The AAPS Journal (© 2013)

DOI: 10.1208/s1 2248-013-9472-8

Regulatory Note

Generic Development of Topical Dermatologic Products, Part II: Quality

by Design for Topical Semisolid Products

Rong-Kun Chang,' Andre Raw,' Robert Lionberger,' and Lawrence Yu'

Received 20 Decenmber 2012; accepted9 March 2013

systematic science and risk-based
Abstract. The emergence of quality by design as a relatively new
approach has added a new dimension to pharmaceutical development and manufacturing. This review
attempts to discuss the quality by design elements and concepts applied for topical semisolid products.
Quality by design begins with defining a quality target product profile as well as critical quality attributes.
Subsequently, this is followed by risk identification/risk analysis/risk evaluation to recognize critical
experiments or other
material attributes and critical process parameters, in conjunction with design of design-of-experiment
appropriate methods to establish control strategies for the drug product. Several
optimization of formulation and
examples are included as practical strategies for the development and
process for topical drug products.
product.
KEY WORDS: dermatologic product: generic; quality by design; semisolid: topical

INTRODUCTION The basic premise of QbD is that a product cannot be

tested or inspected into a quality product and quality
In recent years, the Food and Drug Administration attributes should be designed via the thorough understanding
of raw materials, formulation, and manufacturing process into
(FDA) has emphasized "Quality by Design" (QbD) as a Additionally, under quality by testing (QbT).
current Good Manufacturing Practices (cGMP) initiative for a drug product.
the twenty-first century. Despite numerous presentations and the sample size for each stage of testing is generally
insufficient to ensure acceptable quality attributes for an
publications by FDA officials and leading pharmaceutical
rescarchers (1-10), several pilot programs (11,12), and the entire batch. Hence, under QbD, testing confirms acceptable
adoption of QbD by the International Conference on quality attributes, without extensive sampling. Specifications
Harmonization, misunderstanding about QbD is still preva from raw materials, active ingredients, in-process testing, and
lent throughout pharmaceutical, biotechnology, and medical drug product are only part of the quality control strategy. The
device industries. This article points out the basic components quality needs to be designed into a drug product based upon
attributes
of QbD and application of QbD concepts to develop a systematic understanding of how critical material
formulations and optimize manufacturing processes for (CMAs) of drug substance and excipients, and critical process
parameters (CPPs) used during manufacturing affect critical
topical dermatologic products.
quality attributes of the drug product. This type of under
standing can be achieved using various QbD tools including
The opinions expressed in this review by the authors do not
necessarily reflect the views or policies of the Food and Drug the use of risk assessment, design of experiments (DoE), and
Administration (FDA). process analytical technologies (PAT) and subsequently,
Ofhce of Generic Drugs, Center for Drug Evaluation and Research,
based on this understanding, control strategy as a set of
assurance.
Rock ville control can be adopted for drug product quality
U.S. Food and Drug Administration, 7500 Standish Place. The International Conference on Harmonization (1CH)
Maryland 20855, USA.
guidelines: Pharmaceutical Development (Q8), Quality Risk
To whom correspondence should be addressed. (e -mail:
richard.changafda.hhs.gov) Management (09), and Pharmaceutical Quality System
(Q10) are the corner stone of QbD. The emergence of QbD
ABBREVIATIONS: Q1, Same components as the reference-listed as a relatively new systematic science
and risk-based
drug; O2, Same components in same concentration as the reference approach has added a new dimension to pharmaceutical
listed drug; Q3, Same components in same concentration with he development and manufacturing. Since pharmaceutical sci
same arrangement of matter (microstructure) as the reference-listed ences are multidisciplinary, many scientific disciplines, such as
drug: RLD, Reference-listed drug: QbD, Quality by design; QbT,
medicine, pharmacy, chemistry, engineering, biology, mathe
Quality by testing: CMA, Critical material attribute; CPP. Critical matics, and statistics can all contribute to product develop
process parameter: DoE, Design of experiments; QTPP, Quality
ment, process optimization, and quality improvement.
target product profile: C0A, Critical quality attribute: FMECA, Likewise, it takes a well-organized team to have a successful
Failure mode effects criticality analysis; FMEA, Failure mode effects drug product development using the QbD concept.
analysis.
I50-746130X0XK)-(XXO1/0C 2013 Amencan Assovation of Pharmaceutical Scientusts
aaps
Published onlinc: I0 April 203
 Chang et al.
SYNOPSIS OF QUALITY BY DESIGN
closcly monitored to gain assurance of Q3 microstructure
similarity to the RLD. These CQAs may include rheological
QbD begins with defining a quality target product profile behavior, drug particle size, comparative flux, and type of
(QTPP) as well as critical quality attributes (COAs). This is
emulsion. These CQAs can affect the permeation of the drug
followed by risk identification of the proposed formulation through skin, skin retention of cream, and patient acceptabil
and manufacturing process on drug product critical quality ity. Based upon the OTPP and CQ¢s at the end of
development, meaningful drug product specifications based
attributes and risk assessment to identify putative critical
material attributes and critical process parameters that need on clinical performance can be achieved and an acceptable
to be investigated. Screening DoE can be used as a tool to control strategy can be implemented to ensure equivalent
identify CMAs and CPPs. Then response surface DoE can be safety and efficacy to the RLD.
applied to explore formulation and process spaces for these
CMAs and CPPs. Knowledge gained from DoE studies RISK ASSESSMENT AND RISK CONTROL
inform a control strategy that implements the optimized
manufacturing process within an acceptable range/design Many risk assessment methods mentioned in ICH guideline
space to consistently deliver a high-quality drug product. This Q9 can be considered in pharmaceutical development and
approach also encourages the use of process analytical manufacturing. For example, a failure mode effects criticality
technology tools to monitor and adjust the process to analysis (FMECA) model can be developed to screen process
compensate for raw material or process variability. An risks due to raw material variability. The model examines the
appropriate combination of some or all of the QbD tools various potential failure risks (e.g., particle size, morphology,
may be applicable to formulation optimization, single-unit and polymorph of active ingredient; acid value, viscosity, and
operation optimization, or to the entire manufacturing gelling property of associated excipients) and associated poten
process to ensure process intermediate and end-product tial failure effects (e.g., failure to meet content uniformity, drug
quality assuran ce. Finally, in the commercial production product viscosity, and impurity level requirements). Based on
stage, QbD calls for continual improvement, i.e., tracking, development data and prior knowledge, the severity of impact
trending. and adjusting of process operations and statistical of each failure along with the probability of occurrence needs to
process control, as well as models and design space mainte be evaluated and action plans can be developed to address
nance/update. signifcant factors related to raw material variations, i.e..
critical material attributes (CMAs) to mitigate these risk
QUALITY TARGET PRODUCT PROFILE factors. Table II gives an example summarizing the risks
AND CRITICAL QUALITY ATTRIBUTES identifed through such an assessment with the action
plans to reduce the impact of raw material variability on
The physicochemical characterization of the reference drug product critical quality attributes. Similarly, other
risk assessment methods, such as Ishikawa diagram,
listed drug (RLD), its packaging insert, and labeling provides Failure Mode Effects Analysis (FMEA), what if analysis,
the foundation to define the quality target product profle of
the generic drug product, which in turn forms a development fault tree analysis, process map and flow chart, may be
applied to identify the risk factor(s) for the proposed
goal with specific elements and justifications. Having the
same components (Q1), in same concentration (Q2), with the
manufacturing process. In general, the failure modes
same arrangement of matter (microstructure) (O3), the RLD involved in the manufacturing process for topical dosage
is the most rational approach to formulate a generic forms may include the following:
dermatological product (13). However, topical products are
not generally required to have a formulation that is qualita " Failure of assay test
tively and quantitatively (Q1/Q2) the same as the RLD for " Lack of homogeneity of the drug product
Abbreviated New Drug Application (ANDA) filing. For . Undesirable rheological properties
various reasons, the generic firm may choose to formulate Undesirable globule size (emulsion type preparation
the generic product with a different excipient composition to only)
the brand product. In these cases, it may be necessary to " Undesirable drug particle size (drug dispersion type
perform formulation optimization using experimental designs, preparation only)
depending on the extent of deviation from the RLD formula. Undesirable in vitro drug release rate
In addition to formulation optimization, process development " Physical instability, which may include coalescence of
studies may also be needed to achieve a similar arrangement globules, phase separation, and growth of drug
of matter as the RLD, which provides assurance of similar particles
critical quality attributes to those of the RLD. Microstructure " Chemical instability of the drug and excipients
sameness includes similar rheology, type of emulsion (O/W " Microbiological contamination
emulsion, W/O emulsion, globule size), state of drug in
semisolid system (drug form, solubilized drug vs. dispersed
solid drug, and particle size of drug particles) compared to Appropriate application of risk assessment principles
those of the RLD. lo the proposed manufacturing process can be helpful in
Table I provides an example of QTPP and CQAs for prioritizing and focusing the pharmaceutical development
Flurouracil Cream, USP 5%. COAs, which are a subset of the to enhance product quality attributes (e.g., satisfactory
content uniformity, assay value close to 100% label claim,
QTPP that have the potential to be altered by formulation
to be desired rheological properties, desired drug particle size
and process variables, are essential attributes that need
Part II: Quality by Design for Topical Semisolid Products

