CMC STRATEGY FORUM JAPAN 2017

Transcription

1 CMC STRATEGY FORUM JAPAN 2017 Prior Knowledge in Attribute Based Control Strategies Michael Abernathy Executive Director, Regulatory Affairs Jette Wypych Director, Process Development Amgen 06 December

2 Disclaimer The views expressed herein represent those of the author and do not necessarily represent the views or practices of the author s employer or any other party. 2

3 TODAY, WE ARE IN A BIOTECHNOLOGY ERA WHERE INNOVATION PROGRESSES RAPIDLY Processing Manufacturing Analytical Testing Novel Modalities Timely innovation is integral to our industry 3

4 BIOLOGIC MEDICINES ARE MORE HIGHLY ENGINEERED AND DIVERSE Product quality drivers Supply requirements Financial considerations Regional manufacturing One size does not fit all APPROPRIATE MANUFACTURING TECHNOLOGIES CAN BE MATCHED TO MODALITIES TO DELIVER TO THE QUALITY TARGET PRODUCT PROFILE 4

5 AN INDUSTRY GOVERNED BY cgmp AND REGULATORY GUIDELINES EXPECTATION OF SPONSORS IS TO STAY CURRENT WITH REGULATIONS AND GUIDANCE REGULATIONS STAY CURRENT AND ALIGN WITH TECHNOLOGICAL INNOVATION 5

6 MANUFACTURING IS CHANGING Conventional Flexible Key Enabling Technologies High titer processes Single-use systems Modular design and construction Connected processing Online / At-line analytics Real-time remote monitoring Raw material variation control PATIENT BASED, MODULAR AND DISTRIBUTED MANUFACTURING MAY REPRESENT OUR NEXT PARADIGM SHIFT 6

7 TODAY WE ARE AMIDST AN INDUSTRY REVOLUTION WHERE KNOWLEDGE AND DATA MANAGEMENT ARE CRITICAL Product portfolios continue to evolve and correlated knowledge and data are growing exponentially Small Molecules Therapeutic Proteins Novel Treatment Modalities - Simple - Single defined structure - Predictable chemical /reagent reaction - Production of identical copies - Stable - Easy to fully characterize - Minimal data packet - Large - Complex structure - Bank of living cells - Identical clones unlikely - Sensitive to environmental conditions - Correlation of structure/function elusive - Robust data packet - Vary from small to large - Complex and unique structure - Varied modular manufacturing - Heterogenous sub-populations - Varied stability - Personalized nature difficult and costly to characterize - Limitless data packet HOW DO WE EFFECTIVELY MANAGE THIS KNOWLEDGE AND DATA? 7

8 KEY STEPS IN IMPLEMENTATION OF QbD FOR A BIOTECHNOLOGY PRODUCT (ICH Q8 (R2)) AN OPPORTUNITY TO LEVERAGE PRIOR KNOWLEDGE QTPP forms basis of design for the development of the product Product quality attribute assessment ranks the risk of an attribute having a clinical impact to identify quality attributes with higher risk that need to be within an appropriate range/limit to ensure the desired product quality (critical quality attributes) Product Risk Assessment (s) to link material attributes and process parameters to CQAs Lifecycle management QTPP Product quality attribute assessment Control Strategy designed to ensure that a product of required quality will be consistently produced Integrated control strategy Product risk assessment Lifecycle management and continuous improvement 8

9 AMGEN IS APPLYING PRIOR KNOWLEDGE IN PROCESS DEFINITION PLATFORMED MODALITIES AND ASSOCIATED ATTRIBUTES: Mabs: IgG1s and IgG2s Canonical BiTEs Half-life Extension BiTEs Fc fusion Proteins PROCESS DEFINITION AND MANUFACTURING TECHNOLOGIES: DRUG SUBSTANCE: Harvest, VI and Filtration Batch, Perfusion, Continuous Affinity, CEX, HIC and Mixed Mode UF/DF DRUG PRODUCT: 9

