a control strategy based on data generated for their own product.
On counterpoint, Martin Schiestl (Sandoz GmbH) first suggested that the phrase ‘well-characterized’ biological should be avoided because the term is not defined. He noted that mAbs can be characterized by DNA sequence; identity and amount of variants; glycosylation profile, including identity and content of individual glycans; and relevant bioassays for pivotal Fab and Fc-related biological functions, but whether this means the mAb is ‘well-characterized’ is subjective. The critical question is the clinical relevance of detected differences between the biosimilar and reference product.
Dr. Schiestl then noted that current physicochemical tools are able to detect batch-to-batch differences in reference product mAbs, and that these same tools should be applied to biosimilar mAbs. He emphasized again that the question is not the ability to detect differences, but the determination of their clinical relevance. In addition, a bioassay toolbox is available to characterize the relevant biological properties, and these bioassays complement physicochemical methods for determination of higher order structure. The bioassays help to establish structure-function relationships; they are an essential part of biosimilar comparisons and are equally needed in the holistic evaluation of biosimilarity together with physicochemical, preclinical and clinical data. Indeed, a comprehensive evaluation of multiple functional assays may enhance overall product understanding, and allow a reduction of the preclinical and clinical program.
Regarding substituting quality data for gaps in knowledge in function assays, Dr. Schiestl suggested that comprehensive quality data mitigates the risk of the unknown. As an example, he discussed that the combination of functional binding, complement dependent cytotoxicity (CDC) and antibody-dependent cell cytotoxicity (ADCC) assays, together with sensitive quantitative glycan data may also serve as a surrogate for unknown additional Fc functionality not directly covered by CDC and ADCC. On the question of how similar glycosylation should be, he stated that the identity of the individual glycan structures should be the same, and that the quantitative glycan composition should be comparable. However, the degree of acceptable differences in qualitative and quantitative composition would depend on the relevance of the respective individual glycan.
Dr. Schiestl noted that generally a biosimilar mAb should contain the same variants in comparable amounts as the reference product, but that deviations from this rule are acceptable depending on the level of understanding of the clinical relevance of the variants and differences. For example, differences in levels of terminal lysine variants may not affect the biological function, and glycosylation is known to play a reduced role for some mAbs exhibiting no effector functions. He emphasized that differences between biosimilar and reference products can be accepted based on the level of understanding of clinical relevance. An understanding of the process relative to product and critical quality attributes is needed, and existing public knowledge provides valuable input for risk assessments. However, final justification for remaining differences must be established by the biosimilar sponsor. Differences may be accepted based on the outcome of the overall comparability exercise, including physicochemical, biological, preclinical and clinical data, although a holistic interpretation of overall comparability data is needed. The existing knowledge of the mAb class may increase the level of confidence.
Regarding the question of ICH Q8 and Q9, Dr. Schiestl referred to ICH Q8 concepts of quality-by-design (QbD) and design of experiment (DoE) as of key importance in biosimilar mAb development. These concepts are recommended for development of a process that consistently delivers a comparable product. He stated also that ICH Q9 risk management procedures can be applied to the evaluation of biosimilarity, including the definition of comparability criteria, and evaluation and management of remaining differences. In concluding remarks, Dr. Schiestl noted that the expertise to develop and evaluate biosimilar mAbs is available at companies and regulatory agencies, as demonstrated by recent cases of approved and rejected applications for manufacturing process changes.
In the final presentation of the CMC session, Kowid Ho (Agence francaise de sécurité sanitaire des produits de santé) reviewed the European regulatory guidelines for similar biological medicinal products (Table 2). Dr. Ho briefly outlined the timeline of changes in the legal environment in the EU that ultimately allowed approval of Omnitrope in 2006 under the biosimilar framework. A critical factor was the issuance of an ‘overarching’ guideline in 2005, Guideline on similar biological medicinal products (CHMP/4307/04), that defined key concepts and principles for approval of biosimilar products. He then discussed aspects of guidelines CHMP/BWP/49348, CHMP/BWP/157653, and the availability of guidelines for specific products, including erythropoietins (CHMP/94526/05), granulocyte-colony stimulating factor (CHMP/31329/05), somatropin (CHMP/94528/05), human insulin (CHMP/32775/05), low molecular weight heparins (CHMP/BMWP/118264/07), interferon alpha (CHMP/BMWP/102046/06).
Table 2.
European Medicines Agency guidelines relevant to biosimilar development and approval
Dr. Ho emphasized that the target for quality of biosimilars is the quality profile of the reference product. There are a wide variety of physicochemical and biological characteristics to assess, including deamidation, oxidation, N-terminal pyro-Glu, glycosylation, glycation, constant region differences (e.g., deamidation, oxidation, C-term Lys), binding (e.g., affinity avidity, immunoreactivity), effector functions, epitope immunogenicity, modulatory region (e.g., Tregitope), and pharmacokinetics. Given the battery of available technology to assess these variables, he asked a key question: What differences in structure and function might be acceptable?