Table I. OTPP and COA for Fluorouracil Cream USP. 5%

Elements Justification COA items

Target
Dosage form Cream Pharmaceutical cquivalent requirement
Route Topical Pharmaceutical equivalent requirement
of administration
Dosage strength 5% w/w Pharmaceutical cquivalent requirement
Oil in water emulsion cream with Match RLD
Dosage design
fluorouracil dispersed in the cream base
Appearance White smooth cream with dispersed fluorouracil API Match RILD and for patient acceptability
Identification Positive for fluorouracil Necded for clinical cffectiveness COA
90-110% Needed for clinical effectiveness COA
Assay CQA
Impurities Impurity A: NMT 0.2% Necded for safety
Impurity B: NMT O.2%
Any individual unknown: NMT 0.2%
Total impurities: NMT 0.5%
Needed for clinical effectiveness CQA
Homogeneity Top, middle and bottom of three containers
Nine assay values should be within 90.0%
and tube uniformity
to110.0% label claim and RSD is not
more than 5%.
Needed for clinical effectiveness CÌA
Physical attributes Match RLD
and patient acceptability
Rheological behavior Required to demonstrate Q3
Fluorouracil particle size
Oil globule size Required to demonstrate Q3 CQA
In vitro release test Match RLD
Methyl Paraben: 80.0-110.0% label claim Needed for ensuring antimicrobial
Preservatives content
Propyl Paraben: 80.0-110.0% effectiveness
Microbial limits Meet USP <61> Needed for safety CQA
Meet USP <467> Needed for safety CQA
Residual solvents
Container closure Identical primary packaging to RLD Match RLD and for patient acceptability
system
Package integrity No failure Needed for stability, clinical
effectiveness and safety
Match the shelf life of RLD
Stability No less than 24-month expiration dating period
Note that generic drug products should meet pharmaceutical equivalence, bioequivalence, safety and commercial requirements
physical Atrbutes may includethe following: appearance, odorirritant, rheological attributes (particle size, viscosity, specihc gravity. globule
size, spreadability) and pH as applicable
"Flux assay using either porcine ear or synthetic membrane or cadaver skin
Formulation andprocess variables are unlikely to impact the CQA. Therefore, the CQA will not be investigated and discussed in detail in
subscquent nisk assesment and pharmaceutical development. However, the CQA remains a target element of the drug product proile and
should be addressed accordingly

