Journal of Pharmaceutical Sciences 105 (2016) 168e181

Contents lists available at ScienceDirect

Journal of Pharmaceutical Sciences

journal homepage: www.jpharmsci.org

Pharmaceutics, Drug Delivery and Pharmaceutical Technology

Integrated Application of Quality-by-Design Principles to Drug

Product Development: A Case Study of Brivanib Alaninate
FilmeCoated Tablets
Sherif I.F. Badawy 1, *, Ajit S. Narang 1, Keirnan R. LaMarche 1,
Ganeshkumar A. Subramanian 1, Sailesh A. Varia 1, Judy Lin 2, Tim Stevens 2,
Pankaj A. Shah 2
1
Drug Product Science and Technology, Bristol-Myers Squibb Company, New Brunswick, New Jersey 08903
2
Analytical and Bioanalytical Development, Bristol-Myers Squibb Company, New Brunswick, New Jersey 08903

a r t i c l e i n f o a b s t r a c t

Article history: Modern drug product development is expected to follow quality-by-design (QbD) paradigm. At the same
Received 2 July 2015 time, although there are several issue-specific examples in the literature that demonstrate the appli-
Revised 8 October 2015 cation of QbD principles, a holistic demonstration of the application of QbD principles to drug product
Accepted 10 November 2015
development and control strategy, is lacking. This article provides an integrated case study on the sys-
tematic application of QbD to product development and demonstrates the implementation of QbD
concepts in the different aspects of product and process design for brivanib alaninate filmecoated
Keywords:
tablets. Using a risk-based approach, the strategy for development entailed identification of product
quality by design
tablets
critical quality attributes (CQAs), assessment of risks to the CQAs, and performing experiments to un-
brivanib alaninate derstand and mitigate identified risks. Quality risk assessments and design of experiments were per-
design space formed to understand the quality of the input raw materials required for a robust formulation and the
control strategy impact of manufacturing process parameters on CQAs. In addition to the material property and process
critical quality attributes parameter controls, the proposed control strategy includes use of process analytical technology and
critical process parameters conventional analytical tests to control in-process material attributes and ensure quality of the final
critical material attributes product.
dissolution
© 2016 American Pharmacists Association®. Published by Elsevier Inc. All rights reserved.
porosity

Introduction established acceptance criteria for the critical quality attributes

The principles of Quality by Design (QbD) emphasize the need
for understanding sources of variability in the manufacturing pro-
cess and how it links to product quality as prerequisite for
robustness. Risk-based approach is used to define the goals of QbD
development and to design appropriate development studies to
achieve these goals. Product quality attributes based on patient
needs and clinical relevance are identified, and the risks to these
attributes due to input material variability and manufacturing
process parameters are assessed. The output of QbD development
takes the form of a process and product quality control strategy
which ensures that the final product consistently meets pre-
* Correspondence to: Sherif I. F. Badawy (Telephone: þ1-732-227-6451; Fax: 732-
227-5150).
E-mail address: sherif.badawy@bms.com (S.I.F. Badawy).
(CQAs).1,2 Although QbD principles were introduced more than a
decade ago by the ICH guidelines, there are hardly any published
examples demonstrating how these principles are reduced to
practice in the commercial development of a drug product as well
as transfer and implementation of control strategy in a
manufacturing plant. Although there are some published reports
invoking QbD principles in the study of a specific issue,3-11 none of
them provide a wide view of the application of QbD to a compre-
hensive development program.
Brivanib alaninate (BMS-582664) is a potent, orally active
inhibitor of vascular endothelial growth factor receptors.12 A film-
coated tablet formulation for brivanib alaninate has been devel-
oped using a wet granulation manufacturing process. The
principles of QbD were applied during formulation and process
development of brivanib alaninate filmecoated tablets. Using a
risk-based approach, the strategy for the formulation design and

http://dx.doi.org/10.1016/j.xphs.2015.11.023
0022-3549/© 2016 American Pharmacists Association®. Published by Elsevier Inc. All rights reserved.
 S.I.F. Badawy et al. / Journal of Pharmaceutical Sciences 105 (2016) 168e181 169

understanding was developed by (1) identifying desired product Table 1

attributes, (2) identifying quality risks to the attributes during Experimental Design for the Formulation Optimization Study

development, and (3) adopting the appropriate strategy and per- Batch Croscarmellose Hydroxypropyl Crospovidone Magnesium
forming experiments to mitigate or reduce risks. Quality risk as- Number Sodium (%) Cellulose (%) (%) Stearate (%)
sessments and statistically designed experiments were performed 1 3.0 3.0 3.0 1.0
to understand the quality of the input raw materials required for a 2 3.0 3.0 3.0 0.5
robust formulation and the impact of manufacturing process pa- 3 4.5 1.5 1.5 1.5
4 4.5 1.5 1.5 0.5
rameters on the CQAs of the drug product. The data obtained were
5 1.5 1.5 4.5 1.5
used to establish acceptable material attribute and process ranges 6 1.5 1.5 4.5 0.5
and to define the control strategy for the commercial manufacture 7 4.5 4.5 4.5 1.5
of brivanib alaninate filmecoated tablets. The proposed commer- 8 4.5 4.5 4.5 0.5
9 4.5 1.5 4.5 1.5
cial manufacturing process uses process analytical technology and
10 4.5 1.5 4.5 0.5
conventional analytical tests to control the in-process material at- 11 1.5 4.5 1.5 1.5
tributes. The product quality is controlled and assured through a 12 1.5 4.5 1.5 0.5
holistic approach of input material, in-process attribute, and 13 1.5 4.5 4.5 1.5
parametric controls combined with appropriate finished product 14 1.5 4.5 4.5 0.5
15 4.5 4.5 1.5 1.5
tests. The purpose of this article is to provide an integrated case
16 4.5 4.5 1.5 0.5
study on the systematic application of QbD to product development 17 3.0 3.0 3.0 1.0
and to demonstrate the implementation of QbD concepts in the 18 3.0 3.0 3.0 1.0
different aspects of product and process design. The work pre- 19 1.5 1.5 1.5 1.5
sented demonstrates how QbD concepts, such as product robust- 20 1.5 1.5 1.5 0.5

