continued process verification: an industry position · pdf filepage 4 – bpog continued...

CONTINUED PROCESS VERIFICATION: AN INDUSTRY POSITION PAPER WITH EXAMPLE PLAN

BPOG Continued Process Verification: An Industry Position Paper With Example Plan – Page 3 – BPOG CONTINUED PROCESS VERIFICATION: AN INDUSTRY POSITION PAPER WITH EXAMPLE PLAN

The following people were lead contributors to the content of this document, writing sections, editing and liaising with colleagues to ensure that the messages it contains are representative of current thinking across the biopharmaceutical industry. This document is a consensus view of a model CPV Plan, but it does not represent fully, the internal policies of the contributing companies.

Cynthia Ball (AstraZeneca), Joerg Gampfer (Baxter), John Grunkemeier (Bayer), Madeline Roche (Gallus), Lada Laenan (Genzyme), Dan Baker (GSK), Rajesh Beri (Lonza), Julia O’Neill (Merck), Abe Germansderfer (Novartis), Jeff Fleming (Pfizer), Jenny McNay (Regeneron).

Additionally, excellent editorial support and constructive criticism was provided by:

Ranjit Deschmukh (AstraZeneca), Mark Smith (Genentech/Roche), Beth Junker (Merck), Christelle Pradines (Novartis) Eric Hamann (Pfizer), Paul McCormac (Pfizer), Rajesh Ahuja (Regeneron), Bert Frohlich (Shire).

The work was facilitated by Darren Whitman and Robin Payne of the BioPhorum Operations Group (BPOG).

Though this paper is issued under copyright, © 2014, BPOG - Biophorum Operations Group, it is intended to be readily accessed across the industry, free of charge and can be accessed from the BioPhorum website at the following address:

www.biophorum.com/Page/123/BPOG-CPV-Case-Study.htm

When citing this paper, please use the following form:

BPOG, 2014, Continued Process Verification: An Industry Position Paper with Example Plan, © 2014, BPOG - Biophorum Operations Group

This paper is a response to US Food and Drug Administration (FDA) 2011 process validation guidance on Stage 3, ‘Process Validation: General Principles and Practices’[5]. It describes the approach commonly referred to as ‘Continued Process Verification’ (CPV). As one might expect, manufacturers in the biopharmaceutical sector all wish to respond to this guidance appropriately. A group of 20+ companies felt it would be valuable to work on this topic together, using the facilitation services of the BioPhorum Operations Group (BPOG) (www.biophorum.com). This paper is one of the results of the collaborative effort. It is written as a consensus view of an acceptable CPV program, but it does not fully represent the internal policies of the contributing companies. It is a basis upon which to build and share knowledge further across the industry. The authors believe this is one of the first comprehensive papers on this topic.

EXECUTIVE SUMMARYCPV PAPER LEADING CONTRIBUTORS

SUMMARY

The paper seeks to provide practical developments on the themes: what is CPV, why is it important, and how might it be implemented. It offers some specific recommendations on the content of a CPV Plan, along with associated rationale. These recommendations are based on a typical cell culture production process for making a fictitious monoclonal antibody product, described in the ‘A-Mab Case Study’ [3]. Consequently, not all of the details contained in this paper are going to apply directly to actual products or processes. The authors recognize that the A-Mab Case Study represents only one industry archetype, and that there are a number of others that are important. However, the concepts and principles upon which the content of this paper was derived should help with CPV implementation for a real product. Some of the complications of implementation are addressed, with recommended approaches, but the issue of information technology (IT) systems is not dealt with directly here. The case for IT systems, their design and introduction, is the subject of other collaborative efforts facilitated by BPOG and some of the results of that work may be published in the future.

CPV is fundamentally a formal means by which a commercial manufacturing process is monitored to ensure product quality.

It encompasses a written plan for monitoring a licensed biopharmaceutical manufacturing process, as well as regular reporting and actions based on the results of monitoring the process. CPV reporting provides a basis from which to improve process understanding, risk assessment, the control strategy (CS) [9], and ultimately the process itself. In general, the nature and extent of CPV is aligned with the outcomes of process qualification. Whilst a CPV Plan is likely to include data related to Batch Release (BR), and so may be useful in supporting BR decisions, CPV operates independently from the BR process and is not expected to have any impact on batches that have been previously released.

Adopting or building on an existing system of monitoring manufacturing process performance means more data will be collected over the lifetime of the product. CPV execution may involve examination of existing process control measurements and improved methods for data tracking and analysis. Enhanced monitoring of process performance provides the opportunity to identify and control sources of variation and hence improve process robustness, offering the major benefit of reliable supply to the market.

www.biophorum.com/Page/123/BPOG

-CPV-Case-Study.htm

www.biophorum.com

BPOG Continued Process Verification: An Industry Position Paper With Example Plan – Page 5Page 4 – BPOG Continued Process Verification: An Industry Position Paper With Example Plan

1.0 PURPOSE .............................................................................................................................................................................................................................7

2.0 SCOPE ................................................................................................................................................................................................................................9

3.0 ROLES AND RESPONSIBILITIES ................................................................................................................................................................................ 10

4.0 CPV PLAN REFERENCES ............................................................................................................................................................................................12

5.0 PRODUCT AND PROCESS DESCRIPTION .............................................................................................................................................................13

5.1 BRIEF DESCRIPTION OF THE GENERAL APPROACH USED IN THE A-MAB CASE STUDY ..................................................14

5.2 PARAMETERS TO BE INCLUDED IN CPV ...............................................................................................................................................15

5.3 UPSTREAM PROCESS OVERVIEW ..........................................................................................................................................................16

5.4 DOWNSTREAM PROCESS OVERVIEW ..................................................................................................................................................17

5.5 IDENTIFICATION OF CQAS AND ACCEPTANCE RANGES ................................................................................................................18

5.6 PROCESS PARAMETER CHARACTERIZATION ....................................................................................................................................20

5.7 CONTROL STRATEGY CQAS AND CPPS .................................................................................................................................................22

6.0 DEVELOPING A MONITORING STRATEGY ...........................................................................................................................................................23

6.1 RATIONALE AND BACKGROUND ...........................................................................................................................................................23

6.2 HYPOTHETICAL SCENARIOS AND PLANNED PROCESS CHANGES ............................................................................................24

7.0 CPV PLAN RECOMMENDATIONS FOR THE A-MAB PROCESS ....................................................................................................................28

7.1 STEP 1, SEED CULTURE EXPANSION IN DISPOSABLE VESSELS – CPV RECOMMENDATIONS .........................................29

7.2 STEP 2, SEED CULTURE EXPANSION IN BIOREACTORS – CPV RECOMMENDATIONS ........................................................30

7.3 STEP 3, PRODUCTION CULTURE BIOREACTOR – CPV RECOMMENDATIONS ........................................................................ 31

7.4 STEP 4, CLARIFICATION (CENTRIFUGATION AND DEPTH FILTRATION) – CPV RECOMMENDATIONS ...........................35

7.5 STEP 5, PROTEIN A CHROMATOGRAPHY – CPV RECOMMENDATIONS ...................................................................................36

7.6 STEP 6, LOW PH TREATMENT – CPV RECOMMENDATIONS .........................................................................................................37

7.7 STEP 7, CATION EXCHANGE CHROMATOGRAPHY (CEX) – CPV RECOMMENDATIONS .....................................................39

7.8 STEP 8, ANION EXCHANGE CHROMATOGRAPHY (AEX) – CPV RECOMMENDATIONS ......................................................40

7.9 STEP 9, SMALL VIRUS RETENTIVE FILTRATION (SVRF) – CPV RECOMMENDATIONS ..........................................................42

7.10 STEP 10, ULTRAFILTRATION AND DIAFILTRATION (UF/DF) – CPV RECOMMENDATIONS ..................................................43

7.11 STEP 11, FINAL FILTRATION AND FREEZING OF BDS – CPV RECOMMENDATIONS ............................................................45

7.12 BULK DRUG SUBSTANCE LOT DATA – CPV RECOMMENDATIONS ............................................................................................47

8.0 FREQUENCY AND SCOPE OF CPV ANALYSIS .....................................................................................................................................................49

8.1 SCOPE OF CPV ANALYSIS ........................................................................................................................................................................49

8.2 FREQUENCY OF ANALYSIS .........................................................................................................................................................................50

9.0 ESTABLISHING CONTROL LIMITS .......................................................................................................................................................................... 51

TABLE OF CONTENTSOne of the main technical issues to resolve when implementing CPV relates to the quantity of data required before product commercialization. In a sense, CPV complements the ‘Quality by Design’ (QbD) framework that manufacturers have developed to license and commercialize the product, though a CPV Plan may be constrained to data available in manufacturing. It should be noted that not all products will have a QbD framework but all need a CPV Plan. Also, at the time of commercial product introduction, there is unlikely to be a statistically robust set of data at the scale of commercial manufacture. To manage this situation in practice, it is recommended that short term control criteria are set initially, based on prior process experience and including data acquired at the laboratory and clinical scales of manufacture. This initial period of production would then be used to establish longer term criteria that are more statistically appropriate.

The implementation and ongoing execution of a CPV Plan is likely to require additional effort, beyond what is typically needed to prepare for the Annual Product Review (APR), because significant amounts of additional data are collected and analyzed to improve understanding of process variability. However, it is likely that the benefits accruing from the

improved information available for process improvement will outweigh any additional costs. The actual additional cost depends on the amount of data to be analysed which in turn depends on the outcomes of quality risk assessments that define data collection scope and frequency. The frequency of collection depends on several factors, including: whether production is campaigned or continuous; the extent of variability apparent in the process; whether risks to product quality (and thus product disposition) and process consistency are sufficiently mitigated, and the intended use of the reported data (for example, use in continuous improvement may mean collecting and analyzing certain data on a daily basis).

Given the importance of CPV in both compliance and process improvement terms, the authors encourage executives to read and share this paper with their colleagues. The authors also welcome any comments or questions arising which can be submitted via the following email address: [email protected].

mailto:[email protected]

This document is written with the aim of providing a technical, non-binding, industry consensus response to regulatory guidance. It is not in itself guidance. The objective of this paper is to provide:

(1) an example of key portions of a Continued Process Verification (CPV) plan for a biologics process; (2) relevant industry thinking on CPV plan development and implementation.

This document is different from others on this subject [1, 2] because it is specific to a biologics manufacturing process and provides a comprehensive case-study lifecycle view that leverages antibody manufacturing process development, as described in the A-Mab Quality-by-Design case study [3]. It is worth the reader being familiar with the A-Mab case study and perhaps having a copy available for reference. It should be recognised that the monoclonal antibody process is just one archetype in the industry, though it is a useful one upon which to demonstrate principles, as it is known to many.


10.0 EXAMPLE CPV EXECUTION PLAN FOR DRUG SUBSTANCE ..........................................................................................................................53

10.1 STEP 1, SEED CULTURE EXPANSION IN DISPOSABLE VESSELS – CPV VARIABLES ..............................................................57

10.2 S TEP 2, SEED CULTURE EXPANSION IN BIOREACTORS – CPV VARIABLES ...............................................................................58

10.3 STEP 3, PRODUCTION CULTURE BIOREACTOR – CPV VARIABLES .............................................................................................59

10.4 STEP 4, CENTRIFUGATION AND DEPTH FILTRATION – CPV VARIABLES ...................................................................................62

10.5 STEP 5, PROTEIN A CHROMATOGRAPHY – CPV VARIABLES ........................................................................................................63

10.6 STEP 6, LOW PH TREATMENT – CPV VARIABLES .............................................................................................................................64

10.7 STEP 7, CATION EXCHANGE CHROMATOGRAPHY – CPV VARIABLES ......................................................................................65

10.8 STEP 8, ANION EXCHANGE CHROMATOGRAPHY – CPV VARIABLES .......................................................................................66

10.9 STEP 9, SMALL VIRUS RETENTIVE FILTRATION – CPV VARIABLES .............................................................................................68

10.10 STEP 10, ULTRAFILTRATION AND DIAFILTRATION – CPV VARIABLES .......................................................................................68

10.11 STEP 11, FINAL FILTRATION/BULK FILL AND FREEZING OF BDS – CPV VARIABLES ............................................................. 70

10.12 CPV MONITORING OF BULK DRUG SUBSTANCE LOT DATA ........................................................................................................ 71

11.0 CPV SAMPLING PLAN ................................................................................................................................................................................................73

11.1 TEMPLATE FOR SPECIFIC PROCESS STEPS .......................................................................................................................................76

12.1 IDENTIFYING SOFTWARE ...........................................................................................................................................................................................80

12.2 DESCRIPTION OF TOOLS TO TREND AND EVALUATE DATA .......................................................................................................... 81

12.3 PROCESS CAPABILITY INDEX ....................................................................................................................................................................82

12.4 CONTROL CHARTS ........................................................................................................................................................................................84

12.5 MULTIVARIATE DATA ANALYSIS ..............................................................................................................................................................86

12.6 RESPONSES TO SHIFTS AND TRENDS ...................................................................................................................................................87

12.7 ESTABLISHING INITIAL LIMITS .................................................................................................................................................................88

12.8 ESTABLISHING LONG-TERM LIMITS ......................................................................................................................................................88

12.9 FINDING SIGNALS OF SPECIAL CAUSE VARIATION .........................................................................................................................89

13.0 CHANGE MANAGEMENT ..........................................................................................................................................................................................90

14.0 DATA VERIFICATION ....................................................................................................................................................................................................93

15.0 DISCRETIONARY ELEMENTS OF A CPV PROGRAM .........................................................................................................................................95

16.0 TECHNICAL REFERENCES ..........................................................................................................................................................................................96

17.0 GLOSSARY ........................................................................................................................................................................................................................97

1.0 PURPOSE

The example of a CPV plan shown in this paper describes how to meet expectations [5] for routine monitoring of critical process parameters (CPPs), critical quality attributes (CQAs), key process attributes (KPAs) and key process parameters (KPPs) to demonstrate the state of control over the manufacturing process. N.B. at the time of writing, the European Medicines Agency (EMA) draft guidance on Process Validation is out for consultation, referring to KPAs as 'performance indicators'. The thought processes and examples presented in this document are backed by biotech industry experience with, subject matter expertise

in process monitoring for monoclonal antibody and similar manufacturing processes.

Furthermore, this document describes the thought processes that determine the content for a CPV plan. The plan serves as the procedure governing document for the implementation and maintenance of CPV for a licensed manufacturing process. Various parts of the plan are described in the following sections of the document, as noted in Table 1.1 overleaf:

SECTION 1.0

GENERAL TOPIC SECTION SECTION NUMBER/ TITLE

DESCRIPTION

Manufacturing Process 5 Summary of the A-Mab manufacturing process and the A-Mab product description.

6 Selection of the process monitoring sampling plan backed by the process validation Phase I and II data and the updated risk assessment.

7 The rationale for classification of quality-linked process parameters summarized in the A-Mab case study is reviewed and summarized in the table that presents process performance consistency and robustness. Rationale for what to include in CPV is provided, based on a review and analysis of quality-linked process parameters from the A-Mab case study that affect process performance, consistency and robustness.

BPOG Continued Process Verification: An Industry Position Paper With Example Plan – Page 9PAGE 8 – BPOG CONTINUED PROCESS VERIFICATION: AN INDUSTRY POSITION PAPER WITH EXAMPLE PLAN

Table 1.1. Plan parts referenced by Section Number:

Consistent with the FDA’s 2011 Process Validation guidance document [5] describing three stages of the product lifecycle, CPV implementation discussed in this paper is limited to Stage 3, commercial manufacture of a drug substance, following process design (Stage 1) and process qualification and qualification of the equipment and the facility (PQ, Stage 2, see FDA 2011 Guidance Stage 2 [4]) [12].

Note: Whilst this paper focuses on the drug substance manufacturing process, CPV should be applied all areas of Operations including formulation, fill and finish.

2.0 SCOPE

The application of the principles discussed in this document for new products relies on product and process development and characterization studies (Stage 1) to define the scope of the CPV program. This document is based on the CS presented in the A-Mab bioprocess development case study and is primarily focused on the commercialization of a new product. However, the proposed approach for CPV implementation is also applicable to legacy products where quality attributes and parameters for monitoring can be determined based on a combination of process knowledge and historical performance data.

The BPOG is initiating collaborative work, specifically focused on CPV for established, licensed (or legacy) products and the resulting recommendations may be published in the future. An ISPE group produced an article covering this broadened scope in 2012 [24]; here we believe we address a reduced scope in greater detail, providing deeper development of a model CPV Plan.

SECTION 2.0

GENERAL TOPIC SECTION SECTION NUMBER/ TITLE

DESCRIPTION

Verification process 8 The frequency of CPV data analysis and trend review is discussed. The concept of an initial or short-term CPV phase is introduced, where sufficient process experience is collected to establish the manufacturing control limits for the process attributes identified during validation. A subsequent phase of CPV implementation; that of steady state or long-term process monitoring is also discussed.

9 Statistical and general methods for establishing CPV trend limits are presented.

10 The summary of the monitored attributes and parameters within the scope of the CPV program are presented. The monitoring method and periodicity associated with specific attributes and parameters are also specified.

11 The sampling plan derivation with tabulated examples.

12 Aspects of data analysis and evaluation of results are discussed in this section. The emphasis is on the possible outcomes of routine monitoring.

13 Change management and the impact of CPV on this process.

14 The specific need for data verification.

15 Elements of CPV that are considered discretionary.


SECTION 3.0

3.0 ROLES AND RESPONSIBILITIESThe roles and responsibilities suggested here as examples, are based upon a typical organizational structure of a biopharmaceutical manufacturing company (Table 3.1.).

Table 3.1. Roles and Responsibilities for a CPV Program:

FUNCTIONAL AREA RESPONSIBILITY

Management • Ensure that adequate resources are available to carry out the CPV program and to regularly perform a review of CPV plan summaries or reports.

Development • Provide documentation that defines current process knowledge, quality attributes, process parameters and elements of the overall CS that forms the basis for the CPV program.

• Provide documented scientific justification for parameters, limits, ranges and elements of the CS, based upon development studies or other prior knowledge.

• Provide technical input to develop response actions, including input in prioritization of continuous improvement activities.

• Consider application of CPV outcomes to new processes in development.

Validation/ Quality functions • Provide internal advice on current validation principles and ensure that validation protocols, interim and final reports meet applicable standards.

• Participate in cross-functional teams to review production and QC data as part of the CPV program.

• Review the data, pursue appropriate investigations and make decisions on how to proceed.• May generate CPV plans and summary reports.• Review and approve CPV plan, CPV reports and any changes to the CPV plan.

FUNCTIONAL AREA RESPONSIBILITY

Operations / Manufacturing Science and Technology

(N.B. It is not unusual for a Manufacturing Science and Technology function to be independent of Operations and Quality organisations. An alternative arrangement may be reporting into Process Sciences.)

• Own the manufacturing process and take responsibility to ensure that it is maintained in a state of control throughout the product lifecycle in manufacturing.

• Ensure that all required production and process data are collected as part of executing the CPV plan for the product.

• Performs continued process monitoring activities, including collecting, entering, verifying, reviewing and analyzing process data.

• Generate control charts and document CPV analysis for process data. • Regularly participate in cross-functional teams in order to review production and QC data

as part of the CPV program.• Maintain the process commercial master batch production and control records up to date,

capturing continuous improvements resulting from CPV in documentation as necessary.

Quality Control • Perform quality control testing and document results that are used in CPV evaluations.• Perform continued process monitoring activities, including collecting, entering, verifying,

reviewing and analyzing QC data.• Generate control charts and document CPV analysis for QC data. • Participate in cross-functional teams to review production and QC data as part of the CPV

program.

Quality Engineering / Mathematical Sciences / Non-Clinical Statistics

• Provide internal advice on statistical analyses needed to successfully complete CPV activities.• Act as a Subject Matter Expert (SME) and train personnel in other groups on statistical data

analysis techniques used in CPV.• Provide internal advice on how to develop the data collection plan and help select suitable

statistical methods and procedures that are used to measure and evaluate the process stability and capability.

• Generate procedures that define the way statistical tools and approaches are to be used in routine process monitoring.

• Provide guidance on how to set control limits and define and interpret signal criteria.

Quality Assurance • Review and approve CPV plans and reports.• Review and approve the list of attributes and parameters to be monitored, and control chart

limits.• Participate in cross-functional data review to review production and QC data as part of the

CPV program.• Review CPV reports and establish where signals require formal non-conformance

investigations.• Coordinate assembly of CPV program reports.

Several primary functional areas have important responsibilities required to successfully execute the CPV program. These areas are: Development, Validation, Operations, Quality Control, Quality Engineering and Quality Assurance. Operations, a function which may also be known as Technical Operations, is assumed to contain Manufacturing as well as Manufacturing Science and Technology personnel. Mathematical sciences or non-clinical statistics support is of paramount importance in achieving sound data interpretation. Each functional area has responsibility for specific activities, as shown in Table 3.1.

Outputs of the CPV program can be used by the Regulatory Affairs and Quality organizations for annual agency updates, such as the Annual Product Review (APR) and Product Quality Review (PQR). Terminology for each function may vary by organization.

Note: The responsibilities for continued process monitoring should be clearly defined within the organization and recorded in the CPV Plan. Responsibilities can be tailored to a specific organizational structure, given its maturity and size.


SECTION 4.0

4.0 CPV PLAN REFERENCES

5.0 PRODUCT AND PROCESS DESCRIPTION

The following references are expected to be created in the quality management system and are important when constructing a CPV plan, providing background and critical internal interpretation of regulatory guidance. They should be referenced accurately in a CPV Plan document. Note, a CPV Plan is expected to be product and process specific. It may be advantageous to develop corporate policies and this forms the basis for some of the list of references that follows.

• Quality Policy, Manual or Master Plan on CPV• Company Standard/Guideline for CPV (requirements for

CPV, for e.g. timing, relationship to APRs, etc)• SOP on CPV (Definitions, Abbreviations, responses to

deviations, report generation, etc)• SOP on Statistical Methods for trending, statistical

analysis and identifying special cause variations• Template for CPV Plan• Template for CPV Charts & Graphs• Template for CPV Report • Manufacturing process description

• Control Strategy for the process (version number)• Process risk assessment (version number)• Applicable Risk assessment(s) (version number)

providing basis for rationale of CPV monitoring selection • Previous annual product report(s) if available, otherwise

consider evidence for a similar product*.

Technical references relevant to the detailed sections of this paper are provided in section 16. References 1 to 9 are recommended as initial texts when creating or updating a CPV plan.

In preparing this CPV example plan, it is assumed that Stage 2, was completed successfully for the A-Mab process. The plan described applies to Stage 3 of the process validation lifecycle.

Note: Whilst a QbD approach could be said to provide advantages in terms of process understanding, it is not an approach that has to be applied. However, it is necessary to have a CPV Plan for each product, even if a QbD approach has not been applied.

The A-Mab case study describes a model Quality by Design (QbD) approach for development of a monoclonal antibody (A-Mab) [3, 6]. Considering the FDA process validation guideline [5], the case study includes work covered during Stage 1 (Process Design) but does not include information on Stage 2, Process Performance Qualification (PPQ) [5].

SECTION 5.0

* The authors recognize that the plan illustrated in this paper is written largely with CPV for new products in mind and that there would not be APRs available at the point of product licensure. This bullet point is included as a reminder that historic APRs would provide data for the creation of a CPV plan where established, licensed or legacy products are concerned.


PPQ CPV CPV

TTP QTTP CQA CCP

EQ

ProvenAcceptable

Ranges(Design Space)

Covered in A-Map Study

Short-termPlan

Long-termPlanDevelopment of Control Strategy

PV Stage 2 PV Stage 3PV Stage 1

Figure 5.1.2. Process flow of a QbD based product development according to ICH Q8, 9, 10, 11 and FDA PV guideline January 2011.