and globule size, and desired flux rate) and apply control the drug powder addition rate, rotor stator gap, rotor
strategies to mitigate risks. For example, the risk of speed and homogenization time are relatively critical
factors (i.e., CPPs) that may impact the homogeneity of
microbiological con tamination can be mitigated by micro
bial testing of the finished product and by monitoring the the drug product. Initial risk assessment for process
preservative content to ensure that the preservative level development is displayed in Table I showing the
potential for each unit operation to affect various CQAs.
is higher than the established minimum limit. This risk can
be further minimized by microbial testing of all raw Table IV gives an example of risk assessment for assay
materials and by an environmental monitoring program failure and lack of homogeneity using FMECA which is a
for the manufacturing and packaging areas. If generic more quantitative analysis of how each parameter would
firms put their efforts toward developing the afore afect these CQAs and ranks these upon the severity,
mentionced mitigation strategies for these risks and there probability, and detectability. In using this approach, it is
is no rend and no evidence of failure in these CQAS essential to quantitatively evaluate each risk thoroughly
there may be less need to implement a control strategy in lo identify the risk factors associated with the
batch release to monitor microbiological attributes in the manufacturing process.
finished drug product.
In another example, if a proposed manufacturing SCREENING DESIGN EXPERIMENT TO IDENTIFY
process calls for the emulsification of aqueous and oil CRITICAL FACTORS
phases to form a cream base and subsequent dispersion
of the drug substance into the cream base through Following the initial risk assessment, a screening design
powder eduction and a rotor-stator lype homogenizer, experiment, such as fractional factorial designs (i.e..
 Chang el a
Table I1. Raw Matcrial Risk Assessment and Action Plans
Formulation Potential impact on
component Potential risk drug product CQAs Action plan
Drug substance Particle size Shift in content uniformity, Micronized drug substance with identical solid state
or morphology change drug release and dermal form to the RLD from aqualified source is used
distribution of the drug for the drug product manufacturing and particle
size is measured as part of drug substance release
testing with a tight limit of D90 of not more
than 10 um
Drug concentration in the cream preparation
needs to be monitored to ensure homogeneity
White petrolatum Viscosity variation
of drug distribution in the drug product matrix
Shift in viscosity White petrolatum from a qualified source is used for
the drug product manufacturing. Consistency is
measured as part of every white petrolatum lot
via release testing using more stringent limits than
USP limits (consistency value 100to 200 vs. USP
limits 100 to 300) to ensure product viscosity closely
Propvlene Glvcol matching that of the RLD.
Unidentifed
stearyl alcohol Variation in stearyl Shift in iscosity Each lot of stearyl alcohol should contain not less
alcohol content than 90.0% of stearyl alcohol. Monitor and trend
Polysorbate 60 viscosity of the product.
Methyl and
Unidentified
Possible chemical instability Shift in preservative The antimicrobial properties of the drug product
propyl Paraben of preservatives in the cream content in the cream are studied during the product development
stage through antimicrobial effectiveness test.
Based on the results from these microbial
studies, set an adequate lower limit of preservative
content for drug product release and stability
specifications to reduce the risk of microbial
Purified water Increased water activity contamination.
Drug product microbial limit Quality system, cGMP
and bacteria growth potential