ness and scale-independent design space, were reduced to practice

in the development of brivanib alaninate tablets.
Methods
Formulation Selection
Formulation and process screening studies were conducted to
select a prototype formulation and manufacturing process for bri-
vanib alaninate tablets. Key formulation selection studies focused
on the selection of tablet disintegrant in which performance of both
croscarmellose sodium (CCS) and crospovidone (CPVP) was evalu-
ated. Formulation selection batches were manufactured at 600 g
batch size by a wet granulation process in a 4-L high-shear Diosna
granulator (Diosna Dierks & So €hne GmbH, Osnabrück, Germany).
All batches were granulated using 45%-50% w/w water, relative to
total batch size, at impeller speed of 5.2 m/s. Wet granulation was
screened through 8 mesh screen, and granules were subsequently
dried in a Glatt GPCG-1 fluid-bed dryer (Glatt Air Techniques,
Ramsey, NJ) until the loss on drying (LOD) was 2% w/w. Dried
granulations were milled using a conical mill (Comil 197S, Quadro,
Waterloo, Ontario, Canada) and then blended with the extra-
granular excipients in a bin blender. The final blend was com-
pressed into 400 mg of strength tablets (800 mg tablet weight) on a
6 station Korsch PH 100 tablet press (Korsch AG, Berlin, Germany)
using embossed tooling. Batches were characterized using methods
described below.
Formulation Optimization
After the prototype formulation was identified, a formulation
design of experiment (DoE) study was conducted to assess
formulation ruggedness and optimize composition with respect to
the binder (hydroxypropyl cellulose [HPC]), disintegrant (intra- and
extragranular), and lubricant (magnesium stearate). A split-plot
full-factorial experimental design was used to complete the DoE.
The design entails the manufacture of a larger scale granulation
batch and splitting the batch into 2 parts to study the 2 levels of
magnesium stearate, instead of manufacturing separate batches for
each combination. This design reduces the number of granulation
batches that need to be manufactured by half, while still providing
sufficient information to understand the main effects and in-
teractions of the various excipients. Experimental design batches
are shown in Table 1.
DoE batches were manufactured in a Fuji high shear granulator
with a 25-L bowl at 5 kg total batch size. The impeller speed was
maintained at 240 rpm (equivalent tip speed of 4.8 m/s), chopper
run at low speed (1000 rpm). A peristaltic pump was used to add
water at a rate of 100 g/min/kg over a period of 4 min and 30 s. The
wet massing time was kept constant at 15 s for all batches while
maintaining same impeller and chopper speeds. The wet granula-
tion was passed through a conical mill (Comil 197S) with 0.125-inch
(3.2 mm) round opening screen at impeller speed of ~1350 rpm.
The granulation was then dried using the GPCG-1 fluid-bed dryer at
70 C to a moisture content of 1.50%. The dried granulation was
milled using a conical mill (Comil 197S) with 0.045-inch (1.1 mm)
screen opening at impeller speed of ~1350 rpm. The milled gran-
ulation was split into 2 equal sublots and then placed in a bin
blender (8.3 L) and mixed with the extragranular excipients and
appropriate amount of magnesium stearate per the experimental
design. The final blend was compressed into 400 mg strength
tablets (800 mg tablet weight) on the 6 station Korsch PH 100 tablet
press using embossed tooling. Batches were characterized using
methods described below. Statistical analysis of response data was
performed by regression analysis in SAS JMP (SAS Institute Inc.,
Cary, NC).
QbD Methodology
After formulation composition was established, studies were
conducted to evaluate effect of material attributes and process
parameters on product attributes of the selected formulation. CQAs
for brivanib alaninate tablets were identified as tablet potency,
dosage uniformity, impurities, dissolution rate, and appearance.
Appearance was considered a CQA as it is perceived to affect patient
compliance and hence outcome of therapy.
Subsequent to the identification of the CQAs, failure mode effect
analysis was conducted to identify material attributes and process
parameters to be studied for their impact on CQAs. Although all
material attributes and process parameters that can potentially
affect CQAs were considered as part of the risk analysis, only a
subset of these attributes and parameters were selected for
development studies as warranted by the risk analysis. Studies
were conducted to generate knowledge regarding the effect of the
selected input material attributes and process parameters on CQAs
and identify critical material attributes (CMAs) and critical process
170 S.I.F. Badawy et al. / Journal of Pharmaceutical Sciences 105 (2016) 168e181
Figure 1. QbD methodology applied to the development of brivanib alaninate tablets.
parameters (CPPs) based on these studies. Critical attributes and
parameters are defined as such if their variation within reasonable
range results in a measurable impact on any of the CQAs. Proven
acceptable ranges (PARs) were established for the CMAs and CPPs
based on these development studies. Controlling CMAs and CPPs
within their PARs, together with in-process tests and controls of
intermediate materials, constituted the control strategy for the
brivanib alaninate tablet manufacturing process (Fig. 1).
All non-CPPs which were either not identified for evaluation as
per the risk analysis or found to be noncritical from development
studies were not included in the control strategy. These parameters
were set at a fixed value at batch record level, and no PARs were
established for these parameters.
Physical Characterization
Particle Size Distribution
The particle size distribution of the final granulation was
determined by mesh analysis using an Allen-Bradley Sonic Sifter
(Allan-Bradley, Milwaukee, WI) equipped with a series of 6 screens
