Suzanne Farid (EPSRC Centre for Innovative Manufacturing, UCL) www.ucl.ac.uk/epsrccim www.ucl.ac.uk/biochemeng/industry/vision

advertisement
Suzanne Farid
(EPSRC Centre for Innovative Manufacturing, UCL)
www.ucl.ac.uk/epsrccim
www.ucl.ac.uk/biochemeng/industry/vision
EPSRC CIM - UCL VISION Event 2 Dec 2013
www.ucl.ac.uk/epsrccim
“Connectivity is our strength”
The VISION Programme aims to be the nexus for the leaders of
the biological sciences industry to hear and debate the latest
technological and business developments in the sector.
Contact Dr Karen Smith, Director of Bioprocess Leadership on
+44 (0)20 7679 4411 or email karen.smith@ucl.ac.uk
www.ucl.ac.uk/biochemeng/industry/vision
EPSRC CIM - UCL VISION Event 2 Dec 2013
Chairman’s Introduction
Neil Weir
(UCB Pharma)
EPSRC CIM - UCL VISION Event 2 Dec 2013
Proposed Session Format
17.00
Registration & Refreshments
Posters and guided tours of UCL’s Biochemical Engineering facilities
18.00
Welcome and Introduction
Suzanne Farid (UCL)
18.10
Chairman’s Introduction
Neil Weir (UCB Pharma)
18.25
Implementing QbD in a Large Company
Graham McCartney (Eli Lilly)
18.45
Implementing QbD in a Contract Manufacturing Organisation
Graham McCreath (Fujifilm Diosynth Biotechnologies)
19.05
Open Forum Discussion with Industrial Panel
Richard Francis (Francis Biopharma)
Graham McCartney (Eli Lilly)
Graham McCreath (Fujifilm Diosynth Biotechnologies)
Stephen Ward (Cell Therapy Catapult)
Chairman: Neil Weir (UCB Pharma)
19.30
21.00
Networking Reception
Close
EPSRC CIM - UCL VISION Event 2 Dec 2013
QbD – ‘Cycle of Life’
EPSRC CIM - UCL VISION Event 2 Dec 2013
7
QbD in Discovery vs Empiricism
•
•
•
•
•
•
•
•
Therapeutic concept
Chemical / Biologic starting point
Potency
Selectivity
Pharmacokinetics and distribution
Relative optimization
Pharmaceutical properties
Therapeutic concept in vitro, in vivo , in patients
EPSRC CIM - UCL VISION Event 2 Dec 2013
Implementing QbD in a Large Company
Graham McCartney
(Eli Lilly)
EPSRC CIM - UCL VISION Event 2 Dec 2013
Eli Lilly Kinsale
We Make Medicine.
Implementation of QbD in a “Large”
Pharma Company
R. Graham McCartney PhD
Technical Lead Biotechnology
Eli Lilly Kinsale Ireland
UCL Pre-Conference QbD Workshop
Introduction
 Eli Lilly has recently constructed and qualified a commercial cell
culture biologics manufacturing facility in Kinsale, Ireland.
 Presentation uses the first molecules validated in this facility to
illustrate:
 Approach taken to integrate QbD principles into robust Control
Strategy development.
 Practical Translation of “concepts” into outcomes: a successful
Process Validation start-up
 “There are many paths up the mountain”...this is but one path
.
Kinsale Ireland: Biotech Infrastructure
IE28
Analytical, TS/MS, Micro
capabilities and infrastructure.
Testing, mfg support, late stage R&D
IE43 (2016)
4 x 11,000L bioreactors
1 purification train
65lots/annum. Expandable
IE42
3 x 5,000L bioreactors
1 purification train
44lots / annum. Multi product
IE30
Platforms, SM fixed
LM disposable.
Roll in – out equip
Capable of both NS0 + CHO platforms and adaptable to longer term Lilly pipeline molecules
.