10 THE ULTIMATE ATTRIBUTE BASED CONTROL STRATEGIES INTEGRATE ALL ASPECTS OF PROCESS AND PRODUCT CONTROLS In-process testing (IPCs, process monitoring, validation) End product testing (specifications, comparability, stability) Input controls (raw materials and components) Production Process Procedural controls (process design and facility, equipment and operational controls) AN EFFECTIVE, RISK AND KNOWLEDGE BASED INTEGRATED CONTROL STRATEGY ENSURES CONSISTENT PRODUCT QUALITY 10

11 INPUT CONTROLS: AMGEN ADOPTS THE ICH Q9 CONCEPTS PRIOR KNOWLEDGE IS APPLIED TO E&L STUDIES Identification Extraction studies use a wide array of analytical methods to characterize and identify extractables (> lower limit of quantitation) Qualification Toxicology assessments are conducted on identified extractables. Thresholds (e.g. TTC, PDE) are established on compounds of concern in the qualification of process and product contact materials Action Threshold Thresholds (e.g. TTC, PDE) are converted to concentration limits in drug products to guide methods development in targeting the compounds of concern Action thresholds ( 10 micrograms/dose) are limits above which a quality investigation is conducted to determine potential product impact. Threshold of Toxicological Concern (TTC): A level of exposure to a chemical below which there would be no appreciable risk to human health (FDA CFSAN 2011, Kroes et al. 2004, Patlewicz et al ) Permitted Daily Exposure (PDE): An acceptable intake of an impurity (e.g. residual solvent) in a pharmaceutical product (ICH Q3C) 11

12 AMGEN IS BUILDING AN E&L DATABASE ON A WIDE VARIETY OF MATERIALS The Amgen database has > 169 organic compounds from extractables & leachables studies: Pre-filled syringes Vial stopper systems On-body infusion devices Single-use bioprocess systems and components Process contact materials THE DATABASE IS UNIQUELY RELEVANT TO BIOTECHNOLOGY PRODUCTS 12

13 USE OF PRIOR KNOWLEDGE: PROCESS CHARACTERIZATION A priori determination of potentially high risk process parameters Prior Knowledge Assessments Amgen has accumulated a large knowledge base on cell culture and purification performance, including the impact of process parameters on quality attributes and process consistency This data can be leveraged to focus design studies on higher risk areas: Potential critical process parameters (CPPs) are identified for further evaluation based on prior knowledge or knowledge gaps Non-critical process parameters (ncpps) can be identified without product specific experimentation in scenarios where the process/product of interest is sufficiently homologous with the available prior knowledge data Risk and knowledge based process design facilitates deeper understanding of the impact of the process on product quality Cell Culture Applications For common cell lines and media, in seed train processes where product quality is not directly impacted, parameter ranges can be determined from early development studies without additional characterization Culture performance is evaluated every production lot through in-process testing control, ensuring a continued well controlled process For the production bioreactor, potential CPPs and ncpps can be identified for processes that apply similar cell lines and process conditions, especially in cases where a large body of prior knowledge exists (e.g., Mabs) Focused process characterization studies then enable a detailed understanding of the impact of process parameters on product quality and process consistency Purification Applications For downstream unit operations, potential CPPs and ncpps can be identified for common unit operations (e.g., Protein A chromatography, viral filtration, etc.), especially in cases where a large body of prior knowledge exists (e.g., Mabs) 13

14 EXTENSIVE PLATFORM DATA CLEARLY IDENTIFY HIGH RISK PARAMETERS (RADIAL PLOTS OF NORMALIZED IMPACT) Impurity 1 Impurity 2 Load Rate (g/l resin) Resin lot / ligand density 0.8 Load ph Temperature Flow rate Gradient length Stop Collect Bed Height Load cond Load treatment Equil ph Gradient length Equil Conductivity / concentration Equil volume Stop Collect Temperature Flow rate Load Rate (g/l resin) Bed Height 0.5 Load ph Load cond Load treatment Equil ph Equil Conductivity / concentration Start Collect Elution salt concentration/cond Elution buffer ph Wash volume Elution buffer Concentration Wash ph Wash cond./ concentration Start Collect Elution salt concentration/cond Elution buffer ph Elution buffer Concentration Wash ph Equil volume Wash cond./ concentration Wash volume Same process parameters impact impurities 1 and 2 14