and a pan. Approximately 5 g of the sample was tested with a pulse
setting of “5,” sift setting of “5,” and total sifting time of 5 min. The
geometric mean diameter of the granulation was calculated from
the granulation weight fraction retained on each screen. The
percent fines (<75 mm) was calculated by the summation of the
percent material retained over the 325-mesh screen and dust
collected in the pan of the sonic sifter.
Flow Properties
Flow properties of the lubricated blend were tested using an
Erweka GTB powder flow tester (Erweka GmbH, Heusenstamm,
Germany). The time required for 100 g of the blend to flow through
10-mm aperture was determined and converted to a flow rate
in g/s.
Compaction Properties
Presster compaction simulator (Metropolitan Computing Cor-
poration, East Hanover, NJ) was used to generate the compaction
profile data at 14-ms dwell time using oval-shaped B tooling and
800 mg tablet weight. Different compaction forces were used to
compress the granulation, and tablet hardness obtained at each
compression force was plotted against the corresponding
compaction force to obtain the compaction profile. The Presster
data were used to determine compactibility parameter (initial slope
of the compaction profile), hardness to force ratio at target tablet
hardness, and maximum hardness achieved for each batch.
Granule Porosity
Pore distribution of granulation and tablets was determined by
mercury intrusion porosimetry (MIP, AutoPore III; Micromeritics,
Norcross, GA). A fraction of the tablet or a granulation sample of the
particle size fraction passing through 60-mesh screen and retained
on 100-mesh screen was used for the porosity determination.
Incremental pore volume was determined at different pressures
ranging from 1 to 61,000 psi. The pore diameter corresponding to a
given pressure was calculated by the software of the instrument
according to the Washburn equation using mercury surface tension
value of 485 dyne/cm and mercury contact angle of 130 . Pores in
the 0.1- to 10-mm range were used for the assessment and com-
parison of the different batches.
Tablet Appearance
Compression runs for all batches were monitored for sticking to
the tablet compression tooling and defects on the tablets. Tooling
was removed after each run to assess for sticking, filming, and
filling of the embossing. A scoring system was devised to help
quantify the extent of sticking on punches, and extent and fre-
quency of defects on the tablets. The criterion for tablet picking was
based on the severity and frequency of picking. A score of “0”
indicated no more than minor picking on <30% of tablets, whereas a
score of “5” indicated severe picking in >30% of tablets. A scoring
system was also established to rank order the formulation for
tendency to stick to the punches, wherein no sticking was assigned
a score of “0” and severe sticking that obliterated punch face
emboss was assigned a score of “5.”
Analytical Methods
Tablet Dissolution
Brivanib alaninate is a Biopharmaceutics Classification System
class II compound with pH-dependent solubility and high perme-
ability.13 Dissolution test of brivanib alaninate tablets was con-
ducted in 1000 mL of 50-mM phosphate buffer, pH 6.8, containing
1.0% Triton X-100 at 37 C using USP apparatus II at 75 rpm.
The target dissolution limit was not less than 75% dissolved after
30 min, which was based on the dissolution profiles of formulations
used in the different stages of clinical development.
Moisture Methods of Blends and Tablets
An LOD method with testing temperature at 105 C was used for
blends and tablets to determine moisture at the end of fluid-bed
drying and during tablet coating.
A USP volumetric KarleFischer titration method was used to
determine the water content in the final film-coated tablets. The
tablets were ground to fine powder before adding to the
KarleFischer titration vessel.
An on-line near infrared (NIR) spectroscopic method was also
developed as an in-process control to measure and control moisture
in brivanib granulation during fluid-bed drying using a Brimrose
4030-EX explosion-proof NIR spectrometer (Brimrose Corporation,
Sparks, MD). The spectrometer is attached to a window on the side of
the dryer bowl and is used to measure granule moisture through the
window in real time during the drying process. A chemometric
model was developed and calibrated to water content (as % w/w) in
the granulation using results from LOD measurements of calibration
standards determined by a moisture balance. The on-line NIR
measurements determine when the material has reached the
desired moisture level and provide the end point for the drying step.
 S.I.F. Badawy et al. / Journal of Pharmaceutical Sciences 105 (2016) 168e181 171

Calibration standards were made up of granules equilibrated to

varying moisture levels. During method development, additional
samples were created to test the robustness of the method with
respect to granule density, granule size, excipient and drug sub-
stance concentration, and drug substance particle size.

HPLC Potency and Impurity Method

Tablet potency and impurity were determined by HPLC reverse
phase method. Tablets were dispersed to fine powder in the
diluent, a mixture of acetonitrile, water, and acetic acid. Filtered
sample solutions were analyzed by an isocratic HPLC method for
potency and a gradient method for impurities using Zorbax SB-AQ
column (Agilent Technologies, Santa Clara, CA) and mobile phase
consisting of ammonium trifluoroacetate buffer, acetonitrile,
methanol, and trifluoroacetic acid. The tablet potency was deter-
mined based on the UV response of the active peak from the sample
solution and that from an external standard solution.
HPLC potency and impurity method was validated according to
ICH guideline Q2(R1), Validation of Analytical Procedures. The
validation results demonstrated that the specificity, linearity, ac-
curacy, precision, and robustness of the method were appropriate
for its intended use.