Elements to Consider
 Eli Lilly’s Approach to QbD
for Biotech Products
 Quality Target Product
Profile
 Critical Quality Attributes
 Risk Assessments and
Tools
 Attribute Driven Process
Optimization
 Characterization via DOE
 Multivariate Analysis
 Robust Control Strategy
Development
 Validation and Continuous
Monitoring
.
Drug Product Profile
Product Attribute
Target Profile
Active Pharmaceutical Ingredient
IgGx
Indication

To be considered
Dosage Form

Single Use Sterile Parenteral Solution
Dosage Strength


Two dosage forms
Fixed volume with 2 concentrations
Container closure system

glass prefilled syringe
Shelf life

X months @ 2 to 8 oC
Administration Route



Consider injection route
Injection volume NMT X mL
Acceptable needle

Patient self administration with auto-injector and
manual syringe
Short term unrefrigerated in-use stability
Formulation Delivery

.
Molecule Attributes: Orthogonal Analytical Method
Strategy Required
Site(s)
Modification
N-terminus
Modification
Fc Met and Fc
Trp
Oxidation
Asn in Fc
Domain
Reduced
RP-HPLC
SEC
Nonreduced CESDS
CZE
ThioGLo1 CE-LIF
Reduced
CE-SDS
LC-MS
X
X
X
X
Deamidation
X
X
Asp in Fc
Domain
Isomerization
X
X
Fc Domain
Des
Asn
Glycosylation
Cys-Cys
Free Sulfhydryl
Multiple Sites
Glycation
Aggregation
X
X
X
X
X
X
X
X
X
X
X
.
Overall Risk Formula
RPN = Severity x Occurrence x Detectability
.
CQA Risk Ranking & Filtering Tool
A Continuum of Criticality
 Assess relative safety and efficacy risks using two factors:
– Impact and Uncertainty used to rank risks
 Impact = impact on safety or efficacy, i.e. consequences
– Determined by available knowledge for attribute in question (prior, clinical, etc)
– More severe impact = higher score
 Impact on biological activity, PK/PD, immunogenicity, adverse effects
 Uncertainty = uncertainty that attribute has expected impact
– Determined by relevance of knowledge for each attribute
– High uncertainty = high score (no information with variant or published literature
only)
– Low uncertainty = low score (data from material used in clinical trials)
Severity = Impact x Uncertainty
 Severity = risk that attribute impacts safety or efficacy
.
Severity Assessment:
Knowledge Source
Basis for Impact Assessment
No data available
Literature / Expert Opinion
Platform knowledge
Molecule-specific in vitro or
non-clinical data
Description
No information is available to assess impact of quality attribute (e.g., new variant or no relevant literature data
available on impact of quality attribute).
Impact assessment based upon published external literature for the quality attribute in a related molecule.
Literature must be from representative molecule(s) (e.g., similar MOA to assess impact of quality attribute on
efficacy, similar clearance mechanism to assess impact of quality attribute on PK) with a similar level of the
quality attribute.
Impact assessment is based upon internal data (in vitro, non-clinical or clinical) from a similar class of molecule
(i.e., platform knowledge). Platform knowledge must be from appropriate molecule(s) (e.g., similar MOA to
assess impact of quality attribute on efficacy) and similar level of the quality attribute.
Impact assessment is based upon in vitro or non-clinical data for the specific quality attribute and the specific
molecule (e.g., in vitro potency testing for enriched fractions to assess impact on potency, correlating PK/PD
data in relevant animal data to presence/level of quality attribute to assess impact on PK/PD, toxicology
“margins of safety” for quality attribute to assess impact of quality attribute on toxicity).
Impact assessment is based upon clinical data for the specific quality attribute and the specific molecule (e.g.,
correlating clinical immunogenicity data to presence/level of quality attribute to assess impact on
immunogenicity, observation that clinical exposure of the quality attribute at relevant levels yields good safety
profile to assess impact of the quality attribute on Safety as low).
