OPTIMIZATION TECHNIQUES
(RESPONSE SURFACE METHOD,
CONTOUR AND FACTORIAL DESIGN)
Presented by: Presented to
Pharmacy Seeker team Dr. Vandita Kakkar
Professor Of Pharmaceutics
UIPS, PU
Slides- 1-50
Time duration- 40 min.
2
CONTENTS
• Response surface methodology
• Methods
• Applications
• Contour design
• Types of Contour design
• Factorial Design
• Types of Factorial Design and applications
• Central composite design
• Steps and Types
• Applications
RESPONSE SURFACE 3
METHODOLOGY
Response surface methodology (RSM),
introduced by Box and Wilson, is a collection of
mathematical and statistical techniques whose
purpose is to analyze, by an empirical model,
problems as the one posed.
OBJECTIVES 4
• To generate knowledge in the experimental domain of
interest.
• To reliably estimate the experimental variability (pure
error).
• To guarantee the adequacy between the proposed model
and the experimental data (to make it easy to detect the
lack of fit).
• To predict the observed response, as exactly and precisely
as possible, in points within the experimental domain
where no experiments were done.
. 5
• To propose sequential strategies to carry out the
experimentation with different alternatives according to
the results obtained.
• To maintain a high efficiency with respect to economical
cost, time, and any other practical limitations.
• To make the identification of outlier data easy.
• To make the decision making possible under uncertainty
conditions, reducing the ambiguity.
6
STAGES OF RESPONSE SURFACE METODOLOGY
1. Objective: Fixing the objective of the study.
2. Design of Experiment: in which significant independent
variables are selected with the help of first order response
surface design such as 2v factorial designs (FD), fractional
replicates of the 2v factorial designs (FFD), Simplex
designs, central composite design etc. for the Optimization
REGRESSION MODELLING 7
Start Is there a significant lack of fit
| |
Perform Regression Analysis
/--------------------\
|
Are any effects statistically | |
significant
Yes No
|
/--------------------\
| |
| | Use a higher-order (second- First-order
Yes No
model is
| |
Form a regression model Model is inadequate; order) model
using significant consider alternative (polynomial) adequate.
effects (first- models or data review.
| |
order terms) |
| End End
Check for significant lack of fit End
8
4. Experimentation using response surface design: Select the
appropriate second order rotatable design according to the
selected experimental matrix (based on the number of factors
selected, the level, and the number of runs). The most commonly
used second order response surface designs are
(i) 3vfactorial design
(ii) Box Behnken design (BBD)and
(iii) Central (face) Composite design (CCD/FFCD).
5. Model building and validation: Evaluate model adequacy
using criteria like residual analysis, R². If adequate, apply
optimization; if not, refine with a higher-order model.
.
9
6. Optimization of response using graphical
(response surface plot and contour plot) and
numerical approach. In case of multi-response
optimization desirability function approach is used.
7. Verification of results: To verify the desired
optimum conduct confirmatory trial.
METHODS OF RSM
1) The method of steepest
ascent
The steepest ascent method is a
numerical optimization technique
that finds the direction of fastest
change in a function
The steepest ascent method works by:
1) Calculating the gradient
2) Using the gradient to determine
the direction of the next point
3) Taking a step that relates to the
distance between the new and old
points
2) The method of steepest
descent
The steepest descent method is a
technique used in optimization
and mathematics to find the
minimum of a function or to
approximate integrals
Application of response surface method
• RSM is used to optimize the composition of pharmaceutical
formulations, such as tablet formulations, liquid solutions, and
creams.
• RSM can be used to optimize parameters affecting the
solubility and dissolution rate of poorly soluble drugs, which
directly impacts their bioavailability.
• RSM helps to improve the sensitivity, selectivity, and accuracy of
assays used for the quantification and stability testing of drugs.
• RSM is a core tool in the Quality by Design (QbD) approach,
which aims to design pharmaceutical products and processes
with built-in quality.
CONTOUR DESIGN
INTRODUCTION
15
Contour design is a graphical technique used to visualize data by
plotting lines connecting points of equal value. This method is
particularly useful in biostatistics for analyzing multivariate or spatial
data, providing insights into relationships between variables, trends, or
spatial patterns in biological, epidemiological, and clinical research.