5.2

PARAMETERS TO BE INCLUDED IN CPVAll types of parameters should be considered for inclusion in CPV. Typically those included will be weighted more in favor of CPPs and WC-CPPs because of their importance to the control strategy, but non-critical “key” and general parameters should not be overlooked if they are indicative of process performance and/or measurably impact process variation. Parameters to be included should be based on the current understanding of the manufacturing process and may be subject to change over time.

Parameter types described in A-Mab study are as follows:(1) Critical Process Parameter (CPP) and (2) Well-Controlled Critical Process Parameter (WC-CPP): CPPs and WC-CPPs are process parameters whose variability impact a critical quality attribute and should be monitored or controlled to ensure the process achieves the required product quality.

• A WC-CPP has a lower risk of falling outside the specified limits.

• A CPP has a higher risk of falling outside the specified limits.

The assessment of risk is based on a combination of factors that include severity of impact to quality, equipment design considerations, process control capability and complexity, the size and reliability of the proven acceptable range and/or design space, ability to detect/measure a parameter deviation, etc.

(3) Key Process Parameter (KPP): An adjustable parameter (variable) of the process that ensures operational reliability when maintained within a narrow range. A key process parameter does not affect critical product quality attributes but rather impacts process consistency.

(4) General Process Parameter (GPP): An adjustable parameter (variable) of the process that does not have a meaningful impact on product quality or process performance.

Typically the parameters included in CPV will be weighted more in favor of CPP and WC-CPP because of their importance to the control strategy. But, non-critical “key” and general parameters should not be overlooked as they may be indicative of process performance and/or measurably impact process variation. Definitions of A-Mab terms used to define categories of process parameter are provided in a Glossary at the end of this document.

Note: Throughout this paper the A-Mab classification of process parameters is used for consistency with the structure of that case study, but it must be recognised this is not the only scheme used in the industry; a situation arising in part no standard approach is recommended by the regulators. Consistency with ICH Q8 and Q11, where definitions exist seems prudent. A recent informal communication by FDA/EMA counseled against using “key parameter” for describing lower levels of criticality in formal submissions and stated that: ‘all parameters potentially impacting product quality should be classified as critical process parameters’ [23]. The use of KPPs in internal systems and documentation seems not to contravene this statement.

In general, it is the responsibility of the biopharm company to establish a categorization and nomenclature fitting with their development approach and risk evaluation tools. The company’s approach should be clearly explained and followed over the life cycle of the product.

5.1

BRIEF DESCRIPTION OF THE GENERAL APPROACH USED IN THE A-MAB CASE STUDY Principles outlined in the ICH guidelines Q8, Q9, Q10 and Q11 [7-9, 22] provide the basis for the methodology used for this case study, even though Q11 was published after the A-Mab case study.

One principle of a QbD approach is to develop a Target Product Profile (TPP). As a natural extension of a TPP, a Quality Target Product Profile (QTPP) is built to describe quality characteristics (attributes) of the drug product.

The process of systematic development follows a general roadmap that includes the following steps:

• Identification of Quality Attributes (QA) based on a QTPP;

• Risk Evaluation to identify CQAs;• Upstream/ downstream/ drug substance and product

development;• Risk based approaches and potentially, multivariate

analyses [25] (see Section 12.5 for a description of multivariate analysis), to classify process parameters and other variables linked to product quality (e.g. identification of Critical Process Parameters, CPPs);

• Univariate and multivariate approaches to define Proven Acceptable Range (PARs) or limits;

• Rational approach to define a CS that reflects product/ process knowledge and risk mitigation;

• Process (and Equipment) Performance Qualification to verify the CS established in Stage 1 of development.

• Facility design qualification of Stage 2 [5].

In creating this CPV plan it is assumed that all deliverables up to establishment of a CS and PQ are available based on the work described in the A-Mab study (see Figure 5.1.2 /green boxes). For the A-Mab process, it is assumed that PPQ was completed successfully, after investigating and resolving deviations.

PPQ and Equipment Qualification (EQ) are part of Stage 2 and are therefore presumed to have been completed before Stage 3 where CPV guidance applies. They are a pre-requisite for Stage 3 CPV. See guidance for Industry [5].


ThawWorking Cell Bank

Seed Culture Expansionin disposable shake flasks

and/or bags

Seed Culture Expansionin fixed stirred tank reactors

N-1 Seed Culture Bioreactor3000L WV

Production Bioreactor15,000 L WV

HarvestCentrifugation & Depth Filtration

Clarified bulk

Step 4

Step 3

Step 2

Step 1

Nutrient feed

Glucose feeds

Seedmaintenance

Seedmaintenance

Seed cultures are expanded through multiple passagesby increasing the volume and/or number of disposableculture vessels. Seed cultures may be maintained foradditional culture passages or used to generateadditional inoculums trains.

Additional seed expansion in fixed stirred tankbioreactors. Cultures obtained from Step 1 areexpanded to increase the volume of culture to meet thetarget initial cell density for the production bioreactor.

Production bioreactor is inoculated with the seedculture prepared in Step 2 to achieve an initial ViableCell Concentration (VCC) and is cultivated atcontrolled conditions for temperature, pH anddissolved oxygen (DO). A bolus addition of nutrientfeed (NF-1) and multiple discrete glucose feeds areused to maintain the glucose concentration at > 1.0g/L. Antifoam solution is used for foam control ofthe agitated mixture. VCC, culture viability andresidual glucose concentration are monitoredperiodically. The fermentation reaction produces amixture containing the A-Mab drug substance.

Cultures are clarified by a primary continuouscentrifugation step using a disk-stack centrifuge toremove the bulk of suspended cells and cell debris.A secondary clarification using a depth filtrationsystem removes remnant solids and smaller debris toprovide a clarified bulk solution of A-Mab.

Step 5Protein A Affinity Chromatography

Step 6Low pH Incubation

Step 7Cation Exchange Chromatography

Step 8Anion Exchange Chromatography

Step 9Small Virus Retentive Filtration

Step 11Final Filtration, Fill and Freeze

Clarified Purpose

Purpose of step

• Capture monoclonal antibody from clarified harvest liquid• Removal of process-related impurities (HCP, DNA and small molecules)

Step 10Formulation:

Ultrafiltration and Diafiltration A

A-Mab drug substance

• Inactivate enveloped viruses that are potentially present in therapeutic protein products derived from mammalian cell culture

• Reduce aggregate to acceptable levels for drug substance• Reduce HCP to acceptable levels for subsequent processing by AEX chromatography

• Remove HCP, DNA, Protein A and endotoxins to levels that meet drug substance acceptance criteria• Virus removal

• Removal of small parvoviruses such as minute virus of mice (MVM) and larger viruses such as murine leukemia virus (MuLV), potentially present in product derived from mammalian cell culture

• Formulation and concentration of mAb to drug substance specifications (e.g. 75 g A-Mab/L)

• Sterilize filtration and dispensing for drug substance storage

5.3

UPSTREAM PROCESS OVERVIEW 5.4

DOWNSTREAM PROCESS OVERVIEW The upstream commercial manufacturing process for A-Mab comprises 4 steps and is summarized below and in Figure 5.4.

The A-Mab cell culture process uses a proprietary, chemically defined, basal medium formulation. The medium is essentially protein free with recombinant human insulin (1 mg/mL) being the only protein component added. The growth medium also contains 1 g/L pluronic and 50 nM methotrexate, which are added up to the N-2 seed bioreactor. The N-1 and production bioreactor steps do not contain methotrexate.

Figure 5.4. Upstream process flow diagram. [Adapted from A-Mab case study, Page 62 (Figure 3.1)]

The downstream manufacturing process for A-Mab comprises 7 steps which are summarized in Figure 5.5.

The downstream process captures A-Mab from the clarified bulk and purifies the antibody by a combination of chromatography unit operations [11]. Also included in the process are two orthogonal steps dedicated to virus inactivation and removal. The antibody is formulated through an Ultra-Filtration/Dia-Filtration (UF/DF) step to a composition and concentration suitable for drug product manufacturing. The formulated product is 0.2 μm filtered, filled into the appropriate storage containers and stored frozen.

Figure 5.5. Downstream process flow diagram. [Adapted from A-Mab case study, Pages 113 (Figure 4.1) and 114 (Table 4.1)]

“PRO

DUCT

QU

ALIT

Y AT

TRIB

UTE

S”

RAW MATERIAL CONTROLS

STEPS 1 & 2: SEED CULTURE EXPANSION

STEP3: PRODUCTION BIOREACTOR

STEP 4: CENTRIFUGATION AND CLARIFICATION

STEP 5: PROTEIN A CHROMATOGRAPHY

STEP 6: LOW PH TREATMENT

STEP 7: CEX CHROMATOGRAPHY

STEP 8: AEX CHROMATOGRAPHY

STEP 9: NANO-FILTRATION (SVRF)

STEP 10: ULTRA-FILTRATION (UF/DF)

STEP 11: FINAL FILTRATION AND FREEZING

BDS OR DP TESTING FOR THIS CQA?

IDEN

TITY

Form

Form

BDS,

DP

PRO

TEIN

CO

NCE

NTR

ATIO

NFo

rmAl

ter

Alte

rAl

ter

Alte

rAl

ter

DP

IPC

ADCC

ACT

IVIT

YFo

rmD

P

SEC

MO

NO

MER

Form

BDS,

DP

SEC

AGG

REG

ATES

Form

Rem

ove

Form

Rem

ove

Rem

ove

Form

Form

Form

BDS,

DP

COLO

RIn

trod

uce

Alte

rAl

ter

DP

CLAR

ITY

& S

UB-

VISI

BLE

PART

ICLE

SIn

trod

uce

Alte

rRe

mov

eRe

mov

eD

P

DEAM

IDAT

ED IS

OFO

RMS

Form

Rem

ove

Rem

ove

Rem

ove

BDS,

DP

OTH

ER A

CIDI

C VA

RIAN

TSFo

rmRe

mov

eRe

mov

eRe

mov

eBD

S, D

P

CHAR

GE

VARI

ANTS

Form

Rem

ove

Rem

ove

Rem

ove

BDS,

DP

“OLI

GO

SACC

HAR

IDES

: AF

UCO

SYLA

TED

GLY

CAN

SG

ALAC

TOSY

LATE

D G

LYCA

NS”

Form

“GLY

COSY

LATI

ON

REL

ATED

:SI

ALIC

ACI

D CO

NTE

NT,

MAN

NO

SE C

ON

TEN

T,N

ON

-GLY

COSY

LATE

D H

EAVY

CH

AIN

”

Form

OSM

OLA

LITY

Alte

rD

P IP

C

PHAl

ter

Alte

rAl

ter

Alte

rAl

ter

Alte

rD

P IP

C

MET

HO

TREX

ATE

Intr

oduc

eIn

trod

uce

Rem

ove

Rem

ove

Rem

ove

non-

rout

ine

ANTI

FOAM

CIn

trod

uce

Intr

oduc

eRe

mov

eRe

mov

eno

n-ro

utin

e

PRO

TEIN

A L

IGAN

DIn

trod

uce

Intr

oduc

eRe

mov

eRe

mov

eRe

mov

e

HO

ST C

ELL

PRO

TEIN

(H

CP)

Form

Form

Rem

ove

Rem

ove

Rem

ove

Rem

ove

non-

rout

ine

DNA

Form

Form

Rem

ove

Rem

ove

BIO

BURD

ENIn

trod

uce

Intr

oduc

eIn

trod

uce

Intr

oduc

eIn

trod

uce

Intr

oduc

eIn

trod

uce

Intr

oduc

eIn

trod

uce

Rem

ove

DP

ENDO

TOXI

NIn

trod

uce

Intr

oduc

eIn

trod

uce

Intr

oduc

eIn

trod

uce

Intr

oduc

eIn

trod

uce

Intr

oduc

eBD

S, D

P

“ADV

ENTI

TIO

US

VIRA

L AG

ENTS

(AV

A)”

Intr

oduc

eIn

trod

uce

Intr

oduc

eIn

activ

atio

nRe

mov

eRe

mov

est

ep 3

IPC,

BDS

rele

ase

impa

cted

by

CPP=

im

pact

ed b

y W

C-CP

P=

impa

cted

by

KPP=

n

o ke

y im

pact

cla

imed

=

ent

ry te

st o

r pre

p co

ntro

l=


5.5

IDENTIFICATION OF CQAS AND ACCEPTANCE RANGES

The product quality attributes and the points where they are impacted in the A-Mab drug substance process are summarized in the Table 5.6.2 below.

Table 5.6.2. A-Mab drug substance Product Quality Attributes and the points where they are impacted in the process (see A-Mab Case Study (3), Section 2.3.2, Page 29). BDS is Bulk Drug Substance, DP Drug Product and IPC in-process control.

Table 5.6.1 provides the QTPP of the A-Mab drug product, as defined in the A-Mab case study. The QTPP describes quality characteristics (attributes) that the drug product should possess in order to reproducibly deliver the therapeutic benefit promised in the label. Attributes in the red box are determined during Drug Substance (DS) manufacturing. Therefore, these attributes guide determination of DS CQAs [22] relevant for establishing a CPV strategy.

Table 5.6.1. QTPP for A-Mab (reference 3, Page 180). DS relevant product attributes are marked with a red box.

The DS QAs related to the QTPP are identified as summarized in Table 5.6.2. Criticality Analysis was performed using a risk ranking approach (as in ICH Q9 [8]) and CQAs were identified as attributes of high or very high risk regarding their potential impact on patient safety.

PRODUCT ATTRIBUTE TARGET

Dosage Form Liquid, single use

Protein content per vial 500mg

Dose 10mg/kg

Concentration 25mg/mL

Mode of administration IV, diluted with isotonic saline or dextrose

Viscosity Acceptable for manufacturing, storage and delivery without the use of special devices (for example, less than 10 cP at room temperature

Container 20R type 1 borosilicate glass vials, fluro-resin laminated stopper

Shelf life ≥ 2 years at 2–8°C

Compatibility with manufacturing process

Minimum 14 days at 25°C and subsequent 2 years at 2–8°C, soluble at higher concentration during UF/DF

Biocompatibility Acceptable toleration on infusion

Degradants and impurites Below safety threshold, or qualified

Pharmacopeial compliance Meets pharmacopoeial requirements for parental dosage forms, colorless to slightly yellow, practically free of visible particles and meets USP criteria for sub-visiable particles

Aggregate 0–5%

Fucose conent 2–13%

Galactosylation (%G1+%G2) 10–40%

HCP 0-100 ng/mg


5.6

PROCESS PARAMETER CHARACTERIZATION In reviewing the A-Mab process information while preparing the CPV plan, members of the BPOG team questioned the completeness of the CPPs, KPPs and Key Process Attributes (KPAs) identified in the case study. Specifically, it was felt two steps of the downstream process (step 10 UF/DF, and step 11 final filtration and freezing of the Bulk Drug Substance, BDS) were not addressed in sufficient detail in the case study for the purpose of developing a CPV Plan, so typical characterization outcomes for these steps were assumed and CPPs, KPPs and KPAs were identified based on that characterization [10]. In addition, two more KPPs and KPAs were identified for process steps 3 and 7, based on typical outcomes for similar monoclonal antibody processes. The following table summarizes all CPPs, in-process quality attributes (IPQAs), KPPs and KPAs identified for the process in preparation for CPV.

Table 5.7. Critical and key process parameters and key process attributes identified during process characterization. Lists were amended during planning for CPV (bold entries)

PROCESS STEP CRITICAL PROCESS PARAMETERS

IN-PROCESS CONTROLS KEY PROCESS PARAMETERS

KEY PROCESS ATTRIBUTES

Step 1: Seed Culture Expansion in disposable shake flasks and/ or bags

None None Temperature,Culture duration,Initial VCC/ split ratio

VCC (viable cell conc), Culture viability

Step 2: Seed Culture Expansion in fixed stirred tank reactors

None None Temperature,pH, Dissolved oxygen, Culture duration,Initial VCC/ split ratio

VCC, Culture viability

Step 3: Production Bioreactor 15,000L WV

Temperature,pH,Max partial pressure of CO2 (pCO2),Culture duration,Medium Osmolality

Bioburden,Mycoplasma,MMV and AVA(murine minute virus and adventitious viral agents)

Antifoam conc.,Time of nutrient feed, Volume of nutrient feed, Time of glucose feed, Volume of glucose feed,Dissolved oxygen

Product yield (titer),Viability at harvest,Turbidity at harvest,Peak VCC, Remnant glucose concentration

Step 4: Harvest Centrifugation & Depth Filtration

None None Flow rate, Pressure,Duration of clarification

Step yield,Turbidity

PROCESS STEP CRITICAL PROCESS PARAMETERS

IN-PROCESS CONTROLS KEY PROCESS PARAMETERS

KEY PROCESS ATTRIBUTES

Step 5: Protein A Affinity Chromatography

Protein load ratio, Elution buffer pH

Bioburden, Endotoxin

End collection,Step duration

Step yield

Step 6: Low pH Incubation

pH,Time,Temperature


Quantity of acid added

Step 7: Cation Exchange Chromatography

Protein load ratio,Wash conductivity,Elution pH,Elution stop collect


Step duration Step yield,Eluate volume

Step 8: Anion Exchange Chromatography

Equilibration/ Wash1 buffer conductivity, Protein load ratio,Load conductivity,Load pH,Flow rate


Step duration Step yield

Step 9: Small Virus Retentive Filtration

Operating pressure,Filtration volume


None Step yield

Step 10: Formulation: Ultrafiltration and Diafiltraion

Number of dia-volumes,pH,Step processing time, Protein conc. prior to fill


Protein conc. prior to Diafiltration,Recirculation flow rate

Step yield,Permeate flow rate

Step 11: Final Filtration, Fill and Freeze

None Bioburden, Endotoxin

Filtration volume,Filtration time,Maximum (inlet) pressure

Bulk fill step yield


5.7

CONTROL STRATEGY CQAS AND CPPS

6.1

RATIONALE AND BACKGROUND

Risk-based criticality assessment, along with process characterization studies, allows a CS to be established which is subsequently verified during PPQ. Table 5.7 summarizes the CS established for the A-Mab upstream and downstream process steps for A-Mab production. The CS consists of CPPs and WC-CPPs, KPPs, KPAs and IPQAs. The CS should ensure required product quality and a consistent and robust process.

Here, CPPs must be controlled within limits and in-process controls (specifically microbial and viral safety) must be within specified ranges to ensure drug safety and efficacy. Although KPPs and KPAs have been shown not to impact

product quality, they are included in the CS because their monitoring and control ensures that the process is operated in a consistent and predictable manner. The control of KPPs and KPAs also ensures that commercial success criteria such as cycle time and yield are met.

Product quality and safety are ensured by controlling all quality-linked process parameters (CPPs and WC-CPPs) within the limits. Process consistency is ensured by controlling KPPs within established limits and by monitoring relevant process attributes.

In general, the points in the process to be monitored during CPV should be comparable to, but not necessarily include all of those selected during the initial validation. If limited data results are available at the time of PPQ completion, prior to execution of the CPV plan, a short term sampling plan may be established to continue sampling based on the PPQ protocol until sufficient data results are gathered in preparation for CPV. Additional considerations that influence the determination of which points in the process are monitored during a CPV exercise are summarized below.

(1) The final classification of attributes should be revisited.

(2) The process risk assessment, which is typically performed prior to the initial PPQ, should be revisited and updated to develop the CPV plan. The revised risk assessment should reflect learning obtained during PPQ, any additional laboratory process characterization information, and key findings from historical manufacturing experience. In revisiting the process risk assessment prior to commercial manufacture, late stage clinical manufacturing knowledge is particularly important. Levels of risk, and indeed the range of risks, that apply in the manufacturing environment might

be quite different to those anticipated from the early stage development environment.

(3) The control strategy should be updated as necessary and hence the CPV Plan.

The selection of points in the manufacturing process that are to be monitored for CPV purposes may be either a subset of those selected during PPQ or include additional monitoring points beyond those included in the initial PPQ to reflect new learning obtained since the initial validation was conducted. This includes but is not limited to:• New CS elements• Process elements that have proved challenging but

may not have been covered during the initial process validation

• Changed or additional analytical capabilities, including the availability of online data collection systems and improvements in assay or instrument capabilities

• If a parameter has been shown to have good control and consistency, it may not be necessary to continue monitoring this parameter in subsequent CPV evaluations.

CPV is a formal activity enabling the detection of variation in the manufacturing process that might have an impact on the product quality or process consistency. CPV should evaluate whether the process consistently delivers product with acceptable QAs and continues to operate robustly, within the validated state. It should also identify any new sources of variability in the process that may have arisen since the initial Stage 2 PQ was performed. For this case study it has been determined that PPQ batches will be included in CPV data collection and analysis; indeed, all appropriate batches should be considered. CPV efforts should, where appropriate, also focus on areas that have proved challenging or may have shifted since the initial validation. A risk based approach to process monitoring should be used to direct these efforts. For products with a legacy history, a defined time period or number of batches should be set to determine how much of the historical experience will be considered. The assessment interval chosen should be sufficient to establish a solid production history and also reflect the frequency of production. For example a product that is produced frequently may permit a shorter time period to be used relative to a product that is produced infrequently.

6.0 DEVELOPING A MONITORING STRATEGY

SECTION 6.0


(4) CPPs, WC-CPPs, KPPs and GPPs should be clearly specified. These parameters and established release specifications, additional product characterization testing, and KPAs should be appropriately considered during CPV. Any changes since the initial validation should be explained and justified.

(5) All changes implemented should be assessed in the context of potential impact on process validation. Process changes which may have occurred after the PPQ, such as vendor initiated change in a raw material, should be handled a change control process including but not limited to data trending and risk assessments, to determine if the change has any impact on process performance and/ or product quality. These changes may potentially require additional testing beyond that performed as part of PPQ to ensure full characterization. Such testing may be incorporated as part of CPV or may be handled separately as part of the company’s change control process, depending on the nature of the change and the potential for product impact.

(6) Appropriate regulatory reporting of CPV outcomes, such as inclusion in the Annual Product Review (APR), must be made for any conclusions related to process assessment conducted during CPV. The CPV reports should be consistent with regulatory reporting standards, so that CPV charts may be copied and pasted directly into the regulatory submissions or included as an attachment. The regulatory submissions then provide context and unify the information presented in the attached CPV reports.

(7) Other elements of Good Manufacturing Practice (GMP) applicable to biopharmaceutical production operations are assumed to be handled by appropriate quality systems and are therefore outside the scope of this document, and will not be discussed further in the context of process validation. In particular, acceptable microbial control is a critical element for any biopharmaceutical process and is typically demonstrated via initial validation efforts and then monitored as part of routine operations.

Scenario 1: Supplier change notification - culture medium change.

A supplier converted to a new process to manufacture a cell culture medium ingredient that may alter its performance in the A-Mab process without impacting the material procurement specifications. No intentional changes to composition, test requirements or certificate of analysis were made. The following justification for the change was provided:

(1) Improved control of temperature during blending reduces potential for degradation of the heat labile components; (2) Equipment cleaning will use robust validated cycles to reduce ingredient carryover risks;(3) Equipment is located in an Animal Origin Free area to reduce cross contamination risks.

The following strategy was employed to introduce the revised cell culture medium:

• Determining process and quality impact for the material change through the change control process was electively agreed to by process experts and quality representatives via verification testing of culture performance and the ability to operate within the established parameters and attributes;

• A study was thus completed in the small-scale model from thaw through the production bioreactor to provide additional process characterization data and establish confidence in expectations of process control when the new material lot is introduced into the commercial scale process;

• Minor but statistically significant differences for KPPs normal operating ranges and attributes (e.g. VCC, and cell density, titer and turbidity at the end of the bioreactor production) were identified at small scale;

• Medium qualification attributes should be assessed in the change control evaluation to determine if/ how these attributes may be impacted. The supplier was requested to demonstrate if a detectable mean shift in any of their output tests could be identified with respect to their change.

Small scale production bioreactor material was purified downstream. No structural modifications to the protein, or shifts in CQAs were observed. Based on the outcome of the small scale studies, a comparison should be made to evaluate the product quality obtained at full scale, to verify that no unexpected quality change has occurred and to provide further verification of process control ranges and performance outcomes.

