25-10-2023

QbD

Continual improvement and lifecycle management

Product CQA CMA Risk Systematic Control

profile & Analysis experiments Strategy
CPP
Design of Experiments (DoE)
Quality critical Critical Risk Empirical or Control
Target quality material analysis enhanced strategy
product attributes and and (Design of based on
profile process Impact Experiment) design
(QTPP) attributes analysis to generate a space
/ that of CMAs design space that
impact and CPPs includes
Target CQAs on CQAs specificat
Product ions
profile
(TPP) Iterative
process

QbD Design of Experiments (DoE)

Continual improvement and lifecycle management Design of Experiments/

Systematic Optimization Approach
Product CQA CMA Risk Systematic Control
profile & Analysis experiments Strategy
CPP

DoE term first introduced by Sir Ronald Fisher in 1925

Quality critical Critical Risk Empirical or Control
Target quality material analysis enhanced strategy
product attributes and and (Design of based on
profile process Impact Experiment) design Pharmaceutical Optimization: Application of statistical
(QTPP) attributes analysis to generate space
that of CMAs a design that technique to get as perfect, as effective, as functional as
/
impact and CPPs space includes
Target CQAs on CQAs specificat
possible pharmaceutical product.
Product ions
profile
(TPP) Iterative
process
 25-10-2023

Design of Experiments (DOE)

Empirical - Old Concept Enhanced - New Concept
COST/OVAT Approach/ Computer assisted statistical design  In the past a traditional approach based on
Trial & Error Approach
Advantages:
univariate approach or a trial and error based
Disadvantages:
•Univariate approach •Fewer experiments and hence approach was used to develop quality
•Time consuming, less expensive pharmaceutical products
•Strenuous •Significant saving of time, effort,  Execution of traditional approach required large
•Unpredictable materials and cost
•Unable to predict interaction •Able to estimating interactions
number of experiments without achieving optimum
•Expensive parameters.
Other drawbacks include :
• Inability to predict interactions
• Non reproducibility
• Streneous
• Expensive
• Time consuming

DOE essential to QbD DOE contd …

• DOE is one of the essential elements of QbD
• It is a statistically organized experimental plan that DoE exhibits wide application in
provides information with high precision on the variation • Screening of important factors
of product/process response as a function of change in the
input. • Optimization of
• This technique involves the application of experimental  Formulation variables
design based on suitable variables, followed by its  Manufacturing Process
statistical evaluation and the search for an optimum  Analytical Instruments
solution using a graphical or a mathematical approach.
• DOE has 3 basic objectives :
Robustness testing of
1. Screening – identification of critical variables and their levels  Formulation
2. Optimisation – identification of optimum input variables to  Process
achieve optimum response  Methods
3. Robustness – identification of sensitivity of response to small
changes in the factor
 25-10-2023

DoE DoE
Levels Value assigned to factor is termed as Level

Variables are the parameters i.e. product or process which

Variables
affect the quality of finished product Effect Change in response obtained by change in input variables
Independent Under the control of formulation or
Variables process scientist Orthogonality Change in response with change in input variable
i.e. Linear effect
Qualitative e.g. Type of emulsifier, machine
type, organic base, reactor
Non-linear effect; represents lack of orthogonality;
Quantitative e.g. concentration of disintegrant, Interaction generally occurs due to inherent quality of data
suspending agent, catalyst loading etc. and can be assessed quantitatively
Not under the control of formulation
Dependent Non-linear effect; represents lack of orthogonality;
scientist/ Characteristics of finished
Variables generally occur due to improper selection of factors,
product e.g. particle size, drug release, Confounding
Yield, Purity level, experimental design and data analysis and can
be assessed qualitatively.
Input variable which influence formulation/process
Factors
characteristics termed as factor

DoE DoE

Experimental Experimental Domain: Part of factor space that is

Domain investigated experimentally

Experimental
Domain

Design Space

Transforming natural variable into a nondimensional coded

Coding variables. Coding depicts effect and interaction using sign
(+) or (-). e.g. (-1, +1) (0, 1, 2) Represents an optimum region and generated by mapping
Design Space
of response obtained from different input variables
 25-10-2023