The CMC session group discussion was led by Jean-Hugues Trouvin. All participants were invited to make comments on CMC-related issues. First, there was general agreement that the ‘well-characterized’ term should be avoided. Dr. Kresse noted that the term was introduced by FDA in 1995, but abandoned in 1996,10 because it was difficult to achieve a sufficiently clear and specific understanding of the term. It was noted that variability is also a problem for innovators, and that it is the duty of all companies to maintain product consistency and eliminate uncertainty in the use of products. However, key questions remained for participants: What does it mean if differences are detected between biosimilar and reference products? What are the critical attributes of the product?
From the biosimilar industry perspective, the main point was that the manufacturing process will not be identical to that of the innovator, but the process will be controlled, and the product will be characterized sufficiently to support a claim of similarity to a reference product through linking physicochemical data with bioassays and clinical study results. Quality limits for a biosimilar attribute will be the range for the reference product. From the innovator industry perspective, the main point was that although mAbs have platform processes for manufacturing, the processes can still give variable products. The innovator has the experience to know how the variations might affect product attributes, including clinical results, but a biosimilar company does not.
The disconnection between knowledge about the product prior to administration to patients, and what happens in the patient was also discussed. Analytical methods can provide precise information about product characteristics, but what happens to the product in the patients is less well-understood. An example was given of a product that was carefully controlled for the presence of deamidation during manufacturing, but was found to be completely deamidated after administration to patients, suggesting that patients can and do tolerate product variants. In general, it was agreed that clinical studies are a blunt instrument for assessment of product differences.
The session was summarized by Jean-Hugues Trouvin. There was general agreement on the following points: (1) there is no need for new guidelines specifically for biosimilar mAbs, but application of the current provisions should be more consistent; (2) it is not possible to exactly reproduce a mAb product, which likely changes somewhat over time anyway; (3) the product and impurity profile of the reference product is the target for the biosimilar mAb; (4) different expression systems can be used, but problems are more likely to arise due to the increased difficulty in matching profiles; (5) better links between physicochemical analysis, bioassays and clinical data are needed; and (6) further work is needed to understand quality attributes and what they mean.
Non-Clinical Issues Session
EMEA posed the following questions on non-clinical issues: What non-clinical studies should be requested, given that the studies often need to be done in monkeys to be relevant, and thus the number of animals per group will be limited? How can pharmacodynamic (PD) measures (‘fingerprinting’) be supplementary to quality development? For antitumoral mAbs, to what level would a comparison on the functional level besides ADCC/CDC (if relevant) be required or feasible? What is the impact of formulation on in vivo behavior (injection site and infusion rate comparability), and how could it best be studied?
The innovator industry representative, Danuta Herzyk (Merck), initially discussed the role of non-clinical assessment of biosimilar mAbs. She emphasized that non-clinical pharmacology, pharmacokinetic and toxicology studies are key components of an integrated assessment of comparability between biosimilar and reference products. For comparative pharmacology, the equivalence of biological endpoints in response to both products needs to be demonstrated, i.e., in vitro potency assays at a functional level. Such comparative evaluations might include ligand binding as assessed by ELISA or Biocore, Fc receptor binding, cell-based assays (e.g., mitogenesis, flow cytometry, apoptosis), bioassays and in vivo animal models (e.g., murine xenographs, transgenic animals). For comparative pharmacokinetics (PK), the equivalence of PK parameters for both products in relevant animal species needs to be demonstrated. For comparative toxicology, the lack of toxicologically meaningful differences between toxicity profiles of the biosimilar and reference products needs to be demonstrated.
On the question of appropriate use of relevant species, Dr. Herzyk noted that comparative PK/PD obtained in a relevant species should be mandatory, but, where possible, PK, PK/PD (including dose response) studies should be combined to reduce the number of animals used. A head-to-head comparative PK/PD evaluation in an adequate animal model, if feasible, should be done to understand how in vitro PD results translate into in vivo effects. Toxicology studies should include one repeat dose study of minimal, but sufficient, duration to evaluate the toxicity profile in relation to that of the reference product. In principle, a comparator arm should be included unless exclusion is justified, but there is a need to balance extensive animal use against the ability to detect potential unexpected toxicity of a biosimilar relative to the described toxicity (or lack of it) for the reference product. A repeat dose toxicity study, typically done in non-human primates, that includes PD markers should be done, if feasible. The treatment duration should be adequate to detect potential differences between products. Recovery groups generally should be included, with control and high-dose recovery groups generally sufficient. However, if toxicity is known to be reversible, then there is no need to evaluate. Immunogenicity should be included to explain potentially unexpected PK/PD profiles or toxicity. Safety pharmacology should be included on a case-by-case basis, e.g., cardiovascular endpoints should be included in a repeat dose toxicology study. Injection sites should be evaluated to determine local tolerance.