A CPV plan is expected to take account of this type of scenario, providing the internal policies and procedures upon which decisions related to changes in process verification should be based. The change described in this scenario can be addressed through the change management system and does not require additional sampling in the CPV plan, as routine sampling is already in place to monitor the upstream cell growth impact of this scenario (Tables 7.1, 7.2, 7.3 and 10.1, 10.2, 10.3). Potential downstream impact could be included in the monitoring plan, e.g. the KPAs of inlet pressure to depth filters and duration of the broth clarification, which are suggested as optional items for CPV in Tables 7.4, 10.4.

Note: Attributes should only be considered optional after their impact on the process has been risk assessed and any lack of monitoring fully justified.

Scenario 2: High Protein A leachate observed in chromatography eluate, Step 5.

A PPQ batch contained 123 mg of protein A/g A-Mab in the Protein A pool, which exceeded the control limit for this process-related impurity. Investigation revealed that:

• Protein A ligand released from the chromatography resin (‘Resin A’ from Supplier A) and entered the process stream during product elution. R&D and Supplier A confirmed that elevated amounts of Protein A can leach from the bead surface during an initial elution after extended resin storage, even when storing under recommended conditions;

• Extended storage can cause increased Protein A leaching in the next use cycle. The resin storage time of more than 12 months between the last clinical manufacturing batch and first PPQ batch was longer than previously experienced and was not represented in small scale trials used to establish PPQ limits;

• In-process testing of the Protein A clearance will be performed to further demonstrate downstream process capability of control of this product quality attribute (AEX Table 7.8, 10.8);

• The level measured in the Protein A step eluate for the batch implicated by this scenario was orders of magnitude below the impurity safety limit for final drug product. Also, at full scale in the affected PPQ batch, downstream clearance of Protein A below the detectable level was demonstrated which is consistent with small-scale observations that the subsequent chromatography steps are capable of removing Protein A (The possibility that limits or controls on extended storage time, conditions, and/ or resin treatments may need to be considered if data indicates the clearance capability of the process is not sufficiently high enough for the reader’s situation).

An additional Design of Experiments (DOE) study was conducted after PPQ to determine the potential for Protein A leaching relative to storage time, resin age (use cycles) and storage conditions. Spiking study confirmation of clearance capabilities in the downstream process steps was achieved and is discussed in the amended CS revision completed after the PPQ experience, where the new CPPs to control clearance are clearly identified. Within CPV, results will be monitored to detect any departures from the expected behavior observed during development; monitoring tools such as ‘tool wear charts’ or ‘residuals charts’ may be useful, and consultation with a statistician is recommended. These tools are mentioned again in Section 12.4.

6.2

HYPOTHETICAL SCENARIOS AND PLANNED PROCESS CHANGESFive hypothetical scenarios and planned changes are provided below to illustrate how the CPV monitoring plan might be affected by events encountered during commercialization of a product such as A-Mab. In this example it is assumed that the PPQ campaign proceeded smoothly and that the expected results were achieved. In particular, CPPs, WC-CPPs, KPPs and GPP are defined and achievable and the process CS is appropriately established. The process CS is assumed to include input raw material controls, procedural controls, process parameter controls and monitoring, in-process testing, and product specification testing (see Figure 5.1.1). These scenarios are accounted for in the CPV plan:

BPOG Continued Process Verification: An Industry Position Paper With Example Plan – Page 27PAGE 26 – BPOG CONTINUED PROCESS VERIFICATION: AN INDUSTRY POSITION PAPER WITH EXAMPLE PLAN

Scenario 3: High elution volume from CEX, Step 7.

Because of resin capacity limitations, elution of the product stream through the CEX resin requires processing a batch in multiple portions (sub-batches). The column eluate streams are then pooled. With one PPQ batch, an unexpected additional volume of buffer solution was required to complete the elution of A-Mab from the CEX column resin for one sub-batch. The prior wash of process impurities from Cation Exchange (CEX) Chromatography resin proceeded without incident but there was a delay while the additional elution buffer was prepared (during which the product loaded column was idle) before proceeding to complete the product elution operation to recover all the A-Mab from the resin. • No impact on A-Mab quality was detected, which

involved a deviation for a KPA (elution buffer volume); • The investigation did not reveal a definitive root cause.

Performance of the flow meter was not implicated as a cause of the unusual observation from review of GPPs and instrument calibration checks;

• Flow channeling through the resin was the initial suspected cause, but no similar observation was made during earlier or later PPQ sub-batches;

• Delay in starting the elution operation may have played a role, but this could not be confirmed because it had not been specifically studied, nor did delays after load prior to elution occur in historical small-scale studies;

• Similar incidents have not been observed with other A-Mab batches at any scale studied; A change in the buffer (e.g. conductivity which is not a CPP, or pH which is a CPP) as a result of the delay has not been conclusively eliminated as a cause, but no deviation associated with the buffer was apparent from careful scrutiny of the batch record (BRc) and interview with process operators.

Investigation of elution buffer stability data is also suggested. If insufficient hold time and buffer attribute data exists to determine the potential for buffer stability to be a contributing cause, this may be pursued as an independent study, rather than including buffer chemical stability in the CPV Plan. Tracking of buffer volume used to elute A-Mab from the CEX column is included in the CPV recommendations for this step (see Tables 7.7, 10.7) because it has demonstrated variability and there is a theoretical potential for increased aggregates with extended processing time (not observed in any studies as of yet) that may result from the need for additional elution to recover A-Mab from the CEX resin.

Scenario 4: UF/DF measurements exceeded action limits.

During preparation of one PPQ batch, the starting UF/DF concentration measurements did not meet the PPQ control limits and step yield was above the expected PPQ range. The starting UF/DF concentration has not been classified as a KPA in the A-Mab case study.• A change prior to PPQ revised the in-process UV

absorbance (A280) test method, which led to an apparent upward shift in yield results. While a bridging study was conducted to determine the suitability of the revised test method, evaluation of the change did not consider the impact to the limits used during PPQ that were calculated based on earlier experience. Limits in place during PPQ were based on measurements from the previous version of the method used for in-process monitoring.

• Change control improved the accuracy of the measurement and also removed a bias error when compared to the final bulk drug substance concentration which uses a different method performed in the QC release testing laboratory.

• The implemented change in the test method involved improvements to both the precision and accuracy of the in-process measurement system; there has been no change to the UF/DF process. Analytical SME’s decided it would be inappropriate to compare new results to a set of limits based on data measured using a different/ altered procedure, or simply adjust previous results for a fixed bias correction (due to potential proportional variance, see section 12.4).

• The corrective action being implemented will supersede the original PPQ limits with new CPV limits calculated using data from the revised test method procedure.

No monitoring recommendations for CPV are proposed as a result of this scenario. Care should be taken not to include data generated prior to the method change in calculating long-term limits.

Scenario 5: Environmental monitoring during Bulk Drug Substance (BDS) fill, Step 11.

During environmentally controlled open system filling of one BDS batch, the routine sentinel plates indicated environmental monitoring (EM) bioburden was above the PPQ action limits. Investigation determined that:• Based on organism identification, the likely source was

skin flora shed by an operator who conducted the final filtration and filling of the BDS;

• The bioburden samples of each post-filtration product container (for the PPQ) met the acceptance criteria with results of 0 CFU/ 10 mL. This confirmed that the 0.2 μ m filtered BDS was not impacted and the routine criteria were met for batch release (BR);

• Following the filling operation, BDS is frozen within 24 hrs, and once thawed, the material is pooled, mixed, and sampled for bioburden prior to sterile filtration when initiating drug product manufacturing;

• Corrective and preventive actions have been implemented, including a review of personnel practices, skills and training, and changes to operating procedures to alert operators to use appropriate practices when working in the controlled filling environment.

No additional monitoring recommendations for CPV are proposed as a result of this scenario because, even with this incident, no impact to the BDS was found and corrective actions have been implemented to prevent its recurrence. Routine monitoring is sufficient. No addition to the enhanced monitoring plan is needed because it is not reasonable to expect from a single incident that there will be variability in bioburden results due to the processing of this step. Note: Whilst attributes and parameters that are included in a CPV Plan are likely to include some that are relevant to BR, a CPV program is expected to operate independently of BR processes and procedures. Analysis of data within the CPV program is not expected to have an impact on product that has been previously released. The release of batches compares batch quality and performance to a specific set of pre-determined specifications and other measures. In contrast, the focus of CPV is to reveal trends and sources of variation in batch quality and performance that already fall within the predetermined criteria for BR.


SECTION 7.0

7.0 CPV PLAN RECOMMENDATIONS FOR THE A-MAB PROCESS This section describes, for each of the A-Mab process steps, what to include in the CPV plan and the justification for its conclusion. This justification is primarily based on process knowledge and process experience. A table is provided for each step to summarize the recommendations for CPV. Discretionary items are also included that may be needed in a CPV program depending on the assurance of process understanding or that provide additional depth to the monitoring plan.

No recommendation for including in-process product pool hold times in CPV is proposed, because the hold times were validated as part of the basis for controls within the Master BRc. In the event that a hold time is exceeded this one-off event would trigger a deviation within the Quality System, under which impact to product quality would be determined.

In the steps with elution of product from resin beds (i.e. steps 5 and 7), several resin loading/ elution cycles are used to process each batch. No controls have been identified for resin regeneration operations in either of these steps. For these steps, concurrent validation of the resin use lifetime includes periodic sample testing of appropriate quality attributes for continued verification of packed resin effectiveness during its use lifetime. Effectiveness of resin regeneration conditions is included in the ongoing resin use validations. Therefore monitoring of CQAs for this purpose need not be included in the CPV plan. Continued monitoring, and further verification of effective process controls, should be considered for CPV when resin use lifetime monitoring ceases, if further data are needed for understanding of impurity clearance.No recommendation for including in-process hold times in CPV is proposed because ongoing study of hold times during commercial manufacturing is conducted using a separate hold time qualification study.

Steps that have in-process quality attributes related to microbial control (bioburden, endotoxin) are sampled and

tested as routine in-process controls. The nature of test results in this case (approximately 0 cfu/ sample, and ≤ Limit of Quantification, LOQ, respectively) do not permit meaningful Statistical Process Control (SPC) analysis in CPV. QC microbiology laboratory review of these results against action and alert limits will provide appropriate monitoring for drift in microbial control of the process and management of deviations, so monitoring, data analysis and any response to bioburden and endotoxin results are not included under this CPV plan.

Note: It could be seen as best practice that the quality system for bioburden and endotoxin monitoring and the CPV system are connected, so that any deviations would be reflected in CPV Reports. Statistical criteria that may be applied to analyses of data are discussed in section 12.

The A-Mab case study did not identify any critical raw materials or address CS or risk assessment for input material controls. However, as a result of a hypothetical culture medium change described in section 6, one monitoring recommendation related to material variability is provided as a recommendation for the CPV plan. Additional monitoring of materials used in the bulk drug formulation is also included as an option.

The process risk assessment established that steps 1 and 2 of the A-Mab process do not entail risk of impact to product quality in the production bioreactor because no product is accumulated at these stages. Specifications for raw materials, such as cell banks and media components, assure use of the intended genetic cell line to produce A-Mab and control introduction of endotoxins which could affect cell metabolism.

CPV for this step should focus on process consistency and obtaining sufficient data to calculate long-term control limits (see Sections 9.0 and 10 for further discussion and examples of control limits, and Section 12 for information on the statistical basis for control limits. which account for normal process variability. As stated in the A-Mab case study and demonstrated in the PPQ, BR procedures, SOPs, automated process controls and use of alarms all ensure the

seed expansion steps are routinely monitored and operated within established limits. Therefore, monitoring of non-critical parameters in this step such as temperature, pH, and dissolved oxygen need not be included in the CPV plan. This is shown in Table 7.1 below.

Environmental Monitoring (EM) is routinely performed for open (under appropriate ISO classified conditions) process manipulations (including use of Rodac and settling plates) to demonstrate microbial control and the existing QC laboratory program is established for reporting results and assessing trends. Therefore inclusion of this EM monitoring plan in the CPV plan is unnecessary. As noted previously, these systems need to connect as it would be best practice to ensure deviations are present in CPV Reports.

7.1

STEP 1, SEED CULTURE EXPANSION IN DISPOSABLE VESSELS – CPV RECOMMENDATIONS

Table 7.1. Step 1 CPV recommendations

VARIABLE CLASS CQAS IMPACTED

CPV RECOMMENDATION & JUSTIFICATION

DETERMINATION METHOD AND/OR SOURCE

TYPE OF DATA EXPECTED/ ANALYTICAL APPROACH

VCC (each passage end)

KPA – Include, to verify process consistency

Routine batch documentation for each passage.

Discrete value,univariate

OPTIONAL ELEMENTS TO INCLUDE IN CPV

Initial VCC/split ratio (each passage)

KPP – Optional, to verify process consistency

Calculation from routine batch documentation for each passage, ratio of passage ending cell density over initial cell density of next passage.

Discrete value, multivariate

Culture duration(each passage)




Culture viability (each passage end)

KPA – Optional, to verify process consistency




7.2

STEP 2, SEED CULTURE EXPANSION IN BIOREACTORS – CPV RECOMMENDATIONSAs noted in section 7.1 for step 1, cell growth is complex and it is difficult to comprehensively define or predict all sources of variability. Expansion culture conditions may impact cell biology which in turn can impact product quality during product expression. CPV for this step should focus on process consistency and obtaining sufficient data to resolve long-term control limits which account for normal process variability.

Inclusion of EM in the CPV plan is unnecessary because an existing QC program is established for reporting and trending of EM results.





VCC (each passage end)





Initial VCC/ split ratio (each passage)


Calculation from routine batch documentation for each passage, ratio of passage ending cell density over initial cell density of next passage.


Culture duration (each passage)




Culture viability (each passage end)



Discrete value, Multivariate

The process risk assessment established the BDS CQAs that may be impacted by this step. As stated in the A-Mab case study and demonstrated in PPQ, BR procedures, other SOPs, automated controls, and an alarm system all ensure the production step is routinely monitored and operated within established limits for many of the related parameters.

Turbidity at harvest (a KPA known to vary in response to bioreactor culture conditions) is not included in the CPV plan because of the confidence that centrifugation and depth filtration can accommodate variability in the harvest material (low differential pressure across depth filters). However, if there is a filter change, or the medium component change introduced between PPQ and commercial manufacturing indicates a shift in other monitored variables for this step or process performance of the next step, establishing control limits for turbidity at the end of the production bioreactor should be added to the CPV plan.

The CPPs for medium osmolality and culture duration are included in the CPV plan. For these CPPs, a large tolerance for variation has been shown in development during process characterization studies. Maximum pCO2, bioreactor temperature and bioreactor pH are other identified CPPs to be included in the CPV plan.

At the time this protocol is initiated, the PLS model is classified as a KPA; it is a predictor of A-Mab oligosaccharide structure CQAs and acidic variants. Model input parameters of temperature and pH, and model input attributes of titer, VCC, and viability are separately included in the initial monitoring while the bioreactor model is qualified.

Remnant glucose concentration is not included in the CPV plan because it is assumed to be a fixed value CPP which triggers additions of glucose feed. However, as an attribute of the culture, it is measured daily and when the glucose concentration drops below a particular level, a discrete volume (assumed to be a fixed KPP) of a glucose solution is added as a bolus to ensure the glucose concentration remains ≥ 1.0 g/L. A fixed volume of nutrient feed is added at a defined time under automation and routine batch document controls, therefore trending of the KPPs nutrient feed volume and

timing of nutrient feed does not provide value because they are not subject to random variation. Harvest attributes of titer, viability, and culture duration are also included for trend monitoring to verify process outcome consistency.

The Partial Least Squares (PLS) multivariate model generated during process characterization in the A-Mab case study [3, Section 3.10, Page 108] [25], includes other CPPs (e.g. dissolved oxygen, pressure, gas addition rates) and KPAs (e.g. VCC, viability), noted in Table 7.3.

The CPV plan may optionally include selected KPPs and KPAs to provide additional measurements of robust process consistency and to obtain sufficient data to resolve long-term control limits that account for normal process variability. Two suggested discrete KPAs, peak VCC and culture viability at harvest, are optional in Table 7.3 for Step 3. Other KPAs (glucose and lactate concentrations) are also included as part of the PLS model described in the A-Mab case study.

In-process quality attributes for this step, namely bioburden, Murine Minute Virus (MMV), mycoplasma, and Adventitious Viral Agents (AVA) are controlled as routine in-process specifications linked to drug substance BR. Their binary pass/ fail nature does not permit meaningful SPC trend monitoring, and does not provide prospective warning of pending batch failures. These routine control measures are sufficient for maintaining the process in its validated state and deviations detected will trigger investigations for out of control situations/events.

Regarding the productivity of the production culture step, whether to include the Antibody-Dependent Cellular Cytotoxicity (ADCC) bioassay in the CPV plan or not, is an interesting and somewhat complicated question. ADCC is correlated with afucosylation in vitro. Thus measurement of potency by ADCC is an indicator for this quality attribute that might impact Fc effector function. However, this bioassay is not qualified to test crude production bioreactor material just prior to harvest due to broth interference. Fundamentally, that would not prevent reliable results that correlate with the potency of the purified material. But, confirmation of

7.3

STEP 3, PRODUCTION CULTURE BIOREACTOR – CPV RECOMMENDATIONS


VARIABLE CLASS CQAS IMPACTED CPV RECOMMENDATION & JUSTIFICATION



Culture duration CPP Aggregates, glycosylated glycans, HCP, DNA; can also impact turbidity at harvest, yield variation

Include, to establish SPC capability and large tolerance for variation

Routine batch documentation

Discrete value, univariate

Maximum (dissolved) pCO2

CPP Glycosylated glycans, deamidated isoforms; also product yield

Include, to establish SPC capability and correlate with in-vitro cell age (IVCA)



(Bioreactor) temperature

CPP Glycosylated glycans, deamidated isoforms

Include, to demonstrate appropriate range is established


Continuous datastream, univariate

(Bioreactor) pH CPP Glycosylated glycans, deamidated isoforms

Include, to demonstrate that appropriate monitoring and automated adjustments are established



Afucosylated glycans

CQA – Include, to verify process consistency

Will require non-routine test, record results in Laboratory Information Management System (LIMS)


Galactosylated glycans

CQA – Include, to verify process consistency

Will require non-routine test, record results in LIMS


PLS model employing pH, DO, temperature, pressure, gas rates, weight, VCC, viability, titer, glucose, lactate

KPA Isoforms, variants, DNA, monomer, aggregates, HCP oligosaccharides

Include, to verify process consistency


Continuous datastream, multivariate

Product yield (titer at harvest)


QC ELISA results in LIMS Discrete value, univariate


functional activity is relevant to the finished dosage form since it is the drug product that is provided to the patient. So, monitoring of bioreactor harvest for potency is not recommended since ADCC activity has a drug product release specification for CPV trending and is a stability indicating assay included in routine stability testing protocols [derived from A-Mab case study section 6.4.2, Page 247].

The CS categorized the antifoam ingredient to be a critical raw material (CRM), probably because it is a process residual CQA. However, there is no particular critical material attribute (CMA) that requires enhanced monitoring. The addition of antifoam varies as needed up to a maximum 100 mg/L concentration (section 5). As a single antifoam lot is used for multiple bioreactor batches, lot change points in the batch genealogy will be traceable to correlate with any process shifts in trends for this step, the clarification step, or the Protein A chromatography step (steps 3, 4 and 5 in the process). Because only one lot of antifoam was introduced in the PPQ, three BDS batches during the initial CPV period, which employ different antifoam lots in the upstream process, are tested to provide evidence of robust clearance of the process residual. Due to the low turnover in antifoam lots, routine but periodic batches being tested for stability may also be selected for this extra testing in BDS, i.e. at ‘time zero’. This does not suggest that clearance of antifoam is a stability indicating attribute.

The oligosaccharide profile (a CQA) is solely influenced by the production bioreactor. Input material and procedural controls are in place to ensure the quality of raw materials and the cell line. Control of step 3 CPPs (temperature, pH, dissolved carbon dioxide, culture duration, and medium osmolarity) within their limits ensures consistent glycosylation. No process clearance or further glycan modification occurs in downstream processing, and the oligosaccharide profile is not regarded as stability indicating. Routine testing is not part of the drug substance lot release specification based on the development process design history, process risk assesments, CS, and PPQ. The risk that exists is that no process clearance or further modification is expected in downstream processing. An oligosaccharide profiling method utilizing Capillary Electrophoresis-Laser Induced Fluorescence (CE-LIF) was developed and qualified for characterization of the oligosaccharide profile. There is also an in vitro cell-based

bioassay qualified to enable collection of biological activity data related to ADCC functions, as a means to assess Fc-oligosaccharide structure-function relationships.

CPV trend monitoring of afucosylated and galactosylated glycans in the bioreactor for step 3 is recommended to build confidence in process consistency (see Table 7.12, 10.12). Note that sialylation, high mannose content (also afucosylated) and non-glycosylated heavy chain were also determined to be CQAs but recommended only as optional elements to include in CPV monitoring (see Table 7.12, 10.12). The frequency of lifecycle monitoring of glycans will be reviewed and adjusted based on trends. Characterization of the oligosaccharide profile will be conducted to confirm comparability when needed to support process changes [derived from A-Mab case study section 6.4.5, Page 250 and section 6.6.1, Page 251-253].

The mechanism and conditions conducive to formation of deamidated isoforms are widely known and well understood. This knowledge, in conjunction with the level of risk associated with the quality attribute in the post-PPQ risk assessment, negates the need for in-process CPV testing. Process control includes testing with a routine CEX HPLC method at lot release, of both drug substance and drug product, to confirm the identity of A-Mab, monitor charge heterogeneity and detect shifts in deamidated isoforms [derived from A-Mab case study section 6.6.4, Page 259]. The method separates the main charged isoforms of A-Mab that are considered to be product-related substances as defined in ICH Q6B. The resulting chromatographic profile is specific to A-Mab and unambiguously distinguishes it from other monoclonal antibodies. The spectrum of isoforms contained in the reference chromatogram for A-Mab represents acidic and basic isoforms. The chromatogram is inspected to ensure a consistent profile with the reference standard and the absence of any new peaks. A quantitative definition of new peaks is included in the CEX test method. Charged isoforms of A-Mab do not increase when stored at recommended conditions; therefore, the attribute is not monitored on stability [derived from A-Mab case study section 6.4.2, Page 247].

The KPA of titer (yield) is included in CPV for trend monitoring for process consistency.


VARIABLE CLASS CQAS IMPACTED CPV RECOMMENDATION & JUSTIFICATION




Antifoam lot CMA Residual antifoam C

Include, to track lot changes. Test clearance at BDS for 3 different lots

Routine batch document genealogy

Qualitative text/ label, univariate

(Medium) osmolality

CPP Glycosylated glycans, deamidated isoforms

Include; large tolerance for variation has been shown. Monitor by exception a



Mannose content CQA – Optional, to verify process consistency



Sialic acid content

CQA – Optional, to verify process consistency



Non-glycosylated heavy chain

CQA – Optional, to verify process consistency



Time of glucose feeds (hrs since inoculation)




Peak VCC KPA – Optional, to verify process consistency



(Culture) viability at harvest




a: The term “monitor by exception” means that reported data outside of established alert or action limits will be reported as incident(s); for CPV, a review of reported incidents will examine the occurrence of any events outside of established limits and determine the collective impact of these events.


The process risk assessment established that the clarification step is unlikely to impact product quality. Monitoring for CPV is limited to process consistency for the purpose of examining data to establish long-term control limits that account for normal process variability (per statistical confidence criteria stated in section 12). Therefore, the KPA of yield is included in CPV for trend monitoring as a process performance indicator. One KPA, turbidity of filtrate, is also recommended to confirm process consistency following the culture medium change (see section 6.2, scenario 1).