Molecule-specific clinical data Note –Based upon the range of quality attributes of the batches utilized in clinical studies as well as the design
of the clinical studies, it is often not possible to ascribe specific clinical results to specific quality attributes of
the molecule. Therefore, caution is needed when utilizing clinical data to assign impact scores to specific
quality attributes (e.g., the observation of adverse events should not be utilized to assign a high impact Safety
rating to a specific quality attribute unless the adverse events can be specifically attributed or correlated to the
presence/level of the quality attribute).
.
Severity: Overall Scoring
Knowledge Source
Moleculespecific in
vitro or nonclinical data
Moleculespecific
clinical data
No data
available
Literature
Platform
knowledge
9 (CQA)
9 (CQA)
9 (CQA)
9 (CQA)
9 (CQA)
Moderate
Impact
9 (CQA)
9 (CQA)
3 (CQA)
3 (CQA)
Low Impact
1
1
1
1
High Impact
.
Overall Risk Formula
RPN = Severity x Occurrence x Detectability
.
Occurrence Assessment
•
Potential sources for assessment:
– Experimental data/design space:
• Univariate ranging study
• Multivariate DOEs
• Challenge studies
– Historical data from lab, demo and GMP campaigns
– Phase appropriate risk assessments
– Scientific judgment, external literature/experience
– Platform knowledge (in-house experience)
– Parameter ranges intended at commercial scale/site
Holistic assessment based on the overall body of data/scientific
judgment
.
Risk Assessment Tool
Fishbone – Production Bioreactor
.
Process
- Risk Assessment
IdentifyOptimization
interactions to include
in DOE studies:
Example Cell Culture Development
Mitigation:
- None
- DOE (parameter)
- Linkage in study
Assess >>>> Prioritize >>>>>Mitigate (Actions)
Process Bone
Aggregation
N-Terminal
Modification
Isomerisation
Deamidation
AA
Oxidation
Formation
of Single
Chain
Host Cell
Proteins
DNA
6
6
6
6
10
6
2
2
Analytical
Method
SEC
CZE
CZE
LC-MS/
CZE
ELISA
PiPicogre
en
Facilities:
Scale Effect
0
0
0
0
0
0
0
Operations:
Culture
Duration
0
9
9
0
9
0
Operations:
Temp control
0
9
0
0
9
Operations:
Temp shift
0
0
0
0
Operations:
DO control
0
9
0
Operations:
pH control
0
3
Operations:
pCO2 control
0
0
Process
Sub-Bone
Relative Rank
Production
Bioreactor
Operation
Score
Mitigation
0
0
None
0
0
252
Time Course Study,
Operational Procedures
0
0
0
198
DOE Study,
Control Strategy
0
0
0
0
0
Operational Procedures
0
9
0
0
0
198
DOE Study,
Control Strategy
9
0
9
0
0
0
180
DOE Study,
Control Strategy
0
0
3
0
0
0
30
DOE Study,
Operational Procedures
.
CESDS
1
Block
Early
Gen
33.5
Temp
36
N-1 Temp
4.5
Inoc Age
.
6.8
pH
6.8
N-1 pH
2.5
[Media]
0.2
Curv
98.289
Mean CO2
90
70
50
160
30
120
80
1.25
40
.75
.25
-0.25
2.7
2.5
2.3
6.95
6.85
6.75
6.95
6.65
6.85
6.75
5
6.65
4.75
4.5
4.25
37
4
36.5
36
35.5
34.5
35
34
33.5
33
32.5
Late
Early
2
1
d14 Lac
±19.07497
30.96819
±0.1229
1.705595
D14 titer
±0.946038
7.815795
Oxy
% Trp(25)
±1.72782
±0.124822
0.94337
Pho
15.16178
% Pyruvylation
% Ser(46)
Critical Quality Attributes
±0.25605
2.768565
% des H/HG
DOE Studies to Define Design Space
Bringing Together Process and Product Attributes
Example – Production Bioreactor
Prediction Profiler
5
4
3
1.75
2
1.25
0.75
0.25
20
17.5
15
12.5
10
7.5
12
10
8
6
2.5
4
1.5
2
100
1
60
-20
20
58.813
Mean Nova
DO (D3-D14)
DOE Study to Characterize CQA Impact of
pCPP’s - Central Composite Design
 Execute the corner points of
the design space - to confirm
with data the model
predictions of acceptability.