Dr. Herzyk addressed the question of PD measures by explaining that PD markers for biosimilars should be chosen appropriately to demonstrate equivalent target binding or capture and other relevant functional endpoints. PK-PD characterization may utilize downstream markers from primary target binding based on known, relevant biology. Either single or multiple PD markers (a fingerprint) may be relevant to profile a biosimilar. However, broad spectrum ‘-omics’ approaches should be considered exploratory.
With regard to non-clinical evaluation of anti-tumoral mAbs, Dr. Herzyk stated that comprehensive, comparative (i.e., head-to-head) functional activity, in vitro characterization is needed. The need for such studies done in vivo in appropriate animal models should be considered based on results of in vitro characterization and the PK profile of the biosimilar mAb. These studies would be warranted when ADCC/CDC comparison results in significant differences, or the impact of the differences is not understood, and when PK profiles and in vivo findings in non-tumor animal models are significantly different. She further noted that the feasibility of the evaluation of anti-tumor mechanism of action-related endpoints, e.g., target dependent signalling pathways, is product dependent. In addition, she suggested that comparative evaluation might be enhanced if use of relevant endpoints in pharmacology studies generated with newly emerging methodology is considered.
Concerning the impact of formulation on in vivo behavior, Dr. Herzyk suggested that the pivotal non-clinical study for a biosimilar should mimic the injection site and infusion rate intended for use in clinical studies. However, if the injection site or infusion rate for a biosimilar is different from that used for the reference product, then a clinical study using the new conditions is warranted.
Dr. Herzyk summarized by noting that non-clinical pharmacology, PK and toxicology studies for biosimilar mAbs need to be adequately designed to detect potential relevant differences in therapeutic and safety profiles. The assessment criteria should be product-specific, and formulated in the context of full understanding of the product's structural, biochemical and bioactivity attributes, e.g., potency, PK/PD relationship, safety. She also explained that the extent of the non-clinical studies will be dependent on the nature of the pharmacology, as well as the nature of adverse effects and the dose-response relationship for known adverse effects. Her final point was that some aspects of biosimilarity, e.g., product label statements regarding immunogenicity, can currently only be addressed in properly designed clinical studies.
The biosimilar industry's perspective on non-clinical issues was provided by Alexander Berghout (Sandoz Biopharmaceuticals). He opened his presentation by emphasizing that the innovator has already established key factors for successful mAb development, including the availability of appropriate bioassays, the selection of appropriate animal species, antigen cross-reactivity, PK and PD, dose selection and treatment schedule, drug interactions, toxicity and safety profile, immunogenicity, and the history of clinical experience. Therefore, broad experience with the reference product will allow focused preclinical development of a biosimilar mAb. Dr. Berghout then noted that clinically-relevant PD effects of biosimilar and reference product should be compared in appropriate species. The non-clinical toxicity evaluation should be one repeat dose study that includes toxicokinetic measurements and local tolerance assessment. In addition, antibody titers and neutralizing capacity should be determined, and the study duration should be appropriate to allow detection of relevant differences in toxicity or immune responses.
On the question of appropriate use of relevant species, Dr. Berghout referred again to the fact that the innovator has already established key factors, including relevant species and the toxicity profile. He stated that dose-response is more suitably compared in non-clinical studies, rather than clinical trials, and, importantly, unnecessary duplication of toxicity studies with the reference product should be avoided. Dr. Berghout encouraged exploration of new methodologies, e.g., modeling, simulation, use of biomarkers, to optimize study design.
Dr. Berghout noted that functional bioassays to measure the principle mechanisms of action are indispensible in the target-directed development of a biosimilar, and are utilized throughout the process of engineering, selecting the desired clones and the final drug product development. All established effector functions should be investigated. Regarding antitumoral mAbs, Dr. Berghout commented that functional bioassays will usually be sufficient to establish the comparability of mAbs because the reference mAbs were generally selected by such assays. So, it may be expected that identical Fab binding to the target cell receptor will control signaling events in the same way. He also noted that in the case where modulation of signaling is the predominant function, respective analysis may be required.