DoE METHODOLOGIES DoE Methodologies: Factorial Design

DoE Methodologies Simultaneous Statistical modeling and search for an optimum is

Optimization performed after the experimental part is concluded

Simultaneous Optimisation Sequential Optimisation Factorial Design

Factorial design • Introduced by Fischer 1946

Sequential Simplex • Simplest method based on first degree
Fractional Factorial design
Method mathematical model for linear
Plackett – Burmann responses
Design • Commonly used for optimization study
Central Composite Design • Study of each factors at each level
Box-Behnken Design • The total number of experiments as xn
n is number of factor
Mixture Design
x is level
D-optimal design
• The number of levels limited to 2-3

SIMULTANEOUS OPTIMISATION METHOD SR Metformin Tablets

Factorial Design: Composition –
 Simplest method for simultaneous optimization of MCC diluent
pharmaceutical product Ethyl Cellulose (SR polymer) Independent Variables
 Primarily based on first degree mathematical model PVP K-30 (binder)
Mag Stearate (lubricant) Dependent Variable –
for linear responses Drug Release Profile
Aerosil (Glidant)
 It simultaneously studies the effect of each factor at
each level as well as interaction between factors at INDEPENDENT VARIABLES LEVELS
different levels
Sr. No Notation Independent Factors mg/tab -1 +1
 The number of experiments is designated as Xn
1 X1 Microcrystalline Cellulose 80 100
where X represents the number of level and n
represents the number of factors 2 X2 Ethyl Cellulose 5 10

 In general the factorial design can be performes 3 X3 PVP K-30 3 5

either at 2 or 3 levels
Input Variable
 25-10-2023

Experinmental Plan for 23 Factorial Design Experinmental Plan for 23 Factorial Design
Experiment MCC EC mg / tab PVP K-30 Drug Release %
No. mg / tab mg / tab
Main Effect 2 Factor Interaction 3 Factor Interaction
1 80 5 3 80
Experiments Notation X1 X2 X3 X1 X2 X2 X3 X1X3 X1 X2 X3 Drug
Release 78
2 100 5 3
(%)
1 (Null) (-1,-1,-1) -1 -1 -1 +1 +1 +1 -1 80
3 80 10 3 65
2(X1) (+1,-1,-1) +1 -1 -1 -1 +1 -1 +1 78
3(X2) (-1,+1,-1) -1 +1 -1 -1 -1 +1 +1 65
4(X3) (-1,-1,+1) -1 -1 +1 +1 -1 -1 -1 64 4 100 10 3 64
5(X1X2) (+1,+1,-1) +1 +1 -1 +1 -1 -1 +1 72
6(X1X3) (+1,-1,+1) +1 -1 +1 -1 -1 +1 -1 71 5 80 5 5 72
7(X2X3) (-1,+1,+1) -1 +1 +1 -1 +1 -1 -1 62
8(X1X2X3) (+1,+1,+1) +1 +1 +1 +1 +1 +1 +1 60 71
6 100 5 5

7 80 10 5 62
Effect of factor X1 =(78+72+71+60)/4-(80+65+64+62)/4
8 100 10 5 60
= 6 8 . 2 5 - 6 9 . 7 5 = -1.5

DoE Methodologies: Fractional Factorial Design DoE Methodologies: Fractional Factorial Design

Fractional Factorial Design

• High number of factor associated with

high number of runs and highest order • Degree of fractionation is an
of interaction have no significant effect. important aspect which may
result in Confounding effect
• Hence, to overcome this, systematic
design is introduced i.e. FFD/ Partial
factorial design • The degree to which main
• 2 Level Design effect is Confounded is
• The total number of experiment will be estimated by Resolution
xn-r
n is number of factor
r is degree of fractionation
 25-10-2023

DoE Methodologies: Fractional Factorial Design DoE Methodologies: Fractional Factorial Design

DoE Methodologies: Fractional Factorial Design DoE Methodologies: Fractional Factorial Design
Experimental Number of Factors
Runs 2 3 4 5 6 7
4 2 23-1
(Res III)
8 23 24-1 25-2 26-3 27-4
(Res IV) (Res III) (Res III) (Res III)
16 24 25-1 26-2 27-3
(Res V) (Res IV) (Res IV)
32 25 26-1 27-2
(Res V) (Res IV)

Resolution III Design: main effect confounded

Resolution IV Design: no main effect; two factor interaction confounded with two factor
interaction.