15 AMGEN IS APPLYING PRIOR KNOWLEDGE FOR ANALYTICAL MEASUREMENTS PLATFORMED MODALITIES AND ASSOCIATED ATTRIBUTES: Mabs: IgG1s and IgG2s Canonical BiTEs Half-life Extension BiTEs Fc fusion Proteins MEASUREMENTS, ANALYTICS, ADVANCEMENT OF PAT: ANALYTICAL METHODS PAT and MAM Advancement 15

16 PLATFORM APPROACH TO PRODUCT PURITY AND PROCESS IMPURITY METHODS ATTRIBUTE(S) TYPE PLATFORMED PARAMETERS APPLICATION Product Concentration Spectroscopy System, controls Multi-product HMWS SE-UHPLC Chromatography system. Column, load, buffers Post translational modifications and clips Multi-attribute Method LC-MS system, column, load, buffers, digestion conditions Charge variants CEX-HPLC Chromatography system. Column, load, buffers Product specific or Universal Product specific Product specific Clips CE-SDS CE system, capillary, sample preparations, run conditions Product specific DNA qpcr System, sample preparation and run conditions Multi-product Host cell Protein ELISA System, sample preparation and run conditions Multi-product Potency Bioassay System, sample preparation and run conditions Product Specific Identity Raman/ELISA System, sample prep Product Specific Using platform approaches for measurements optimizes development 16

17 PLATFORM APPROACHES TO ANALYTICS PROVIDE FIRST INTENT AND IS LARGELY BASED ON PRIOR KNOWLEDGE Product Quality Attribute Assessment and Quality Target Product Profile Molecule Assessment Sequence Variant Analysis Platform Release and Stability Methods Impurities Testing Testing Strategies for Process Reagents Stability Indicating Properties of Methods Analytical Target Profile (ATP) Biological Characterization Product Characterization and CMC Higher Order Structure and Particles Comparability Biosimilarity Specifications Control Strategy Rapid Analytics Support for Process Development GxP Testing Sample Plan Reference Standard Stability and Expiry Method Qualification and Validation Method Technology Transfer Method Remediation Attribute Impact Forensics Extractables and Leachables Process Analytical Technologies and Product Attribute Control Predictive Modeling and Digital Analytics Hardware and Software Platform Platform Adherence and Technical Performance Reviews Against 17

18 START WITH A WELL DEFINED TARGET PRODUCT PROFILE (TPP) Patient centric life cycle management Target Product Profile (TPP) Product Quality Attribute Assessments (PQAA) Quality Target Product Profile (QTPP) Indication & use Dosage & administration Tolerability Dosage forms & strength Efficacy Safety/side effects Value & access Attribute definition Product quality attribute assessment Potential impact for safety/efficacy Critical quality attribute selection Attribute range determination Attribute focused molecule & process design & development 18

19 PRODUCT QUALITY ATTRIBUTE ASSESSMENT (PQAA): IDENTIFY ATTRIBUTES AND IMPACT Target Product Profile Identifying Attributes Scoring impact on safety and efficacy Target ranges Product Quality Attribute Assessment Quality Target Product Profile CDR modifications Oxidation, Deamidation, Isomerization (molecule specific) Loss of potency Low, < x % Fc binding regions Methionine oxidation PK and efficacy Low, < x % ± y% Glycan structure Other backbone modifications and aggregated forms High mannose variants (IgG class) PK and efficacy Low, < x % ± y% Sialylation PK high x- y% Disulfide variants (IgG2, IgG4) Potency Depends on criticality Truncated/clipped forms Potency and PK due to missing functional regions high, < x% Host Cell Protein Immunogenicity xppm 19