resolution IIJ and IV), Plackett-Burman designs, or Taguchi

effects (i.e., powder eduction rate, rotor speed,
orthogonal arrays can be used to evaluate the relative and
importance of the process variables. The following is a homogenization time) that affect blend uniformity and
need further investigation.
hypothetical six-factor problem concerning a cream product When invoking these screening designs, several DoE
using a 12-run Plackett-Burman design. The variables for
the experiments along with actual two-level experiment
software packages, such as Statgraphics, Design Expert.
Design-Ease, and JMP, can simplify complicated problem
conditions and code are listed as follows: mixer speed solving and
(100 rpm (-1) and 300 rpm (+l), mixing time (10 min factor effect interpretation of the results. For example, the
(-1) and 20 min (+1)), rotor-stator gap (10 mm (-1) and that analysis via JMP-9 software similarly reports
powder eduction rate, rotor/stator gap and homoge
20 mm (+1), powder eduction rate (slowest machine nization time are
three significant factor effects through
setting (-1) and fastest machine setting (+1), rotor speed the statistics table, Pareto chart, and half normal
(1,000 rpm (-1) and 2,000 rpm (+1), and homogenization chart
(14). By invoking this computation, the process parame
lime (10 min (-1) and 30 min (+1)). Various responses ters that show strong relationships to CQAs (ie., %RSD)
(e.g., content uniformity, viscosity, globule size, and drug are likely CPPs and should become a key focus point of a
particle size) can be measured after each randomized thorough study during prOCess optimization. Other CQAS
experiment is carried out. Table V illustrates a factor such as viscosity and globule size can be similarly
effect computation for this 12-run Plackelt-Burman evaluated through the same procedure and it was found
screening design to rank the relative importance of that there is no significant factor effect for these re
various input prOcess parameters on bulk uniformiy sponses. This is suggestive that unlike bulk uniformity,
(defined as %RSD, calculated by the standard deviation these CQAS are not significantly impacted by the process
of 10 assay values divided by mean assay value ). When parameters investigated within the range of the studies. It
analyzing this data, a 95% confidence interval for a factor should be pointed out that although other screening
effect including zero means that the factor effeet is designs such as a fractional factorial design can be used.
essentially zero and not significant statistically. Such Plackett-Burman designs have been mentioned here
results are suggestive of only three important factor because they are available in more convenient sizes (e.g..
Part Il: Quality by Design for Topical Semisolid Products
Table III. Initial Risk Assessment for Process Development

Manufacturing operation
Drug product CQA Pharmacy Aqueous phase Oil phase Emulsification Drug powder eduction phase
Appearance Low Low Low Medium Medium
Assay High Low Low Low Low
Impurities Low Low Low Low Low
Content uniformity Low Low Low Low High
Drug particle size Low Low Low Low High
High
Viscosity Low Low Low Medium

Note for relative risk ranking system: Low broadly acceptable risk: no further investigation is needed. Medium risk is accepted: further
investigation may be needed in order to reduce the risk. High risk is unaceptable: further investigation is needed to reduce the risk