TYPES OF CONTOUR DESIGN
17
ISOQUANT TOPOGRAPHIC RESPONSE
CONTOURS CONTOURS SURFACE
• Represent levels of • Show spatial variation CONTOURS
constant response or in a biological or • Represent outcomes
effect. Example: epidemiological from multivariate
Contours showing parameter. Example: experiments. Example:
equal survival rates in Mapping malaria Contours showing
clinical trials under prevalence in a region. optimal conditions for
varying doses of two microbial growth.
drugs.
18
THRESHOLD-
PROBABILITY HEATMAP-BASED
BASED
CONTOURS CONTOURS
CONTOURS
• Indicate regions of • Combine gradient • Focus on specific
equal probability for colors with contour critical values or
events. lines to enhance thresholds.
interpretability.
19
GRADIET FIELD STATISTICAL
CONTOURS DENSITY
• Indicate gradients or rates of CONTOURS
change in data across space or • Represent probability density
conditions. functions for multivariate data.
CLUSTER CONTOURS
• Delineate groups or clusters
based on similar traits or data
patterns.
20
KEY ELEMENTS OF
CONTOUR DESIGN
DATA SOURCE
• The type of data used, such as spatial, multivariate, or
experimental.
AXES AND DIMENSIONS
• Axes represent the independent variables or factors
influencing the contour plot.
CONTOUR LEVELS
• Lines or regions connecting points of equal values.
COLOUR INTERPOLATION
GRADIENTS METHOD
• Gradual color • The mathematical
transitions technique used to
representing data estimate values
intensity or between known
magnitude. data points.
SCALE AND
RESOLUTION RANGE
• The granularity • The numerical
of the grid used range and scaling
for plotting. of the data on
the axes and
contour levels.
22
ANNOTATIONS BOUNDARY
• Labels, legends, and markers for CONDITIONS
clarity. • Constraints or limits on the data
being represented.
STATISTICAL AND SOFTWARE AND
BIOLOGICAL TOOLS
CONTEXT • Programs used to generate contour
• The scientific or statistical rationale designs.
underlying the contour design.
23
ROLE OF CONTOUR DESIGN
❖ Contour designs are employed to identify the optimal combinations of different
formulation variables that yield the desired drug delivery outcomes.
❖ These designs are particularly useful in multivariable systems, where multiple
independent variables influence the formulation's properties, such as drug release rate,
stability, bioavailability, and therapeutic effect.
❖ Contour design is a technique , which systematically explores the relationship between
input variables (such as excipients, formulation ingredients, processing conditions) and
output responses (such as drug dissolution, particle size, or release profile)
24
•
EXAMPLE
A pharmaceutical company wants to develop a controlled-release tablet for a drug like
Diclofenac sodium.
• The main challenge is to balance polymer concentration and compression force to
get the desired release profile.
• Independent Variables (Factors):
• Polymer Concentration (X1)- (HPMC), which controls the drug's release.
• Compression Force (X2): The force used to compress the tablet, which influences
tablet hardness and porosity.
• Dependent Variable (Response):
• Drug Release Rate (Y): The rate at which the drug is released from the tablet over
time, particularly at the 12-hour mark.
25
• Step 1: Experimental Design - The experimental runs are designed to measure the drug release
rate under these varying conditions.
• Step 2: Data Collection and Mathematical Model
• A regression model is then built to describe how the polymer concentration and compression
force affect the drug release rate.
• Step 3: Creating Contour Plots - The contour plot would show how the drug release rate varies as
the two independent variables change.
• The X-axis represents Polymer Concentration (e.g., 5%, 10%, 15%).
• The Y-axis represents Compression Force (e.g., 5 kN, 7 kN, 9 kN).
• The contour lines represent different drug release rates
26
APPLICATION OF CONTOUR DESIGN
• Pharmaceutical formulation - Researchers use contour designs to optimize the excipient
concentrations (e.g., binders, disintegrants, lubricants) in tablet formulations. The aim is to
achieve a desirable dissolution rate, mechanical strength, and ease of manufacture.
• Food Product Development - In a research study developing a soft drink, contour design could
be used to optimize the concentrations of sugar, citric acid, and carbon dioxide for the desired
taste profile and carbonation level.