The PPQ demonstrated that BR procedures, SOPs, automated process controls and alarming ensure the centrifuge and filtration step are routinely monitored and operate within established limits. Temperature, centrifuge feed rate and rpm, and filter flow rate are not CPPs and are tightly controlled engineering or fixed design parameters that are not subject to random variation and therefore do not merit inclusion in CPV.

Evaluation of a change in the manufacturing method of the culture medium used upstream (see Section 6, Scenario 1) could include additional monitoring of downstream KPAs of inlet pressure to depth filters and duration of the broth clarification, which are noted in the Table 7.4 as optional items for CPV. The small scale model evaluation of new lots of the medium material showed that product quality attributes of Host Cell Protein (HCP), DNA, and product structural characteristics are not impacted, so monitoring of these KPAs is not included in the CPV recommendations. The decisions not to include these KPAs in CPV could be re-examined pending the results of the change management evaluation of the culture medium change.

7.4

STEP 4, CLARIFICATION (CENTRIFUGATION AND DEPTH FILTRATION) – CPV RECOMMENDATIONS





Turbidity(of filtrate)

KPA – Include, to confirm process consistency following medium change

Non-routine testing needed


Step yield (product in filtrate)

KPP – Include, to verify process consistency

QC ELISA test results in LIMS



Duration of broth clarification

KPP – Optional, to confirm process consistency following medium change

Routine batch documentation, elapsed time from start of harvest (opening of bioreactor bottom valve) to end of filtration (closing or filtrate vessel inlet valve)


Inlet Pressure to filters KPP – Optional, to confirm process consistency following medium change


Continuous datastream or discrete value, univariate






Protein load (ratio) (in HCP model), each sub-batch

CPP HCP, DNA, process impurities

Include, to establish SPC capability

Routine batch documentation, calculated using packed resin volume


Elution buffer pH (in HCP model)

CPP HCP, DNA, process impurities

Include, to verify process consistency



Residual Protein A in eluate pool

CQA Protein A Include, to establish SPC capability

Non-routine testing needed, results recorded in LIMIS


Step duration KPP – Include, to verify that process can capably control this CQA

Routine batch documentation, elapsed time from closing of vessel inlet valve to eluate pooling is completed


Step yield KPA Include, to confirm process consistency

Routine batch documentation, calculation using in-process A280 test result

Discrete value, multivariate likely to exhibit normal distribution


The process risk assessment established that protein A purification may impact product quality (aggregate; charge variants; leached Protein A; clearance of HCP, DNA, and methotrexate) and links to performance of chromatography steps 7 and 8 (CEX and AEX). Platform and prior process knowledge negate the need for specific process studies except as noted below for HCP and leached Protein A.

CPV for step 5 (the first of the downstream DS process steps) should focus on process consistency to obtain sufficient data to establish long-term control limits that account for normal process variability (see section 12 for statistical confidence criteria). Variables recommended for inclusion in the CPV plan are shown in the Table 7.5 below, including CPPs identified for this step (protein load ratio and elution buffer pH). Elution buffer pH is closely controlled by batch procedure and buffer is not released for use if pH is out of range. This variable is included in the CPV plan to monitor the extent of buffer pH variability incorporated in the HCP model prediction. Step duration, a KPP, is included in the CPV recommendations to establish capability on processing time for this step.

The key process attribute of yield is included in the CPV recommendations for trend monitoring of process consistency.

CPV need not include monitoring of flow rate through the resin, nor the end collection point (column volume or A280 absorbance) for the eluate because the PPQ verified the expected control and minimal variation for these key parameters. Operating temperature and other GPPs were

shown in characterization studies to not impact product quality or process consistency when controlled within easily achieved design ranges. The automated continuous process controls and alarm system, as well as BR sequencing and SOPs, ensure the step is routinely monitored and operated within its established limits.

A linkage model study was proposed in the A-Mab case study to examine HCP levels at different points within the process (after each chromatography step – steps 5, 7 and 8). This is included in the CPV plan and involves non-routine analysis to provide data on measured HCP levels at the final point in the process covered by the multivariate model (after AEX chromatography, step 8) against the predicted outcome.

Storage of the Protein A resin (section 6, scenario 2) is expected to potentially introduce a variable amount of leached Protein A into the product stream. This will be monitored via residual protein A (leached from the resin) testing and trending of the results to establish process capability for controlling this CQA.

Process control deviations for this step should evaluate the case-by-case potential impact on these attributes (and viral clearance), as process streams continue further downstream for purification. For deviation investigations, it may be appropriate to review the risk assessment justification for any low risk CPPs. Note that the process has high Impurity Safety Factor (ISF) (for a definition, see A-Mab Case Study [3] Section 4.10.3, Page 167) clearance (>5x104) for all process related impurities for normal processing.

7.5

STEP 5, PROTEIN A CHROMATOGRAPHY – CPV RECOMMENDATIONS

The process risk assessment established that the low pH treatment step for viral inactivation impacts two product CQAs (aggregate and viral inactivation). There is no claim for removal of process related impurities (HCP, DNA, methotrexate or leached Protein A) but some incidental reduction in these impurities may be achieved in this step, which includes precipitation and downstream filter clearance. In general, CPV for this step should focus on obtaining sufficient data to resolve long-term control limits related to viral inactivation, so CPPs for inactivation time and pH should be included in CPV. The viral safety risk CQA (inactivation of particular AVA) for the A-Mab process has been validated in the small scale model during stage 1 process validation. Inactivation time and pH are readily controlled within desired limits for the process as shown by PPQ. However, inclusion of both these parameters in CPV is recommended, because they are manually controlled and

susceptible to variation within their PARs. One additional test, for aggregates, is also recommended for CPV to establish process capability for this CQA.

CPV need not include inactivation temperature and agitation mixing because PPQ verified the expected control and minimal variation for these key parameters. BR sequencing, automated process controls and the alarm system will ensure the step is routinely monitored and operated within its established limits for these parameters.

The limit for maximum protein concentration in the Protein A pool is bound by the pH inactivation step requirements, but trending of the protein concentration is not recommended as the information will provide little benefit in process understanding. However, a related optional inclusion for CPV is to trend the amount of acid added, to ensure the

7.6

STEP 6, LOW PH TREATMENT – CPV RECOMMENDATIONS



A-Mab process does not drift or shift toward the edge of the qualified conditions of the platform process without this being recognised.

Step yield is not included as a CPV recommendation because yield is not expected to be impacted at this point in the process

and variability around the expected 100% has been associated with measurement uncertainty rather than process variability, therefore it does not merit CPV trending or monitoring. The basis for yield for the next step (7, CEX) begins from the eluate pool of the previous step (5, Protein A).





pH (during inactivation)

CPP AVA, aggregates

Include, to confirm process consistency

Routine batch documentation, integrated average of online pH values during inactivation time


Post-inactivation aggregates

CQA – Include, to establish SPC capability

Non-routine testing needed, results recorded in LIMS

Discrete, multivariate


(Inactivation) time CPP AVA Optional, to establish SPC capability

Routine batch documentation, calculate from completion of acid addition to start of titration


Quantity of acid added KPP – Optional, to establish SPC capability

Routine batch documentation, change in supply vessel weight






Protein load (ratio) (in HCP model)

CPP Aggre-gates, HCP


Routine batch documen-tation, calculation using packed resin volume


Wash conductivity (in HCP model)

CPP HCP Include, to confirm process consistency

Routine batch documenta-tion


Elution pH CPP HCP, DNA, Protein A, aggregates


Routine batch documenta-tion


Aggregates in CEX eluate pool

CQA – Include, to verify process performance

Will require non-routine in-process test, results recorded in LIMS


The process risk assessment established that the CEX step primarily impacts two product CQAs, aggregate and residual HCP. Characterization studies showed that DNA and protein A clearance were not impacted by this step, and there is no claim of viral clearance for this step. Trending of CPPs identified as potentially impacting these CQAs are included as recommendations for CPV.

CQAs for HCP (as supporting evidence for the linkage model prediction) and aggregate are included in CPV to establish process capability, and the KPA of yield is included in CPV for trend monitoring as a process performance indicator.

Univariate monitoring is not required for flow rate through the resin, elution buffer pH, load buffer pH, wash buffer pH, re-equilibration buffer pH, eluate volume, nor the starting or end collection point (A280/A320) for the eluate because PPQ verified the expected control and minimal variation for these

key parameters. BR sequencing, automated process controls, and the alarm system will ensure the step is routinely monitored and operated within its established limits for the independent parameters. Development studies concluded a wide operating temperature range had no impact on product quality or performance/ process consistency, so monitoring of temperature beyond that routinely done for each batch is not included in CPV recommendations.

CEX eluate volume (each sub-batch) is included in CPV to determine a higher confidence range of the normal variation due to the special cause event that occurred during PPQ (see section 6.2). Data obtained will be used to show process consistency with respect to this parameter.

The basis for yield for this step begins from the step 5 eluate pool.

7.7

STEP 7, CATION EXCHANGE CHROMATOGRAPHY (CEX) – CPV RECOMMENDATIONS




The process risk assessment established that the AEX step impacts several CQAs (viral clearance, aggregate, endotoxin, and clearance of protein A, charge variants, HCP, DNA, and methotrexate). Trending of three CPPs identified as potentially impacting these CQAs (protein load ratio, equilibration buffer conductivity and load pH) are included as recommendations for CPV.

Monitoring of other CQAs impacted by this step is not recommended because trending for HCP and Protein A are sufficient to represent the performance and establish the capability of this step. Since the step 5, 7, 8 linkage model (see Section 12) is for predicting an impurity CQA (residual HCP), the output of the model is classified as a KPA, and is also included for monitoring against CPV control limits (not BR acceptance criteria).

Monitoring of step duration (a KPP) is suggested as an optional inclusion for measuring process capability. Inclusion of other process parameters including flow rate (a CPP) and KPPs such as starting or end collection UV for the eluate, or pH of the prepared equilibration/wash 1 buffer are not suggested because PPQ verified the expected control and minimal variation for these parameters. BR sequencing, automated process controls, and the alarm system will ensure the step is routinely monitored and operated within its established limits for these independent parameters. Development studies concluded that a wide protein concentration range had no impact on product quality or performance/process consistency, so trending of protein concentration is also not included in CPV recommendations.

Monitoring of step yield will serve as an indicator of any drift in process control for this step.

7.8

STEP 8, ANION EXCHANGE CHROMATOGRAPHY (AEX) – CPV RECOMMENDATIONS





HCP content in CEX eluate pool

CQA – Include, to verify process performance

Will require non-routine in-process test, record results in LIMS


CEX Eluate volume (each sub-batch)

KPA – Include, to confirm process consistency



Step yield KPA – Include, to confirm SPC capability

Routine batch document calculation using field A280 test result



Step duration KPP – Optional, to confirm SPC capability

Routine batch documentation; elapsed time from end of step 6 (vessel inlet valve closes)






Protein load (ratio) CPP HCP, viral clearance


Routine batch documentation, calculation using packed resin volume


Load conductivity CPP Viral clearance


Routine sample and test for buffer use


Load pH (in HCP model)

CPP HCP, viral clearance


Routine batch documentation, sample test


Equilibrium/ Wash 1 buffer conductivity (in HCP model)

CPP HCP, viral clearance

Include, to verify process performance

Routine sample and test for buffer use


Linkage model output for HCP content in AEX eluate (predicted)

KPA – Include as outcome of HCP linkage model, to demonstrate understanding of HCP clearance through multiple processing steps

Calculated from six variable terms logged in batch documents for step 5,7,8


HCP content in AEX eluate (measured)

CQA – Verify model of HCP clearance through multiple processing steps

Non-routine test, results recorded in LIMS


Residual Protein A in eluate

CQA – Include, to confirm process consistency



Step yield KPA – Include, to confirm SPC capability




Step duration KPP – Optional, to confirm SPC capability

Routine batch documentation; elapsed time from end of step 7 (vessel inlet valve closes) until product elution completed in step 8




SVRF is a physical size separation step that is critical for viral clearance. Filter function is confirmed after each batch by standardized integrity testing. Introduction of leachate from the filter is minimized by a routine pre-use rinse of the filter with a validated quantity of AEX elution buffer.

Operating pressure is a WC-CPP recommended for including in the CPV plan. Some variation in pressure has been observed; trending of pressure data will increase predictability and confidence in knowledge of the natural variation and what, if any, impact this variation may have on process consistency. Correlation of operating pressure with filtration volume (the other CPP to be monitored) and protein concentration will also ensure consistent viral Log Reduction Value (LRV) and serve as a basis for future process improvements/change controls. Small-scale studies have shown that the likely variation in these parameters does not represent a BR risk for product safety or quality. Including filtration load volume in CPV provides an alternate measure of process consistency, given its impact on processing time for this step.

Verification of the filter integrity testing was included in PPQ and will be included upon completion of the filtration of every process batch. Re-filtration has also been validated and details are registered in regulatory licenses. Detectable impact of re-filtration is a decrease in the measured protein concentration due to dilution by a hold-up recovery flush after filtration to optimize step yield. Incidents of failed filter integrity and/ or when re-filtration is performed are tracked with the change control system as incidents and are trended as part of Annual Product Review, so will not be included in CPV.

Step yield is not included as a CPV recommendation because yield is not expected to be impacted by this step. Yield after step 10 will include step 9.

Rinse or processing flow rate through the filter and the flush volume used are not recommended for inclusion in CPV, because PPQ demonstrated tight control and minimal variation of these KPPs. Although these variables are manually controlled, the BR instructions and sequence will ensure the step is routinely monitored and operated within its established limits.

The A-Mab case study does not provide sufficient detail to trace CPV rationale back to a CS, risk assessment, or process development for this step. Therefore, six well known process parameters typical for the operations of a UF/DF step are considered for CPV:

• Trans-Membrane Pressure (TMP);• Temperature of the product containing stream;• Permeate and recirculation flow rates;• Number of dia-volumes to complete the buffer

exchange;• Product concentration (prior to and after buffer

exchange);• Step processing time.

Platform knowledge was leveraged to define an initial membrane life limit controlled via batch documentation and equipment logbooks. A specific membrane lifetime monitoring protocol is expected to be in place alongside the CPV plan, to verify filter performance.

Flow rates are response variables that automatically adjust to maintain a fixed TMP set point (pressure controlled operation). Because flow rate profiles tend to vary over the re-use lifetime of the membranes, an optional choice for CPV includes monitoring of permeate and recirculation flow rates via a trajectory profile of the continuous dynamic data. Reference standard profiles (3SD tunnels, i.e. control charts with + 3 standard deviation acceptability limits) will be shown for comparison (sourced from the initial use cycles for the membrane and from the PPQ batches).

It is assumed that PPQ demonstrated that BR procedures, automated process controls, and alarming ensure the UF/DF step is routinely monitored and operated within established limits characterized in small scale development DOE studies. Therefore temperature and TMP are dismissed as non-CPPs and are not recommended for inclusion in CPV.

The number of dia-volumes needed to complete the buffer exchange, pH of the AEX eluate solution to be processed in this step, and UF/DF processing time are additional CPPs to include in CPV, because of their potential impact on product

concentration and dia-volumes (affects osmolality), the potential formation of aggregate (from lengthy processing time and/or incorrect pH), and because process validation has not yet provided sufficient data to demonstrate process capability for these parameters.

It is assumed that a risk assessment established that the UF/DF step has potential to affect various product quality attributes. Variation in protein concentration prior to BDS freeze and fill (step 11, post-filtration) may affect downstream drug product manufacturing controls/ capability, so trending of this CQA is included in CPV recommendations. These data will also provide evidence for any correlation with other variables (e.g. dia-volumes needed for buffer exchange). Optionally, CPV may include selected product CQAs, chosen because of knowledge that they may reveal the impact of variability in the process or provide useful information about process capability. The product solution identity, composition, and aggregation could be altered by either post-diafiltration pH or osmolality (or by a trace contaminant in compendial grade raw material), so inclusion of these CQAs should be considered for CPV. The A-Mab case study CS established that impurity clearance capability for residual methotrexate is very high and does not require further verification. The genealogy link between the culture media used in the upstream process batches (impurity clearance of antifoam and methotrexate) and the UF/DF membrane lot will be logged in the enterprise resource planning system for use in investigations.

Protein concentration prior to diafiltration is suggested as an option for inclusion in CPV, to provide data linking this in-process CQA measure to the final protein concentration at completion of this step. Another suggested option is to trend aggregates in the final retentate with the intent of providing additional data to trend this step for ability to control formation of this process-related impurity.

Diafiltration and final formulation buffer ingredients are excipients of the drug product and therefore are critical raw materials. Supplier quality management includes specifications for material purity, content of particular elemental impurities, and endotoxin from inspection of

7.9

STEP 9, SMALL VIRUS RETENTIVE FILTRATION (SVRF) – CPV RECOMMENDATIONS

7.10

STEP 10, ULTRAFILTRATION AND DIAFILTRATION (UF/DF) – CPV RECOMMENDATIONS





Operating (inlet) pressure

CPP Viral clearance


From online data acquisition, plot results with range of acceptable standard profiles. Correlate these data with filtration volume and protein concentration.


Filtration (load) volume

CPP Viral clearance


Routine batch documentation, vessel weight change from pre-rinse tare to filled weight




Certificate of Assay (CoA) summaries and/or in-house verification testing. Reserve samples and CoA summaries for each excipient lot are preserved until drug product expiration to enable investigations as needed. Although not specifically identified here, an un-named Critical Material Attribute (CMA) for ‘excipient 1’ may require characterization of the variability of ‘attribute A’ during CPV as a risk mitigation action for the CS, so this has been included as an option for CPV.

The key process attribute of yield is included in CPV for trend monitoring as a process performance indicator.

Normalized Water Permeability (NWP) and average filtrate flux are monitored and verified by a membrane lifetime protocol and new membrane installation SOP. Sampling for lifetime monitoring, verification, and potentially extension of the number of reuses is managed under this separate protocol and is therefore not considered here for CPV.

Final filtration performed while filling sterile containers provides assurance of microbial control of the drug substance intermediate, but is otherwise unlikely to impact product quality. No provisions are assumed for a validated re-filtration option. Filter function is confirmed after each batch by standardized filter integrity testing. Introduction of filter leachates are minimized by process design and leachable studies, which include a pre-use rinse of the filter with a qualified fixed amount of final formulation buffer.

The KPA of yield was chosen for monitoring this step in CPV, because trend monitoring of yield will provide a good process performance indicator.

Although it is not a CPP, maximum inlet pressure (filter pressure) is known to exhibit product specific batch variation from platform process knowledge and so is suggested as an optional inclusion in CPV. Filtration volume is another KPP that would be a reasonable optional choice for CPV, providing a different measure for assessing processing capability. Correlation of filter pressure with filtration volume, protein concentration, and filtration time are other optional considerations that could be included in the CPV plan, to characterize normal performance and variation for the A-Mab process for future predictability.

7.11

STEP 11, FINAL FILTRATION AND FREEZING OF BDS – CPV RECOMMENDATIONS





UF/DF processing time CPP Aggregates Include, to verify process capability

Routine batch documentation, elapsed time from UF start until defined UF end

Discrete, univariate

Number of dia-volumes

CPP Product concentra-tion and several others

Include, to establish process consistency

From online data acquisition, include in batch documentation


UF/DF retentate final pH

CPP Aggregates Include, to establish process consistency



Protein concentration prior to BDS fill step

CPP Protein conc. of BDS

Include, to establish process consistency

Routine batch document recording of field A280 test results


Yield (final retentate) KPA – Include, to confirm SPC capability




Excipient ”1” Attribute “A”

CMA – Optional, to examine variability of materials used

Released by compendia testing or COA, results recorded in LIMS


SEC aggregates in final retentate

CQA – Optional, to confirm consistency of mixing and foam control



Protein concentration prior to dia-filtration

KPP – Optional, to establish process consistency

Routine batch document recording of field A280 test results


Recirculation flow rate KPP – Optional, to establish process consistency



Permeate flow rate KPA Optional, to establish process consistency



Table 7.10. Step 10 CPV Recommendations


The BDS freezing rate profile and container seal integrity has been validated at commercial scale and the freezing time and conditions are well controlled and documented to support any investigations. The freezing equipment is included in the periodic validation maintenance and instrument preventative maintenance programs. CPV monitoring is not proposed for several related operating variables because PPQ demonstrated tight control and minimal variation of these variables, including: bulk mixing after UF/DF and during the fill, the fixed flow rate through the filter, the flush volume used, verification of the filter integrity testing and product intermediate freezing temperature. BR sequence and instructions, automation monitoring and alarm systems will

ensure the step is routinely monitored and operated within its established limits. The time the intermediate is stored frozen prior to shipment for drug product manufacturing, could be considered as a means of identifying any potential correlation with data from the stability program, but this is not included in the CPV recommendations here.

EM is a supporting quality system subject to periodic monitoring, so it is not included in CPV recommendations, despite a related incident report for this step (see Section 6, scenario 5).

The QC group ensures each drug substance batch meets specifications for lot release to drug product manufacturing. Some specifications, such as the identity attribute of consistency with reference standard and inspection for new peaks, are not amendable to trend analysis for CPV. The QC microbiology laboratory monitors all drug substance lot data for endotoxin and bioburden against alert and action limits to provide appropriate monitoring of the process and management of microbial control deviations.

In the A-Mab case study, routine BR specifications proposed for the drug substance were intentionally minimized to show a potential application of QbD development for process validation Stage 1. Endotoxin testing, a BDS and drug product release requirement is reviewed as per the QC laboratory SOP, with SPC based alert limits which are assessed for suitability during each annual product review. For the CPV plan, additional BDS CQAs were selected for continued verification and enhanced monitoring, to demonstrate consistency over a longer period during which more process variation may be observed. Content of various oligosaccharide structures were selected as high risk CQA examples from the A-Mab case study (refer to section 7.3 and 10.3). Control limits are set inside the claimed acceptable range (see section 5) based on statistical analysis of data to provide early warning during trend monitoring.

Some additional process and product related impurity parameters are included in the CPV plan for this process step. For example, monitoring of antifoam C rather than methotrexate clearance was chosen. Antifoam additions vary batch-to-batch to control foam and clearance is combined over steps 4 and 5. In contrast, methotrexate is a fixed addition prior to the N-1 seed bioreactor, resulting in significant dilution as the process scales up to 15,000L and a high log reduction factor was demonstrated for the Protein A step 5 alone. HCP is not included for CPV at BDS because it is monitored for CPV at step 8 (AEX), as both a special sample test with a control limit well inside the 0 to 100 ng/mg acceptable range (based on the similar X-mAb process) and via the multivariate model for linked chromatography parameters [25]. No particular deamidated isoforms (which incidentally, were not designated as CQAs) or other charge

variants are required for CPV monitoring. The routine drug substance specification confirmation of A-Mab identity includes a CEX HPLC method which separates isoforms (product-related substances as defined by ICH Q6B) and both a consistent profile with the reference standard and absence of new peaks are part of the acceptance criteria, so this would not be included in the plan.

It is recommended for CPV, that the routine lot release Size Exclusion Chromatography (SEC) results for percent monomer and aggregates be trended and long-term control (alert) limits defined within their release and stability specifications. The data for these parameters will not form a normal distribution, so control will involve QC review of the results against action and alert limits. Note that samples for SEC testing are collected from the product intermediate prior to bulk freezing and the effect of freezing, storage, and shipping conditions on aggregation should be considered for inclusion in CPV as inputs to the DP process.

Additional CQAs are listed here as optional for inclusion in the CPV plan. Those CQAs that should perhaps be included in trending more often than annually for the Annual Product Review, as a result of relatively frequent manufacture, should be included in CPV. Content of three oligosaccharide structures (sialic acid, mannose and non-glycosylated heavy chain) and two process impurities (DNA and methotrexate) were selected as options for CQAs to be added to CPV trending. As noted earlier, trending of results for two other oligosaccharide structures will be done at step 3. Methotrexate is a raw material used in steps 1 and 2 and there are no specific controls for its removal but since there is a high safety clearance limit for this residual process impurity, testing for it in the BDS is optional.

