1. Connecting People, Science and Regulation ®
Risk Assessment and DoE must be used
in Synergy for the success of QbD
Alain Poncin
Process Development Unit Manager
LFB Biotechnologies
Quality by Design, Frankfurt 1
2. Connecting People, Science and Regulation ®
Quality by Design
Science
Statistics Knowledge
Risk Management
- of
the production
process
- of the product
summarized in the Risk
Assessment Report
2
3. Connecting People, Science and Regulation ®
Risk Management
Technical report 42, PDA (2005)
Pharma : ICH Q8 : Pharmaceutical Development (2005) and annex (2007)
ICH Q9 : Quality Risk Management (2005)
Compliance
ICH Q10 : Pharmaceutical Quality System (2008)
ICH Q11 : Development and Manufacture of Drug Substance (concept
paper, 2008, draft expected in 2009)
Focused on Product Quality
Medical Device ISO13485 : Medical devices – Quality management systems – Requirements
Compliance for regulatory purposes
ISO14971 : Medical devices – Application of Risk Management to medical
Economy devices
Focused on Product Quality + Product Availability
3
4. Connecting People, Science and Regulation ®
Risk Management at LFB Biotechnology
When
Standard initial Risk analysis (FMEA) as soon as a first lab process gives
satisfactory results
Updated during product/process development, clinical development and post
approval.
How
Based on standardised ‘’blocks’’ (Upstream, Harvest, Chromatography,
Ultra/diafiltration,…)
‘’Personnalised’’ using what is known : protein stability, ease/difficulty of steps,
occurrence of hasard,…
Ranks the work to be performed during development and process characterisation
Reflects and summarises all what is known about the protein and the process
4
6. Connecting People, Science and Regulation ®
2- Table of content
N° System Sub system Justification
5 Capture by Capture of target
Chromatography product after harvest
5.1 Cleaning before
chromatography
5.2 System assembly
5.3 Storage of intermediates
5.4 Sample preparation
5.5 Purification
5.6 Cleaning after purification
6
7. Connecting People, Science and Regulation ®
3- Risk Analysis
Based on previous experience, building of new blocks for this
first Risk Analysis at LFB Biotechnologies
Copy and paste to a ‘’white’’ Risk Assessment
Missing blocks (viral inactivation, nanofiltration,…) specially
for plasma product
Not adapted to LFB Biotechnologies (history,…)
Final analysis not homogeneous
7
8. Connecting People, Science and Regulation ®
Lack of homogeneity :
N Product, Part, Possible Possible S Possible cause P Risk control D RPN
# System, Hazard/ Effect
Fonction Failure (harm)
2.2 Buffer Wrong buffer Ultra/diafiltrati 6 Defective 3 Qualified 3 54
preparation (pH, on failure, equipment (pH, equipment,
conductivity) contamination scale,…) preventive
or degradation maintenance
of target Trained staff
product
Calibration before
use
5.5 Purification Wrong buffer Contamination 4 Human error 2 Trained staff 3 24
for of target (inversion of Written
equilibration/ product buffers) SOP/method of
wash and or production
elution
6.8 Ultra/ Wrong buffer Contamination/ 7 Human error 3 Written SOP 5 105
Diafiltration for loss of sample Trained staff
equilibration
8
9. Connecting People, Science and Regulation ®
To increase homogeneity and quality of RA:
Define risks
For the Product safety/efficacy/availability
-contamination (physical, chemical,…)
-degradation (lower yield, production failure, immunogenicity)
Identify risks for a general ‘’process’’
Identify Possibles causes (human, material,…)
Identify the measures to reduce the risk
Adapt for each kind of process : fermentation, chromatography,
ultrafiltration,…
9
10. Connecting People, Science and Regulation ®
N° System Sub system Justification
5 Capture by Capture of target
Chromatography product after harvest
5.1 System assembly and calibration
5.2 Cleaning before
chromatography
5.3 Storage and expiration time of
intermediates
5.4 Sample preparation
5.5 Purification
5.6 Product recovery
5.7 Cleaning after purification
5.8 Storage of Matrix/column
10
11. Connecting People, Science and Regulation ®
2- Table of content
N° System Sub system Justification
6 Ultra/diafiltration Buffer Exchange
6.1 System assembly and
calibration
6.2 Cleaning before
ultra/diafiltration
6.3 Storage and expiration time of
intermediates
6.4 Sample preparation
6.5 Ultra/diafiltration
6.6 Product recovery
6.7 Cleaning after
ultra/diafiltration
6.8 Membrane/carter Storage 11
12. Connecting People, Science and Regulation ®
5- Quantification of Residual Risk
N# Product, Possible Possible effect S Possible cause P Risk Control, D RPN
Part, hazard/ (harm) of the Measures of
System, failure hazard/failure Risk reduction,
Function, Tests
Process
5.4.5 Sample Wrong Purification 5 Reagents 2 Description/QC of 3 30
preparation preparation failure, identity/Quality raw material
(salt Production Approved
addition, … stopped suppliers