20 QTPP: PQA RANGES CAN BE DESIGNED INTO THE PRODUCT DURING DEVELOPMENT Peptide modifications Deamidation Succinimide Oxidation N and C-terminal variants Amino acid substitution Disulfide isoforms Glycosylation G0, G1, G2 Core fucosylation High mannose Size Truncation Half molecules Dimer/Multimers Particles Category DS Attributes Target Range Ranges Achieved Strength Concentration mg/ml mg/ml Asp Isomerization 2% % Trp Oxidation 5% 0.1% Met Oxidation 5% % Met Oxidation 5% 0.4% Met Oxidation 1% 7% % Met Oxidation 5% % High Mannose Glycans 2% 12% % Quality Protein Dimer/Oligomers (SEC HMW) 1% % Protein Fragmentation (rce LMW+MMW) 1% < 0.6% Glycation (LC K) 5% % Hydroxylysine (HC K) 2% < 0.1% Hydroxylysine (HC K) 2% % Osmolality mosm/kg mosm/kg Polysorbate % 0.015% % ph Host Cell Protein 100 ppm ppm Safety Residual Protein A < 6 ppm < 1 ppm Endotoxin 0.25 EU/mg EU/mg Bioburden 10 CFU/10 ml 0 20

21 EVOLVING OUR BUSINESS PROCESS AND TECHNOLOGY PLATFORMS TO PROVIDE MEANINGFUL ATTRIBUTE FOCUSED SPECIFICATIONS Pre-Pivotal Specification setting is based on a combination of prior knowledge of molecule attributes and process performance in combination with a risk based approach through the PQAA and QTPP to provide relevant ranges Evolution of technology platforms for attribute measurement enables collection of site specific attribute data to help inform relevant specifications MAM provides specific attribute measurement Allows control over levels of individual molecular CQAs OVERALL CONTROL STRATEGY IS BASED ON CRITICAL MOLECULE ATTRIBUTES AND OUR CLINICAL EXPERIENCE 21

22 IDENTIFYING QUALITY ATTRIBUTES WHERE PRIOR KNOWLEDGE APPLIED TO MONOCLONAL ANTIBODIES CAN BE USEFUL Methionine oxidation in Fc region High molecular weight species Heavy Chain C-terminal modification C-terminal lysine C-terminal proline amidation Glycan structure in N-linked Fc region Mannosylation Galactosylation Fucosylation Deamidation in Fc region Tri-sulfides Glycation in non-cdr regions 22

23 EXAMPLE QUALITY ATTRIBUTES WITH PRIOR KNOWLEDGE FOR MABS Attribute Immune Safety Non-Immune Safety Efficacy: PK Efficacy: Potency Prior Knowledge Sources Fc Methionine Oxidation 1 NA 3 NA Based on general understanding of attribute impact, not specific molecule information Gao et al., JOURNAL OF PHARMACEUTICAL SCIENCES 104: , 2015 Stracke et al. mabs 6, , 2015 Bi et al. JOURNAL OF PHARMACEUTICAL SCIENCES 102, , 2013 Aggregates Dimer 5 NA 5 7 Includes aggregates that form subvisible and visible aggregates Carpenter et al. JOURNAL OF PHARMACEUTICAL SCIENCES 98, , 2009 Singh et al. JOURNAL OF PHARMACEUTICAL SCIENCES 99, , 2010 For most Mabs immune safety is evaluated as part of clinical studies. For many Mabs, dimers have reduced potency Heavy Chain C-terminal Lysine Heavy Chain C-terminal Proline Amidation 1 NA NA 1 1 Based on general understanding of attribute impact, not specific molecule information. Cai et al., BIOTECH BIOENG 108, Antes et al. J Chromatography B , 2007 Van den Bremer et al. mabs 7, , 2015 Based on general understanding of attribute impact, not specific molecule information Johnson et al. Anal. Biochem. 360, (2007). Kaschak et al. mabs 3, (2011). Tsubaki et al. International Journal of Biological Macromolecules 52, (2013). 23

24 KEY EXPERIMENTS WERE PERFORMED TO ASSESS THE IMPACT OF MET OXIDATION ON SAFETY Stimulation index (above monomer) In Vitro Comparative Immunogenicity Assessment (IVCIA) Assay IL-10 IFN -γ Early Response mab2 Oxidized IL-1β M C P-1 M IP -1α Late Stage Response mab2 Oxidized IL-10 IL-13 IL-2 TNF-α IL % responding donors A D A (signal / noise) mab1 monomer mab1 oxidized Xeno-het Mouse mab2 monomer mab2 oxidized mab1 stir-20h 200µ g/ml TCE-KLH-mAb Conclusions: Safety: Met oxidation does not appear to increase immunogenicity risk as shown by the in vitro cell-based assays and the in vivo Xeno-het mouse model Clearance: Oxidation at the conserved Fc met 252 and 428 under reasonable conditions has negligible impact on FcRn binding and subsequent PK clearance (Stracke et al., mabs, :5, ) % ADA Positive Mice Immunogenicity CQA team THESE RESULTS STRONGLY SUGGEST THAT OXIDATION OF MET RESIDUES IN AMGEN ANTIBODY PRODUCTS DOES NOT POSE AN INCREASED RISK OF IMMUNOGENICITY OR IMPACT ON PK 24