Shipping of the drug substance has been validated. Monitoring of in-shipment time and maximum temperature during shipment is routinely verified to be within qualified limits. Trending of shipping conditions should be considered for monitoring of the shipping process, though these parameters are considered out of scope for Stage 3 CPV, so they are not included in the plan.

7.12

BULK DRUG SUBSTANCE LOT DATA – CPV RECOMMENDATIONS





Bulk Fill step yield KPA – Include, to establish process consistency

Routine batch document determination using field A280 test result



Filtration volume KPP – Optional, to verify process capability



Maximum (inlet) pressure

KPP – Optional, to establish process consistency

Routine batch documentation, max. inlet pressure from online data acquisition during fill


Filtration time documentation, elapsed time

KPP – Optional, to define normal range

Routine batch Discrete value, univariate



8.1

SCOPE OF CPV ANALYSIS To address the problem of limited data when commercial production starts, it is recommended that CPV analysis is performed in two phases, the initial CPV phase and the long-term CPV phase.

Phase 1: Initial CPV Phase The initial CPV phase is considered pre-SPC and provides the ability to analyze process performance based on a limited data set to gain understanding of the normal process variability in the commercial facility. This phase should include enough batches to provide data to reflect the range of potential variability and allow statistical process ranges to be established. During this phase, charts are run using the specifications based on PPQ, clinical and process characterization information. Data collected will be used to identify possible trends and to demonstrate that the process remains in a state of control. For A-Mab, the initial CPV phase will continue until at least 30 batches have been produced (this is assuming one upstream cell culture batch feeds one downstream purification batch). It is worth noting that, though 30 batches are suggested as the minimum number to form a representative data set, this should not be regarded as a ‘magic number’. Many introductory statistical texts cite 30 as a reasonable start for independent data that

fit approximately the description of a Normal distribution. But, the actual sample size needed to establish variation with a good level of confidence could involve a larger number of batches. It is recommended that a statistician is consulted in the context of a particular data set.

At the conclusion of the initial CPV phase, alert limits for the monitored parameters should be established where applicable, if they do not already exist, or to justify the alert limits that have been set. Additionally, the risk assessment performed following completion of the PPQ batches should be reviewed to determine whether the additional process experience has changed the risk score for the monitored parameters. Trends in process related non-conformances should also be included in the review of the risk assessment, and this should involve considering whether parameters not originally included in the plan for the initial CPV phase ought to be added. Should there be an increase or decrease in risk for the monitored parameters, or a noted non-conformance trend for a parameter which was not previously monitored for CPV analysis, the plan may be revised to reflect the updated process understanding and risk analysis prior to initiation of the long-term CPV phase.

CPV analysis commences with commercial production, following successful completion of the PPQ batches. The start of data collection and analysis begins with the first representative commercial batches produced at the commercial scale facility. Due to potential scale and facility differences, as well as modifications in the process control or adjustments to test methods prior to PPQ, CPV monitoring will not include data from clinical batches, though experience gained in these project phases are likely to help in assigning initial control limits. As a result, the amount of directly relevant data available to set appropriate monitoring limits will be limited at this point. This poses a problem, at least until significant quantities of data have been gathered.

8.0 FREQUENCY AND SCOPE OF CPV ANALYSIS

SECTION 8.0





Monomer (by SEC) CQA – Include, to establish SPC LIMS results from routine testing of BDS


Aggregates (by SEC) CQA – Include, to establish SPC LIMS results from routine testing of BDS


Galactose content CQA – Include, to establish SPC LIMS results from routine testing of BDS

Discrete value

Afucosylation CQA – Include, to establish SPC LIMS results from routine testing of BDS

Discrete value


DNA CQA Include, to establish SPC LIMS results from routine testing of BDS

Discrete value

Methotrexate and/or antifoam C

CQA Include, to establish SPC LIMS results from routine testing of BDS

Discrete value

Sialic acid content CQA Include, to establish SPC LIMS results from routine testing of BDS

Discrete value

Mannose content CQA Include, to establish SPC LIMS results from routine testing of BDS

Discrete value

Non-glycosylated heavy chain

CQA Include, to establish SPC LIMS results from routine testing of BDS

Discrete value



Phase 2: Long-term CPV Phase The long-term CPV phase is the statistical process control phase. This phase has the following objectives:

• Ongoing verification of the process over the lifetime of the product to demonstrate the process remains in control;

• Identify trends which may be within the normal process variability, but indicate a potential to trend outside the alert limits;

• Continue to build understanding of the sources of variability in the process and their impact.

Section 10 provides detail proposing how monitored items fit into the plans for short-term and long-term CPV.

8.2

FREQUENCY OF ANALYSISA documented analysis and conclusion as to whether the process remains in a state of control (a CPV Report) may be performed based on the production schedule. For example, the CPV plan might include the following conditions for a particular product like A-Mab:• Campaign (< 10 batches) – Minimally at the conclusion

of the campaign;• Campaign (> 10 batches) – Minimally every 10

batches, and at the conclusion of the campaign, or at a predetermined time interval (e.g. quarterly);

• Continuous – Minimally every 10 batches, or at a predetermined time interval;

• A frequency preference of every 10 batches has been selected to enable trend identification via typical tests for special causes of variation in control charts. Note that analysis will be performed as described in section 9 per the requirements of the phase of CPV analysis. Frequency of documented analysis and conclusion may be increased when greater than desired process variability is noted or if conclusions are needed to support product disposition.

It is important to note that these statements are given as an example for a product like A-Mab, being manufactured at a frequency of the order of two to ten batches a month. Even so, formal CPV Reports are only likely to be created up to four times a year. For products where the frequency of batch manufacture is low e.g. once a year, it wouldn’t make sense to have more than one CPV Report a year.

To initially establish control limits a documented business process should be in place to address collecting, analysing, reporting and storing of data for the process at the manufacturing scale. Additionally, data generated during development, scale up, as well as small scale data, can be used to set control limits. By evaluating process performance, the initial control limits would help provide an early indication of a lack of control in the process for certain process parameters or quality attributes, by establishing the anticipated range of expected variation. During the evaluation process, such indications may need timely intervention to drive process

consistency. Initial control limits in a CPV plan should not be interpreted as acceptance limits (i.e. a specification for the product).

Through the initial control limits evaluation, the strategy for process control should be identified and applicable limits established based on process and measurement system capability. The Process Capability Index (Cpk) and Process Performance Index (Ppk) provide useful indicators of the level of control likely to be achievable for the process.

Note: Control limits (for parameters and attributes selected for the CPV Plan) need to be established initially. However, they are likely to be re-established at some point and this may require change control (see Section 13). This section focuses on the principles involved in establishing initial and long-term control limits, in preparation for the example of a CPV execution plan (Section 10). More detailed, mathematical considerations are covered in Section 12.

9.0 ESTABLISHING CONTROL LIMITS

SECTION 9.0

Figure 9-1. A Mab case study example indicating different limits during monitoring of certain process parameter


The means of establishing SPC limits for an initial period and for the long-term, is described in detail in Section 12.4. Additionally, an SPC chart showing different limits: upper and lower control limits (UCL, LCL), being the most commonly used and arguably important for many data sets, is presented in the Figure 9-1. If insufficient data are available, either with a new process or after a major process/assay change, initial, temporary limits may be proposed, based on available development, initial small scale data and process knowledge. If so, small scale models should be appropriately developed and qualified in order to guarantee the scale process is representative and predictable.

Statistical and scientific rationale should pre-determine what data set is required. Once sufficient at-scale data are available long-term SPC limits can be established.

Understanding which elements (e.g. raw materials, operators, facility etc.) contribute to common-cause variation may depend on the relevance and knowledge of the specific process, and will help to set relevant and appropriate SPC limits. This requires the inclusion of sufficient data (initially determined or statistically relevant) to capture long-term common-cause variation. Factors that may lead to variation include for example: pack-to-pack variation in chromatography columns, measurement system recalibrations, raw material lot-to-lot variability, etc.

The calculation of control limits depends on an assumption that data is normally distributed and each datum point is independent. This may not always be the case, and data transformation can be helpful in making data meet the

assumptions of normality. Process knowledge may help in transforming data to a more SPC-amenable form. For example when a known and codified relationship exists between process parameters and QAs, normalizing the data (taking into account the available process knowledge) can lead to a more relevant and reliable value for trending.

Data distribution should be considered when selecting analysis tools. The method of reporting each data set should be defined and approved in applicable GMP documentation. All excursions outside approved documentation should be further investigated, justified and documented in appropriate GMP documentation.

When more data are available, calculated SPC limits can be identified. The SPC limits should be periodically reviewed to capture process variability and be brought into line with any new regulatory or quality guidance or additional CPV Plan requirements. Established SPC limits should be reviewed in light of process changes to confirm their continuing validity and may be adjusted in response to generation of additional data. The process monitoring procedure, as well as process capability review, should be established in applicable documentation (e.g. the CPV Plan). With a given frequency of analysis, further statistical examination is required to determine if the results suggest a potential impact on the product. This is described further in Section 10. Multiple data sources and applicable analysis should be organized and integrated in appropriate process data analysis tools. Subsequent statistical tools should be appropriate for the data to be analyzed (see Section 12).

Note: This plan sub-section is neither a minimalistic nor comprehensive listing of variables expected for CPV monitoring. Rather we attempt to maintain a reasonable consistency with the A-Mab case study to provide an example of likely CPV variables associated with a product launch (where in this case, understanding of variability is not evident immediately after completing a platform-based Process Validation (PV) Stage 2 PPQ with only two commercial scale batches of the particular A-Mab molecule). This is meant to demonstrate reliable process control and ability to detect process drift. Commercial scale process data for legacy processes would likely be available and may justify a smaller set of CPV monitored variables. It is emphasized that this is an untested example package for consideration, not general guidance or proven best practice approach.

Continued assurance of consistent process performance and identification of potential out of trend results is achieved by applying SPC rules and capability analysis (Ppk) as discussed in Section 12. CPV datasets enable process capability predictions with higher confidence, deepen process understanding, and improve process robustness by increasing the likelihood of detecting sources of process variability before they cause batch failures. The CS is updated based upon reviews of related risk assessments, as a part of assessing accumulated CPV data in summary CPV Reports. CPV should be integrated into the organization’s development process and quality system. A CPV Master Plan may be used across a corporation, to guide development of product specific CPV procedures including the incorporation of outputs from Stage 1 and Stage 2 (e.g., CQA, CPP). CPV output (from the executed plan) will be documented and summarized at a frequency defined by the plan. Figure 10.1 is a schematic showing the continuity of review in the product lifecycle.

After completion of the PPQ, continued process verification should demonstrate that the process is in control. In the words of the FDA guidance from 2011, it should offer: ‘assurance that the process is reasonably protected against sources of variability that could affect production output, cause supply problems, and negatively affect public health’ [5]. A monitoring and trending program for the A-Mab drug substance process parameters and attributes is outlined in this section, but the reader should view the discussion and content of the tables as recommendations and ensure that the parameters and values they use are appropriate for their product and process. It is applicable for both initial and long-term monitoring of the drug substance manufacturing process. Selection of variables for monitoring is based on information and rationale in Sections 5 and 7.

10.0 EXAMPLE CPV EXECUTION PLAN FOR DRUG SUBSTANCE

SECTION 10.0


Figure 10.1: Continued review as part of the product lifecycle.

Process UnderstandingCPP/CQA’s

Risk Assessment ReviewProcess Knowledge Report

Process AnalysisInitial Process Performance

Evaluation Acceptance & ReleaseOngoing Process Monitoring

CpK Statistics DatabaseAnnual Product Review

ContinuousQuality

Monitoringand Feedback

Process Control Strategy

Batch Record DataSpecifications

ContinuousProcess

ImprovementChange

ManagementDocumentation

ProductQuality

12

34

The A-Mab case study (and assumed post-PPQ CS) contains information on the linkage between product quality attributes and CPPs across the eleven process steps. These relationships and the relationships between KPPs and KPAs are specific to each step. Therefore, this section is divided into tables for each (Steps 1 through 11, plus BDS) in order to list process parameters and attributes to be monitored to verify process control over the entire lifetime of manufacturing the drug substance. Non-routine sampling and specific data gathering will augment routine sampling and data recording to generate the data for trending and monitoring under this CPV plan.

For the tables presented in this section, CPV process variables and their classification are listed in Columns A and B, as recommended earlier in Section 7. Column C includes information on any data treatment required before graphing to monitor trends. ‘Unadjusted’ raw data are measurement results (source data) that are directly charted. ‘Converted’ data indicate that monitoring the process variable involves treatment of measured results and either combining with other process data (e.g. yield is a ratio of combined raw data) or standardizing to match the intent of control limits (e.g. weight measurements converted to volume or converting totalized flow through volume to column volumes). Raw (or converted) data that is mathematically ‘transformed’ is a third type of data treatment that may be required before charting for trend monitoring.

Column D specifies the recommended SPC tool for monitoring performance trends against control limits. The tool listed is selected based on subject matter expert experience with the process development history. The chosen tool provides a means to visually review the data and may be revised when the nature of the data is better understood. The options included in the plan include: ‘individual run chart’ for data without initial control limits; ‘individual measurement chart‘ (a control chart) for data that can be plotted with an expected fixed mean and proposed control limits; ‘EWMA chart’ (Exponentially Weighted Moving Average control chart) (see Section 12 for application); ‘3SD tunnel’ (a control chart with +3 standard deviation control limits) for data that has a dynamically changing mean during the batch processing time (such as a VCC profile, UV chromatogram, or UF/DF flow rate); or ‘exception flag’ which uses routine process monitoring for process parameters and reporting of any out of range result (exception). Any custom correlations

that are developed during CPV would also be shown in this column (e.g. VCC versus dCO2, or step processing time versus step yield).

Column E identifies plan monitoring requirements for an initial short-term CPV period of manufacturing which follows completion of PPQ in order to obtain sufficient data to set long-term control limits (unless limits already exist with sufficient confidence and understanding of the expected long-term normal variation). This period is based on a minimum number of independent batch experiences, e.g. 30 as mentioned in Section 8.1 (with a reminder to consider that raw material lot impact experience may lag process lot experience), or achieving a target Ppk for the variable’s range of control. Proposed initial control limits to use when starting CPV baseline monitoring are given in column F and are based on assumed PPQ criteria for A-Mab, or a fictitious control range proposed after completing the process validation Stage 2 effort. See Section 9 for more information on establishing initial control limits.

Note: Due to various strategies for combining batches in manufacturing, 30 completed batches may not necessarily be sourced from 30 independent vial thaws; or use 30 uniquely prepared lots of the involved solutions; or employ 30 different raw material lots (which could be sub-lots of fewer supplier bulk lots); or may not produce 30 drug substance or drug product batches. Awareness and tracking of different lot counts for different variables is important information during CPV.

Column G is dedicated to the CPV plan after the initial baseline period ends and the frequency of sampling and/or trending is subject to long-term monitoring during commercial manufacturing. At this future point in the CPV lifetime, it is assumed that sizeable specific commercial manufacturing experience and knowledge of the A-Mab process exists and sources of variation are understood well enough to support a reduced frequency of testing or review of data trends with lowered risk for undetected process drift. In some cases it may be possible to gain enough confidence in the behavior of certain parameters and their relationship to the process, that they may be removed from the CPV Plan and only reconsidered for enhanced monitoring to evaluate future process changes. In column H, ‘Initial limits’ is an abbreviated placeholder term for documenting the dates that particular life-time control limits apply which would

Table 10.1. Step 1 CPV variables

A B C D E F G H I

VARIABLE CLASS DATA TREATMENT PRIOR TO ANALYSIS

MONITORINGTOOL

INITIAL BASELINEMONITORING(SHORT-TERM)

INITIAL BASELINECONTROL LIMITS(SHORT-TERM)

PERIODIC MONITORING(TIME/CYCLE-BASED)

LIFETIMECONTROL LIMITS(LONG-TERM)

FOR CAUSEMONITORING(CHANGE-BASED)

Viable Cell Conc., each passage end

KPA Unadjusted (raw data)

Individual chart with long-term 3-sigma limits

Every batch until long-term limits set

0.7 to 2.8 x 106 (vc/mL)

Every batch LT range1 TBD (from date X to Y) LT range2 (from date Y to Z)

Cell bank or growth medium changes

OPTIONAL ELEMENTS

Initial VCC split ratio, each passage

KPP Converted (ratio)

Individual Run chart


Char-acterize (No PPQ limits)

Once annu-ally

LT Period1 Range1 TBD LT Period2 Range2


Culture duration, each passage

KPP Raw data Individual chart with long-term 3-sigma limits


3 to 4 (days)

While PpK < 1.0, Oth-erwise not required

LT range TBD (dates: TBD)


Culture viability, each passage end

KPA Converted (ratio)

EWMA chart Every batch until long-term limits set

88 to 98 (%)

not required LT range TBD (from date X to Y)



be documented as ‘Range1’. This information would be populated after the initial baseline monitoring is complete and short-term control limits are superseded by long-term limits. Long-term limits and ‘Range2’ are included in the early tables to demonstrate the historical nature of CPV lifecycle management. Control limits may change at a given time for a particular justified reason (such as appearance of very long-term variation or change factors), and past data profiles should not necessarily be assessed (or displayed) against more recent control limits. However, the ability to review historical data ranges along with changes in more recent predecessor control limits can enhance process understanding over the product lifespan, especially if these changes are associated with a set of diagnosed root causes.

Collection of data may at some point provide sufficient demonstration of control of a variable that it may be removed from the long-term monitoring plan, or that the frequency with which a particular variable is monitored can be reduced to an occasional (audit) basis. Examples of variables that might be removed from long-term monitoring include CPPs that have been identified as being well-controlled (WC-CPP). Initial (short-term) monitoring data may verify that the expected control is routinely achieved, and that there are select CQAs for which monitoring data shows little variability. Responding to signals in the data in this way allows adjustment of the long-term monitoring plan to tailor it for monitoring those elements of the process most likely to exhibit variability and hence need the greatest attention.

Besides time-based risks to maintaining the validated state of process control, the other type of risk that requires verification and monitoring involves change-based risks.

These assumed known ‘for-cause’ events are shown in column I and new SME knowledge can be added as it is gained. These changes (e.g. critical raw material lot changes or process improvements) may have an impact that extends beyond the change implementation and may make previous data and trend characteristics obsolete and invalidate previous short-term or long-term control limits. When available, collected monitoring data should be provided with the resolution recorded in its raw data form, rather than reflecting any rounding to the significant figures included in the control limits. This enables more accurate statistical analysis and determination of capability.

Note: Situations that would result in duplicating information across a table are occasionally presented with alternative proposals to offer the reader different options to consider for CPV. Various charting options are presented as examples and different life-time plans shown for the variables. Rationale would be subject to SME justification for each individual variable, and as not shown below, may actually result in the same monitoring tool and life-time monitoring plan.

Note: Since the contents of column H are subject to more frequent updates than the other plan elements, the reader could consider migrating or referencing the column H lifetime control limits for each variable (as they become available) in a separate document for efficient review and approval of revised ranges to maintain both the historical control ranges with current control ranges.

10.1

STEP 1, SEED CULTURE EXPANSION IN DISPOSABLE VESSELS – CPV VARIABLES

The CPV plan for process variables in this step are shown in the following table and align with the rationale in Section 7. The qualification of new cell banks is beyond the scope of this CPV plan and subject to change control management by registered regulatory agreements. Monitoring and verification of the commercial scale impact of a change in a medium component presented earlier are included in

this example. In this example, it was assumed the split ratio for this step did not have PPQ acceptance criteria (VCC, % viability, and duration ranges employed in PPQ) nor was there sufficient process data from the A-Mab working cell bank to adopt initial CPV control limits.


A B C D E F G H I


MONITORINGTOOL






Culture duration CPP Unadjusted (x.x resolution)



16 to 18 (days)

Every Batch LT range TBD (dates: TBD)

Change in cell bank, culture medium, or process setpoint

Maximum pCO2 CPP Unadjusted (raw data)


Every batch until long-term limits set, also correlate w/ IVCA

45 to 140 (mmHg)




A B C D E F G H I


MONITORINGTOOL






Viable Cell Conc., each passage end




3.9 to 6.0 x 106(vc/mL)

Every batch LT range TBD(from date X to Y)


Optional elements

Initial VCC split ratio, each passage

KPP Converted (ratio)



3.0 to 4.1 Once annually

LT range TBD(from date X to Y)


Culture duration, each passage

KPP Unadjusted (x.x resolution)



3 to 5 (days)

While PpK < 1.0, Otherwise not required

LT range TBD(from date X to Y)


Culture viability, each passage end

KPA Converted (ratio)



90 to 99 (%)

not required LT range TBD(from date X to Y)



The CPV plan for process variables in this step is shown in the following table and follows the rationale in Section 7. Monitoring and verification of the commercial scale impact of a change in a medium component presented earlier is

included in this example. It was assumed the cell culture split ratio for this step had sufficient data for the expected bioreactor expansion performance to adopt initial control limits.

10.2

STEP 2, SEED CULTURE EXPANSION IN BIOREACTORS – CPV VARIABLES

To be consistent with the status of the A-mab case study when this plan is initiated, the option of creating a time-dependent multivariate PLS (partial least squares) bioreactor model is a CPV objective, based on previous successful experiences [25]. The A-mab case study [3, Section 3.10, Page 107-109] describes a principle components bioreactor model as a predictor of acceptable oligosaccharide structure and acidic variant CQAs. The parameter inputs to the model include temperature and pH, and attribute inputs to the model include titer, VCC, and viability. All these variables are included individually for CPV monitoring while the bioreactor model is qualified. The output of the model for each batch is classified as a KPA. The potential added value in using the model is in ensuring internal correlations among different variables are considered. In the future, the values generated from this model may provide a multivariate output for trend monitoring that is predictive of process performance, with its own alert and action limits.

A fixed volume of nutrient feed is added at a defined time under automation and routine batch document controls, therefore trending the amount and timing of the nutrient feed addition does not provide value because they are not subject to random variation. Production cultures are harvested within an acceptable duration based on viability and titer considerations.

Temperature and pH are continuously feedback controlled to set points, during the 16 to 18 day culture but measured values are dynamic over that time period. Therefore, 3SD tunnels (the range defined by the mean + 3 standard deviations) for the parameter profiles will be developed during the initial CPV period, to generate an expected ’conduit’ for results when tracking consistent control of the CPP.

10.3

STEP 3, PRODUCTION CULTURE BIOREACTOR – CPV VARIABLES

A B C D E F G H I


MONITORINGTOOL






OPTIONAL ELEMENTS

Antifoam lot CMA Unadjusted(raw data)

Exception Flag (new lot)

Test BDS clearance for 3 different lots

< LOD Not required < LOD Change in material ID or supplier

Medium osmolality

CPP Unadjusted(raw data)


Track OOR Exception flags

365 to 435(mOsm)


375 to 425 mOsm(dates: current)

Change in medium prep

Mannose content

CQA Unadjusted(raw data)


Required for model qualification only

5 to 8(%)

Not required at this time

LT range TBD(dates: TBD)

Changes in cell bank or step 3 setpoints

Sialic acid content




NMT 1.6(%)




Non-glycosylatedheavy chain




0 to 2.4(%)




Time of glucose feeds(hrs since inocula-tion

KPP Unadjusted(raw data)


Every batch untillong-term limits set

Feed 1: a to b hrsFeed 2: c to d hrsFeed 3: e to f hrs




Peak VCC(Viable Cell Conc.)