On the topic of formulation, Dr. Berghout stated that, in general, the formulation, injection site and infusion rate will be similar for the reference product and the biosimilar product, and comparability will be confirmed in human studies. He noted also that the best way to explore the impact of formulation on in vivo behavior will be in a relevant animal model, and that the use of new methodologies such as modeling and simulation should be considered. In summary, Dr. Berghout reiterated his points that mAbs, like all biosimilars, follow the same principles of focused preclinical development, mAbs are multifunctional proteins requiring an extended set of bioassays for evaluation, use of new methodology should be explored to optimize study design, and unnecessary duplication of toxicity studies comparing biosimilar to reference products should be avoided.
The non-clinical session chair, Beatriz Silva-Lima (SWP), then presented her remarks on the non-clinical issues questions. Regarding non-clinical studies in relevant species, she emphasized that only informative studies should be requested, but that alternatives such as other models, e.g., in vitro or tissue cross-reactivity studies, may be considered and may be more informative. In any case, a thorough justification for the model used should be presented. On the question of PD measures as supplement to quality development, she noted that PD and quality are inter-related when relevant differences are identified, e.g., receptor-target interaction as assessed by potency, Emax, binding site, off-target characteristics, cellular cascades. Assessment of these qualities may reveal relevant differences and indicate when the products are not similar. For comparison of the functional activity of antitumoral mAbs, she stated that the feasibility level is dictated by the approaches taken for previous characterization of the reference product, taking into consideration relevance suggested by the biosimilarity exercise. Regarding formulation, she noted that the impact may be local, and these effects would be application-site and vehicle dependent. However, systemic variation, e.g., different kinetics, enhanced activity, modified immunogenicity, would need to be assessed in the case of different formulations.
Dr. Silva-Lima also raised some additional questions for consideration. These were: (1) If non-similarity of the biosimilar product is concluded, due for example to a different glycosylation pattern compared to the reference product, but the basic molecule and mechanism of action are the same, then how should the development of the product proceed? (2) What about non-human primate (NHP) developmental and reproductive toxicology (DART) studies for biosimilar mAbs or a ‘non-similar’ product when there is experience, presumably in humans also, with the reference product?
The discussion on non-clinical issues was moderated by Beatriz Silva-Lima. The point of only repeating studies that were indicated as relevant in innovator development programs was reiterated. The role of comparative toxicology, considering the small number of NHPs used, was questioned since understanding the resulting outcomes could be challenging. A representative of the innovator industry suggested that creation of animal models, e.g., transgenic mouse model, might be a solution to the limited size of NHP studies.
Numerous exchanges occurred concerning the non-clinical evaluation of potential impurities. It was suggested that if toxic effects of impurities needed to be assessed, then perhaps a non-relevant species such as the rat or even human tissues could be used for these studies. A rodent model might also be used to assess toxicity of glycoforms. However, representatives of the biosimilar industry questioned the basic assumption. If CMC-related studies and PK/PD studies have indicated that the biosimilar is comparable to the reference product, then why are toxicity tests of putative variants and impurities necessary? Process-related impurities are in fact not an issue specific to biosimilars, but a general problem that is dealt with prior to the non-clinical evaluation stage. There are no known examples of toxic or negative clinical outcomes that have been specifically linked to a product variant. Regarding DART studies for mAbs, the question of relevance was raised, since other biosimilar products are not required to undergo reproductive toxicology evaluation. Also from the perspective of the biosimilar industry, head-to-head comparability studies in toxicology do not make sense.
The final question discussed during the session related to how much information should be asked for at the non-clinical stage. For example, regulators could request studies on mechanism at the target, i.e., the effects of the signaling pathway. Studies like this might not have been done by the innovator. It was noted that, from the regulator's perspective, teasing apart what an innovator has done and what biosimilar companies should have to do might be a challenge. However, from the biosimilar industry perspective, the fundamental question remains the same: if similarity has been established, then why are studies that should provide expected results necessary? Addressing the specific example, if binding to the target has been shown to be comparable to the reference product, then one would expect that the downstream effects of binding by either the biosimilar or reference product would be the same. Dr. Silva-Lima summarized the main points of the discussion as: (1) it was generally agreed that comparative PD studies are useful; (2) a case-by-case approach for non-clinical evaluation is justified; and (3) dedicated studies for effects of impurities are not needed.
Clinical Issues Session
The clinical issues session included topics relating to PK/PD, extrapolation of efficacy and safety, and outcome measures. The PK/PD-related questions to be considered were: (1) What role could new methodologies such as simulation, modeling and biomarkers play in clinical studies? (2) In which population(s) should PK/PD be measured? Questions relating to the extrapolation of efficacy and safety were: (1) To what extent can efficacy be extrapolated from one indication to another in different scenarios, provided that physicochemical and biological characterization has been shown to be comparable? For this question, EMEA requested that three specific cases (immunomodulatory mAbs that might involve extrapolation from psoriasis to rheumatoid arthritis, antitumoral mAbs and antitumoral mAbs that are also indicated in inflammatory conditions) be considered; (2) To what extent can safety be extrapolated, and what can be done post-marketing?; and (3) For antitumoral mAbs, what would be acceptable as patient subpopulations for studies in different indications?
