Purpose of statistics

1. Role of Statistics in
Pharmaceutical Development
Using Quality-by-Design
Approach – an FDA Perspective
Chi-wan Chen, Ph.D.
Christine Moore, Ph.D.
Office of New Drug Quality Assessment
CDER/FDA
FDA/Industry Statistics Workshop
Washington D.C.
September 27-29, 2006

2. 2
Outline
 FDA initiatives for quality
 Pharmaceutical CGMPs for the 21st Century
 ONDQA’s PQAS
 The desired state
 Quality by design (QbD) and design space (ICH Q8)
 Application of statistical tools in QbD
 Design of experiments
 Model building & evaluation
 Statistical process control
 FDA CMC Pilot Program
 Concluding remarks

3. 3
21st Century Initiatives
 Pharmaceutical CGMPs for the 21st
Century – a risk-based approach (9/04)
http://www.fda.gov/cder/gmp/gmp2004/GMP_
finalreport2004.htm
 ONDQA White Paper on Pharmaceutical
Quality Assessment System (PQAS)
http://www.fda.gov/cder/gmp/gmp2004/ondc_
reorg.htm

4. 4
The Desired State
(Janet Woodcock, October 2005)
A maximally efficient, agile,
flexible pharmaceutical
manufacturing sector that
reliably produces high-quality
drug products without
extensive regulatory oversight
A mutual goal of industry,
society, and regulator

5. 5
FDA’s Initiative on Quality by Design
 In a Quality-by-Design system:
 The product is designed to meet patient
requirements
 The process is designed to consistently meet product
critical quality attributes
 The impact of formulation components and process
parameters on product quality is understood
 Critical sources of process variability are identified
and controlled
 The process is continually monitored and updated to
assure consistent quality over time

6. 6
Quality
by
Design
FDA’s view on QbD, Moheb Nasr, 2006

7. 7
Design Space (ICH Q8)
 Definition: The multidimensional combination and
interaction of input variables (e.g., 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

8. 8
Current vs. QbD Approach to
Pharmaceutical Development
Current Approach QbD Approach
Quality assured by testing and
inspection
Quality built into product & process
by design, based on scientific
understanding
Data intensive submission – disjointed
information without “big picture”
Knowledge rich submission – showing
product knowledge & process
understanding
Specifications based on batch history Specifications based on product
performance requirements
“Frozen process,” discouraging
changes
Flexible process within design space,
allowing continuous improvement
Focus on reproducibility – often
avoiding or ignoring variation
Focus on robustness – understanding
and controlling variation

9. 9
Pharmaceutical Development &
Product Lifecycle
Candidate
Selection
Product Design & Development
Process Design & Development
Manufacturing Development
Product
Approval
Continuous Improvement

10. Design of
Experiments
(DOE)
Model Building
And Evaluation
Process Design & Development:
Initial Scoping
Process Characterization
Process Optimization
Process Robustness
Statistical Tool
Product Design & Development:
Initial Scoping
Product Characterization
Product Optimization
Manufacturing Development
and Continuous Improvement:
Develop Control Systems
Scale-up Prediction
Tracking and trending
Statistical
Process Control
Pharmaceutical Development
& Product Lifecycle

11. 11
Process Terminology
Process Step
Input Materials Output Materials
(Product or Intermediate)
Input
Process
Parameters
Measured
Parameters
or Attributes
Control Model
Design
Space
Critical Quality Attributes
Process
Measurements
and Controls

12. 12
Design Space Determination
 First-principles approach
 combination of experimental data and
mechanistic knowledge of chemistry, physics, and
engineering to model and predict performance
 Statistically designed experiments (DOEs)
 efficient method for determining impact of
multiple parameters and their interactions
 Scale-up correlation
 a semi-empirical approach to translate operating
conditions between different scales or pieces of
equipment

13. 13
Design of Experiments (DOE)
 Structured, organized method for determining
the relationship between factors affecting a
process and the response of that process
 Application of DOEs:
 Scope out initial formulation or process design
 Optimize product or process
 Determine design space, including multivariate
relationships