Resolution V Design: no main effect and no two factor vs two factor while
two factor confounded with three factor interaction.

Pharmaceuticals: Resolution III, IV, V design preferred

Orthogonality: Infinite Resolution

 25-10-2023

DoE Methodologies: Fractional Factorial Design DoE Methodologies: Plackett-Burman Design

Plackett-Burman Design
• Two level fractional factorial
design/Hadamard design
• Preliminary used for screening purpose
• The number of experiments is multiple
of four i.e. 12, 16, 20, 24 The main effect is orthogonal.
Two factor interaction are
• The total number of experiment will be partially confounded. (Differ
will be one more than factor and is from Resolution III)
expressed as N= K+1
Where, N is number of experimental run
and K is the number of factor

DoE Methodologies: Plackett-Burman Design DoE Methodologies: Central Composite Design

Experimental runs in Plackett Burman design at Coded Variables for Plackett-Burmann design
different factor
Number of experimental Number of factor N First line of design
run 8 +++-+---
8 7
12 ++-+++---+-
12 11
16 ++++-+-+--+---
16 15 Central Composite Design
20 19 20 ++--++-+-+-+----++-
24 23 24 +++++-+-++--++--+-+---- • CCD is response surface methodology
Factors used For factor with non-linear
X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 response Axial point

+1 -1 +1 -1 -1 -1 +1 +1 +1 -1 +1 • Preliminary used for optimization

+1 +1 -1 +1 -1 -1 -1 +1 +1 +1 -1 purpose
-1 +1 +1 -1 +1 -1 -1 -1 +1 +1 +1
+1 -1 +1 +1 -1 +1 -1 -1 -1 +1 +1 • CCD is an amalgamation of FD, Star
+1 +1 -1 +1 +1 -1 +1 -1 -1 -1 +1 Design and Center Point
+1 +1 +1 -1 +1 +1 -1 +1 -1 -1 -1 • Distance between centre point and star
-1 +1 +1 +1 -1 +1 +1 -1 +1 -1 -1 • The total number of experiment will be point is axial point (± α) and it is
-1 -1 +1 +1 +1 -1 +1 +1 -1 +1 -1 will expressed as 2k+2k+n. generally square root of number of
-1 -1 -1 +1 +1 +1 -1 +1 +1 -1 +1 factor. For two factor (± α)=1.414 while
+1 -1 -1 -1 +1 +1 +1 -1 +1 +1 -1
• k is number of factor
for three-factor (± α) = 1.732
-1 +1 -1 -1 -1 +1 +1 +1 -1 +1 +1 • Run at minimum five level
-1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1
 25-10-2023

DoE Methodologies: Central Composite Design DoE Methodologies: Box-Behnken Design

Box-Behnken
Box-BehnkenDesign
Design

Experiment Component Factor X1 Factor X2 • BBD is second order quadratic model,

for factor with non-linear responses
1 Center Point 0 0 • Preliminary used for optimization
purpose
2 -1 -1
3 Factorial Point -1 +1
4 +1 +1 Sr. No. Central Composite Design Box-Behnken Design
1 Experiments are run at five levels Experiments are run at three levels
5 +1 -1 2 Contains embedded FD and star FD with central point form cube.
6 Star Points -1.414 0 points
3 Works when the factor space is Works when the factor space is
7 +1.414 0 outside the experimental domain within the experimental domain
8 0 +1.414 4 Domain volume is 20 Domain volume is 6.
Factors Number of runs
9 0 -1.414
2 13 (5 center point runs) ---
3 20 (6 center point runs) 15
4 30 (6 center point runs) 27
5 48 (6 center point runs) 46

Flow Chart of Optimisation Methodology EXPERIMENTAL DESIGN

Experimental Design
Selection of Influential Variables

Quantification of Influential Variables

Model Application

Selection of Experimental Design

Linear Model Non linear Model Screening Purpose Optimisation

Formulation based on Experimental Factorial Central Factorial Factorial Design

Design Design Composite Design
Design Fractional
Fractional Factorial Design
Factorial Plackett –
Statistical Evaluation Burmann
Design Box- Central Composite
Design Design
Behnken
Plackett – Design
Burmann Mixture Box- Behnken
Validation Design Design Design
Mixture Design
 25-10-2023