NIR Tablet Potency Method

The at-line NIR method was developed and validated in the
reflectance mode using a Thermo Antaris NIR spectrometer
(Thermo Fisher Scientific, Waltham, MA) in conjunction with a
chemometric model calibrated to brivanib alaninate concentration
(% w/w). Using prepared calibration tablet standards, the NIR
measurements were calibrated with results from the primary HPLC
tablet potency method computed from the average of 2 prepara-
tions of 5 tablets each. An appropriate partial least squares che-
Figure 2. Flow diagram of the manufacturing process for brivanib alaninate filme
mometric model was selected to ensure that the spectral variation
coated tablet.
was covariant with the brivanib concentration. The NIR method
computes % concentration of brivanib alaninate from 30 individual
compression tooling. Previously reported study showed that wet
tablet measurements, which is then adjusted to % label claim using
granulation reduces sticking of the blend to tablet compression
the individual tablet weights. The batch potency is computed from
tooling compared with a roller compaction process.14 The lower
the average of the 30 tablet results. Because the NIR method
sticking tendency of wet granulation over roller compaction was
measures only the surface of the tablet, a statistical analysis was
attributed to the differences in the ability of the granulation process
conducted to determine the appropriate number of tablets that
to modify exposed particle surface through the agglomeration of
would provide a similar assurance of batch potency as measuring
particles. The presence of greater proportion of primary drug par-
potency of 2 sets of 5 tablets by HPLC analysis.
ticles in the formulation is expected to exacerbate sticking, whereas
Tablet standards were prepared to build the calibration model for
the presence of agglomerated drug particles reduces sticking. Fines
the NIR method. The composition of the tablets was varied using a
in the wet granulated formulation are expected to be agglomerates
structured design so that obtained spectra would encompass ex-
of primary drug particles, in contrast to fines in roller-compacted
pected variations that may be encountered during routine use. The
samples which are largely unagglomerated primary drug particles.14
calibration standards were developed using a development granu-
lation batch and then through DoEs, extragranular excipient levels
were uniformly varied to achieve a drug substance calibration in-
dependent of extragranular excipient level. All calibration tablets
were manufactured using pilot-scale processes that mimicked the
commercial manufacturing process. Robustness was further built
into the model by identifying sources of method variation and either
testing these variations directly or incorporating them into the
calibration set of prepared tablets. Sources of method variation that
were identified and assessed included drug substance particle size,
granule size, tablet moisture content, tablet hardness, laboratory
humidity, and instrument-to-instrument variability.

Results

Formulation Selection and Optimization

Wet granulation process (Fig. 2) was selected to reduce the risk of Figure 3. Impeller current profile during granulation of formulations with different
drug substance adhesion to manufacturing equipment and disintegrants.
172 S.I.F. Badawy et al. / Journal of Pharmaceutical Sciences 105 (2016) 168e181

Table 2 The formulation of DoE study showed that higher level of HPC
Composition of Brivanib Alaninate FilmeCoated Tablets improves the formulation and process robustness by reducing the
Ingredient Function Prototype Final amount of fines in the granulation, which lowers its sticking
Formulation Formulation potential to tablet tooling and picking of tablets in compression. In
Concentration (% w/w) addition, higher amount of the lubricant (magnesium stearate)
reduced the risk of sticking and did not affect the dissolution rate.
Brivanib alaninate Active 50.00 50.00
(BMS-582664-02) The negative effect of magnesium stearate on compaction proper-
Microcrystalline Filler/binder 39.50 38.25 ties was minimal and of little practical significance. In addition, a
cellulose (PH102) 2-way interaction was observed between HPC and magnesium
Croscarmellose sodium Disintegrant 3.00 3.00
stearate (Fig. 4), which showed that formulations with lower binder
Hydroxypropyl cellulose Binder 3.00 4.00
Crospovidone Disintegrant 3.00 3.00 content have higher lubricant requirement. At the higher HPC level,
Colloidal silicon dioxide Glidant/compaction aid 0.50 0.50 the powder is better granulated resulting in magnesium stearate
Magnesium stearate Lubricant 1.00 1.25 being less critical to reducing sticking. However, at the lower level
Total, core tableta 100.0 100.0 of HPC, the powder is less granulated with more fines, and the
Opadry® (weight gain) Film coat 3.00 3.00
amount of magnesium stearate is critical to reduce sticking of
a
Total core tablet weight ¼ 800 mg. powder to the punch surface, hence resulting in the 2-way inter-
action between HPC and magnesium stearate. The outcome of the
study indicated that the formulation needs to be well-granulated to
Figure 3 shows impeller power consumption versus amount of minimize sticking risk. Accordingly, concentrations of HPC and
water added for formulations with CCS or CPVP as the intragranular magnesium stearate were increased in the optimized formulation
disintegrant. As water is added, impeller power increases as granule selected for the commercial product (Table 2).
growth takes place. A rapid increase in impeller power around
granulation end point is generally undesirable as it indicates rapid
CQAs, Process Parameters, and Material Attributes
uncontrolled granule growth which makes the process more diffi-
cult to control. The use of CCS as the intragranular disintegrant
Details for each CQA are presented below with respect to ma-
showed steady controlled granule growth, which increases the
terial attributes and process parameters assessed in development
robustness of the manufacturing process. In contrast, the use of CPVP
studies and those identified as critical based on development
as the intragranular disintegrant showed rapid granule growth and
studies. Fishbone diagrams, 1 for each CQA, are depicted below
“sharp” end point, which increases sensitivity to changes in process
showing the different parameters and attributes evaluated for their
parameters. This was shown by impeller power consumption mea-
impact on the CQA. Those identified as critical based on develop-
surement (Fig. 3) and visual observations during granulation.
ment studies are colored red, whereas noncritical attributes and
On the other hand, the use of CCS as a disintegrant was associ-
parameters are colored green.
ated with incomplete in vitro drug release. A previous study
showed that this was due to pH-dependent and reversible ionic
binding of brivanib alaninate by CCS. A pharmacokinetic study in In Vitro Drug Release (Tablet Dissolution)
monkeys indicated that bioavailability of brivanib alaninate is not Process parameters of individual unit operations and material
negatively impacted by presence of CCS.15 This in vitro binding attributes studied based on risk assessment of dissolution are
phenomenon was not observed with the use of CPVP as a dis- shown in Figure 5.
integrant. Although binding of brivanib alaninate by CCS was not
expected to be biorelevant based on this study, bioavailability risk Material Attributes and Dissolution. As shown in Figure 5, drug
posed by CCS still remained. Based on the findings from the wet substance particle size was evaluated for its effect on the dissolu-
granulation and CCS binding studies, CCS was selected as an tion rate for this BCS II compound. Tablet batches were manufac-
intragranular disintegrant to enhance robustness of the wet gran- tured at target process parameters using drug substance lots with
ulation process. However, CPVP was selected as the extragranular different particle size and evaluated for their dissolution behavior.
disintegrant to reduce total amount of CCS in the tablet and mini- Tablet dissolution rate was found to be dependent on drug sub-
mize the risk of potential in vivo brivanib alaninateeCCS binding, stance particle size. Tablets manufactured with drug substance
which could compromise bioavailability. Composition of the pro- particle size around the upper limit of the range typically produced
totype formulation is shown in Table 2. by the drug substance manufacturing process (D90 ¼ 133 mm) still