(and ranges) to avoid failure of blend uniformity and to

12-, 20-, 24-, 28-, 36-, and 24-run foldover design, etc.) and
achieve the optimized %RSD for the drug product. The
generally allow the experimenters to effciently screen a
large number of variablcs in a small number of trials. One design space at the batch scale for DoE can be defined in
this case as the three-dimensional combination and
should be aware that there is amajor drawback associated
with these designs, i.e., all interactions are confounded interaction of these process parameters that have been
with main effects. demonstrated to provide assurance of product quality, e.g.,
blend uniformity % RSD less than 5. As a consequence,
any process falling within the design space should be
RESPONSE SURFACE OPTMIZATION, DESIGN acceptable and gain regulatory flexibility, provided QbD
SPACE, AND CONTROL STRATEGY experimentation is carried out at the manufacturing scale.
Furthermore, the developed mathematic model would
Subsequently, QbD calls for response surface optimi need to be verifed and validated using an independent
zation using identified significant process variables from data set to gain assurance of the multivariate predictive
Screening experiments. Many response surface designs, such properties. For optimization of multiple responses, trade
as central composite and Box-Behnken designs can be offs are often necessary to find a set of independent
utilized to find the optimum process conditions for drug variable levels that yield the best set of responses.
product. In general, central composite and Box-Behnken Table VII displays the application of control strategies
designs are effcient, compared to three-level factorial based upon the identified design space to reduce the
designs. When deciding upon a particular DOE design, it identified risks in process parameters.
is imporlant to have an understanding of critical design In accordance with QbD principles, sponsors should
issues (e.., variables and levels, center points, randomiza have a mechanistic understanding of the major degrada
tion, blocking, replication, responses, and transformations) tion pathways of the API in their formulation. The forced
and an understanding of advantages and drawbacks to cach degradation studies carried out during the method valida
approach. Central composite designs are usually preferred tion activities for the related substances are useful to
because many types of central composite designs can be pinpoint the degradation pathway. Based upon this
selected, the quadratic effects are much better estimated, understanding, sponsors should take appropriate steps to
the designs are more robust against missing data, and it is limit this degradation with appropriate excipient control,
formulation design, manufacturing process, and container/
possible to use the corner and center point trials from
previous conducted factorial experiments. Box-Behnken closure choices, especially given that stability problems are
design can simplify cxecution of the experiments due to signihcantly exacerbated when dealing with topical drug
its three levels of cach factor and its missing corner may be products. The need to control pH, add a chelating agent.
useful when the experimenter should avoid combined factor incorporate an antioxidant, and provide protection from
extremes. One should be aware that larger number of light via container/closure selection in the RLD, give a
center points are necessary for uniform precision and face hint of instability of the drug in the formulation matrix.
center designs and a single missing data point may lead to Other than the formulation strategies used for the RLD, it
the inability to estimate an interaction for Box-Behnken is prudent to put special attention to the grade of
design (15-17). excipients, including possible reactive residues of excipi
The following is a hypothetical three-variable central ents (e.g., peroxides) and extractables/leachables of con
composite design to explore the effects of powder lainer/closure system, because these minor differences may
eduction rate, rotor speed, and homogenization time on Irigger significant chemical degradation. Furthermore,
blend uniformity. Table VI shows a JMP-9 face -center variation of raw materials, for example acid value
cube central composite design with three center points variation of excipients, may shift the chemical stability of
along with %RSD response values; Fig. 1 shows param drug product. When chemical stability is a concern DoE
eter estimates by JMP-9 software along with the predic can be utilized as a tool during formulation optimization.
tion equation and contour plots for the study. These plots The following is a hypothetical 2 factorial design to
can be used to guide the selection of process parameters investigate the effect of acid value variation for two
 Crank
SxPxD
RPN current the
Chan,
7 6 3 4 2 1 homogeneitv.)
in
the (i.e..
in state
(i.e..
6 6
12 16 12 24 32
state of molten
lack
(U-5) molten
CQA,or
solution
D" 1 1 2 2 2 2 2 or
solution on
P(1-5) parameter
in base.
2 are
2 2 2 cream
Jomogencity
2 3 4 in ingredients
are
(1-5) base process
ingredients the
5 3
S 3 4 3 4
cream to
of
the thesubstance
maytemperature
a maymixing
entrapment;
a may andlower of
Lack form cause non
a uniformitydrug ofuniformity
of thethe of all
range range uniformity criticality
ensure
range
entrapment
time all to
for to the substance drug
the may ensure
Analysis cause cause
failure
of
Effectin thematerials; the the mixing of to
range content uniformity detectability)
probability) the
the materials
materialthan maymixture.
than mixturc.
non-honmogencous
non-homogeneous contentcontent to
than materials suggestsof
manufacturing may
materials; drug
Criticality higher range higher
air the
excessive
cause air mixture.
homogeneous dispersion
of higher on on
impact
Potential the
of
amount instability
than excessive
range
impact content on mosthighest
RPN
raw
the
Temperaturc speed impact rawof the
lowertime ofinstability dispersion
Effects than
the Potential on the the High lorbefore
batch drugPotential
Wrong
lower Mixing speed Mixing
cause cause
drug
the
Impact the the
drugfor withwith points
Mode than
points
detectability.
°C,
before casecase
case) 30
melting
Failure melting thethe to
30°C, means °C
means and
severe
probability,
severity,
Using failure
Causc
of the25
ofLack Lack
cGMP ofprocess process of
Lack
process processprocess
Lack ofLack
procesS of
process of theto
monitoring monitoring monitoring Lack
of
25°C 5 on to
monitoringmonitoring monitoringon
monitoring and5
most and cooled
training based
Assessmcnt of basedsizethe detectability
Lack Lack
to
particleprobability is
cooled means is base
is phases
error
ofWeighing
phases
Risk out out of eduction
Powder
1he stator of Homogeni-zation is and cream
failure
ofMode gap
Rotor the S5
Temperaturespeed out out
range andleastlowest non-aqueous
baseviscosity
non-aqueous of
IV. of
the
range range time
ofthe of speed of creamcasethe product The
Table material the
range out the therange out with
Mixing Mixing severe
with °C).
Tange of Rotor
timerange The
example,
Iate casea and75
out
and75°C). case is
least thenumber at
Raw
material
weighing and aqueous
mixing the aqueous
largeted
aqueousfor
Process
parameter
rate
eduction
Powder Homogenization
time the means
means
for
Temperaturc categories,
at
non-aqueous Rotor
stator
gap means priority
targeted
speed
Mixing mixing
is
for
mixing
temperature 1
Mixing
time Rotor
speed 1 scale, range
phases 1 scale,
scale,1-5
is other (risk for
range 1-5 temperature
numbertemperature
to 1-5the the
temperature
expand
the(in (in Criticality
homogeneity rank
Detectability
priority
Category (inProbability case,
that
ofLack Note to Severity
that
"Need Risk FNote current
case,
Part II: Quality by Design for Topical Semisolid Products

Table V. Factor Effect Computation for 12-Run Plackett-Burman Design

Mixing Mixing Rotor Eduction Unassigned Response
speed time stator gap rate Rotor speed H. time factors (%RSD)
Trial # Mean X, X X. X X10 X11

+ 4 + 4
4 4.5
5.1
2
4.2
3 +
4.8
4
5.9
5
1.2
6.9
6.2
8 2.5
3.8
10 6.2
+
11 7.2
+
12 28.2 27.7 30.5
33.1 22.3 30.1 22.6
Sum + 58.5 29.0 28.0 27.1 34.7
36.2 28.4 35.9 30.3 30.8 28.0
0 29.5 30.5 31.4 23.8 25.4
Sum 58.5 58.5 58.5 58.5
58.5 58.5 58.5 58.5
Check Sum 58.5 58.5 58.5 58.5 -0.7 -2.1 3.1 2.5
10.9 7.7 -13.9 1.70
Difference 58.5 -0.5 -2.5 -4.3 0.42
-2.32" 0.28 -0.12 0.35 0.52
Effect 4.99 -0.08 -0.42 -0.71 1.82º 1.28
-0.65 -1.05 -3.03 -0.41 -0.51
-1.50 1.25 1.28 -3.01
Lower CL -1.01 -1.34 0.81 1.28 1.45 1.35
2.45 2.48 -1.81 1.21
0.85 0.51 0.22
Upper CL
aStatistically significant factor effect =0.36-4.
+ UFE +UFE; +UFE) /q)=(0.28) +(-0.12)' +(0.35)° +(0.52)° +(0.42)°/s
Standard error of a factor effect, SpE = (UFE; + UFE; unassigned
where UFE is unassigned factor effect and q is number of factors
I value for 95% confidence and 5 degree of freedom=2.57 CL=factor effect -0.93 and upper CL=factor effect +0.93
Plus or minus part of CL (95%)=0.364 x 2.57=0.93: lower

excipients (cetyl ester wax and glyceryl monostearate)