14. 14
DOE Methodology
(1) Choose experimental design
(e.g., full factorial, d-optimal)
(2) Conduct randomized
experiments
(4) Create multidimensional
surface model
(for optimization or control)
(3) Analyze data
Experiment Factor A Factor B Factor C
1 + - -
2 - + -
3 + + +
4 + - +
A
B
C
www.minitab.com

15. 15
 Models for process development
 Kinetic models – rates of reaction or degradation
 Transport models – movement and mixing of mass or heat
 Models for manufacturing development
 Computational fluid dynamics
 Scale-up correlations
 Models for process monitoring or control
 Chemometric models
 Control models
 All models require verification through statistical
analysis
Model Building & Evaluation -
Examples

16. 16
 Chemometrics is the science of relating
measurements made on a chemical system or
process to the state of the system via application
of mathematical or statistical methods (ICS definition)
 Aspects of chemometric analysis:
 Empirical method
 Relates multivariate data to single or multiple responses
 Utilizes multiple linear regressions
 Applicable to any multivariate data:
 Spectroscopic data
 Manufacturing data
Model Building & Evaluation -
Chemometrics

17. 17
Statistical Process Control -
Definitions
 Statistical process control (SPC) is the application
of statistical methods to identify and control the
special cause of variation in a process.
 Common cause variation – random fluctuation of response
caused by unknown factors
 Special cause variation – non-random variation caused by
a specific factor
Upper Control Limit
Lower Control Limit
Target
Upper Specification Limit
Lower Specification Limit
Special cause variation?
3s

18. *Percent out of specification beyond the high risk specification limit.
σ
3
)
SL
X
(
min
Cpk


2.28%
2s
0.7
15.9%
1s
0.33
0.135%
3s
1
0.003%
4s
1.33
0
5s
1.7
0
6s
2
Expected Avg. OOS%*
|X - SL|
Cpk
2.28%
2s
0.7
15.9%
1s
0.33
0.135%
3s
1
0.003%
4s
1.33
0
5s
1.7
0
6s
2
Expected Avg. OOS%*
|X - SL|
Cpk
Industry Practice is to
consider processes with
Cpk below 1.33 as “not
capable” of meeting
specifications.
Cpk = 1.33 Cpk = 0.33
Process Capability Index (Cpk)

19. 19
Quality by Design & Statistics
 Statistical analysis has multiple roles in
the Quality by Design approach
 Statistically designed experiments (DOEs)
 Model building & evaluation
 Statistical process control
 Sampling plans (not discussed here)

20. 20
CMC Pilot Program
 Objectives: to provide an opportunity for
 participating firms to submit CMC information based on QbD
 FDA to implement Q8, Q9, PAT, PQAS
 Timeframe: began in fall 2005; to end in spring 2008
 Goal: 12 original or supplemental NDAs
 Status: 1 approved; 3 under review; 7 to be submitted
 Submission criteria
 More relevant scientific information demonstrating use of QbD
approach, product knowledge and process understanding, risk
assessment, control strategy

21. 21
CMC Pilot - Application of QbD
 All pilot NDAs to date contained some elements of
QbD, including use of appropriate statistical tools
 DOEs for formulation or process optimization (i.e.,
determining target conditions)
 DOEs for determining ranges of design space
 Multivariate chemometric analysis for in-line/at-line
measurement using such technology as near-infrared
 Statistical data presentation and usefulness
 Concise summary data acceptable for submission and
review
 Generally used by reviewers to understand how
optimization or design space was determined

22. 22
Concluding Remarks
 Successful implementation of QbD will require
multi-disciplinary and multi-functional teams
 Development, manufacturing, quality personnel
 Engineers, analysts, chemists, industrial
pharmacists & statisticians working together
 FDA’s CMC Pilot Program provides an
opportunity for applicants to share their QbD
approaches and associated statistical tools
 FDA looks forward to working with industry to
facilitate the implementation of QbD