Figure 4. (a) Interaction plot of magnesium stearate and hydroxypropyl cellulose on tablet picking. (b) Contour plot of the effect of magnesium stearate and hydroxypropyl cellulose
on sticking to compression tooling.
 S.I.F. Badawy et al. / Journal of Pharmaceutical Sciences 105 (2016) 168e181
Figure 5. Material attributes and process parameters impacting tablet dissolution.
met dissolution limit (Fig. 6). However, dissolution of tablets
manufactured with this large drug substance particle size was
lower compared with the smaller drug substance particle size (%
dissolved at 30 min was 89% and 99% for drug substance D90 ¼ 133
mm and 30 mm, respectively). Hence, drug substance particle size
was considered a CMA as variation in particle size was reflected in a
change in the tablet dissolution rate.
Binder (HPC) particle size and solution viscosity, disintegrant
(CCS) particle size, and lubricant (magnesium stearate) surface area
were also evaluated for their potential impact on tablet dissolution
rate. This was achieved by using alternate vendors for these ex-
cipients exhibiting differences in the above attributes of interest.
Variation in these excipient attributes resulted in minimal change
in dissolution rate, and hence they were considered noncritical
(Fig. 5).
Process Parameters and Dissolution. Based on risk assessment,
development studies were conducted to assess the impact of pro-
cess parameters in the following unit operations on tablet disso-
lution rate: wet granulation, wet milling, dry milling, and
lubrication.
Wet Granulation Unit Operation. A preliminary screening study
was conducted to select wet granulation parameters to be studied
in a full-factorial DoE study. The results of the screening study
identified the following process parameters for the granulation
process DoE study: (1) water amount, (2) impeller tip speed, and (3)
Figure 6. Dissolution profiles of brivanib alaninate tablets manufactured using drug
substance with different particle size. Dissolution in 50-mM phosphate buffer at pH 6.8
with 1% Triton-X; USP Apparatus II; paddle speed of 75 rpm. Error bar represents
1 standard deviation.
173
wet massing time. A 23 full-factorial design with 3 replicate center
points was used to study the effect of these parameters on tablet
dissolution at 2 kg scale in a 10-L high-shear granulator using a
single batch of drug substance. Details of study design and results
are reported elsewhere.16 Regression analysis showed that all main
effects of the 3 granulation parameters and 2-way interactions
were statistically significant (p < 0.05) for the amount of drug
dissolved at 30 min. Regression model fit to the DoE data and model
prediction for in-process material attributes and tablet dissolution
are shown in Figure 7. All 3 wet granulation process parameters
affected tablet dissolution in the same direction. Higher level of
water, impeller speed, or wet massing time resulted in a decrease in
dissolution rate. Tablet dissolution rate increased with the increase
in granule pore diameter (determined by MIP; Fig. 8). In addition,
granule pore diameter showed similar dependence on granulation
parameters as tablet dissolution rate (Fig. 7b), suggesting that the
effect of granulation parameters on tablet dissolution is achieved
through their impact on granule porosity. In contrast, dissolution
showed poor correlation with granule size (mean granule diam-
eter; Fig. 8). In addition, tablet disintegration did not correlate with
granulation parameters or tablet dissolution rate. This indicated
that granule disintegration (rather than tablet disintegration) is the
rate controlling step for dissolution. Hence, the 3 wet granulation
parameters highlighted in Figure 5 were considered critical for
dissolution through their impact on granule porosity. These data
are consistent with recent studies highlighting that granule
porosity is the overriding mechanism impacting drug product
patient-centric CQAs such as drug release (dissolution)17 and
bioavailability. Efforts have been made to study the kinetics of
granule consolidation during wet granulation and to make real-
time measurements of granule densification during wet granula-
tion processes.18 These reports and our present study indicate that
particle size alone may not be an adequate response parameter to
assess granulation outcomes.19,20
Milling Unit Operation. A milling study was conducted to eval-
uate 3 milling parameters: (1) screen size for wet milling, (2) screen
size for dry milling, and (3) impeller speed for dry milling. A split-
plot DoE design was used for the study by making a single wet
granulation batch, which was split into 2 sublots for wet milling;
each was then used for a full-factorial study of the 2 dry milling
parameters. There was no impact of the change in milling param-
eters on tablet dissolution despite the small changes observed in
the granule particle size distribution resulting from the different
milling conditions. This is consistent with the findings from the wet
granulation study, which concluded that granule size does not
impact tablet dissolution rate. Milling parameters were, therefore,
designated as noncritical with respect to tablet dissolution as
shown in Figure 5.
Blending (Lubrication) Unit Operation. A magnesium stearate
blending (lubrication) study was carried out to assess the risk of
overblending of magnesium stearate on tablet dissolution. Increasing
blending time of magnesium stearate is known to reduce tablet
dissolution rate in some cases on account of the increased shearing of
magnesium stearate layers, which subsequently coat and impart
more hydrophobic characteristics to granule surfaces (a phenomenon
commonly referred to as overlubrication).21-23 In the manufacturing
process for brivanib alaninate tablets, magnesium stearate is blended
for 250 revolutions in a bin blender. Increasing number of revolutions
to 1000 did not have any impact on tablet dissolution. To further
assess the risk of overlubrication, the concentration of magnesium
stearate was also increased to 1.5% w/w (from 1.25% w/w in the
selected formulation) and also blended for 1000 revolutions. These
changes did not impact tablet dissolution which confirmed low risk of
overlubrication. Magnesium stearate blending time (number of
blender revolutions) was, hence, designated as noncritical.
174 S.I.F. Badawy et al. / Journal of Pharmaceutical Sciences 105 (2016) 168e181