KPA Unadjusted(raw data)


Every batch untilPLS qualified

20 to 30 x 106(vc/mL)

Not required, included in PLS model

See PLS model,LT range TBD(dates: TBD)


Culture viabilityat Harvest

KPA Converted(ratio)


Every batch untilPLS qualified

40 to 61(%)


See PLS model,LT range TBD(dates: TBD


A B C D E F G H I

VARIABLE CLASS DATA TREAT-MENT PRIOR TO ANALYSIS

MONITORINGTOOL

INITIAL BASE-LINEMONITORING(SHORT-TERM)





Culture duration CPP Unadjusted (x.x resolution)



16 to 18 (days)



Bioreactor pH CPP Unadjusted (raw data)

3SD tunnel Every batch until long-term limits set for reference

6.75 to 6.95 (-log [H+])


See PLS model, LT range TBD (dates: TBD)


Afucosylated glycans CQA Unadjusted (raw results)



5 to 10 (%)

Once annually (time 0 of annual stability batch)

LT range 5 to 10 Initial to current date


Galactosylated glycans

CQA Unadjusted (raw results)



15 to 35 (%)

Once annually (time 0 of annual stability batch)

LT range 15 to 35 Initial to current date


PLS model employ-ing pH, DO, tem-perature, pressure, gas rates, weight, VCC, viability, titer, glucose, lactate

KPA Converted & Transformed

CustomPLS model PCA t1

Every batch until model is > 95% predictive

Trajectoryversus time ± 3 StDev

Every Batch LT range TBD(dates: TBD


Product yield (titer) at Harvest



Every batch untillong-term limits setand PpK >1.0

4.0 to 5.5(g/L)


See PLS model,LT range TBD(dates: TBD)




A B C D E F G H I


MONITORINGTOOL






Protein Load Ratio (in HCP model)Each sub-batch

CPP Converted(ratio)



15 to 40(g A-mAb/L resin)

Every Batch LT range TBD(dates: TBD)

Resin, or processchange to this or previ-ous step

Elution buffer pH (in HCP model)

CPP Unadjusted (x.xx resolu-tion)


Every batch until HCP model limits set

3.4 to 3.8(-log [H+])


LT range 3.4 to 3.8(dates: initial to current)

Buffer formu-lationscale-up

Residual Protein Ain eluate pool


w/ upper limit, & correlationvs. storage age


≤ 1234(ng/mg A-mAb)

Per resin/column lifetime protocol


First two cycles afterextended storage(≥ 3 months)

Step duration KPP Converted (elapsed)



Charac-terize(No PPQ limits)

Once annu-ally

LT range TBD(dates: TBD

Scale in-creases

Step yield KPA Converted(ratio)



68 to 88(%)


Reset range for process change tothis or previ-ous step


A B C D E F G H I


MONITORINGTOOL






Turbidity of filtrate




< 2(NTU)

Not required LT range TBD(dates: TBD)

CRM or pro-cess change to this or previous step

Step yieldFiltrate





Every batch LT range TBD(dates: TBD)

CRM or pro-cess change to this or previous step

OPTIONAL ELEMENTS

Duration of Broth Clarification

KPP Converted (end minus start time)




Once annu-ally


Upstream scale-up,Centrifuge feed or flowrate setpoint changes

Inlet pressure, depth filters



Not required None Not required ± 3 StDev of mostrecent 30 batches(pre-change)

Upstream scale-up, filter changes, Centrifuge feed, or filter flow rate setpoint changes, shifts in turbidity or filtration time.


For Step 4, Critical Raw Materials (CRM) that impact the step include the working cell bank, upstream growth medium components, and depth filters. Changes to these items would require a period of enhanced monitoring of filtrate turbidity and step yield, to demonstrate that the process can

attenuate upstream process variability prior to purification. If turbidity or the duration of depth filtration shifts upward, monitoring the inlet pressure parameter or the attribute of differential pressure across the filters may need to be added to CPV.

10.4

STEP 4, CENTRIFUGATION AND DEPTH FILTRATION – CPV VARIABLES

Initial control limits for load ratio and elution buffer pH are assumed, based on a simulated normal range within the DOE PAR provided in the A-Mab case study. Likewise, the short-term residual protein A control limits are based on the outcome of a fictional study done after PPQ at small scale, which examined the capacity for high cycle count AEX

resin to clear protein A (spiking study) and, characterization of potential Protein A leachate from both high and low use cycle counts with respect to resin storage time. Note that elution buffer pH data is included while the limits for the HCP model are generated.

10.5

STEP 5, PROTEIN A CHROMATOGRAPHY – CPV VARIABLES

x.xx


The viral safety risk CQA (inactivation of particular AVA) for the A-Mab process has been validated in the small scale model during Stage 1 process validation. Initial control limits for inactivation time and pH are assumed, based on a fictitious normal range within the DOE PAR provided in the A-Mab case study. 3SD tunnel monitoring of pH during inactivation ensures the parameter remains in range throughout the inactivation time and it is not monitored as a single point measurement. Aggregate results are assumed

to include a sum of all quantifiable non-monomer SEC peaks (dimer, trimer, etc). Yield is not expected to be impacted by this step and variability around 100% has been associated with measurement uncertainty, therefore it does not merit CPV trending or monitoring. The basis of yield for the next step (7, CEX) begins from the eluate pool of the previous step (5, Protein A).

10.6

STEP 6, LOW PH TREATMENT – CPV VARIABLES


A B C D E F G H I


MONITORINGTOOL






pH during inactivation


3SD tunnel Every batch until long-term tunnel set

3.4 to 3.8 (-log [H+])


Scale in-creases

Post inactivation aggregates

CQA Converted(additive)


Every batch until long-term limits end and PpK >1.3

≤ 3.0(%)

Once annu-ally


Process changeto this or previous step

OPTIONAL ELEMENTS

Inactivation time

CPP Converted (elapsed)


Every batch until along-term sigma set

80 to 120(minutes)



No known risk events

Quantity of acid added





Not Re-quired


Scale in-creases

Initial control limits provided in the table below are assumed, based on a fictitious normal range within the DOE PAR provided in the A-Mab case study. Step duration has a batch document control limit based on platform experience, but for

A-Mab CPV a typical range will be determined for historical reference. Aggregate results are assumed to include all quantifiable non-monomer SEC peaks (dimer, trimer, etc).

10.7

STEP 7, CATION EXCHANGE CHROMATOGRAPHY – CPV VARIABLES


A B C D E F G H I


MONITORINGTOOL






Protein Load Ratio (in HCP model)







Wash Conductivity(in HCP model)

CPP Unadjusted (x.x resolution)



4 to 6(mS/cm)



Buffer formulationscale-up

Elution pH CPP Unadjusted (x.xx resolution)

3SD tunnel Every batch untillong-term tunnel set

5.9 to 6.1(-log [H+])


Process change tothis or previous step

Aggregates in CEX eluate pool



Every batch until long term limits set and PpK >1.3

≤ 1.0(%)

Per resin/ column lifetime protocol



HCP contentin CEX eluate pool




≤ 130(ppm)

Per resin/ column lifetime protocol



CEX eluate volume (each sub-batch)




3.7 to 4.7(CV)

Not Required


Column pack or repack

Step yield KPA Converted(ratio)



83 to 97(%)


Process change to step

OPTIONAL ELEMENTS

Step duration KPP Converted (elapsed

IndividualRun chart



Not Re-quired


Process change to step

x.xx

A B C D E F G H I


MONITORINGTOOL






HCP content in AEX eluate(measured)




≤ 15(ng/mg)

Per resin column Lifetime protocol


Process change to any chromatogra-phy step

Residual Protein Ain eluate



Not Required ≤ 10(ng/mg A-mAb)

Per resin columnLifetime program


Align moni-toring with ProA process changes

Step yield CQA Converted(ratio)



≥ 87(%)

Trend every batch vs. long-term limits


Reset range for Process change to step




Feed 1: a to b hrsFeed 2: c to d hrsFeed 3: e to f hrs




OPTIONAL ELEMENTS

KPP Converted (elapsed)




Not Required


Reset range for Process change to step


A B C D E F G H I


MONITORINGTOOL






Protein Load Ratio







Load Conductivity CPP Unadjusted(raw data)



4.0 to 7.0(mS/cm)



Load pH(in HCP model)

CPP Unadjusted (x.xx resolution)



7.4 to 7.7(-log [H+])



Process change toprevious step

Equilibration/ Wash 1 Buffer Conductiv-ity (in HCP model)




2.0 to 3.2(mS/cm)



Buffer formu-lationscale-up

Linkage model output for HCP (predicted)

KPA TransformedA-mAb case studymodel equa-tion 6(Pg 158)


Every batch until AEX eluate HCP is predictive with 95% confidence

99.5% predic-tioninterval for meanof HCP output


Re-qualify model for process changes in ProA, CEX, or AEX steps

BPOG CONTINUED PROCESS VERIFICATION: AN INDUSTRY POSITION PAPER WITH EXAMPLE PLAN – Page 67Page 66 – BPOG CONTINUED PROCESS VERIFICATION: AN INDUSTRY POSITION PAPER WITH EXAMPLE PLAN

Initial control limits provided in the table below are assumed, based on a fictitious normal range within a DOE PAR provided in the A-Mab case study. The CEX eluate is adjusted to a target pH and conductivity (CPPs) before loading for flow-through mode chromatography. The fictitious HCP

clearance linkage model for the three chromatography steps (Protein A, CEX, and AEX) is included here for monitoring the trend in the algorithm output, using the 6 parameters (two from each step) for a given batch.

10.8

STEP 8, ANION EXCHANGE CHROMATOGRAPHY – CPV VARIABLES

x.xx


A B C D E F G H I


MONITORINGTOOL






OperatingInlet pressure


3SD tunnel Every batch untillong-term limits set

Tunnel derived from PPQ and platform batches at same scale




Filtration loadvolume




640 to 1040(L)

Not Re-quired




3SD tunnel monitoring of the operating inlet pressure ensures the parameter (and filtration flow rate) remains consistent throughout the filtration; it is not monitored as a single point measurement. Initial control limits provided in the table below are assumed based on a fictitious normal

range within a DOE PAR, provided in the A-Mab case study. Platform knowledge of SVRF operations indicates this step does not impact yield, therefore consideration is limited to any future qualification of re-filtration.

10.9

STEP 9, SMALL VIRUS RETENTIVE FILTRATION – CPV VARIABLES

To address the risk that future incoming buffer component lots test within specification, but outside a typical experience to cause an undesired or unpredicted variation not controlled by the process, we propose monitoring a trend of the CoA reported attribute for at least 30 batches from each approved supplier (manufacturer/ packager/ distributor/ vendor) as an optional item (see ‘Excipient 1’ in Table 10.10). If the actual range for this material attribute does not exhibit a Ppk > 1.3 against its ICH Q3D limit, further monitoring and control actions may be necessary. The initial control limits provided, are based on fictitious normal ranges to serve as early warning triggers. Buffer pH, osmolality, and conductivity are controlled (and adjusted as necessary), before the solutions are released for use in

this process step. The A-Mab case study includes IPC tests for pH and osmolality in drug product manufacturing, but does not include BDS specifications for these attributes. It will be assumed, based on the case study information, that the process stream after UF/DF (rather than after Step 11) is subject to probe monitoring for pH (PAT), to show performance prior to final filtration, as an input to the drug product CPV effort (which occurs after BDS storage, shipping, and thawing).

TMP is a design parameter with a fixed setpoint even as membrane use cycles increase. The permeate flow output attribute varies and the input retentate flow parameter responds to achieve and maintain the TMP setpoint.

10.10

STEP 10, ULTRAFILTRATION AND DIAFILTRATION – CPV VARIABLES


A B C D E F G H I


MONITORINGTOOL






UF/DF processing time

CPP Converted(elapsed)



≤ 7(hours)



Scale or processchange to step

Number ofdia-volumes




9.5 t 10.4(DV)

Not Required


Buffer formulationor protein conc. changes

UF/DF retentate final pH




5.2 to 5.4(-log [H+])

Once annually


Step 10 or 11 process or raw material changes

Protein Conc. A280Prior to BDS fill step




65 to 85(mg/mL)



Process changeto this step

Product yield,Final Retentate




Per mem-branelifetime program



When extending re-use cycles

OPTIONAL ELEMENTS

Excipient1AttributeA

CMA Unadjusted(raw data)

IndividualRun chart

Each new lot until 30 lots tested


Not Required


Qualify new suppliers,RM or CoA changes

SEC aggregates in final retentate




≤ 1.4(%)

Not Required


Process change to this step, extensions of membrane lifetime

Protein Conc. A280Prior to diafiltration

KPP Unadjusted(raw data



40 to 60(mg/mL)

Not Required



Recirculation flow rate



Profile within tunnel limits

Not Required



Permeateflow rate



Profile within tunnel limits

Verify when new membranes installed




A B C D E F G H I


MONITORINGTOOL






Bulk FillStep yield




≥ 98(%)

Once annually



OPTIONAL ELEMENTS

Filtration volume




200 to 300(L)



Reset range for scale orprocess change to thisor previous step

Maximum Inlet pressure




≤ 2(psig)

Not Required



Filtrationtime

KPP Converted (elapsed)



≤ 3(hours)



Scale or processchange to step

The initial control limits provided in the next table are assumed, based on a fictitious normal range and PPQ

consistency limits, or in the case of yield, an assumed limit from fictitious platform knowledge.


10.11

STEP 11, FINAL FILTRATION/BULK FILL AND FREEZING OF BDS – CPV VARIABLES

As with the process steps, BR depends on a rational assessment of lot data. This is shown in the next table.

10.12

CPV MONITORING OF BULK DRUG SUBSTANCE LOT DATA

Table 10.12. CPV variables for BDS lot release data

A B C D E F G H I


MONITORINGTOOL






Monomerby SEC

CQA Peak integration(raw data



NLT 98(%)



Per BDS stability monitoring protocol

Aggregatesby SEC




NMT 2(%)



Per BDS stability monitoring protocol

GalactosylatedGlycans




15 to 35(%)

Not required, Not stability indicating


Confirm comparability for changes in cell bank or step 3 setpoints

Afucosylated Glycans




5 to 10(%)




A B C D E F G H I


MONITORINGTOOL






OPTIONAL ELEMENTS

DNA CQA Unadjusted(raw data)


Every batch for 30 BDS batches

None detected

Per column resin Lifetime protocols



Methotrexate and/or Antifoam C



Every batch for 30 BDS batches

None detected



One BDS batch for 3 raw material lots

Sialic acid Content




NMT 1.6(%)



Confirm comparability forcell bank or process changes

Mannose Content




5 to 8(%)



Confirm comparability forcell bank or process changes

Non-glycosylatedheavy chain




0 to 2.4(%)





Considerations for non-routine sampling and testing include: • Sample plan for routine monitoring, baseline

monitoring, time-based periodic monitoring, or special event / change based monitoring;

• Sample frequency (e.g. every day during a reactor run, every batch, or every 5 batches, etc);

• Specific sampling location within the process. Provide specific and clear instructions for collection of the required sample, e.g. BRc, BRc step number, sampling device such as manual sample valve or automated sample valve. It is critical to ensure that samples collected at the selected sampling points are representative of the drug substance or process intermediates;

• Sample container and container size (e.g. polypropylene tubes);

• Sample volume, number of aliquots, and retains;• Sample labelling (may be driven by site SOP);• Sample storage temperature and transport conditions;• Tests to be performed (including any additional

sampling due to assay needs (e.g. a blank solution as reference for the test);

• Testing acceptance criteria.

A sampling plan makes it clear which samples need to be taken to meet the data acquisition requirements of CPV. This section shows how this might be done giving examples and using templates. A sampling plan matrix is presented in Table 11.0.1. Please note, this table is not intended to be complete, or entirely consistent with information contained previously. It is an example of a visual tool that can usefully summarize an otherwise complex plan, making it easier to develop and communicate.

11.0 CPV SAMPLING PLAN

SECTION 11.0


Tabl

e 11

.0.1

. Sam

plin

g Pl

an M

atrix

SAM

PLIN

G P

LAN

FO

R PR

OCE

SS A

TTRI

BUTE

S

ON

-FLO

OR

TEST

SQ

C M

ICRO

PRO

DUCT

QUA

LITY

TES

TIN

G

RETAIN

A280 CONC

PH

CONDUCTIVITY

BIOBURDEN

ENDOTOXON

RP-HPLC

CIEF

CGE

SEC

HCP

PRODUCT CONC.

BIOASSAY

DNA

AFFINITY LIGAND

METHOTREXATE

ANTIFOAM

TRUNCATED IMPURITY

PRODUCT VARIANT

Vol r

eq'd

for a

ssay

Stor

age

tem

p<

-40

°C

Test

ing

offs

ite

Test

ing

onsi

te

Met

hod

SOP

Stag

e 1:

Cel

l Exp

ansi

on

Afte

r las

t ope

n m

anip

ulat

ion

prio

r to

tran

sfer

to p

rodu

ctio

n st

age

Stag

e 2:

Pro

duct

Exp

ress

ion

EOR

prio

r to

harv

est

Stag

e 3:

Cla

rifica

tion

Cent

rate

Clar

ified

filte

red

brot

h•

Step

4: A

ffini

ty c

hrom

atog

raph

y

Load

pos

t hol

d pe

riod

Load

flow

-thr

ough

Elut

ion

pool

••

•ο

οο

οο

ο

Strip

flow

thru

ON

-FLO

OR

TEST

SQ

C M

ICRO

PRO

DUCT

QUA

LITY

TES

TIN

G

RETAIN

A280 CONC

PH

CONDUCTIVITY

BIOBURDEN

ENDOTOXON

RP-HPLC

CIEF

CGE

SEC

HCP

PRODUCT CONC.

BIOASSAY

DNA

AFFINITY LIGAND

METHOTREXATE

ANTIFOAM

TRUNCATED IMPURITY

PRODUCT VARIANT

Step

5: C

EX c

hrom

atog

raph

y

Load

pos

t hol

d pe

riod

Load

flow

-thr

ough

Elut

ion

pool

••

•ο

ο

Strip

flow

thru

Step

6: I

EX c

hrom

atog

raph

y

Load

pos

t hol

d pe

riod

Load

flow

-thr

ough

Elut

ion

pool

••

•ο

οο

Strip

flow

thru

Step

7: V

iral F

iltra

tion

Non

e

Step

9:

UFD

F

Load

poo

l pos

t hol

d pe

riod

Perm

eate

dur

ing

conc

entr

atio

n

Rete

ntat

e po

ol p

ost d

iafil

trat

ion

••

ο

Step

10:

BD

S fil

trat

ion

an

d Fr

eezi

ng

Post

hol

d pe

riod,

pr

ior t

o fil

ter

BDS

sam

ple

prio

r to

free

ze•

••

••

••

Sym

bols

are

use

dto

indi

cate

whi

chs

ampl

ing

isre

leva

ntto

eac

hA-

Mab

pro

cess

ste

pso

the

inte

ntio

nof

sam

plin

gis

cle

ar:•

(Rou

tine

test

),ο

(CPV

test

), ∆

(Re

tain

),

(Sta

bilit

y),

(R

euse

Life

time

Perf

orm

ance

),

(Cle

anin

g Ve

rifica

tion)


For each sampling activity, it can be helpful to have a breakout template which shows what needs to occur (thus a symbol on Table 11.0.1 may become a template table in itself. Examples of templates for sample collection and testing are presented (Table 11.1.1 and 11.1.2). The rows represent the considerations for sampling and testing and suggested alternatives. The yellow highlighted cells in Table 11.1.1 are the choices for the peak viable cell concentration tested at Step 3 in the process (Production Culture Bioreactor. In Table 11.1.2, the yellow highlighted cells relate to aggregates tested at process Step 7, CEX. It is worth noting that, though justification for the choices made is not given fully in this example, justification would be expected as part of the CPV plan.

Note: Whilst a sampling template may be regarded as good practice, it is not mandatory. Readers may feel it is worth creating such plans for CPV as a priority and extending the

activity to elements of process monitoring not included in CPV on a risk managed basis.

11.1

TEMPLATE FOR SPECIFIC PROCESS STEPS

Table 11.1.1. Template for Sampling and Testing (CPP). Completed example is for Step 3, Production Culture Bioreactor

OPTIONS

Process Step

Step 3: Production Culture Bioreactor

Variable Peak Viable Cell Conc.

Classification (Product Quality Attribute Or Process Parameter)

CQA CPP KPP KPA

Assay MethodCapacitance Probe, Per Sop

Sample PlanRoutine Monitoring

Baseline Monitoring

Time-Based Periodic Monitoring

Special Event Or Change Based Monitoring

Sample Frequency

Multiple Times In A Batch (E.G., Every Day For Bioreactor)

Once Per Batch Every X Batches

Sample Collection

Sample Location Bioreactor

Sampling DeviceManual Sample Valve

Automated Sample Valve

Automated Sampling Device

In-Line Sensor

Container Materials Of Construction (Moc)

Pp Tube, SterileNo Sample - On-Line Instrument

Container Size 10 Ml N/A

Sample Volume 5 Ml

Sample Replicates

No 2

Sample Retain No Yes

Sample Handling No AliquotedPrepared For Shipping

Sample Labeling Driven By Sop N/A

Sample Storage Temperature Ambient 2-8 C -20 C -70 C N/A

Location Manufacturing Qc Lab Development Lab Sample Control N/A

Sample Transportation

Transport Yes (Per Sop) No

Assay Testing Who Manufacturing Qc DeptDevelopment Dept

External LabIn-Situ (On-Line Sensor)

Reference/Blank Reference Blank Solution Control No

Table 11.1.2. Template for Sampling and Testing (CQA). Completed example is for Step 7, Cation Exchange Chromatography (CEX)

OPTIONS

Process StepStep 7: Cation Exchange Chrom.

Variable Aggregation

Classification (Product Quality Attribute Or Process Parameter)

CQA CPP KPP KPA

Assay Method SEC

Sample PlanRoutine Monitoring

Baseline Monitoring

Time-Based Periodic Monitoring

Special Event Or Change Based Monitoring

Sample Frequency

Multiple Times In A Batch (E.G., Every Day For Bioreactor)

Once Per Batch Every X Batches

Sample Collection

Sample Location

After Eluate Is Well Mixed At The Eluate Collection Tank

Sampling DeviceManual Sample Valve

Automated Sample Valve

Automated Sampling Device

In-Line Sensor

Container Moc Pp Tube, SterileNo Sample - On-Line Instrument

Container Size 10 ml

Sample Volume 5 ml


OPTIONS

Sample Collection

Sample Replicates

No 2

Sample Retain No Yes

Sample Handling

No AliquottedPrepared For Shipping

Sample Labeling

Driven By Sop

Sample Storage Temperature Ambient 2-8 C -20 C -70 C N/A

Location Manufacturing Qc LabDevelopment Lab

Sample Control N/A

Sample Transportation

Transport Yes (Per Sop) No

Assay Testing Who Manufacturing Qc DeptDevelopment Dept

External LabIn-situ (on-line sensor)

Reference/Blank

Reference Blank Solution Control No

BPOG Continued Process Verification: An Industry Position Paper With Example Plan – Page 81

Fundamentally, the aim of setting values for capability criteria and analyzing data against those criteria is to establish a plan that provides sound rationale for decision making. This section provides a basic introduction to the statistical tools likely to be used in support of analytical elements of a CPV plan.

12.0 HOW DATA WILL BE ANALYZED

Statistics is a mature though complex mathematical discipline and it is recommended that specialists are consulted when applying the approaches described in this section. A number of references are provided, but given the maturity of the subject, the authors recognize this list is far from exhaustive. The focus here is on recent regulatory guidance and what might be regarded as a few useful, standard texts.

Note: In this paper we are describing some ways in which data can be analysed, but these are not intended to be exclusive of any others. Other analytical methods may be better suited to particular data sets.

CPV Reports focus on quantitative attributes and parameters which can be monitored using statistical process control charts, to identify shifts, trends, and unusual results. Typical practice is that the attributes and parameters in a CPV Report include all the quantitative attributes and parameters which are included in the APR. CPV Reports can also include additional attributes or parameters, which are useful for building process understanding and identifying sources of variation. Whilst the scope of quantitative attributes included in CPV is equal to, or broader than, the scope included in APR, the APR includes additional sections not required for CPV. The APR will provide comprehensive assessments of product performance that are only made once a year.

SECTION 12.0

Page 80 – BPOG Continued Process Verification: An Industry Position Paper With Example Plan

Software supporting CPV should make it as easy as possible for CPV reports to be consistent with the APR, in order to avoid redundant effort by the technical and quality staff. The charts of quantitative results included in CPV should meet all requirements of the APR, such that the CPV charts may be copied and pasted directly into the APR or included as attachments. The APR would provide context and tie together the information brought forward from the CPV Reports.