• Cosmetic Formulations- Researchers can use contour design to optimize the ratio of oil phase
(such as emollients) and water phase (such as humectants) to achieve a desired viscosity and
stability in the cream.
• Chemical and Polymer Formulations - In the development of a new adhesive, contour design
can be used to optimize the ratio of polymer, solvent, and curing agent to achieve the best
bonding strength and curing time.
• Advanced Contour Design Approaches - Contour design is often combined with DOE
methods, like Central Composite Designs (CCD) or Box-Behnken Designs (BBD), to explore the
effect of multiple factors and their interactions more efficiently
FACTORIAL DESIGN
28
INTRODUCTION
• Factorial design is a powerful experimental strategy used in
scientific research to study the effects of two or more factors
simultaneously.
• It enables researchers to explore not only the individual (main)
effects of the factors but also the interactions between them.
• Experimental design valuable: when multiple variables are expected
to influence an outcome, as it provides a comprehensive
understanding of their combined effects.
• Factorial design Representation: 2^k, where k is the number of factors
and 2 is number of levels.
EXAMPLE 29
•Objective: To optimize a tablet formulation by studying the effects
of two factors:
•Concentration of Binder (Factor A): Low (2%) and High (5%).
•Compression Force (Factor B): Low (5 kN) and High (10 kN).
•Response Variable: Tablet hardness (measured in Newtons).
Hypothetical Results
Binder Concentration Compression Tablet
Run
(%) Force (kN) Hardness (N)
1 2 5 30
2 2 10 45
3 5 5 50
4 5 10 70
TYPES OF FACTORIAL DESIGN
31
FULL FACTORIAL DESIGN
• Advantages:
•Includes all possible combinations of • Provides detailed information about
factor levels, giving a complete picture of main effects and all interactions.
main effects and interactions.
• Ensures accurate results due to full
•Example: coverage of factor levels.
•For two factors (A and B), each at 2 • Disadvantages:
levels (low and high), the design will
have 2^2 = 4 combinations: • Resource-intensive as the number of
factors and levels increases (e.g.,
•(A Low, B Low), (A Low, B High), (A 2^4=16, 2^5=32 etc.).
High, B Low), (A High, B High).
• Time-consuming for large-scale
experiments.
32
• FRACTIONAL FACTORIAL DESIGN
A subset of the full factorial • Advantages:
design, used to reduce the • Reduces experimental cost and
number of experiments while time.
maintaining essential
• Useful for screening factors to
information. identify the most significant ones.
1. Example: • Disadvantages:
2. For 2^3=8 possible • May omit higher-order
combinations, only 4 interactions, potentially leading to
experiments may be incomplete conclusions.
conducted to assess key • Less reliable if the omitted
effects. interactions are significant.
33
TWO-LEVEL FACTORIAL DESIGN
(2^K DESIGN)
• Advantages:
Each factor has two levels (e.g., high • Simple and easy to execute.
and low), making it ideal for
screening studies. • Effective for identifying
significant factors.
Example:
• Disadvantages:
For three factors (A, B, C), the design
will have 2^3=8 combinations. • Limited to linear relationships;
may not capture non-linear
effects.
• May require additional
experiments to refine results.
34
THREE-LEVEL FACTORIAL DESIGN (3^K DESIGN)
Advantages:
Each factor has three levels • Captures non-linear
(e.g., low, medium, high), relationships between factors
providing a more detailed and response.
analysis.
• Useful in optimization studies.
Example: Disadvantages:
For two factors (A, B), each • Requires more experiments
with three levels, the design compared to 2^k designs.
will have 3^2=9 combinations.
• Higher cost and complexity.
35
MIXED-LEVEL FACTORIAL DESIGN
• Advantages:
1. Factors have different • Flexibility in designing
numbers of levels.
experiments for factors with
2. Example: varying levels.
3. Factor 1: Dose (2 levels: • Reduces unnecessary
Low, High).
combinations.
4. Factor 2: Type of Drug
Delivery (3 levels: Tablet, • Disadvantages:
Capsule, Liquid). • Analysis can be more complex
5. One factor (1) has 2 levels, due to unequal factor levels.
and another factor (2) has 3
levels, resulting in 2×3=6 • Difficult to interpret
combinations. interactions between factors.