5.5.1 Purification Wrong Contamination 5 Human error 2 Trained staff 3 30
buffer (pH, of Drug product Written
conductivity) SOP/method of
production
Automation
5.5.6 Purification Ineffective Contamination 7 To be determined 5 Identification of 7 247
purification of Drug Product critical factors
12
13. Connecting People, Science and Regulation ®
Risk Priority Number
RPN = Severity x Probability x Detectability
Require Internal Policy Definition
At LFB Biotechnology : 4 levels :
o 1 to 100 : broadly acceptable region
o 101 to 150 : as low as reasonable practicable region (ALARP), part I
o 151 to 250 : as low as reasonable practicable region (ALARP), par II
o 251 to 1000 : intolerable region
13
14. Connecting People, Science and Regulation ®
6- Initial Risk Analysis conclusions
Step System Current status/notes
3 Clarification Need further comparability studies to assess
starting material equivalency
4 0.22 µm filtration Final choice of filter type
5 Capture by Not used in standard conditions, need identification
Chromatography of critical parameters to obtain a reproducible
process
6 Viral Inactivation Need first evaluation of viral clearance before First
in Man
7 Chromatography Well known and controlled step, yet close to the
optimum (load, wash and elution)
8 0.22 µm filtration Final choice of filter type
14
15. Connecting People, Science and Regulation ®
From initial Risk analysis : identification of capture
purification step as a high risk Step
To reduce this risk : identification of critical factors and
critical quality attributes.
Experience (int./ex.) Process development Initial Risk Assesment
Publication review First lab scale production
Patent review
Identification of
critical factors by
DoE
15
16. Connecting People, Science and Regulation ®
Particularity of DoE in Process Development
Protein folding : protein concentration, pH, Salt, Organic solvent,… 14 factors
Chromatography : pH and conductivity for equilibration and elution, sample load, matrix and
column resolution,… 10 factors
Ultrafiltration : pump speed, Pin, Pout, membrane, temperature, volume of buffer,… 8 factors
Design space highly multidimensional
2 n experiments (full factorial) cannot be used,
2 n-k experiments (semi factorial) used for identification of critical factors
16
17. Connecting People, Science and Regulation ®
Even 2 n-k experiments are a huge amount of work (process
+ analytics)
Before starting experiments :
Are Analytic tools sufficient ?
Which Design : (semi) factorial ?
Which factors to test ?
Which Range (Design Space) ?
17
21. Connecting People, Science and Regulation ®
Output (Quality attributes)
capture step : high yield, reproducible
Quantification of target protein biological activity
Quantification of total proteins
Purity by SDS-PAGE
On Load, Flow Through and Peak (23 samples)
Calculation of step yield and specific activity
21
23. Connecting People, Science and Regulation ®
But in another region of the space…
Design-Expert® Software
Design-Expert® Software
Yield Yield
92 92
10 10
100 100
X1 = B: Conductivity sample X1 = B: Conductivity sample
X2 = D: pH elution
75 X2 = D: pH elution
75
Actual Factors
A: pH sample = 7.07 Actual Factors
50
A: pH sample = 7.07
Y ie ld
C: Load = 30.00 50
Y i e ld
E: Gradient = 15.41 C: Load = 10.00
25 E: Gradient = 15.41
25
0
0
8.50 10.00
8.00 8.13
7.50 6.25 8.50 10.00
7.00 4.38 8.00 8.13
D: pH elution B: Conductivity sample
6.50 2.50 7.50 6.25
7.00 4.38
D: pH elution B: Conductivity sample
6.50 2.50
23
24. Connecting People, Science and Regulation ®
Results : critical factors : half normal plot of effects
Design-Expert® Software Half-Normal Plot
Yield
Error from replicates
H a lf-N o rm a l % P ro b a b ility
99
E Shapiro-Wilk test
W-value = 0.962
p-value = 0.794 95
A: pH sample E
B: Conductivity sample 90
C: Load
D: pH elution
80 A
E: Gradient
Positive Effects 70
C
Negative Effects
A 50
30
20
10
0
C 0.00 7.63 15.25 22.88 30.50
|Standardized Effect|
- +
Factor Effect = slope/2
24
25. Connecting People, Science and Regulation ®
Or pareto chart
Design-Expert® Software Pareto Chart
Yield E
6.2 4
A: pH sample
B: Conductivity sample A
C: Load
D: pH elution
t-V a lu e o f |E ffe c t|
4.6 8
C
E: Gradient Bonf erroni Limit 4.38176
Positive Effects
Negative Effects
3.1 2
t-Value Limit 2.57058
1.5 6
0.0 0
1 2 3 4 5 6 7
Rank
25
26. Connecting People, Science and Regulation ®
Are they significants ? ANOVA
ANOVA for selected factorial model
Analysis of variance table [Partial sum of squares - Type III]
Sum of Mean F p-value
Source Squares df Square Value Prob > F
Model 4129 3 1376 28,8 0.0014
A-pH sample 1301 1 1301 27,2 0.0034
C-Load 968 1 968 20,3 0.0064
E-Gradient 1861 1 1861 38,9 0.0015
Curvature 960 1 960 20,1 0.0065
Residual 239 5 48
Lack of Fit 207 4 52 1,6 0.5244
Pure Error 32 1 32
Cor Total 5328 9
- +
Factor
26
27. Connecting People, Science and Regulation ®
ANOVA : Statistical analysis : can you do it ?