Table VL. Central Composite Design for Investigating Three Process used in a cream formulation on chemical stability of a
Variables to Minimize %RSD drug. Table VIII shows a 2 factorial design based upon
variations in these two excipients along with percent
Powder impurity A detected for stability samples stored at 40°CI
Design Homogenization eduction Rotor
rate (X,) speed (X3) % RSD (Y)° 75% relative humidity (RH) for 6 months. Estimates of
pattern time (X, ) factor effects by computation and JMP-9 software are in
30 -1 1,000 3.6 agreement that acid values for both excipients are
2 000 20 0 1,500 2.0 statistically significant and interaction of these two
a00 10 1,500 5.8
factors can be ignored. Figure 2 shows the contour plot
of percent impurity A after 6-month storage at 40°C75%
4 30 -1 2,000 1.0

RH versus acid value of cetyl ester wax and acid value of

0A0 20 1 1,500 3.2
2.6
glyceryl monostearate. These studies indicate that a
6 00a 20 1,000
10 1 2,000 7.2 movement toward region with low acid value of cetyl
-1 1,000 5.3
1.6 ester wax and glyceryl monostearate lots is critical to
000 20 ( 1500
10 -1 2,000 5.9 ensure the end product having the level of impurity A
000 2 ( 1,500 1.1 below the specification limit of not more than 1% for 6
Such
12 00A 20 ( 2,000 0.9 month samples stored at accelerated conditions.
13 30 1 1.000 54 studies set the foundation for informing a control
strategy on input excipients to ensure acceptable product
0a0 20 -1 1,500 2.2
|4
10 1,000 7.5
15
4.5
stability.
16 30 2,000 Finally, at scale up and commercial batch production,
2.8
it may be necessary to verify and compare the studies
17 A00 30 1,500
used development in terms of manufacturing processes,
"Other response values, such as viscosity and APl particle size should
be evaluated in the same study. The objective of the present study is e.g., assessmentl of equipment used at development stage
to minimize %RSD and the acceptable range for %RSD is NMT 5% V. commercial equipment, as well as the adjustments
bDuring the outlier checking using Cook's distance test, these points needed for previously defined process parameters at
are identified to be suspicious from a statistical perspective. The commercial scale, lo inform whether experimentation is
regression both with and without these outliers is performed. It is
determined that there is no significant influence on the results. needed for conformation or exploration of design space.
Therefore, the regression with these data points is presented here In other words, the design space and understanding of
 Contour Profer
Chang et at
MortVert E Contour Profiler Contour Profiler
OO Homogenzaton Tme Har Vert f ctr Cuet Noruven Ector uent)
Powder Educon Rate Homegerston Tm 20 2 9 Hemogenten Tme
O Roor Soeed OPouder Edien Rat
1500 Pder Fsen Rat
Rotbr Spd
SRSD Cortour CurremtY Lo Um Mi Lim
150
GRotor Speed 1500
4 1694S kepoese Contour Curreat Y mt Response Contour CurrentY Lo Lmt
4 164e
4 1 6943
2000

1900

1800

1200
170 125

1606
150

1400
1300

1200
1200
1100
110
15
1000
eton Tme
RS75 43
Homogenlon Tme 05
Parameter Estimates Powder Educson Rate
Tom
Intercept Estimate Std Error t Ratio
Prob>|tl
Homogenization Time(10,30) 1.6943662 0.24363 O.0002
Powder Eduction Rate -1.44
0.180048 -8.00 <.0001
Rotor Speed(1000,2o0o) 0.98 0.180048 5.44 0.o010
Homogenization TimePowder Eduction Rate -0.49 0.180048 -2.72 0.0297"
Homogenization TimeRotor Speed 0.225 0.2013 1.12 0.3006
Powder Eduction Rate"Rotor Speed -0.475 0.2013 -2.36 0.0504
Homogenizat1ion Tim 0. 0.2013 0.50
Homogenization Time
Powder Eduction RatePowder 2.5098592 0.347842
O.6346

Eduction Rate 7.22

Rotor SpeedRotor Speed O.9098592 o.347842 2.62 O.0346"
The prediction equation is as follows: Y= 1.694 - 1.44 -0.040141 0.347842 O.9114
X, +0.98 X,-0.49 X, + 0.225 X,X, - 0.475
-0.12

where Coded X, = (homogenization time - X,X, +0.1

= (rotor specd - 1500)500. 20)y10, Coded X,= (machine setting 0)y1, fastest machine setting X,X, +2.510 X' +0.910 X² -0.0401 X,,
and slowest machine setting= -l and Coded X,
Fig. 1. Contour plot of %RSD versus
homogenization time, powder eduction rate, and rotor speed along with
prediction equation parameter estimates and the

process sensitivities developed in the R&D stage form

the not batch size related),
foundation for scale-up and production batch to verify the validity of confirmation
run(s) is/are needed
ing process development. If manufactur the design space at commercial
are within the design space projected process parameters
obtained in
scale. On the other hand, if projected
process parameters
(i.e.. for situations as the scale-up façtor isthe R&D stage are far off from the design space
generated from R&D
very low with batches, experimentation to scrutinize
no equipment change or the
manufacturing unit operation is eter effect is encouraged to create a process param
design space for