• Multivariate (interaction)
confirmation.
 Execute the Axial points to
expand the information
around the variation of a
single variable.
• Univariate (main effects)
confirmation
.
Range Comparison
Knowledge
Space
• The space explored by experimentation and may include areas of
failures.
Design
Space
• The multidimensional combination and interaction of input variables
(e.g., material attributes) and process parameters that have been
demonstrated to provide assurance of quality. (ICH Q8)
Normal
Operating
Range
(BR
Range)
• Region within the design space that define the operational limits (for
process parameters and input variables) used in routine
manufacturing. (A-Mab :a case Study)
.
Control Parameter (Input) Criticality Decision Tree
Complete Following Characterization DOEs
Process
Parameter
Risk Assessment
Probable Risk
to CQAs?
No – do not evaluate in
empirical study
Yes - evaluate in empirical study
Statistical
Significance?
No
Yes
Critical
(CPP/CIPC)
Yes
Practical
Significance?
.
No
Not Critical
(OPP/IPC)
Control Parameter Criticality Assessment
Assuring Significance and Relationship to CT Lots
 Link to clinical manufacturing experience:
• Approach considers variability in development data (5 L to 500 L)
compared to clinical scale experience (5000 L in this example)
 Impact on CQA consists of:
• Is the effect of the parameter on a CQA statistically sig (P<0.05)
• If yes, a comparison is then made between lab DOE CQA data and
large scale CT lot CQA data, to assess whether the prospective
manufacturing ranges yield CQA performance that exceeds clinical
experience. If so, then CPP. If not, then OPP.
 Approach is reliant on/limited by requirement for a reasonable amount
of clinical manufacturing experience
.
Documented Summary of Key Outcomes:
Integrated Control Strategy (ICS)
Analytical
• Origin of critical quality attributes
• Batch release specifications
• In process testing for process validation (PVAC)
Parametric
•Control parameters for drug substance manufacturing
•Rationale for criticality assessment
• Why are specific parameters critical, and why are others noncritical?
Microbiological
•Risk assessment and evaluation
•Selection of control points and limits based on facility Environmental
monitoring history
ICS – the foundation for Process Monitoring and
Process Validation
.
Developing Process Validation Acceptance
Criteria
Parametric Control Strategy (CPPs Maintained in
Range)
Analytical Control Strategy (Demonstrate Unit
Operation Functional Claims)
Drug Substance Specifications
Impurity Profile
In Process Microbiological/Viral Specifications
.
Summary and Commentary (1)
•
Possible to translate QbD “concepts” into a Robust Control Strategy
that leads to successful Process Validation and (hopefully) fewer
process issues through the lifecycle (20+ year investment)
•
Potential Benefits of the approach
1. Increased confidence in validation success (PV complete on 4
processes/molecules)
2. Reduced numbers of deviations in PV (number and severity)
3. Foundation for Post PV process monitoring plan (lifecycle
approach)
.
Summary and Commentary (2)
•
QbD is just they way you do process development and technology
transfer and validation in the 21st century....increasing expectation to
see “elements of QbD” in BLA/MAA
•
QbD focus is not “regulatory relief” focus is a systemic approach that
leads top greater process/product understanding and robustness for
the long term
•
We have not called out our submissions as “QbD” to date and have
not clamed “design space”
•
Regulatory Agencies...where is their head at?
.
Acknowledgements
Ciaran Brady
Matthew Osborne
Sarah Demmon
Kristi Griffiths
Steve Galvin
Pete Lambooy
Tom Black
Graham Tulloch
Marie Murphy
Mark Milford
Bryan Harmon
.