In the outcomes measures area, EMEA's questions for discussion were: (1) Of the following, which endpoints should be used as a general strategy—endpoints that measure patient benefit, but might be less sensitive for detecting product differences; endpoints, e.g., activity endpoints, that measure similarity more sensitively; or, if similarity endpoints are used, should these conform to guidelines or could these be newly developed endpoints? (2) What role could new methodologies such as simulation or modeling play? (3) to what extent would a risk-based approach to immunogenicity be applicable, given that mAbs do not have endogenous counterparts?
The perspective of the innovator industry was presented by Jay Siegel (Johnson & Johnson). He noted that once high similarity had been demonstrated in laboratory and non-clinical testing, clinical similarity may then be tested head-to-head. Extrapolation across endpoints, populations or diseases should be justified scientifically. However, he emphasized that applications of the principles should take into account particular properties of mAbs, such as the fact that multiple features of mAbs determine the clinical activity, critical structure-function relationships are often not well-understood, and mAbs are generally used to treat serious or life-threatening diseases.
A key point was that extrapolation of efficacy would likely be difficult to justify. Dr. Siegel stated that mAbs have diverse functional activities and may be used in diverse indications. However, different indications can require different activities and receptors (or combinations of these) in different sites over different time courses, and in different pharmacologic milieu. As a consequence, mAbs with similar effects in one disease may have different effects in a second indication if the second indication involves a different mechanism of action, action at a different site, a longer time frame, a change in the amount of target antigen expressed or use of different concomitant medications.
On the question of endpoints that measure patient benefits, Dr. Siegel suggested that the science-based principles presented in current EMEA guidelines will, for many mAbs, dictate study of clinical benefit endpoints. He explained that biomarkers may not reflect all relevant activities of mAbs, relevant activities of mAbs often are not fully understood, and dose-response relationships of competitive inhibitors are often complex. As a consequence, differences between biosimilar and reference products may impact the effect on clinical outcomes without impacting the effect on biomarkers. In fact, markers rarely provide quantitative prediction of efficacy. Modest differences in efficacy could have a significant, irreversible impact on many diseases treated by mAbs. Dr. Siegel also noted that where clinical outcomes data are needed, biomarker data can supplement those data, potentially decreasing the amount of clinical outcomes data needed and increasing confidence in the clinical similarity of the biosimilar and reference products.
Dr. Siegel continued by noting that biomarkers and activity endpoints can often be measured faster, cheaper and with more precision than can clinical outcome measures. A ‘highly similar’ biosimilar mAb should in fact be highly similar in all effects in patients. However, he reiterated the point that similarity in effects on biomarkers will not always predict similarity of effects on clinical outcome. The regulatory implications of these points are that head-to-head comparisons of effects on biomarkers will be powerful tools in identifying or excluding some clinical differences, and may prove valuable in supporting extrapolation to other indications, although demonstration of similar effects on easily measured biomarkers should be considered necessary, but not usually sufficient, to establish equivalence.
Regarding the question of immunogenicity data, Dr. Siegel emphasized that increased or altered immunogenicity in any biosimilar mAb has the potential for significant clinical implications. He recommended that all mAbs should be assessed for immunogenicity as described in EMEA guideline CHMP/BMWP/14327/2006, and suggested that biosimilars should be studied head-to-head with the reference product. He cautioned that similar incidence of immunogenicity does not necessarily mean similar immunogenicity.
In conclusion, Dr. Siegal noted that strong CMC and non-clinical data that limits potential differences between the biosimilar and reference product are critical. EMEA guidelines relevant to biosimilar biotechnology-derived proteins (Table 2) serve as a good starting point for clinical requirements for mAbs. However, key properties of mAbs have important implications for how EMEA guidelines should be applied. Extrapolating data between indications should only be done when mechanisms of action in both indications are understood and highly similar, bearing in mind that implications of immunogenicity for mAbs are always potentially substantial. Immunogenicity cannot be predicted, so it must be measured directly.
The biosimilar industry's perspective on clinical issues was presented by Islah Ahmed (Hospira), who first emphasized the point that prior to entering clinical development, biosimilar mAbs will have already demonstrated comparability to the reference mAbs in physicochemical characterization, non-clinical studies (e.g., PK, PD, toxicity profiling), and in vitro functional characteristics. The goal of the clinical development program is then to complement the comparability exercise by demonstrating therapeutic equivalence within an abbreviated pathway.