36
SPLIT-PLOT DESIGN
• Advantages:
• Some factors are applied to • Efficient when certain factors are harder
larger experimental units (whole to randomize.
plots), while others are applied
• Reduces variability in hard-to-change
to smaller subplots.
factors.
• Example: • Disadvantages:
• In agricultural experiments, soil • Complex statistical analysis.
type (whole plot) and fertilizer
type (subplot) are varied. • Requires careful planning to avoid
confounding effects.
37
APPLICATIONS IN PHARMACY
Formulation Development
• Drug Delivery Systems
• Oral Dosage Forms
• Transdermal Systems:
Drug Stability Studies
Bioavailability and Bioequivalence Studies
Preclinical and Clinical Studies
Cosmetic and Nutraceutical Product Development
CENTRAL
COMPOSITE DESIGN
39
INTRODUCTION
Central Composite Design (CCD) is a powerful optimization technique
used in response surface methodology (RSM) to:
1. Model complex relationships between input variables and responses
2. Identify optimal operating conditions
3. Minimize or maximize a response variable
40
COMPONENTS OF CCD
1. Full Factorial Design (2^k): Examines interactions between variables
2. Star Points (2k axial points): Explores curvature and non-linearity
3. Center Points (n_c): Provides replication and estimates pure error
41
STEPS TO IMPLEMENT CCD
1. Define problem and objectives
2. Select variables (k) and levels
3. Determine number of center points (n_c)
4. Generate CCD using software (e.g., Design-Expert, Minitab)
5. Conduct experiments and collect data
6. Analyze data using RSM techniques (e.g., ANOVA, regression)
7. Interpret results and optimize conditions
42
SOFTWARE TOOLS
1. Design-Expert
2. Minitab
3. JMP
4. R (with relevant packages)
43
TYPES OF CCD
Circumscribed design Inscribed design (CCI) Face cantered (CCF)
(CCC)
• Levels of the factors • CCI design is a modified • For each face of the factorial
eventually stand on the edge. version of CCC design, where space, star points are the
• Always magnate with corner CCC design has been divided center point
points (Red Dots) by α to generate the CCI • This variable requires 3 levels
• From the center point (blue), model. of each factor
the extract points •Here also each factor studied • The face cantered designs
are constrained from the sides at level 5. (CCF) are a non-rotatable
(green dots). • It is also a rotational model. design
• Each factor would have 5
levels
• These designs having circular,
spherical or hyperspherical
symmetry and required 5 levels
for each factor
• The Circumscribed (CCC)
was found to be a rotatable
design
COMPARISON OF THE THREE 44
TYPES OF CENTRAL COMPOSITE
DESIGNS
CCD Model
45
PURPOSE OF CCD
• Provides good predictions in the middle of the design.
• Estimates quadratic effects, interactions and curvature between
factors and responses.
• Improve process understanding.
ADVANTAGES DISADVANTAGES 46
• Star points are outside the
hypercube, so the number of
• Extension of 2 level factorial or
levels that have to be adjusted
fractional factorial design.
for every factor is five instead of
• Estimate nonlinearity of three.
responses.
• Inability to estimate individual
• Estimate curvature in obtained interaction terms, i.e., linear by
continuous responses . quadratic or quadratic by
• Maximum information in a quadratic.
minimum experimental trial. • Depending upon the Design, the
• Reduction in the number of trials squared terms in the model will
required to estimate the squared not be orthogonal to each other.
terms in the second-order model.
COMPARISON 47
Contour Factorial
Aspect RSM CCD
Design Design
Study
Fit quadratic
Optimization Visualization individual &
Purpose models for
of processes in RSM interaction
optimization
effects
A design
A A graphical A design
Type method in
methodology tool method
RSM
Input Usually 2 (for
Multiple Multiple Multiple
Variables plots)
Uses models Tests Includes
Contour plots
Key Feature and surface all/fractional factorial +
of response
plots combinations axial points
Initial stages
During During data of For quadratic
When Used
optimization interpretation experimentati modeling
on
48
REFERENCES
• Dr. Vinod Kumar; Dr. Sanjay Sharma; Dr. Deepak Kumar, Biostatistics and
research methodology, Pee Vee Publisher, pp- 227- 252.
OUR TEAM 49
Prabhnoor Singh
Priya
Priyanka
Raj Arihant
Rishika
Ruhani
Sanjana
THANK YOU