Implementing QbD in a CMO
Graham McCreath
(Fujifilm Diosynth Biotechnologies)
EPSRC CIM - UCL VISION Event 2 Dec 2013
Implementing QbD in a
Contract Manufacturing
Organisation
ESPRC
Annual Research
Meeting Monday 2nd
December 2013, UCL
Graham McCreath, PhD
Head of Process Design
…alternatively……
Adventures in Time
and (Design) Space
ESPRC
Annual Research
Meeting Monday 2nd
December 2013, UCL
Graham McCreath, PhD
Head of Process Design
Presentation Outline
►
►
►
►
►
►
Brief introduction to Fujifilm Diosynth
Biotechnologies (FDB)
Brief introduction to Quality by Design (QbD) and
components employed at FDB
Some challenges and benefits of QbD development
programs
How does FDB as a CMO equip itself to deliver
critical components of QbD
Tools and considerations for QbD components
Conclusions
Fujifilm Diosynth Biotechnologies
Billingham,
UK (FDBK)
MERCK /
MSD
Combining >25 years
of Biologics CMO
experience and track record
Acquired April 2011 by
Fujifilm Corp (80%)
Mitsubishi Corp (20%)
RTP, North Carolina,
USA (FDBU)
Fujifilm Diosynth Biotechnologies
EU and USA operations
► >160 development products and
processes
► 5 Commercial Products
(4 Microbial, 1 MCC/BV)
► 900 Staff
►
Billingham, UK
Manufacturing & Process
Development
►
RTP, NC, USA
Manufacturing & Process
Development
Inspection history:
ICH Q8 introduced the concept of
“Quality by Design, (QbD)” for pharmaceuticals
QbD definition (ICHQ8) is:
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
• Scientific, risk-based, holistic and proactive approach to
pharmaceutical development
• Deliberate design effort from product conception through
commercialisation
• Full understanding of how product attributes and process
relate to product performance
In other words….
DESIGN
UNDERSTAND
CONTROL
QbD Elements Employed at FDB
Quality Target Product
Profile (QTPP)
To address patient requirements
such as: Patient Safety,
Efficacy, Dosage and route of
administration
The multidimensional combination
and interaction of input variables
and process parameters that have
been demonstrated to provide an
assurance of quality –ICHQ8
“Process Validation:
General Principles and
Practices” 2011, FDA.
Critical Quality
Attribute (CQA)
Risk
Assessment
Defines desired product
performance (CQA and
KPA)
Identify potential sources
of variability affecting the
CQAs (e.g. Process steps,
Raw Materials) –ICHQ9
Design
Space
Control
Strategy
Continuous
Improvement
Control manufacturing
processes to produce
consistent product quality
over time –ICHQ10
Lifecycle approach to
continuously monitor and
evaluate the process –ICHQ10
What challenges do CMOs face with the
application of QbD
►
►
Varies with customer – generally large pharma / large biotech
expect QbD elements to be included in development
programs
Small to medium sized customers can be less aware of QbD
and therefore require education; particular concerns around
• Time – perception that QbD adds significant time into a development
program (requires a rational staged approach to avoid!).
• Many small / virtual biotechs have a horizon that stretches only as far
as the next clinical or financial milestone.
• Expectation that costs and resource requirements are high.
• Unaware or unconvinced of the benefits including regulatory flexibility.
►
CMOs require a flexible approach to deliver critical elements
of QbD
What are the benefits of a QbD directed
development program ?
►
There can be substantial business benefits to both small and large customers
•
Process understanding allows the most appropriate control strategy to be developed
•
Lower manufacturing costs, higher efficiency, right-first-time, etc
►
QbD development programs reduce the risk of failure in process performance
qualification (PPQ) campaigns and commercial manufacturing - drug shortages
are a concern to FDA.
►
Additional data with which to assess deviations and RCAs.
►
Can increase value to the development and product package especially if
subsequently licensed onto big pharma / biotech who often have QbD expectation.
►
There can also be regulatory benefits
•
Ability to meet increased expectations of regulators
•
Lest post approval updates
– Recent citations by regulators
o Unacceptable levels of process understanding
o Un-indentified factors causing process/product variability
•
Regulatory “flexibility” may not always be considered a primary benefit
QbD = sound science based PD
Does science beat stainless (and plastic)….?