Regarding the question of new methodology applied to PK/PD studies, Dr. Ahmed noted that the PK of mAbs is well-understood, and basically follows the PK of human IgG. PK/PD are based on the type of mAb target, and he provided several examples: (1) if the target is a soluble antigen with low endogenous levels, then PK is often independent of PD, the PK is linear and the half-life is long; and (2) if the target is a soluble antigen or cell bound, then PK often depends on PD, PK is non-linear and the half-life is short for low dose and long for high dose. He also mentioned that the general clearance of mAbs is based on catabolism and the renal clearance is negligible. PK/PD modeling might be designed based on data from the reference mAb if such data are available. Dr. Ahmed emphasized that PK/PD is pivotal to the comparability evidence, and with validated PK/PD models, the comparability study design can be optimized to minimize the number of patients and samples.
Dr. Ahmed addressed the question of which population should be studied by stating that there should be flexibility in the population selection on a case-by-case basis, although the selection would be done by mutual agreement with regulators. For example, a PK study in healthy volunteers may be technically preferable, but not acceptable because of ethical reasons. In other cases, a patient population in an approved indication with low variability in PK will be most suitable. Dr. Ahmed noted that patients in PK/PD trials must be treated for full clinical benefit, and not only for PK/PD comparability. He suggested that PK/PD can be combined within an efficacy and safety study. He also pointed out that PK/PD sampling depends on the PK/PD profile of the reference mAb, e.g., in selected cohorts or as sparse sampling for population kinetics.
Concerning the question of extrapolation from one indication to another, Dr. Ahmed stated that extrapolation of efficacy is acceptable, provided that the mAb has demonstrated comparability to the reference mAb in the parameters already discussed (physicochemical, non-clinical, and in vitro functional characteristics; bioavailability and clinical PK/PD; clinical efficacy in one indication). He emphasized that once comparability has been demonstrated in one indication, there is no scientific reason to expect that the response of the host to the biosimilar product should differ from that of the reference mAb in other indications.
Although noting that safety risk profiles may differ in different indications because of variables such as concurrent conditions or concomitant medications, Dr. Ahmed stated that extrapolation of safety is also acceptable. He noted that safety comparability could be demonstrated in an indication that was judged to have a high sensitivity toward detection of differences and extrapolated to all other indications. In addition, he suggested that a risk management program implemented to collect safety data for low frequency safety risk, e.g., immunogenicity as observed in the post-approval patient population, will usually be similar for biosimilar and reference products. With regard to patient subpopulations, Dr. Ahmed suggested any subpopulation that is most sensitive toward detection of differences between biosimilar and reference mAbs would be acceptable, and that subpopulations with high response rates, those who are more homogeneous with regard to disease stage, or those in which validated biomarkers could be used may be selected for comparability trials.
On the question of efficacy end points, Dr. Ahmed observed that the primary objective of the clinical study program is comparability, and not generation of new evidence of efficacy. Therefore, an abbreviated data package that includes minimal patient exposure to research is appropriate for biosimilar mAbs. A flexible approach would be suitable, with use of validated surrogate end points, if these are available. He suggested that long-term survival based on patient benefit end points is not always necessary, even for antitumoral agents. For example, the liposomal formulation of the antitumoral agent doxorubicin (Myocet) was approved based on an objective response rate as the primary end point. He then noted that an objective response can serve to demonstrate comparability, at least in some mAbs.
Regarding new methodology, Dr. Ahmed suggested that, as the state of science progresses, new methodologies should be applied wherever possible for biosimilar mAbs to achieve abbreviated clinical data packages and alternate statistical models, e.g., Bayesian statistics, can help to optimize clinical trial sample size. He briefly addressed the immunogenicity question by stating that a risk-based approach to immunogenicity should be applied to all biologics, including both biosimilar and reference products.
In conclusion, Dr. Ahmed restated his points that clinical development of biosimilar mAbs should be flexible in the design of PK/PD, efficacy and safety studies to demonstrate comparability in the most sensitive model; modeling, simulations, and statistical methods are applicable to achieve an abbreviated approach to the demonstration of comparability; a risk management program to monitor low frequency safety risk should be the same as for the reference product; and efficacy and safety demonstrated in one sensitive model can be extrapolated for all indications approved for the reference product.
The final speaker, Christian Schneider (EMEA BMWP) presented the regulators perspective on clinical issues. He began by acknowledging the oft-repeated observation that mAb mechanisms of action can be complex, then went on to ask what he referred to as the frequently asked questions' of a heretic: Can the mechanism of action be understood solely as a ligand-receptor interaction (or inhibition, as the case may be)? Is it important to know what comes after? Does the mechanism of action have to be known? Using efficacy and safety of marketed anti-TNF antibodies as an example, he asked how one would design a biosimilar development program that might allow licensure in the seven indications for which these mAbs are now approved? Is therapeutic equivalence or non-inferiority suitable? Should all indications be approved? Should extrapolation of efficacy or safety be allowed? What end points should be used—activity or benefit? What are appropriate for Phase 2 or Phase 3 endpoints? He pointed out that the ‘Guideline on Similar Biological Medicinal Products containing Biotechnology-Derived Proteins as Active Substance: Non-clinical and Clinical Issues’ (EMEA/CPMP/42832/05) already discusses extrapolation of therapeutic similarity shown in one indication to other indications of the reference product.