The CPV analysis may be performed using well established software such as MINITAB, JMP, Discoverant or Statistica. The software vendor should provide documentation of their quality and validation procedures if data is used to support GMP functions [18-20]. ‘Homegrown’ analysis routines should be similarly validated if used to support GMP functions. If specific calculations outside the base software packages are used to support GMP functions, they should be validated also.

12.1

IDENTIFYING SOFTWARE

Tools to trend and assist in the evaluation of CPV data include types of charts and mathematical treatments. The rationale and approaches to evaluation should be documented in a standardized way. The three key documents are a Risk Assessment justification, CPV Plan and CPV Report. Process Risk Assessments that support CPV plan development and implementation have the important purpose of justifying the scope and frequency with which CPV reports are required. They are performed at the start of CPV plan preparation. For new processes they should draw on the outcomes of PPQ and it is recommended they also take into account a Process Failure Mode Effects Analysis (FMEA), which may have been conducted during Stage 1 process design and updated after the Stage 2 PPQ experience.

For legacy products, risk assessments should include the following sets of data: process capability, analysis of campaign trends, historical causes of discards, customer complaints and failure investigations.

A CPV Plan should be written with the purpose of specifying what must be monitored to provide for CPV and how data should be interpreted. Interpretation should involve: how data will be collected, transformed, and evaluated. It is strongly recommended that data driven rules by which

decisions will be made, and the expected outcomes, are recorded in the Plan. A CPV plan should be created and issued following the risk assessment and before the start of Stage 3. As stated previously, the definition of an initial phase and long-term phase of CPV Plan development may be helpful. This approach is designed to ensure that variation in process performance is understood before long-term control limits are established in the control strategy.

Regular assessments of process performance should be documented in CPV Reports. The frequency with which these are created will depend on the assessment of risk as described previously. CPV Reports should detail any control alerts, as defined by the CPV Plan, whether these have led to a formal Quality investigation or not. Justification supporting the response made to these alerts should be given, including any outcomes for the control strategy. Typically, CPV Reports will contain calculations of process capability, statistical process control (SPC) charts, any other chosen charts, control limits and alerts arising from the data.

In combination, these tools help make any changes in the performance of the process obvious, revealing any non-random patterns. The following sections expand on these tools.

12.2

DESCRIPTION OF TOOLS TO TREND AND EVALUATE DATA


Process capability assessment evaluates the risk that an attribute will fail to meet specifications; in other words, it quantifies the likelihood that an attribute will routinely meet specifications.

Any assessment of process capability requires the assumption that the same sources of variation that affected previous results will continue to affect future results, and the expected range of variability does not change.

There are many indices that measure process capability, but two are especially popular: Ppk and Cpk . Both indices compare the width of the specification range to the width of the typical variation range. The key difference is that Cpk uses a short-term estimate of variation, whereas Ppk uses a long-term estimate of variation [13, Chapter 7]. These indices take into account the centering of the process within the specification range, and higher values of either index indicates higher process capability (or lower risk of missing specifications).

The short-term estimate of sigma is a best-case value that represents the minimum variation that could be achieved if all longer-term sources of variation were eliminated from the process. The long-term estimate of sigma includes both the short-term and the longer-term sources of variation. For many manufacturing processes, these longer-term sources of variation are an expected part of the total, routine process variation.

Cpk provides a more optimistic estimate of the potential process capability if all longer-term sources of variation are removed; Ppk provides a more realistic estimate of the process capability that has been achieved in routine production. For this reason, Ppk is preferred here over Cpk as an indicator of expected process capability; although either measure could be used, depending on process circumstances. Table 12.3 contains guidelines rating the level of control over the process based on Ppk values. In this paper, we recommend

an initial period in which control limits are estimated from Stage 1 and Stage 2 and experience with similar processes. Given the nature of Ppk, it becomes most useful as actual manufacturing experience develops and robust control limits are established for the long term.

In cases where longer-term variation exceeds short-term variation, and the longer-term shifts cannot be tolerated, Cpk can be used. However, Cpk should not be calculated until the SPC charts provide evidence that the process is in a state of statistical control, such that no signals of non-random variation are present. This means there should be no evidence of shifts, trends, or results outside of control limits. This is a prerequisite to calculate a meaningful Cpk. Ppk can be calculated even when some non-random signals are present, as long as those signals are considered to be part of the routine, expected, longer-term variation inherent to a process.

The index also assumes the data follow a normal distribution. Transformations are sometimes needed to enable non-normal data to be analyzed in a rigorous statistical manner. A common source of non-normal data is where negative values cannot arise, and the most frequent values are close to zero. Such data sets may be termed ‘log-normal’ and taking the logarithm of the data can transform it into a normal distribution. There is more discussion on this topic in Section 12.4 that follows, but in the context of this paper, we refer the reader to Reference 13, recognize there are many other texts on this topic and recommend consulting a trained statistician.

When reporting process capability, a control chart of the same data should always be constructed to provide a visual check that the index is reasonable.

For discussion of frequency and scope of CPV analysis see Section 8.

12.3

PROCESS CAPABILITY INDEX

Table 12.3. Rating of PpK Index.

PPK INDEX RATING

>1.33 Limited Opportunity: Attribute meets specifications with a very high level of consistency. It may be more valuable to look for opportunities in other areas of the process or other processes.

1 - 1.33 Some Opportunity: Attribute is routinely meeting specifications but there are indications that it might not always do so consistently.

0.68 – 1.00 Considerable Opportunity:Attribute typically meets specifications, but does not do so to the extent that might be expected.

≤ 0.67 Significant Opportunity: Attribute/process cannot be expected to routinely meet specifications.

Note: Ppk values are recognized indicators of process capability. Although quantitative ranges have been selected and matched with the potential for improvement opportunities, these ranges should not replace reasonable investigation efforts to determine the factors influencing the Ppk value obtained. The levels presented in Table 12.3 reflect one example of a commonly accepted set of ranges for a monoclonal antibody manufacturing process.

Other approaches to both assessing process capability as well as matching this assessment to the potential improvement opportunity may be more suitable for specific processes.


Control charts consist of a few simple elements:• Results plotted in time order;• A centerline, usually at the average of the results;• Statistical control limits.

There are many varieties of control charts. The type of chart should be selected based on the type of data to be plotted [13, Chapter 11]. Most attributes plotted for CPV are individual continuous measurements.

One important consideration for control charts is the choice of time order. Results may be plotted in date of manufacturing order (upstream or downstream) or in date of test order. Any of these orders can be informative, since each order highlights sources of variation that occur in the related process steps. When processes are carried out in strict first-in, first-out sequence from upstream to downstream to assay, the time order will be the same and the choice of time variable has no impact on the charts. For CPV, generally one time order will be selected and used; other time orders may be plotted as needed for investigations and process understanding.

Another important consideration is the method for calculating control limits. Control limits should always be set at plus and minus three sigma (standard deviations), but sigma may be estimated using either a short-term formula or a long-term formula, the same as for Cpk and Ppk. For most charts, the long-term estimate of sigma is recommended; this corresponds to the use of Ppk rather than Cpk. Control limits based on the long-term estimate of sigma will encompass the longer-term sources of variation that are an expected part of total routine process variation; short-term estimates will generally be narrower, and may produce false statistical signals when longer-term variation impacts process results.

Standard Shewhart control charts are simple to set up, easy to understand and explain, and good at detecting large shifts quickly. However, the Shewhart charts are not as good at detecting small shifts (relative to variation), and do not build any memory of previous observations. For quick detection of small shifts, EWMA and Cumulative Sum (CUSUM) control charts are recommended. These are easy to set up

using software, and have the advantage of using previous observations to filter noise and detect small shifts more quickly. However, they are slower than Shewhart charts to detect large shifts, and are more difficult to explain to shop floor personnel and business leaders.

Statistical control charts are based on some assumptions about the process results. When these assumptions are not met, the probability of false signals can rise dramatically. The most important assumptions are that the results are independent over time and that the underlying results are approximately normally distributed.

In real manufacturing processes, results are rarely independent over time. Instead, there is some correlation across sequential results (autocorrelation). This can be a natural result of operating conditions, such as raw material lots that are used for several upstream lots in sequence. The presence of autocorrelation can produce many small shifts and trends within the control limits; the science and technology process support and statistical personnel should document in the CPV Plan whether these types of shifts and trends will be addressed as signals of unusual variation, or treated as expected routine variation.

The assumption of normally distributed data is also important, and should be checked using a histogram, box plot, and normal quartile plot. Statistical tests for normality are not generally recommended, since they will be triggered by other issues in the data, such as occasional outliers, or the very shifts and trends that the control chart is intended to identify.

One common violation of the normality assumption is proportional variance. Proportional variance is variability that is proportional to the level of results. For example, measurements of concentration often have higher variability at higher concentrations, and lower variability at lower concentrations. When variation is expressed as a percentage or Relative Standard Deviation, RSD) instead of a simple standard deviation, this is an indication that the variance may be proportional, and should be checked.

12.4

CONTROL CHARTS

Most common SPC tools were designed for results with constant variance, not proportional variance. When proportional variance is present, a simple solution is to transform results from the original scale to a log scale. Proportional variance on the original scale is stabilized to constant variance by transforming to the log scale. The choice of natural or base 10 log scale does not matter, but one or the other should be used consistently. It is worth noting that significant data are required to make transformation a justifiable approach. If there is any doubt about there being sufficient data, transformation is better avoided and a statistician should be consulted.

The control charts below illustrate the importance of finding the appropriate scale for results before identifying special causes of variation or estimating capability. In the chart on the left, log-normally distributed data are treated assuming they are actually normally distributed. The results are not symmetrically distributed within the control limits, and there are several results outside the upper control limit.

These signals would require a response from the process owners in CPV. In the chart on the right, the same results are transformed to the log scale. The log transformation expands the scale at the lower end of the range, and compresses the scale at the upper end of the range. Now the results are more symmetric within the limits, and no results exceed either upper or lower control limits. Capability indices are best calculated on the transformed data, using specifications, mean, and sigma in the transformed scale.

In many situations, the average of a data set is expected to remain fairly constant over time. However, there are circumstances when the mean is expected to change over time, based on process knowledge and experience. When this is the case, additional tools may be used to monitor for departures from the expected behavior. These include “tool wear charts” and other similar charts based on monitoring the residual result (actual – predicted by a model). These tools will not be expanded upon further in this document.

To summarize the recommendations for trending and evaluating data:• Start simply – always plot the data;• Use long-term estimates of sigma for most situations,

both to set the control limits, and to evaluate process capability (Ppk);

• For some special cases when standard control charts are not providing satisfactory assessments, consult a statistician and other process experts. These special cases include:– Results with low resolution (few distinct values are

possible within the specification range);– The original scale does not provide normally

distributed data, and a scale transformation may be needed;

– EWMA charts or CUSUM charts may be preferred when it is important to detect relatively small shifts in data with high variation [14, Chapter 9];

– Statistical methods may not produce meaningful estimates, for example when a very small number of results are available, so technically justified limits must be used either initially or permanently.

An algorithm can be used to support the selection of control charts (Figure 12.4.2).

Figure 12.4.1. Control charts, showing the effect of different scales


Figure 12.4.2. Guide to selecting control charts

1

Obtain data for an

attribute

At least 30

results available

?

At least 5 distinct values ?

Detecting small

shifts in noisy data ?

Many related results

available ?

Plot on a run chart (no

limits) MVDA

no yes

CUSUMEWM

yes Plot on a run

chart (no limits)

no

Shewhar t chart

yes

no yes no

Multivariate Data Analysis (MVDA) [25] combines multiple parameters to provide greater power for detecting changes in results, and to develop deeper understanding of the sources of variation in a process. MVDA requires access to larger bodies of data than univariate approaches. When MVDA is feasible, it should be considered as a powerful extension of CPV; it is most effective when it can be applied

to process data in real-time, while there is still an opportunity to adjust and improve batch results. MVDA is often used to evaluate and improve within-batch performance, while CPV is most commonly focused inter-batch monitoring. CPV using univariate approaches represents the most common approach within the industry at the moment; MVDA represents where the industry is heading in the future.

12.5

MULTIVARIATE DATA ANALYSIS

CPV is only one part of an overall CS. It is envisaged to function at a supervisory level and is not typically tied directly to lot release. There are other basal level control elements such as alarms and alerts, in-process testing and release analytics that are designed to be more immediate controls of product quality. Any result that is Out Of Specification (OOS) should be investigated under existing Quality procedures for handling deviations. CPV is intended to serve as an ‘early warning system’ where process drift can be detected before it can cause an OOS or failure that could otherwise have product quality impact. Thus, responses to shifts and trends discovered during CPV typically include those that remain within specifications. Investigations or other activities may be triggered to identify the root cause of the process and/or quality shift or perturbation; however, closure of the investigation triggered in this way, is not typically tied to lot release; unless an ‘out-of-specification’ (OOS) situation has also occurred. Shifts and trends that remain within specifications should be evaluated by trained personnel who are most familiar with the process and assay; typically floor support engineers and laboratory scientists. The response to the shift or trend may be determined by the local engineering or technology function, with consultation from the quality, operations, and statistical functions. These investigations form part of the

CPV Plan and in most cases, a formal quality investigation/deviation will not be required, as an OOS situation will not normally have occurred alongside a CPV trend.

A typical path for root cause analysis in response to a signal includes:• Establish that the results are valid; • Check for any indications of inconsistency, e.g. within

a laboratory, during the timeframe the result was obtained;

• Evaluate any other attributes that typically correlate with the result, to determine if all attributes trended together as expected, or if the particular result was exceptional;

• Walk the process upstream from the sample point, and collect process performance data to understand any unusual patterns in process operations during the timeframe the result was obtained.

The explanation of within-specification shifts and trends should be documented in the routine CPV Report. If the reason for a shift or trend cannot be identified during CPV, it may be escalated to the status of an official quality deviation, for further investigation. It may be advantageous to define a ‘tiered’, risk-based approach linked to anticipated actions when a shift or trend is observed. This approach could be developed over time.

12.6

RESPONSES TO SHIFTS AND TRENDS


Note: Establishing initial control limits and the choice of analysis tools is also discussed in Sections 9 and 10. This section provides further thoughts on their application to CPV, given the importance of establishing them as part of a CPV Plan.

In general, statistical control limits should be set at the centerline plus and minus 3 sigma (standard deviation). Sigma may be estimated from short-term variation [13, Chapter 11] or long-term variation, using the formula for standard deviation.

Long-term variation is preferred for two reasons. The most important is that long-term sigma includes all the sources of variation that are expected to be inherent to the process. This provides more realistic control limits, which will be better able to distinguish between expected and unusual instances

of variation in results. The second reason is that during initial production, when fewer than 30 results are available for calculating sigma, the long-term estimate stabilizes more quickly than the short-term estimate. For independent, normally distributed results, the long-term and short-term formulas for sigma will provide very similar values.

During initial production or after a process change, when fewer than 30 results are available to estimate sigma, it is recommended that limits are set, based on technical knowledge of the process. If statistical approaches must be used to set initial limits, use the long-term sigma formula. Set temporary limits to be updated once 30 or more results are available. If longer-term sources of variation occur over extended timeframes, more than 30 results may be needed to capture typical long-term variation within sigma and the control limit values.

Once sufficient process history is established, long-term control limits should be established against which the performance of the process can be assessed. Long-term control limits are sometimes said to be ‘fixed’ meaning they should not be changed without a sound, recorded justification. Fixed limits should be based on a minimum of 30 batches, and should reflect all expected sources of variation. If there are potential sources of significant longer term variation, such as a change of raw material lot, it is important to gather data over a sufficient time period to account for that variation.

Initial OOS results which are determined to be invalid may be excluded when setting limits. Examples of invalid results may include laboratory errors. Such data points should be excluded from sigma calculations and charts. However, if the root cause is unknown or representative of the process or testing method, the data points should be retained in the sigma calculation and chart.

When a process change or improvement shifts the mean or changes the variability, control limits should be re-set based on a minimum of 30 results following the change.

12.7

ESTABLISHING INITIAL LIMITS

12.8

ESTABLISHING LONG-TERM LIMITS

Additional rules may be applied to a control chart to increase the sensitivity of detecting shifts and trends that may remain within the control limits. The Western Electric [15] and Nelson [16] rules are implemented in commercially available SPC software. Note that these tests were designed on the assumptions of independent successive results (no autocorrelation) and a normal distribution. The possibility of false alarms may rise dramatically when either of these assumptions is violated. Low-resolution data in particular will generate many false alarms, and the rules should not be applied for data not meeting these assumptions. Also, the number of rules applied should be limited, at least in the initial phase of CPV, since the probability of false alarms increases with each additional test applied.

Low resolution results occur frequently in regulated processes, often because results have been rounded before they are charted. Whenever possible, use the unrounded result for SPC charting and estimation of sigma. It is recommended that the validation procedure and history of measuring systems is checked for repeatability and reproducibility. Such data helps decide whether rounding is appropriate and to what extent. If in doubt, a statistician should be consulted. When low resolution data is to be evaluated, the recommended approach is to plot the data on a run chart in time order, but avoid setting statistical control limits. Limits may be set using technical judgment and process experience [17, 18].

12.9

FINDING SIGNALS OF SPECIAL CAUSE VARIATION


Good change management is critical to getting the greatest possible benefits from a CPV system. Some important general principles of change management are presented here, along with some more specific scenarios related to CPV observations.

13.0 CHANGE MANAGEMENT

Change management is complementary to effective CPV. Product and process-related changes may impact the CPV plan (including any CPV limits document that might exist separately from the plan), and the RA and CS documents that form the basis for the CPV plan. Changes to these controlled documents should proceed per the company’s normal change control procedures, which should include the description, rationale and justification for all significant changes. Changes that do not impact monitored parameters, sampling points or control limits do not require revision of the CPV plan; however the change control should consider whether evaluation of normal process variability for a particular parameter should start again at the initial CPV phase for analysis. After noting in the change control documentation, this may be noted in the next CPV analysis / report.

Process changes or experience with special or common cause variation may require investigations and a revision of the CPV execution plan during the lifetime of a process. Adjustments to the frequency of periodic monitoring, or a return to collecting baseline data before setting (or re-setting) future long-term control limits may be necessary to

manage control of changes to reduce variability or achieve an improved outcome. A deviation which leads to a corrective or preventative action to begin monitoring a new variable trend not previously monitored, or to re-activate monitoring a trend discontinued for CPV may also need to be added to the plan.

Changes to the control strategy need to consider the potential impact to the current status of the CPV plan. Facility, process, equipment, field measurements, or analytical laboratory method changes (examples of normal change controls) require a review of the CS, any related risk assessments, and the current monitoring plan for the steps being changed as well as the downstream steps that have linked parameters or attributes. If a process variable is re-classified in the CS based on new process understanding, changes to the CPV monitoring plan may also be needed and any impact to registered details of regulatory licenses need to be addressed. Below, a decision flow outlines a process for managing changes needed due to ongoing CPV monitoring (Figure 10.13).

SECTION 13.0

Figure 13. Decision Flow for changes to CPV Execution Plan for Drug Substance Lifetime. OOT stands for Out-of-Trend, NOR is Normal Operating Range and PAR is Proven Acceptable Range.

Assessments of CPV trends and CPV plan deviations are to be documented in CPV reports. This should indicate actions taken and capture rationale to justify any recommended CPV plan revisions. These reports are then used as inputs into a periodic APR / Product Quality Review package. Assessments may be as simple as chart status, or include an evaluation into whether a capability index value demonstrates it is possible to reduce or stop monitoring a particular parameter or attribute. If sufficient data has been obtained to calculate and implement long-term limits with high confidence, this should be documented in the CPV Report, before the CPV Plan is changed. An assessment may conclude that monitoring will continue to a new planned milestone without change setting new limits. If the capability index is low, the monitoring plan may need to be changed (e.g. to obtain data more frequently, or a Corrective Action and Preventive Action (CAPA), might be considered to improve control. In any case the decision should be documented in a CPV Report with its rationale.

In this A-Mab example, for cases where a change would cancel out capability indices or remove confidence in continued use of existing control limits, a plan deviation would be used to document suspending or making these control limits obsolete. The process performance trend can then be monitored against a new provisional control range (justified via the deviation and based on PARs, recent history, equipment capability, or qualified small scale model studies). Data collected prior to the change may be used for comparison to assess the impact of the change.

The following table presents several cases where CPV-related changes may occur, the impacted documents and required actions. The change management process is very similar to the initial setup of the CPV plan, as presented in preceding sections of this paper. An assessment of the RA document and its impact on the CS document is always required, even if not explicitly stated in table 13-1 below.

R e vie w C ontrol S tra te gie s & R is kA s s e s s me nts for actua l or

pote ntia lly impacte d controlpa rame te rs and pe rformance

indica tors . . .and if c la s s ifica tionra tiona le is s till jus tifie d or if

de cis ions ne e d to be adjus te d

U ne xpe cte d incide nt re porte d/impe nding change notifica tion

communicate d, C P V data O O T

D oe s Incide nt/ C hange indica tes e tpoints or range s for anyfixe d or re s pons e variable sne e d additiona l monitoring,ve rifica tion, re qua lifica tion,

re s e tting, re de finition(S e tpoints , a le rt/ action limits ,

or NO R / P A R controls )

No, C urre nt s ta tus cancontinue (routine control orC P V e nhance d monitoring

as curre ntly in-place )

Is it comple te ly ne w? Notaddre s s e d in e xis ting ris kas s e s s me nts or control

s tra te gie s ye t

No, but it is not in C P Vmonitoring, jus tify addingor not adding to programin inve s tiga tion or change

impact as s e s s me nt

Y e s , de te rmines ample te s t/data capturere quire me nts

G e ne ra te informationne e de d to e s tablis h initia lproce s s unde rs tanding/

product knowle dge(R &D , S S M, s upplie r data ,

his torica l data ana lys is )

S us pe nd any e xis ting C P Vcrite ria / limits , and re place

with lowe r confide ncecontrol range s via Q MSme chanis m of protocolchange manage me nt

comme rcia l datacolle ction/ as s e s s me ntto e s tablis h long te rm

confide nce in ne wunde rs tanding, andas s urance of control

Y e s . C omple te ne w ors upple me nt e xis ting ris k

as s e s s me nt, re vis e controls tra te gy

C onfirm O O T , inve s tiga te tode te rmine if a s pe cia l caus e ande xcluding data from futre natura l

va ria tion s igma ca lcula tion


Table 13-1. Change Management Examples

REASON FOR CHANGE IMPACTED DOCUMENT(S) ACTIONS

CHANGES WITHIN CPV PLAN

Transition from initial control limits to long-term limitsEnough data available (e.g. n = 30) to establish statistically based control limits

CPV Plan / Limits Document

Document (e.g. in a CPV limits document) how statistically based control limits (CL) were determined. Start routine monitoring with CL’s and run rules

OOT results (e.g. control limit or run rule violations) due to newly experienced, but normal, variability (e.g. due to a new raw material lot)

CPV Limits Document

Justify removal of current CPV limits (if applicable), and continue to collect data until sufficient to recalculate new control limits. During the extended data collection period, continue to monitor for trends (e.g. average ± 3 SD), but without run rules.

Shift in mean or a change in variability (e.g. due to a process improvement)

CPV Limits Document

Justify removal of current CPV limits, and reset counter for the number of runs required to set new control limits for the impacted attribute / parameter. Continue to monitor for trends (e.g. average ± 3 SD).

Add / Remove control elements based on process knowledge gained through CPV, e.g. after periodic monitoring. This data could point to ways to improve the product or optimize the process. Elements may be deleted (or have reduced sampling frequency) if their process capability index is high and variability is well controlled. Elements may be added if a new source of process variability is discovered.

RA / CS Update to reflect the new knowledge


Update. New elements may be monitored (e.g. using average ± 3 SD) until statistically based control limits can be established (n ≥ 30).