25 PRODUCT-RELATED HMW SPECIES DID NOT INDUCE A SIGNIFICANT IMMUNE RESPONSE IN MODEL SYSTEMS Xeno-het Mice: ADA PBMC: Cytokine Secretion Cell Line: PRR Activation Cell Signaling Fold Increase ADA Signal Fold Increase Response Fold Increase (Post:Pre-Injection) % HMW 5% HMW 25% HMW Antigen Injections IL-1α IL-1β IL-1rα IL-6 IL-8 IL-10 MCP1 M IP-1α M IP -1β TNF-α TNF-β THP-1 XBlue Response RAW-Blue Response Ramos-Blue Response All HMW samples All HMW Samples All HMW Samples Stirred aggregates HMW Samples (IgG2) IL-1α IL-1β IL-1rα IL-6 IL-8 IL-10 MCP1 M IP-1α M IP -1β TNF-α TNF-β Stirred aggregates % HMW 25% HMW Stirred aggregates Positive Controls IL-1α IL-1β IL-1rα IL-6 IL-8 IL-10 MCP1 M IP-1α M IP -1β TNF-α TNF-β Antigenic Ligands [Inducer] (µ g/m l) max min y = min+ x 1+ EC 50 TCE-KLH-mAb IL-1α IL-1β IL-1rα IL-6 IL-8 IL-10 M C P-1 M IP-1α M IP -1β TNF-α TNF-β HillSlope Fold Response % Donor with Process-related IgG1 did not induce ADA in the mini-repertoire mouse model (Bessa et al, Pharm Res, (7); ) HIGH LEVELS OF HMW SPECIES, WELL ABOVE THAT OBSERVED IN PRODUCTS, DO NOT POSE AN INCREASED RISK OF IMMUNOGENICITY IN MODEL SYSTEMS 25

26 C-TERMINAL LYSINE, THE MOST COMMON C TERMINAL VARIANT, HAS MINIMAL IMPACT ON SAFETY AND EFFICACY Heavy Chain C-terminal: SLSLSPGK SLSLPG Minimal safety risk due to natural occurrence in humans and short exposure in circulation Naturally occurring, endogenous human antibodies are expressed with a C-terminal K Rapidly cleaved by endogenous serum carboxypeptidase B (CpB) in vivo after intravenous injection with half-life of about an hour Minimal impact on efficacy due to spatial distance from CDR, FcRn, and Fc gamma receptor binding regions; and short exposure in circulation The C-terminus of the HC is not within the known binding sites of CDR, FcRn or Fc gamma receptors Most C-terminal Lys is clipped shortly after patient administration C-TERMINAL LYSINE IS NATURALLY OCCURRING AND NOT NEAR ANY KNOWN BINDING SITES MAKING IT UNLIKELY TO IMPACT SAFETY OR EFFICACY 26

27 C-TERMINAL PROLINE AMIDATION : NO KNOWN IMPACT ON SAFETY OR EFFICACY Minimal safety risk due to natural occurrence in human After HC C-terminal Lysine is processed by carboxypeptidase to yield Glycine at the terminus, enzymatic activities of peptidylglycine-α-hydroxylating monooxygenase (PHM) and Peptidyl- α-hydroxyglycine α-midating lyase (PAL) can generate the proline amidated C-terminus. Heavy Chain C-terminal: SLSLSPGK SLSLPG: Levels in therapeutics Mabs: < 1% for Amgen products; % in 6 of 10 IgG1 Mabs; % in 2 IgG4 Mabs Minimal impact on efficacy due to spatial distance from CDR, FcRn, and Fc gamma receptor binding regions The C-terminus of the HC is not within the known binding sites of CDR, FcRn or Fc gamma receptors Preferential clearance has not been demonstrated Johnson et al. Anal. Biochem. 360, (2007). Tsubaki et al. International Journal of Biological Macromolecules 52, (2013), Kaschak et al. mabs 3, (2011). C-TERMINAL PROLINE AMIDATION IS NATURALLY OCCURRING AND NOT NEAR ANY KNOWN BINDING SITES MAKING IT UNLIKELY TO IMPACT SAFETY OR EFFICACY 27