Dr. Schneider discussed the spectrum of uncertainty that is traversed when considering ‘biosimilars’ of peptides, non-glycosylated proteins, glycosylated proteins, mAbs, blood products and finally advanced therapy medicinal products, e.g., cell therapy. He pointed out that EMEA has already begun to address cases involving complexity and advanced degrees of uncertainty. He directed the workshop participants to the ‘Guideline on non-clinical and clinical development of similar biological medicinal products containing low-molecular-weight-heparins’ (EMEA/CHMP/BMWP/118264/2007) and also to the reflection paper ‘Non-clinical and clinical development of similar medicinal products containing recombinant interferon alfa’ (EMEA/CHMP/BMWP/102046/2006).
Dr. Schneider concluded by raising additional questions regarding immunogenicity and practical issues. He pointed out that mAbs do not substitute for endogenous proteins like other recently approved biosimilars such as epoetin and filgrastim. So, is the perception of risk different? Antibodies against mAbs are mostly anti-idiotype, not anti-isotype and endogenous IgG is abundant. While not suggesting that immunogenicity is unimportant, should immunogenicity be the ‘highest’ safety concern? On practical issues, he wondered about the extent to which the biosimilar philosophy is known to patients and physicians, which leads to questions regarding the acceptability of biosimilar mAbs, especially in an oncological setting. His final questions concerned how to practically deal with Phase 1 PK/PD studies in patients: These are usually single dose studies—should cross-over be used? How should treatment be continued—should patients be switched to the reference product?
Lively exchanges then followed in the clinical issues group discussion session moderated by Christian Schneider. Addressing the ‘heretical’ questions first raised by Dr. Schneider, the conservative response that efficacy is dependent on much more than blocking a ligand or receptor, and so knowing the downstream effects of disrupting the signaling pathway is important was given. However, from the biosimilar industry perspective, the determination that a biosimilar mAb binds to the same epitope as the reference product with same binding constant is part of preclinical evaluation, and, if non-clinical data shows similar results, then it follows that clinical results would also be similar. If differences are seen at the non-clinical stage, then decisions regarding clinical studies must be data-driven.
Further discussion was based on the assumption that hypothetical biosimilar mAb and reference products had demonstrated similarity in quality and non-clinical aspects. Questions and comments regarding extrapolation of results from one indication to another were made by participants. The extension of the use of rituximab from oncology to rheumatoid arthritis patients was given as an example. If extensive clinical studies of the biosimilars in both indications are not done, then there may be a risk of under-treating patients. In general, there may be different responses in different patient populations due to such factors as different receptor levels. However, the point was made that the evidence would have to suggest that patient safety is not at risk, and it is the task of regulators to determine risk to patients. If CMC and non-clinical data show similarity, then the science supports initial clinical studies, although at least one clinical study of each indication would likely be needed. The cases when extrapolation would be a challenge were enumerated: low dose to high dose, combination with other therapeutics and less severe to more severe indications. Other cases, such as extrapolating from first-line to second-line treatment, might be acceptable.
Other aspects of clinical studies were then discussed. Questions arose over how to map PD markers to efficacy since the PK/PD relationship to efficacy is weak, how to choose the most sensitive patient population, and selection of endpoints most sensitive to differences. It was noted that PK is non-linear (i.e., dose dependent), time dependent, and can differ across patient populations. Clinical testing should be comparative, and designed to asses these characteristics of mAb PK. Populations used for PK/PD measurement should be carefully chosen because PK or PK/PD can be different due to mechanism of action, patient age, other medication or disease state. For example, PK and immunogenicity of mAbs are different in pediatric and adult patient populations.
Regarding endpoints, it was noted that the primary endpoint used for clinical studies of the reference product would likely be suitable since regulators have dossiers for reference products. If alternate endpoints are used during the clinical development of a biosimilar product, then it would be difficult to compare with results for the reference product. A key question from the innovator industry was then posed: Do the clinical study results have to show similarity or rule out any possible difference? However, from the biosimilar industry perspective, practicality has to be taken into consideration. Comparison studies might be lengthy and require a large number of patients. Alternate or surrogate endpoints defined by regulatory agencies might be needed to ensure the clinical studies are feasible. For example, study of an antineoplastic mAb that used survival as an endpoint would take many years. The possibility of use of a conditional approval mechanism, whereby marketing approval was given but survival data for clinical study participants continued to be collected, was mentioned.