Design-Expert® Software Normal Plot of Residuals
Yield
Color points by value of
Yield: 99
N o rm a l % P ro b a b i lity
82
95
5 90
80
70
50
30
20
10
5
1
-1.94 -0.97 0.00 0.97 1.94
Internally Studentized Residuals
27
28. Connecting People, Science and Regulation ®
What is the value of the model obtained ?
Design-Expert® Software Predicted vs. Actual
Yield
82.00
Color points by value of
Yield:
82
2
62.75
5
P re d ic te d
43.50
24.25
5.0 0
5.00 24.25 43.50 62.75 82.00
Actual
28
30. Connecting People, Science and Regulation ®
Capture step by DoE and ranges
Design-Expert® Software PAR Factor
One
AS
79
Design Points
NOR
X1 = D: Conductivity ? Edge of
Actual Factors
A: Contact Time = 90
B: pH load = 7.5
61.25
failure
C: Column Volume = 8
E: Elution temperature = 20
AS
43.5
25.75
Definition of
25 + 5 mS/cm
Criticals 8
Parameters
20 25 30 35 40
Nice results isn’t ?
D: Conductivity 30
31. Connecting People, Science and Regulation ®
But following purification results did not correlated well
with prediction from this model.
A new analysis was performed with more powerfull tools.
Two main raisons were implied in the failure of the first
model
One critical factor not identified
High collinearity of some factors
31
32. Connecting People, Science and Regulation ®
How many factors to select ?
Adjusted R-Squared or Mallows’CP statistics
Selection of number of factor
100 2000
80
1500
R Squared
60
Cp
1000
40
500
20
0 0
0 1 2 3 4 5
Number of factor
32
33. Connecting People, Science and Regulation ®
High collinearity : regression by least square not efficient
Use of Ridge statistics
Ridge parameter Factor 1 Factor 2 Factor 3
0.0 (classical regression) 4.2637 -1.5614 -2.9287
0.01 (ridge regression) 0.6741 -0.1870 -0.2684
33
34. Connecting People, Science and Regulation ®
Lack of detection of one critical factor
- +
Factor
Effect = 0 D : pH elution
The new Model is predictible for the purifications performed (up to now)
34
35. Connecting People, Science and Regulation ®
Finally, update of the initial Risk Analysis
N# System, Hazard Possible effect S Possible cause P Risk Control, D RPN
5.4 Sample Wrong Purification 5 Reagents 2 Description/QC of 3 30
preparation preparation failure, identity/Quality raw material
(salt Production Approved
addition, … stopped suppliers
5.5.6 Purification Ineffective Contamination 7 To be determined 5 Identification of 7 247
purification of Drug Product critical factors
5.4 Sample Wrong Increase of 5 Reagents 5 Quantification of 5 125
preparation preparation proteolytic identity/Quality proteolytic
(buffer activities activities
addition) Production Written SOP
failure Automation
5.5.6 Purification Ineffective Contamination 6 Contact Time too 4 Qualification of 5 120
purification of Drug Product short (< 2 min) equipment,
Other factors preventive
under control maintenance
Trained staff
35
36. Connecting People, Science and Regulation ®
Synergy of Risk Management and DoE
Experience (int./ex.) First lab scale Initial Risk Assesment
Publication review Production
Patent review Process
Identification of critical
factors by DoE Process Development
Risk Assesment Update First in Man
Process optimisation GMP for Phase I
Process capability indices Risk Assesment Update Phase I
(Six Sigma)
Process Characterisation
Risk Assesment update Phase II/IIII
Process Validation
CTD
Statistics Risk Assesment update Phamacovigilance Phase IV
(trends,…)
Product discontinuation
36
37. Connecting People, Science and Regulation ®
And in the future
Increased number of experiments (// chromatography, 96
wells technology)
Application to ultra/diafiltration
Introduction of additional statistic tools (bayesian
statistics, Monte Carlo simulation, …)
More applications of Six Sigma
PAT
37
38. Connecting People, Science and Regulation ®
Economical impact of QbD : example
Eurogentec s.a.
60
50
40
30
20
10
0
87 88 89 90 91 92 93 94 95 96 97 98 99- 00- 01- 02- 03- 04- 05- 06- 07-
00 01 02 03 04 05 06 07 08
Turnover in millions € Nber of employees x 10
38