28 USING PRIOR KNOWLEDGE TO ESTABLISH AN ATTRIBUTE FOCUSED SPECIFICATION Clinical exposure with highest level of impurity HMWS Qualified Range 0% 25% TI based off of Manufacturing Experience Prior Knowledge based on clinical exposure of the attribute from relevant products Prior Knowledge of Product attribute safety threshold from in vitro and animal data Attribute acceptance criteria = clinical exposure and manufacturing experience + Prior Knowledge (clinical and in vitro) 0% 10% 25% Adjusted Acceptance Criteria ACHIEVE KNOWLEDGE AND EXPERIENCE BASED SPECIFICATION 28

29 FUTURE PRODUCT TESTING PARADIGM Today: End of Process DS/DP Release 30+ assays overlapping DS & DP End point manual testing Complex and resource insensitive Instrument centric, non PQA specific Tomorrow: Real Time Release Multi-Attribute Method (MAM) Online sensor technology Automated online/at-line testing PQA specific CEX SEC Glycan rce-sds Peptide Mapping Aggregate UV BioAssay ID, Glycosylation, Oxidation, Deamidation, Isomerization, Hydroxylysine, Clips Titer Breakloose Extrusion Osmometer ID by Raman 29

30 ATTRIBUTE-BASED STRATEGY WILL ENABLE MOLECULE DIFFERENTIATION TO MEET PATIENT NEEDS Target Candidate Profile Molecule Assessment Differentiated Molecule/Modality By design Molecule Screening & Design Attribute Understanding (PQAA, QTPP) Apply Quality Target Product Profile to meet patient needs defined within Target Product Profile Deliver attribute understanding, methods to test and control them, and ensure supply for patients Advance new attribute technologies for specific, fast and multi-attribute methods 30

31 PRIOR KNOWLEDGE NEXT STEPS Prior knowledge is applicable from molecule design, molecule selection, non-clinical toxicology testing, first in human studies and through product lifecycle Prior knowledge is utilized to develop a risk-based approach for control strategy including specification setting Prior knowledge of product attributes and the development of the associated manufacturing process creates a basis for more flexible regulatory approaches It is important to understand the criticality of attributes and the impact of these attributes across different therapies PRIOR KNOWLEDGE HAS POTENTIAL FOR GREATER USE IN DETERMINING A CONTROL STRATEGY 31

32 FUTURE DIRECTIONS: INTEGRATION OF DS, DP, AND ATTRIBUTE TESTING (AT) SUPPORTED BY PRIOR KNOWLEDGE Current Paradigm Future Paradigm Vial Process Flow Drug Substance Process Drug Product Process Attribute Testing Process Flow AT AT Integrated Bioprocessing AT AT Device 32

33 LEVERAGING PRIOR KNOWLEDGE REQUIRES BALANCED INDUSTRY AND HEALTH AUTHORITY ENGAGEMENT Education of and by sponsors Some reactions: too slow New Approach Some reactions: too risky Education of and by regulators Industry Perspective Industry Intentions Use of Prior Knowledge Regulations, guidances, data, tools, philosophies, knowledge management, industry best practices Range of Potential Perceptions Agency Thinking Regulator Acceptance Ability to Effectively Inform, Communicate and Implement Prior Knowledge 33 Timely Review Acceptance and Implementation

34 ACKNOWLEDGEMENTS Darrin Cowley Tom Monica Marisa Joubert Andrew Lennard Rohini Deshpande Izydor Apostol Nina Cauchon 34

35 QUESTIONS? 35