DESIGN OF EXPERIMENTS (DoE) DESIGN OF EXPERIMENTS (DoE)

Statistical Techniques for data analysis
“Mechanistic understanding of how formulation and  Data treatment – log data??
process factors affect product performance and  CpK (Process Capability Index)
quality”  SPC (Statistical Process Control)
• Helps in development of formulations with all potential factors to be
 Multi-linear regression
evaluated simultaneously, systematically, and quickly.
 First-order equations (Only main effect)
• Can evaluate the effect of each formulation factor on each response  Y = β0 + β1x1 + β2x2 + β3x3
(and possibly the interaction effects between factors) and identify
 Second-order equations (main effect and interaction effects)
the critical factors based on statistical analysis.
 Y = β0 + β1x1 + β2x2 + β12x1x2 + β11 x12 + β22 x22
• Critical factors can be defined to their optimized levels. Statistical packages
“Design space proposed is subject to regulatory  Minitab  DoE Fusion PRO  OPTIMA
assessment and working within the space is not a change”  Excel add ons  MODDE  JMP

QbD does not equal design of experiments (DoE), but the  Design expert  Stavex  Multi simplex
latter could be an important component of QbD.  Statistica  DoE Wisdom

DESIGN OF EXPERIMENTS -Software's DESIGN SPACE

WHAT IS A DESIGN SPACE?

”…the multidimensional combination and interaction of input varables and

process parameters that have been demonstrated to provide assurance of
product quality.”

Working within a design space is not considered as a change.

The Guidelines applicable also to one or two dimensions

VARIABLE INPUT TO CONTROLLED OUTPUT

Traditional Variable Input Fixed Process Variable Output

DRUG
PRODUCT

QbD / PAT Adapted Consistent

Variable Input Process Output
 25-10-2023

Design Space Design Space Determination

• Definition • First-principles approach

– combination of experimental data and mechanistic knowledge of
– The multidimensional combination and interaction chemistry, physics, and engineering to model and predict performance
of input variables (e.g., material attributes) and
• Non-mechanistic/empirical approach
process parameters that have been demonstrated
– statistically designed experiments (DOEs)
to provide assurance of quality – linear and multiple-linear regression
• Regulatory flexibility • Scale-up correlations
– Working within the design space is not considered a – translate operating conditions between different scales or pieces of
change equipment
• Risk Analysis
• Important to note – determine significance of effects
– Design space is proposed by the applicant and is subject to • Any combination of the above
regulatory assessment and approval

Design Space Example

Design Space
Linkage between the process inputs (input variables and process
parameters) and the critical quality attributes can be described in the
100.0 2 design space
95.0
90.0 Surface Plot Contour Plot 1.8

1.6
Dissolution (%)

85.0 Dissolution (%)

80.0 1.4
90.0-95.0
75.0
Parameter 2

1.2
85.0-90.0
70.0
1 80.0-85.0
65.0
60.0 75.0-80.0
0.8
55.0 70.0-75.0
Design Space 0.6
50.0 (linear ranges) 65.0-70.0
40
2 Design Space 0.4 60.0-65.0
(non-linear)
50 1 0.2

0
600 40 42 44 46 48 50 52 54 56 58 60
Parameter 1

• Design space can be described as a mathematical function or

simple parameter range
• Operation within design space will result in a product meeting
the defined quality attributes
 25-10-2023

DOE SUMMARY
 OFAT is an univariate approach: slow, inefficient,
assuming no interaction, leads to rigid manufacture
process with high risk of DP failure.

Thank You
 DOE is a statistically designed multivariate approach.
 DOE is an efficient tool to identify CPPs and CMAs and
establish relationship between CPP/CMA and CQA
 Different DOEs are appropriate for different phases of
pharmaceutical development.
 Use sound science, design space developed at lab or pilot
scale through DOE studies can be proposed for
commercial scale, but it needs to be verified at
production scale.
 Overall, DOE is a valuable tool to facilitate the
implementation of QbD.