CHANGES EXTERNAL TO CPV PLAN

• Add or Remove control elements based on new process knowledge, e.g. from lab scale studies, CAPA’s, complaints, etc.

• Process changes (e.g. new raw material)• Change or add manufacturing site• New equipment (excluding like-for-like changes)

PPQPerform PPQ run(s) if deemed necessary (consult with QA and Regulatory Affairs).

RA/ CS See “Add/Remove control elements” above


See “Add/Remove control elements” above

New or changed analytical capabilities RA/ CS Only if impacted

Three major sources of data are used for CPV:

1. Process data e.g. viable cell concentration in the bioreactor [26], offline pH / conductivity measurements or process volumes which are recorded in either paper or electronic BRcs.

2. Analytical data generated in QC laboratories which are typically recorded in a LIMS or Data Historian.

3. Data recorded by inline instruments e.g. pH of bioreactor, or buffer flow rates for chromatography operations. These data are archived and managed by the data historian component of the plant control system. Examples of data archival systems include Plant Information (PI_System) by OSISoft and InfoPlus 21 by Aspentech.

Data recorded on paper-based systems require transcribing into a secure electronic archiving system (e.g. a database) so that it can be retrieved in the future for data analysis, control charting and CPV. The manual data entry process is prone to human error. Several options are available to ensure data integrity during the data transcription process. These are described below.

1. Blinded data entry: Data are independently entered into the data archival system by two different operators. The data archival system prompts the operators if there is a discrepancy noted between the two separate entries for a process parameter or attribute so corrections can be made. However there is a remote possibility that both data entry operators can make the same mistake.

2. Single data entry and independent verification: The data are entered from the original data source by one operator and independently verified by a second operator using the source documents. This process is used in many companies, but cannot completely eliminate errors because the data verifier may not always catch all data entry errors.

3. SMART Data entry / Error proofing: for a parameter or attribute being transcribed, the data archival system can limit the allowable values that can be entered, thus any errors can be readily pointed out and corrections can be made, for example:

• Offline pH measurements can be restricted to values from 0.0 to 14.0;

• Viable cell concentration data can be restricted to the expected range of values [26].

Alternatively, the data archival system can generate a report of observed minimum and maximum values for a process parameter or attribute. Any discrepant result outside of the normally observed range can be readily detected.

While a SMART Data entry system can detect some errors, it is also not foolproof. Suppose the observed values for chromatography step yield vary from 50.2 to 80.4 %. For a particular batch, if the yield recorded in the source BRc is 62.6%; this could be wrongly entered as 66.2% which would still fall within the expected observed values and may remain undetected.

This section describes in general the types of data sources in a CPV reporting system. It goes on to make recommendations as to how data in these systems should be verified and hence how the systems should be validated.

14.0 DATA VERIFICATION

SECTION 14.0

In summary, there are hidden sources of error in transcribing data from paper to electronic data archival systems. It is important to recognize that verification is critical for some parameters (e.g. CPPs) but may not be necessary for other process parameters. Therefore verification depends on purpose and criticality of the parameter being measured. The intention should to reduce data entry errors to zero. This may require continuous improvement and it is important to know the extent to which any error impacts the process parameter.

Transcribing data from electronic sources e.g. electronic BRcs or plant information into data archival systems, is not likely to introduce errors provided the data transfer system is designed robustly and the ability to transfer data is validated. For either paper or electronic sources, the ability to retrieve data from the data archival system should undergo an initial validation. Similarly if the data from the data archival system is used routinely for generating control charts or data tables for process monitoring and CPV, the procedure for generating control charts or data tables should be validated or as recommended in section 12 of this document, established software should be used. A number of software suppliers will provide a quality statements regarding the validation of their statistical software [e.g. 19, 20, 21].

Whether the source data are from paper systems or electronic systems, a record should be kept (and continually updated) of the BRc step number or the Tag ID for electronic sources. It is recommended (although not an absolute requirement) for the data archival system to have the ability to record data entry operator ID / time stamp, any corrections made; i.e. an audit trail of all entries and changes.

Finally it would be good practice for any process parameter or quality attribute observed to be out of controllable range (trend) or for observed shifts and trends to be independently verified by re-examining the source data.

Thus in conclusion, it is recommended that robust systems and procedures are designed and developed in order to archive data and validate the accuracy of data retrieval in order to minimize errors during CPV. It is of course, the responsibility of each individual organization to apply the recommendations in this section as they see fit.


Figure 15.1. Discretionary Elements of CPV

ELEMENT DESCRIPTION

Operational or Performance Elements Process performance attributes (i.e., bioreactor titer, column elution volumes) or input parameters linked to process performance attributes.

Multivariate Data Analysis MVDA, particularly Partial least squares (PLS) regression or Principal component analysis (PCA) may be useful to identify latent variables within the CPV data set and increase an understanding of the design space of the drug substance manufacturing process.

Column/Resin/UF Membrane Cleaning and Performance Lifetime Verification

The full scale verification of column and/or ultra-filtration membrane cleaning and performance, out to established lifetime limits, may be included within the scope of the operating company’s CPV program.

The Table below shows some elements that may be included within a CPV program based upon the needs and decisions of the individual operating company. The following elements are not considered mandatory for inclusion in CPV, but may provide the operating company with information valuable to the manufacturing of drug substances.

15.0 DISCRETIONARY ELEMENTS OF A CPV PROGRAM

SECTION 15.0


16.0 TECHNICAL REFERENCES 1. Technical Report No. 60. Process Validation: A Lifecycle

Approach. PDA, Inc. 2013

2. ISPE PQLI Guidance Series Part 4

3. A-Mab: A case study in Bioprocess Development. CMC Biotech Working Group. 2009

4. FDA Code of Federal Regulations 21 Part 211, Jan. 2013

5. Process Validation: General Principles and Practices, guidance for industry, FDA (CDER, CBER, and CVM), January 2011

6. A-Mab study; Pharmaceutical Quality by Design: Product and Process Development, Understanding, and Control; By Lawrence X. Yu, Ph. D.; Director for Science; Office of Generic Drugs; Food and Drug Administration

7. ICH Q8 Pharmaceutical Development: http://www.ich.org/fileadmin/Public_Web_Site/ICH_Products/Guidelines/Quality/Q8_R1/Step4/Q8_R2_Guideline.pdf

8. ICH Q9 Quality Risk Management: http://www.ema.europa.eu/docs/en_GB/document_library/Scientific_guideline/2009/09/WC500002873.pdf

9. ICH Q10 Pharmaceutical Quality System: http://www.ema.europa.eu/docs/en_GB/document_library/Scientific_guideline/2009/09/WC500002871.pdf

10. Technical Report No. 15. Validation of Tangential Flow Filtration in Biopharmaceutical applications. PDA, Inc. 2009

11. Technical Report No. 14. Validation of column-based chromatography processes for purification of proteins. PDA, Inc. 2008

12. Biopharmaceutical Manufacturing Facilities- Volume 6. ISPE Baseline Pharmaceutical Engineering Guide. June 2004

13. Pena-Rodriguez, ME. Statistical Process Control for the FDA-Regulated Industry, 2013. ASQ Press: Milwaukee, WI

14. Montgomery DC. Introduction to Statistical Quality Control, 2013. 7th edition, Wiley: Hoboken, NJ

15. 15 Western Electric. Statistical Quality Control Handbook 1956, Western Electric Corporation, Indianapolis, IN

16. Nelson, LS. The Shewhart Control Chart – Tests for Special Causes, Journal of Quality Technology 1984; 16: 237-239

17. Woodall, WH. Controversies and Contradictions in Statistical Process Control, Journal of Quality Technology 2000; 32: 341-350

18. Limpert, E, Stahel, WA, Abbt, M, Log-normal Distributions across the Sciences: Keys and Clues. BioScience 2001; Vol. 51 No. 5

19. www.jmp.com/software/qualitystatement.shtml

20. www.minitab.com/en-US/support/documentation/software-validation.aspx?langType=1033

21. www.statsoft.com/services/validation-services/

22. ICH Q11 Development and Manufacture of drug substances (Chemical Entities and Biotechnological/Biological Entities)

23. http://www.ema.europa.eu/docs/en_GB/document_library/Other/2013/08/WC500148215.pdf

24. ISPE Discussion Paper: Applying Continued Process Verification Expectations to New and Existing Products”, D. Bika, P. Butterell, J. Walsh, K. Epp and J. Barrick, 2012. https://www.google.com/url?sa=t&rct=j&q=&esrc=s&frm=1&source=web&cd=1&cad=rja&ved=0CCkQFjAA&url=https%3A%2F%2Fwww.ispe.org%2Fdiscussion-papers%2Fstage-3-process-validation.pdf&ei=FIEXU8z1Gurx0wGftoGIBg&usg=AFQjCNFvOl1WjtGX-vPHZjFsPq8n44cI3Q&sig2=QzSLW0X3-cQvbUnwYDGYDA&bvm=bv.62286460,d.dmQ

25. Guidance for Industry Analytical Procedures and Methods Validation for Drugs and Biologics, Draft, February 2014 http://www.fda.gov/downloads/Drugs/GuidanceComplianceRegulatoryInformation/Guidances/UCM386366.pdf

26. European Pharmacopoeia, 2.7.29. Nucleated cell count and viability, p233

SECTION 16.0

TERM EXPLANATION SOURCE (S)

Acceptance criteria Numerical limits, ranges, or other suitable measures for acceptance which the drug substance or drug product or materials at other stages of their manufacture should meet to conform to the specification for analytical procedures.

Q6b

Action limits An action limit is an internal (in-house) value used to assess the consistency of the process at less critical steps. These limits are the responsibility of the manufacturer.

Q6b

Batch Records (BRc) A record of specific identifiers for the batch of material being produced, that includes all activities required to prepare for production, produce the material and close down the process. It provides traceability of who did what, when, and the outcomes of those actions, including any observations on or deviations from the process or anticipated results.

[4]

Batch Release (BR) The process by which the product is tested and results reviewed to ensure product quality under cGMP regulations and guidelines.

Q8(R2)

BPOG BioPhorum Operations Group, a collaboration of biopharmaceutical companies, seeking to accelerate the rate at which the industry achieves a lean state.

Bulk Drug Substance (BDS)

According to 21CFR207.3(a)(4) this means any substance that is represented for use in a drug and that, when used in the manufacturing, processing, or packaging of a drug, becomes an active ingredient or a finished dosage form of the drug, but the term does not include intermediates used in the synthesis of such substances.

21CFR207.3(a)[4]

Capability of a Process (Ppk)

Ability of a process to realise a product that will fulfil the requirements of that product. The concept of process capability can also be defined in statistical terms via the process performance index Ppk or the process capability index Cpk (ISO 9000:2005).

Q10

CMC BWG Chemistry, manufacturing and control, biotech working group of the International Society for Pharmaceutical Engineers (ISPE).

Continued Process Verification

A formal process that enables the detection of variation in the manufacturing process that might have an impact on the product. It provides opportunities to proactively control variation and assure that, during routine production the process remains in a state of control.

[5]

Control Strategy A planned set of controls, derived from current product and process understanding that assures process performance and product quality. The controls can include parameters and attributes related to drug substance and drug product materials and components, facility and equipment operating conditions, in-process controls, finished product specifications, and the associated methods and frequency of monitoring and control.

Q10

Critical Describes a process step, process condition, test requirement, or other relevant parameter or item that must be controlled within predetermined criteria to ensure that the API meets its specification.

Q7

Critical Material Attribute (CMA)

A material attribute, whose variability has an impact on a critical quality attribute and therefore should be monitored or controlled to ensure the process produces the desired quality.

Q8(R2)

Critical Process Parameter (CPP)

A process parameter whose variability has an impact on a critical quality attribute and therefore should be monitored or controlled to ensure the process produces the desired quality.

Q8(R2)

17.0 GLOSSARY

SECTION 17.0

http://www.ich.org/fileadmin/Public_Web_Site/ICH_Products/Guidelines/Quality/Q8_R1/Step4/Q8_R2_Guideline.pdf



http://www.ema.europa.eu/docs/en_GB/document_library/Scientific_guideline/2009/09/WC500002873.pdf






www.jmp.com/software/qualitystatement.shtml

www.minitab.com/en

software-validation.aspx

www.statsoft.com/services/validation

http://www.ema.europa.eu/docs/en_GB/document_library/Other/2013/08/WC500148215.pdf

http://www.ema.europa.eu/docs/en_GB/document_library/Other/2013/08/WC500148215.pdf

https://www.google.com/url?sa=t&rct=j&q=&esrc=s&frm=1&source=web&cd=1&cad=rja&ved=0CCkQFjAA&url=https%3A%2F%2Fwww.ispe.org%2Fdiscussion-papers%2Fstage-3-process-validation.pdf&ei=FIEXU8z1Gurx0wGftoGIBg&usg=AFQjCNFvOl1WjtGX-vPHZjFsPq8n44cI3Q&sig2=QzSLW0X3-cQvbUnwYDGYDA&bvm=bv.62286460,d.dmQ







http://www.fda.gov/downloads/Drugs/GuidanceComplianceRegulatoryInformation/Guidances/UCM386366.pdf



Critical Quality Attribute (CQA)

A physical, chemical, biological or microbiological property or characteristic that should be within an appropriate limit, range, or distribution to ensure the desired product quality.

Q8(R2)

Design Space The multidimensional combination and interaction of input variables (eg, material attributes) and process parameters that have been demonstrated to provide assurance of quality. Working within the design space is not considered as a change. Movement out of the design space is considered to be a change and would normally initiate a regulatory post approval change process. Design space is proposed by the applicant and is subject to regulatory assessment and approval.

Q8(R2)

Detect-ability The ability to discover or determine the existence, presence, or fact of a hazard. Detect-ability is a component of a Failure Modes Effects Analysis (FMEA).

Q9

Drug product (Dosage form; Finished product)

A pharmaceutical product type that contains a drug substance, generally in association with excipients. Drug substance (Bulk material): The drug substance is the material which is subsequently formulated with excipients to produce the drug product. It can be composed of the desired product, product-related substances, and product- and process-related impurities. It may also contain excipients and other components, such as buffers.

Q6b

Failure Modes Effects Analysis (FMEA)

One of the first systematic techniques for failure analysis. It was developed by reliability engineers in the 1950s to study problems that might arise from malfunctions of military systems. A FMEA is often the first step of a system reliability study. It involves reviewing as many components, assemblies, and subsystems as possible to identify failure modes, and their causes and effects. For each component, the failure modes and their resulting effects on the rest of the system are recorded in a specific FMEA worksheet. There are numerous variations of such worksheets.

Q8, IEC 60812 Analysis techniques for system reliability—Procedure for failure mode and effects analysis (FMEA).

General process parameter (GPP)

An adjustable parameter (variable) of the process that does not have a critical effect on product quality or process performance. Ranges for GPPs are established during process development, and changes to operating ranges will be managed within the quality system.

CMC-BWG

Impurity Any component present in the drug substance or drug product that is not the desired product, a product-related substance, or an excipient (including added buffer components). It may be either process- or product-related.

Q6b

In-process quality attributes (IPQA)

Parameters used in the A-Mab Case Study model of a control strategy, to provide a link between KPPs and KPAs

[3]

In-Process Control also called Process Control

Checks performed during production in order to monitor and if necessary to adjust the process and/or to ensure that the intermediate or API conforms to its specifications.

Q7

In-process test In-process inspection and testing should be performed by monitoring the process or by actual sample analysis at defined locations and times. The results should conform to established process parameters or acceptable tolerances. Work instructions should delineate the procedure to follow and how to use the inspection and test data to control the process.

WHO Portal http://apps.who. int/medicinedocs /en/d/Jh1792e/ 20.7.3.3.html

In-process tests Tests which may be performed during the manufacture of either the drug substance or drug product, rather than as part of the formal battery of tests which are conducted prior to release.

Q6a

Intermediate For biotechnological/ biological products, a material produced during a manufacturing process that is not the drug substance or the drug product but for which manufacture is critical to the successful production of the drug substance or the drug product. Generally, an intermediate will be quantifiable and specifications will be established to determine the successful completion of the manufacturing step before continuation of the manufacturing process. This includes material that may undergo further molecular modification or be held for an extended period before further processing.

Q5c


Key Process Attribute (KPA)

An important attribute or output measure of the process used in this paper to maintain consistency with the language used in the A-MaB case study. N.B. It is important not to confuse a KPA, which is a measure of process consistency with measures of quality such as CQAs. N.B. at the time of writing, the European Medicines Agency (EMA) draft guidance on Process Validation is out for consultation, referring to KPAs as 'performance indicators'.

A-MaB Case Study [3]

Key Process Parameter (KPP)

An adjustable parameter (variable) of the process that, when maintained within a narrow range, ensures optimum process performance. A key process parameter does not meaningfully affect critical product quality attributes. Ranges for KPPs are established during process development, and changes to operating ranges will be managed within the quality system. N.B. this category of parameter is not recognised by the FDA or the EMA for use in formal submissions and reports, though they do not oppose its use internally. The agencies see all parameters that may have an impact on CQAs as Critical and hence CPPs [23].

CMC BWG

Knowledge Management Systematic approach to acquiring, analyzing, storing, and disseminating information related to products, manufacturing processes and components.

Q10

Multivariate Analysis MDVA, particularly Partial least squares (PLS) regression or Principal component analysis (PCA) may be useful to identify latent variables within the CPV data set and increase an understanding of the design space of the drug substance manufacturing process.

[25]

Normal Operating Range (NOR)

A defined range, within the Proven Acceptable Range, specified in the manufacturing instructions as the target and range at which a process parameter is controlled, while producing unit operation material or final product meeting release criteria and Critical Quality Attributes.

PQRI

Performance Indicators Measurable values used to quantify quality objectives to reflect the performance of an organisation, process or system, also known as ―performance metrics in some regions.

Q10

Pharmaceutical Quality System (PQS)

Management system to direct and control a pharmaceutical company with regard to quality.

ICH Q10

Plan A detailed description of how something is going to be done. Oxford English Dictionary online

Potency Potency is the measure of the biological activity using a suitably quantitative biological assay (also called potency assay or bioassay), based on the attribute of the product which is linked to the relevant biological properties.

Q6b

Prior product knowledge The accumulated laboratory, nonclinical, and clinical experience for a specific product quality attribute. This knowledge may also include relevant data from other similar molecules or from the scientific literature.

CMC BWG

Procedure A written, established way of doing something in the operating environment. Oxford English Dictionary online

Process Analytical Technology (PAT)

A system for designing, analyzing, and controlling manufacturing through timely measurements (ie, during processing) of critical quality and performance attributes of raw and in-process materials and processes with the goal of ensuring final product quality.

Q8(R2)

Process Control See In-Process Control. Q7

Process Robustness Ability of a process to tolerate variability of materials and changes of the process and equipment without negative impact on quality.

Q8(R2)

Process-related impurities Impurities that are derived from the manufacturing process. They may be derived from cell substrates, culture (eg, inducers, antibiotics, or media components), or from downstream processing (eg, processing reagents or column leachables).

Q6b

Process-related impurities These are impurities that develop from, or are introduced by, the biological or chemical processes by which the product is made.

Q3B(R2),

http://apps.who

20.7.3.3.html


Product lifecycle All phases in the life of a product from the initial development through marketing until the product's discontinuation.

All phases in the life of the product from the initial development through marketing until the product‘s discontinuation.

Q8(R2)

Q9

Product-related impurities Product-related impurities are molecular variants of the desired product arising from processing or during storage (eg, certain degradation products) which do not have properties comparable to those of the desired product with respect to activity, efficacy, and safety.

Q6b

Product-related substances

Product-related substances are molecular variants of the desired product which are active and have no deleterious effect on the safety and efficacy of the drug product. These variants possess properties comparable to the desired product and are not considered impurities.

Q6b

Protocol Method for carrying out an experiment and/or creating an official record of scientific or experimental observations. Written GMP Protocols define prospectively the conditions which will be tested, sample testing plan, and acceptance criteria for results.

Oxford English Dictionary online

Proven Acceptable Range A characterized range of a process parameter for which operation within this range, while keeping other parameters constant, will result in producing a material meeting relevant quality criteria.

Q8(R2)

Quality The degree to which a set of inherent properties of a product, system or process fulfils requirements.

Q9

Quality Attribute (QA) A molecular or product characteristic that is selected for its ability to help indicate the quality of the product. Collectively, the quality attributes define the adventitious agent safety, purity, potency, identity, and stability of the product. Specifications measure a selected subset of the quality attributes.

Q5e

Quality by Design A systematic approach to development that begins with predefined objectives and emphasizes product and process understanding and process control, based on sound science and quality risk management.

Q8(R2)

Quality Control (QC) Checking or testing, that specifications are met. Q7

Quality Target Product Profile (QTPP)

A prospective summary of the quality characteristics of a drug product that ideally will be achieved to ensure the desired quality, taking into account safety and efficacy of the drug product.

Q8 (R2)

Quality risk management A systematic process for the assessment, control, communication, and review of risks to the quality of the drug product across the product lifecycle.

Q9

Raw material Raw material is a collective name for substances or components used in the manufacture of the drug substance or drug product.

Q6b

Reference standards/ materials

In addition to the existing international/national standards, it is usually necessary to create in-house reference materials.

In-house primary reference material: a primary reference material is an appropriately characterized material prepared by the manufacturer from a representative lot(s) for the purpose of biological assay and physicochemical testing of subsequent lots, and against which in-house working reference material is calibrated.

Q6b

Risk The combination of the probability of occurrence of harm and the severity of that harm (ISO/IEC Guide 51).

Q9

Risk analysis The estimation of the risk associated with the identified hazards. Q9

Risk assessment A systematic process of organizing information to support a risk decision to be made within a risk management process. It consists of the identification of hazards and the analysis and evaluation of risks associated with exposure to those hazards.

Q9

Risk evaluation The comparison of the estimated risk to given risk criteria using a quantitative or qualitative scale to determine the significance of the risk.

Q9

Severity A measure of the possible consequences of a hazard, which is a component of a Failure Modes Effects Analysis (FMEA).

Q9

Specification A specification is a list of tests, references to analytical procedures, and appropriate acceptance criteria with numerical limits, ranges, or other criteria for the tests described, which establishes the set of criteria to which a drug substance or drug product or materials at other stages of their manufacture should conform to be considered acceptable for its intended use.

Q6b

Specification - Release The combination of physical, chemical, biological and microbiological tests and acceptance criteria that determine the suitability of a drug product at the time of its release.

Q1a(R2)

Statistical Process Control (SPC)

Statistical process control (SPC) is a method of quality control which uses statistical methods. SPC is applied in order to monitor and control a process. Monitoring and controlling the process ensures that it operates at its full potential. At its full potential, the process can make as much conforming product as possible with a minimum (if not an elimination) of waste (rework or Scrap). SPC can be applied to any process where the "conforming product" (product meeting specifications) output can be measured. Key tools used in SPC include control charts; a focus on continuous improvement; and the design of experiments. An example of a process where SPC is applied is manufacturing lines.

Q8 (R2), [13],

Testing plan A determination as to whether routine monitoring, characterization testing, in process monitoring, stability testing, or no testing is conducted as a part of the overall control strategy. Extended testing plans may be put in place to demonstrate that valuable resources can be used more than once. It may also be necessary to establish additional tests to understand sources of variation and to demonstrate that changes to the process have addressed sources of variation that are considered to present appreciable risk to product quality.

CMC-BWG

TPP See Quality Target Product Profile. Q8 (R2)

Viable Cell Concentration (VCC)

A measure of the number of viable cells per unit volume. [25]

Well Controlled Critical Process Parameter (WC-CPP)

A process parameter which is controlled by process design and standardized procedures or automated control systems that ensure it remains within the design space of the process. It is only likely to vary beyond the design space if there is a failure in the control system and failure modes for this situation are likely to be mitigated.

CMC-BWG

BPOG Continued Process Verification: An Industry Position Paper With Example Plan

continued process verification: an industry position · pdf filepage 4 – bpog continued...

Documents