Figure 7. Regression model for the effect of wet granulation process parameters on in-process material and tablet attributes. (a) Model fit to DoE data. (b) Model prediction;
figure represents main effects plots with 95% confidence intervals, showing the change in response as the variable is changed from the low to the high levels while holding other
variables at their center point value.

Tablet dissolution was also found to be independent of tablet

hardness (data not shown), which was not unexpected considering
that the rate limiting step of tablet dissolution is granule disinte-
gration rather than tablet disintegration as described previously.

Figure 8. Correlation of granulation particle size (GMD) and granule mean pore Figure 9. Material attributes and process parameters impacting water content and
diameter with tablet dissolution. impurities.
 S.I.F. Badawy et al. / Journal of Pharmaceutical Sciences 105 (2016) 168e181
Figure 10. Material attributes and process parameters impacting potency and content
uniformity.
Water Content and Impurities
Brivanib alaninate is an ester prodrug which undergoes hydro-
lysis reaction resulting in the formation of parent compound BMS-
540215 (brivanib).24 BMS-540215 is the only degradant observed in
the drug product on accelerated and long-term stability studies.
Hence, this degradant is the major determinant of product shelf life.
The hydrolysis reaction of brivanib alaninate involves water as a
reactant. Thus, the rate of formation of BMS-540215 was found to
be a function of relative humidity (water activity) and the storage
temperature.25
Process parameters and material attributes studied based on
risk assessment for water content and impurities are shown in
Figure 9. Relative humidity inside the package on stability is a
function of initial tablet water content and package design (mois-
ture permeation rate through the pack and the amount of desiccant
in the package, if present). A mathematical model was developed
and previously reported25 which quantitatively predicts the effect
of tablet initial water content and package parameters on the rate
of formation of BMS-540215 on stability. The concentration of BMS-
540215 versus time profile was simulated using a kinetic model for
the formation of BMS-540215 as a function of relative humidity in
the environment. The kinetic model was combined with a sorption-
desorption moisture transfer model which calculates the relative
humidity inside the package at different time points. The combined
model, which simulates the amount of BMS-540215 in the pack-
aged tablets as a function of time, was used to study the effect of
different variables such as package permeability and initial tablet
water content on product stability. Model simulations indicated
Figure 11. Material attributes and process parameters impacting tablet appearance.
175
that BMS-540215 formation on stability is dependent on initial
tablet water content, particularly for the aluminum blister package.
The initial water content in the tablets was, therefore, designated as
a CQA.
Control of water in the final product is achieved by (1) drying of
the granulation and tablets in the fluid bed and coating unit op-
erations, respectively; (2) bulk storage of the tablets in sealed
aluminum bags with desiccants; (3) final packaging of the tablets
under controlled humidity conditions; and (4) package design to
control water ingress into the package on shelf life. Water added
during wet granulation is removed by the fluid-bed drying unit
operation to an in-process limit aligned with the target water limit
of the tablets at release. An in-line NIR method was developed for
the determination of granulation water content in the fluid-bed
dryer. Subsequent to wet granulation and fluid-bed drying, mois-
ture pick-up by the product can take place during milling, blending,
compression, and coating unit operations. Moisture pick-up could
be avoided by humidity control of the manufacturing environment.
Alternatively, tablet drying in the coating unit operation can be
used to remove any water absorbed during process steps subse-
quent to fluid-bed drying. The latter strategy was implemented to
obviate the need for strict humidity control of the manufacturing
environment. An initial drying step of the core tablets was used to
ensure acceptable core tablet moisture control before initiation of
the coating process. This was considered necessary as drying of
tablet core after coating is applied becomes kinetically more diffi-
cult. In addition, a drying step after the application of the film coat
was also implemented to remove any additional moisture pick-up
during coating and to ensure moisture control of the final coated
tablet. Both drying steps were controlled by a tablet LOD method.
Subsequent to film coating, tablet water content was controlled by
the use of protective (moisture impermeable) bulk package and
control of humidity in the packaging environment.
Degradation During Processing. Because BMS-540215 is formed by a
hydrolytic reaction, the risk presented by aqueous wet granulation
to degradation was extensively assessed. Preliminary studies using
target wet granulation parameters showed no increase in the level
of BMS-540215 (or any other impurity) during wet granulation
even when the wet mass was stored for 25 h at room temperature
before drying. In addition, the effect of wet granulation process
parameters on the formation of BMS-540215 during wet granula-
tion was also assessed. Granulation conditions imparting high
levels of water and shear forces were considered potential risk
factors for brivanib alaninate hydrolysis. Therefore, batches man-
ufactured using the high levels of water, impeller speed, and wet
massing time (in the DoE study described previously) were tested