► If it is becoming recognised that investing in good
manufacturing science is as important as
investing in fixed or disposable assets then what
does this mean to us and our clients….?
► How can we offer the best possible science to
recognise the benefits of a QbD development
program…..?
►
QbD Tools & Support areas
Accuracy
Range
Repeatability
Intermediate
Precision
Assay
Qualification
Linearity
1.8
Specificity
1.6
1.4
1.0
0.8
0.6
0.4
0.2
TEMP
A IR F L
pH
PO 2
In le t_ O 2
E x tr a O 2
G ly c e r o l f
RQ
In le t_ C O 2
OUR
S T IR R
CER
E x it_ O 2
E x it_ C O 2
0.0
M e th a n o l f
V IP [7 ]
1.2
LOQ
QbD Tools & Support area
Edge of
Failure
Design of
Experiments
(DOE)
Normal
Operating
Range
Design
Space
Centre Face
RSM Design
Mapping
1.5
B7
B6
B5
B8
0.5
t[2 ]
Multivariate
Data Analysis
(MVDA)
1.0
Batch
Trajectory
Scale Down
Model
Qualification
B7
B8
B6
B5
B8
B6
B6
B5
B6
B7
B5
B8
B5
BB8
6 B7
B5B8
B5
B7
B8B6
B7
B6
B8
B5
B8
B7
B7
B6
B5
B8
B5
B6
B5
B8
B6
B5
B7
B6
B7B7B5
B5
B7 B5
B6
B8 B5
B8B8
B5
B6
B8
B7
B7
B6 B6
B7BB6
B5
8
B8B6
B8
B7B7B6
B5
B6B5B6 B8B8
B5
B7 B7
B8
B6
B7
B7
B8B6B7
0.0
-0.5
-1.0
() 99%
Hotelling’s T2
limit
-1.5
B8
-2.0
-6
-5
-4
-3
-2
-1
0
1
2
t[1]
Statistical
Process
Control
(SPC)
3
4
5
6
(▲) Laboratory scale
data (prediction set)
(▲) Manufacturing scale
data (model set)
Mechanistic
Models
 Fujifilmdb
Control Charts
Slide No: 47
Training for Project Teams (and customers..!)
►
Master class in QbD
–
QbD origins and execution
►Statistics for Technical Professional
– Statistical theory, Outlier detection, Statistical comparison,
►Introduction to Design of Experiments (DOE)
– Advantage of DoE, How applied in FDB
►Master class in Design of Experiments (DOE)
– Details on design and analysis of different type of designs
– DX-8® and JMP® focused
►Statistical Process Control
– SPC theory and application in FDB manufacturing
– Minitab® focused
►FMEA parameter assembly and scoring
– Process parameter identification
– Scoring system for S, O, D
– RAPTA focused
Slide No: 48
Training for Project Teams (and customers..!)
►Statistical Process Control
– SPC theory and application in FDB manufacturing
– Minitab® focused
►FMEA parameter assembly and scoring
– Process parameter identification
– Scoring system for S, O, D
– RAPTA focused
►Identifying and classifying product quality
attributes
– SOP and in-house tools
►Development and validation
activities leading to product
commercialisation.
– SOP and White papers
49
Where do we see most QbD approaches
being applied for our clients ?
Technology Transfer
Process
Creation
Phase I Tox
Strong use of DoE / Risk
assessment
in development …
CQA mostly DS specification
Sample retention…………….
CQA
Confirmation
Phase II
Process
FMEA
Phase III
PPQ
Assay
Process
Qualification Characterisation
Control
Strategy
Critical Quality Attributes
►
A critical quality attribute (CQA) is 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.
►
CQAs typically include those properties or characteristics that
affect the safety and efficacy of a product.
►
Identifying CQAs is important for biopharmaceuticals and is
becoming a regulatory expectation for market submission.
►
The identified CQAs are also used in assessment of process
parameters during risk assessments ahead of process
characterisation studies therefore early identification is useful.