Table VII. Application of Control Strategy to Mitigate ldentified Risks in Process Parameters

Manufacturing operation
Drug
product
Aqueous phase: Oilphase: Emusification
RT mixing and heating to 75°C phase: emulsification
CQA
Pharmacy heating to 75°C Drug powder eduction
and mixing and cool down
phase
Appearance Low Low Low Low Low
Assay Controlled by cGMP Low Low Low Low
training of dispensing
personnel and
manufacturing
workers
Impurities Low Low Low Low Low
Content Uniformity Low Low Low Low Controlled by eduction
rate, rotor speed and
Drug Particle Size Low Low homogenization time
Low Low Low
Viscosity Low Low Low Low Low
Note that temperature for mixing aqueous and non-aqueous phases is
based on the
ingredients are in solution or molten state (ie., in the current case, temperature range is melting points for the raw malerials to ensure all the
targeted at 75C). The cream base is cooled to 25°C to
30°C, before dispersion of the drug substance to the cream base
Based on experience and knowledge obtained during process development, risk rating was changed to low risk
° Response surface optimization was used to provide guidance for the manufacturing parameters to mitigate risk
 Chang et a
Contour Profiler
Contour Profiler Current X
Contour Profiler
HorizVert Factor
CurrentX OO Homogenzaton Time 20
HorizVertFactor
HorizVert Factor Curret X O O Homogenzation Time 20
O O Powder Educion Rate
1500
OHomogenzation Time OPowder Educton Rate O Rotor Speed
Powder Eduction Rate 1500 Currert Y Lo Limit HI Umt
ORotor Speed Cortour
1500 HLimit Response 1 6943662
O Rotor Speed Contour Current Y Lo Limit
Lo Limt Hi Limit Response 5RSD
Response Contour Current Y
sRSD
4 16943662
2000 SRSO
LRSD
2000 RSO
SRSD
1900
1900
1800
1800
1700-1.325
1700
1600
1600
1600
1500
1400
1400
1300
1300
05 1200
1200
1100
1100
1000-7s
1000
1 1
10 15 20 15 20
Homogenizaton Trme Powder Educbon Rate
Homogenization Time

Parameter Estimates
Estimate Std Error t Ratio Prob>it|
Term
1.6943662 0.24363 6.95 O.0002*
intercept -8.00 <.0001
Homogenization Time(10,30) -1.44 0.180048
0.98 0.180048 5.44 0.0010*
Powder Eduction Rate
-0.49 0.180048 -2.72 0.0297*
Rotor Speed(1000,2000)
Homogenization TimePowder Eduction Rate 0.225 0.2013 1.12 O.3006
-0.475 0.2013 -2.36 O.0504
Homogenization TimeRotor Speed
Powder Eduction RateRotor Speed 0.1 0.2013 0.50 O.6346
Homogenization Time"Homogenization Time 2.5098592 0.347842 7.22 O.0002*

Powder Eduction RatePowder Eduction Rate 0.9098592 0.347842 2.62 0.0346

-0.040141 O.347842 -0.12 0.9114
Rotor Speed*Rotor Speed
The prediction equation is as follows: Y- 1.694 144 X; +0.98 X, 0.49 X, +0.225 X,X,- 0.475 XX, +0.1 X,X, +2.510 X' - 0.910X-0.0401 X,
where Coded X, -(homogenization time - 20y10, Coded X, =(machine setting - 0)y1,fastest machine setting-land slowest machine setting -1 and Coded X,
- (rotor speed - 1500)/500.
Fig. 1. Contour plot of %RSD versus homogenization time, powder eduction rate, and rotor speed along with parameter estimates and the
prediction equation

process sensitivities developed in the R&D stage form the not batch size related), confirmation run(s) is/are needed
foundation for scale-up and production batch manufactur to verify the validity of the design space at commercial
ing process development. If projected process parameters scale. On the other hand, if projected process parameters
are within the design space obtained in the R&D stage are far off from the design space generated from R&D
(ie., for situations as the scale-up façtor is very low with batches, experimentation to scrutinize the process param
Part lI: Quality by Design for Topical Semisolid Products

rble VI. Factorial Design to Investigate the Effct of Acid Value Variation for Celyl Ester Wax and Glyceryl Monostearate Used in a
Cream Formulation on Chemical Stability of a Drug

Design Observations"

X, Replicatc 1 Replicate 2 Y: average (% impurity A)

Run X;
0.135
-1 (1.0° -1 (1.0) 0.15 0.12
0.98 0.965
+1 (5.0) -1 (1.0) 0.95
1.95 1.92
-1 (1.0) +1 (6.0) 1.89
2.54 2.59
+1 (5.0) +1 (6.0) 2.64