for their impurity content. No increase in any impurity level was
observed in these batches compared with the drug substance
indicating that wet granulation process parameters are not critical
for BMS-540215 formation in the tablet. The low risk presented by
wet granulation to tablet impurity content is attributed to the low
solubility of brivanib alaninate in the granulating liquid (water).
Initial impurity level in the tablet is hence determined solely by its
level in the drug substance used in manufacturing. Impurity level in
the drug substance, including BMS-540215, is therefore considered
CMA and is controlled by the drug substance specifications.
Potency and Content Uniformity
Process parameters and material attributes studied based on
risk assessment for potency and content uniformity are shown in
Figure 10.
Blending parameters and drug substance particle size presented
low risk on account of the high drug loading of brivanib alaninate in
the tablets (50% w/w). Therefore, they were not studied for their
176 S.I.F. Badawy et al. / Journal of Pharmaceutical Sciences 105 (2016) 168e181
impact on tablet potency and content uniformity. For high drug
load formulations, potency and content uniformity are typically
controlled primarily by average tablet weight and weight variation.
These were, hence, considered critical for the control of these CQAs.
In-process material attributes known to impact tablet weight
variation during compression includes granulation particle size
distribution and flow properties. Wet granulation and milling
process parameters were evaluated for their effect on granulation
particle size distribution, flow properties, and tablet weight varia-
tion in compression in the abovementioned DoE studies. Although
change in wet granulation and milling parameters caused some
variation in final blend particle size distribution and flow rate
(measured by an Erweka® flow tester; Fig. 7), this variation in
granule size distribution and flow rate had no impact on tablet
weight variation in compression studies. Wet granulation and
milling parameters are hence considered noncritical with respect
to their impact on potency and content uniformity.
Tablet Appearance
Picking and sticking tendency during tablet compression caused
by inherent properties of the brivanib alaninate drug substance was
identified early in development as a risk to tablet appearance and to
process yield and cycle time.14 The wet granulation process was
primarily selected to manufacture brivanib alaninate tablets to
mitigate sticking risk. Incorporation of the drug substance particles
in granules manufactured by wet granulation was shown to elim-
inate sticking.14 Moreover, the concentration of the binder (HPC)
was found to be statistically significant with respect to the sticking
response in the formulation DoE study and hence was optimized in
the selected formulation to eliminate sticking as described previ-
ously. HPC reduces sticking by reducing fines and ungranulated
drug substance particles in the final blend. In addition to HPC, the
concentration of the lubricant (magnesium stearate) was found to
be statistically significant in the formulation DoE and was similarly
optimized to mitigate the sticking risk.
Based on the understanding of the sticking risk, drug substance
particle size, HPC and magnesium stearate material attributes, and
wet granulation, milling and lubrication process parameters were
studied for their effect on sticking (Fig. 11).
Material Attributes and Sticking. The effect of drug substance par-
ticle size on sticking was evaluated. Small drug substance particle
size at the lower specification limit did not show sticking when
used for drug product manufacturing. Variation in HPC particle size
and viscosity and magnesium stearate surface area were also
studied for potential effect on tablet sticking during compression.
As mentioned previously, this was accomplished by using alternate
vendors for these excipients exhibiting differences in the attributes
of interest. No effect was observed due to variation in these
excipient properties, and all batches made with the different
excipient properties showed no sticking during compression.
Process Parameters and Sticking. None of the wet granulation or
milling DOE batches described previously showed any sign of
sticking to the compression tooling. Similarly, the shorter blending
time of magnesium stearate did not result in any sticking in the
lubrication study.
Based on these results, no parameters or attributes were found
to be critical for sticking. The selection of the manufacturing pro-
cess (wet granulation) and the optimization of HPC and magnesium
stearate levels in the selected formulation resulted in a robust
formulation that showed no sensitivity to process parameters or
input material attributes with respect to blend sticking tendency
during compression.
Tablet erosion during coating is a typical risk to the appearance
of a film-coated product and hence presents a risk to the appear-
ance of brivanib alaninate tablets. Tablet hardness was therefore
controlled to minimize erosion risk. A lower limit of tablet hardness
was identified which showed low friability in the USP test (<0.2%)
and resulted in no erosion during tablet coating. Analysis of
compaction data from the wet granulation DOE study revealed no
significant main effects of granulation parameters with respect to
compaction responses. All batches in the study had comparable
compaction profiles indicating that variation in granulation pa-
rameters had no impact on the compaction properties of the final
blend. Moreover, the maximum hardness achieved in the study for
all batches was well above the target tablet hardness.
Control Strategy
Based on development studies and knowledge of CPPs and at-
tributes impacting CQAs, a combination of control elements were