►
FDB have tools and guidance's that assist clients in CQA
identification
Risk Assessments
►
►
A systematic process for the assessment, control, communication and
review of risk to the quality of the drug product across the product lifecycle
(ICH Q9)
Risk assessment tools allow you to distil down, assess, rationalise and
prioritise what important in complex manufacturing processes
►
►
►
►
►
Identify parameters & material
attributes most likely to affect
product quality & process
performance
Highlight which steps & RMs
impact one or more CQAs
Link parameters & RM attributes
to CQAs
Steers product and process
development
Informs design space & control
strategy emphasis
Risk Assessment Tools
High RPN or Severity score:
Parameter for further evaluation
and characterization through
formal process ranging
High Occurrence/Detection Score:
Mitigation through control strategy,
facility/equipment improvements,
PAT
Risk
Assessment
planning
Attribute
Impact
Assessment
Risk
Scoring
Risk Scoring dictates
Mitigation Strategy
Risk
Mitigation
Strategy
Initial RPN scores may change in accordance with mitigation
E.g.: Failure Mode Effects Analysis (FMEA) for a process step
Process FMEA tool – RAPTA
Risk Assessment Process Template Application
• Inbuilt lists of unit variables based on previous
experiences in process characterisation and
Manufacturing
• Automated visual plots (heat maps, sorted score
plots)
• Emphasis on impact of each individual parameter
on CQA’s:
• Impact, Detection and Occurrence within the unit
operation.
• Links the various options of Mitigation Strategy
relevant to process and our manufacturing facilities
54
Design Space Development: Using Statistical Design of
Experiments (DOE) Tool
Design Options
►
There are always a number of options available for
experimental design.
►
DoE is one option but sometime single factor studies are also
appropriate.
►
When DoE is employed, it is important that clients recognise
that high resolution options, whilst providing greater levels of
understanding also take more time, generate more materials
for analytical testing and require more feedstock.
►
FDB always produce a range of options to clients outlining
risks and benefits of each approach.
Discussion on DoE options
Slide No: 57
DOE: Process Characterisation Example
Several design
options proposed
for the study
Parameters for
design derived from
FMEA
Blocking ensures
variation due to
different AKTA &/or
Analyst is taken into
account
Risk Assessment
(FMEA) output
• 25-1=16 Run RV design with 3
centre points per block
design option adopted for
Stage-1 Screening study
• Centre points will help to
account and analyse the
variability due to Process (i.e.
batch to batch variation)
• Centre points also help to
identify curvature (nonlinearity)
Similar design option proposed for different steps in the process
Summary from JMP®
Response with 95%
Confidence interval at
centre point conditions
of factors
Factors at centre
point conditions
Slide No: 59
B
D
G
Monte-Carlo Design Space Simulation
Specification of Peak 1
area<15%
Respective failure rates
Slide No: 61
Control strategy evolution (high level)
Development trajectory
pCQA
identification
Process
development
CQA vs
process steps
Design space
Analytical
development
FMEA
Product
characterisation
CQA vs process
parameters
Method
validation
FMEA
FMEA
Specifications
Raw materials
Draft control
strategy
Updated control
strategy
(several iterations)
PPQ control
strategy
Conclusions - What do you get from a QbD
development program ?
►
A robust well controlled process
►
Risk assessments as an iterative description and repository of process
knowledge.
►
Documented description and justification for small scale process models.
►
Proven acceptable ranges for high risk input parameters and raw materials.
►
Comprehensive description of process and analytical controls.
►
A control strategy specifically designed constructed to minimise risk to product
and process CQAs.
►
QbD is a continuum not a step function – Many traditional pharmaceutical
development approaches are consistent with QbD; all applications have some
aspects of science and risk based approaches
Thanks to
►
►
►
►
Simon Hanslip
Mahesh Shivhare
Somi Mohammadi
Carol Fisher
PANEL DISCUSSION:
Graham McCartney, Graham McCreath, Richard Francis, Stephen Ward
Chair: Neil Weir
EPSRC CIM - UCL VISION Event 2 Dec 2013
Download