Additional points were made regarding safety and immunogenicity. In general, if there are safety concerns with the reference product, then biosimilars should be monitored for the same safety problems. It was noted that post-marketing safety might be challenging after biosimilars are approved because patients might not know what product they are taking and so spontaneous reporting of adverse events might not be accurate. The question of the use of a risk-based approach to immunogenicity was then raised. Problems could arise if biosimilar products are more immunogenic compared to the reference product because the immune response could affect dosing with the innovator product. However, it was noted that immunogenicity is part of the comparability exercise and a class effect that applies to all mAbs.
Concluding Discussion
In concluding the workshop, Dr. Schneider raised final two questions that were briefly discussed. The first question was, should the biosimilar framework be expanded to include products with differences in the amino acid sequence? The general consensus was that the two products have to be the same, and avoidable changes such as amino acid substitutions should not be allowed. The second question was, could some concepts relating to biosimilars be applicable to second-generation products, at least those that are functionally equivalent, but may be structurally different? The general consensus was that there is too much uncertainty surrounding potential differences in modes of action or off-target effects that might be seen with molecules that have structural differences. For example, the anti-TNF mAb infliximab is shows effects in Crohn disease patients whereas the anti-TNF fusion protein etanercept does not.
Dr. Schneider also answered a few questions from participants. The first question concerned international non-proprietary names (INNs). Dr. Schneider stated that INNs are assigned by the World Health Organization (WHO), but are requested on a voluntary basis by companies. The WHO majority decision is that different substances should have different names, leaving it likely that the same substance would be assigned the same name. Naming has ramifications for traceability, and whether confusion would result in cases of voluntary reporting of adverse events, i.e., whether the patient was taking a branded product or biosimilar version. The second question on interchangeability could not be addressed by EMEA regulators because there are country-to-country differences in requirements.
In conclusion, Dr. Schneider outlined the next steps for EMEA. Further internal discussion on whether another guideline is needed for biosimilar mAbs will occur. The group consensus of this workshop seemed to be that there is no need for a new guidance on quality, although a guideline for non-clinical and clinical requirements may be needed. However, writing one might be challenging because EMEA might have to anticipate what would be allowable for differences between reference and biosimilar products. If needed, EMEA would develop a concept paper that would be distributed for comment, then follow with a guideline. If done, it is unlikely that this would happen before 2010.
References
-
1.Woodcock J, Griffin J, Behrman R, Cherney B, Crescenzi T, Fraser B, et al. The FDA's assessment of follow-on protein products: a historical perspective. Nat Rev Drug Disc. 2007;6:437–442. doi: 10.1038/nrd2307. [DOI] [PubMed] [Google Scholar]
-
2.US Food and Drug Administration, author. Guidance concerning demonstration of comparability of human biological products, including therapeutic biotechnology-derived products. http://www.bcg-usa.com/regulatory/docs/1996/FDA199604A.pdf.
-
3.EMEA web site. http://www.emea.europa.eu/htms/general/contacts/CHMP/CHMP_BMWP.html
-
4.Carson KL. Flexibility—the guiding principle for antibody manufacturing. Nat Biotechnol. 2005;23:1054–1058. doi: 10.1038/nbt0905-1054. [DOI] [PubMed] [Google Scholar]
-
5.Kaneko Y, Nimmerjahn F, Ravetch JV. Anti-inflammatory activity of immunoglobulin G resulting from Fc sialylation. Science. 2006;313:670–673. doi: 10.1126/science.1129594. [DOI] [PubMed] [Google Scholar]
-
6.Sheridan C. Commercial interest grows in glycan analysis. Nat Biotechnol. 2007;25:145–146. doi: 10.1038/nbt0207-145. [DOI] [PubMed] [Google Scholar]
-
7.Jefferis R. Recombinant antibody therapeutics: the impact of glycosylation on mechanisms of action. Trends Pharmacol Sci. 2009;30:356–362. doi: 10.1016/j.tips.2009.04.007. [DOI] [PubMed] [Google Scholar]
-
8.Jefferis R. Glycosylation as a strategy to improve antibody-based therapeutics. Nat Rev Drug Discov. 2009;8:226–234. doi: 10.1038/nrd2804. [DOI] [PubMed] [Google Scholar]
-
9.Yamane-Ohnuki N, Satoh M. Production of therapeutic antibodies with controlled fucosylation. mAbs. 2009;1:230–236. doi: 10.4161/mabs.1.3.8328. [DOI] [PMC free article] [PubMed] [Google Scholar]
-
10.Federal Register. 1996;61(94) Tuesday. [Google Scholar]