The percent level of Impurity A detected for stability samples stored at 40°CI75% RH for six months
bCodedX, (acid value of cetyl ester wax used in the experiment )
Coded X, (acid value of glyceryl monostearate used in the cxperiment)
value of glyceryl monostearate-3.5)/2.5
Coded AX;-(acid value of cevl ester wax -3)/2 and coded X,=(acid
Intercept=(0.135 + 0.965 + 1.92 + 2.59)/4 = 1.4025
Estimate of main effect of factor X;= J(0.965- 0.135) + (2.59--0.965)1/2-- 1.92))/2=0.75
= 1.71
Estimate of main effect of factor X, |(1.92 - 0.135) + (2.590.135)l/2 = -0.08
Estimate of interaction effect=((2.59-- 1.92) -- (0.965 -- noalc.a/r)xto025, where degree
confidence interval for an effect, based on r replications of a2² factorial design, is estimate effect +(S
A 95% 1/2
of freedom=4(r-1) for lo025 for df-4, we find that o.02s=2.776. Then, (Soled/r)
consulting the t table
The number of replicates is r=2 and effect=0.75+0.086, X, efect=1.71+0.086, and interaction =0.08+0.086. It can be cóncluded that
(0.001925/2 )* x 2.776 =0.086, so that X, is not statistically significant. The prediction
factor effects (i.e., acid value for two excipients) seem to be real and the interaction term
both X,
equation is Y = 1.4025 + 0.375 X; + 0.8525 X, - 0.04 X;

risk management to
size to obtain the from OTPP to COAs, and rom
scale-up and commercial product batch aim to boost productivity and
regulatory flexibility. DoE. These new endeavors understanding of the
product quality in the end. The
compa
CONCLUSION
myriad roles that QbD plays has pharmaceutical
development and
nies looking at drug product
could not a
emphasis is manufacturing issues in a way they simply risk-based
Across the pharmaceutical industry, the few decades ago. This systematic science and
knowledge from raw materials
on QbD and enhanced approach will lead to the development of control
strate
CPPs),
(e.g.. CMAs) to manufacturing prOcesses (e.g.,
Contour Profiler
Current X
HorizVert Factor
O O acid value of cetyl ester
wax
6
acid value of glyceryl monostearate Hi Limit
Contour Current Y Lo Limit
Response 1
2.8 2.59
%impurity A
%impurity A
6

5.5
1.65
monostearat
5

8 4.5 14

glycery4
1.15
of
value
acid
S 2.5
2

1.5

1 4 5
3.5 4
1 1.5 2 2.5 3
acid value of cetyl ester wax

after 6-month storage at 40°C/75%

Fig. 2. Contour plot of percent impurity A for samples of glyceryl monostearate
versus acid value of cetvl ester wax and acid value
RH
 Chang et al.
gies that reliably produce gencric drug products with a 8. Huang J. Kaul G. Cai C, Chatlapalli R, Hernandez-Abad p
Quality by design case study: an integrated multivariate an.
high quality. proach to drug product and process development. IntJ
Pharmaceutics. 2009:382:23-32.
ACKNOWLEDGMENTS 9. Wu H, Tawakkul M, White M. Khan M. Quality-by-desipn
(ObD): an integrated multivariate approach for the component
quantification in powder blends. Int J Pharmaceutics.
The authors thank Drs. Daniel Peng and Yue Teng for 2009:372:39-46.
their constructive comments during the preparation of the 10. Wu H, White M, Khan M. Quality-by-design (QbD): an
manuscript. integrated process analytical technology (PAT) approach for a
dynamic pharmaceutical co-precipitation process characterization
and process design space development. Int J Pharmaceutics.
2011:405:63-78.
REFERENCES 11. Quality by design for ANDAs: an example for modified release
dosage forms, http://www.gphonline.org/sites/default/iles/
DraftExampleQbDforMRTablet %2026.pdf
1. AAPS Workshop (200S). Pharmaceutical quality assessment-a 12. Quality by design for ANDAS: an example for immediate
science and risk-based CMC approach in the 21st century. release dosage forms, http://www.gphaonline.org/sites/defaulu
October 5-7, North Bethesda, MD, USA
2. AAPS Workshop (2009). Utilization of process modeling and
files/ExampleQbDforlIRTabletApril% 2026-revised.pdf
13. Chang R, Raw A, Lionberger R, Yu L. Generic development of
advanced process control in QbD based drug development and topical dermatologic products: formulation development, process
manufacturing. April 27-28, Baltimore, MD. USA development, and testing of topical dermatologic products.
3. GPhA/FDA (2011) quality by design workshop for ANDAs. AAPS Journal. 2012. doi:10.1208/s1 2248-012-9411-0.
May 5-6, North Bethesda, MD, USA 14. JMP-9 Statistical Discovery from SAS, Cary, NC 27513
4. Lionberger R. Lee S, Raw A, Yu L. Quality by design: concepts 15. Montgomery DC. Design and analysis of experiments. 3rd ed.
for ANDAS. AAPS J. 2008:10(2):268-76. New York: Wiley; 1991.
5. Yu L. Pharmaceutical quality by design: product and process
16. Rosas JG, Blanco M, Gonzalez JM, Alcala M. Quality by design
development, understanding, and control. Pharm Research. approach of a pharmaceutical gel manufacturing process, part 1:
2008:25(3):781-91. determination of the design space. J Pharm Sci. 2011:100:4432
6. Raw A, Lionberger R, Yu L. Pharmaceutical equivalence by 41
design for generic drugs: modified-release products. Pharm Res. 17. Xie L, Wu H, Shen M, Augsburger L, Lyon R, Khan M, et al.
2011:28(7):1445-53. Quality-by-design (QbD): effects of testing parameters and
7. Yu LX. Pharmaceutical quality by design: product and process formulation variables on the segregation tendency of pharma
development, understanding and control. Pharm Res. ceutical powder measured by the ASTM D6940-04 segregation
2008:25(4):781-91. lester. J Pharm Sci. 2008:97:4485-97.