implemented to ensure that the drug product met the required
quality standards. The drug product manufacturing control strategy
included (1) process parameter controls (process design space), (2)
input material controls, (3) in-process attribute controls during
manufacturing, and (4) end-product testing. The control strategy is
essentially a combination of limits on all critical starting material
attributes, in-process material attributes, and process parameters
identified for each CQA (Figs. 5, 9, 10, and 11). If a parameter or
attribute is critical for >1 CQA, the limit in the control strategy is
based on the CQA with the tightest requirement.
In addition, the control strategy also included at-line release
tests and finished product release tests. The at-line release tests
included testing of core tablets for potency and uniformity of
dosage units. Release tests of the finished product (film-coated
tablets) include dissolution, water content, appearance, and iden-
tification. The control strategy is listed in Tables 3 and 4 and
described in the following sections.
Process Parameters Design Space
Wet granulation parameters (water amount, impeller speed and
wet massing time) were the only process parameters identified as
critical based on the studies described previously. Wet milling, dry
milling, and lubrication parameters studied were found to be
noncritical as all CQAs showed little sensitivity to these parameters.
The results of the wet granulation DoE study described previ-
ously formed the basis for assigning the 3 wet granulation pa-
rameters as critical parameters on account of their impact on tablet
dissolution rate. Because all batches from the DoE study met the
acceptance criteria for dissolution (not less than 75% dissolved in 30
min) and all other CQAs, parameter ranges in the DoE are consid-
ered multivariate PARs and hence represent the process design
space (Table 3). Parameter values for the center point batches were
established as the target parameters.
The design space for the wet granulation parameters was
established based on the DoE study at 2.5-kg scale. According to the
International Conference on Harmonization guideline, a design
space can be developed at any scale and be applicable to the
commercial scale if the design space is shown to be scale
Table 3
Control Strategy: Process Parameters Design Space
Process Parameter Unit Target (Design Space Range)
Water for granulation % w/w 46.5 (44-49)
Impeller tip speed m/s 4.8 (3.6-6.0)
Wet massing time s 30 (10-50)
 S.I.F. Badawy et al. / Journal of Pharmaceutical Sciences 105 (2016) 168e181
Figure 12. Comparison of brivanib alaninate in-process material and tablet attributes for product manufactured at 3 different scales. (a) Particle size distribution of granulation,
(b) compaction profile of final blend, and (c) tablet dissolution profile.
independent.26 As a result, a strategy to justify the extrapolation of
the design space from the small scale to the commercial scale was
established. The strategy described below provides the necessary
justification, while minimizing the number of batches to be man-
ufactured at the commercial scale.
A prerequisite for the extrapolation among scales is to have the
design space expressed in scale-independent terms. The critical
wet granulation parameters of the brivanib alaninate design space
are expressed in scale-independent terms as shown in Table 3. The
suitability of the scale-independent terms used for the design space
was verified by comparing product quality attributes at 3 scales,
representing laboratory scale, pilot scale, and commercial scale. Key
attributes of these batches manufactured at the 3 scales using
target parameters values were comparable as shown in Figure 12.
In addition, the design space was also verified by manufacturing
2 batches at 2 extreme regions in the design space at the com-
mercial scale (off-target batches). The strategy for selecting the
parameters for those batches was based on selection of worst-case
combination of the 3 critical parameters. Because the 3 critical
granulation parameters affect dissolution in the same direction
(Fig. 7b), combination of the high levels of all 3 parameters repre-
sents the highest risk for slow dissolution rate. Conversely, com-
bination of the low levels of all parameters represents the highest
risk of “undergranulation” and associated risks of poor flow prop-
erties and sticking to tablet compression tooling. Results from the 2
off-target batches met all acceptance criteria, hence providing
additional justification for the applicability of the design space to
the commercial scale.
Input (Raw) Material Controls
Drug substance particle size and impurity content were the only
input material attributes identified as critical to product perfor-
mance or CQAs. None of the excipient-related attributes was found
to be critical as concluded from the development studies. A corre-
lation between the particle size of the drug substance and dissolu-
tion rate was established as previously discussed. Control of the drug
177
substance particle size was therefore necessary to ensure consistent
and acceptable dissolution of the tablets. A particle size specification
of 40-120 mm (D90) was established to ensure acceptable dissolution
rate (not less than 75% dissolved at 30 min). While the upper limit is
established to ensure acceptable dissolution, the lower limit of par-
ticle size was set to minimize risk of sticking to compression tooling
resulting from fine drug substance particles.
Because both wet granulation parameters and drug substance
particle size are critical to dissolution rate, it was necessary to
establish that the upper limit of the drug substance particle size
range shows acceptable tablet dissolution even at the edge of the
process design space. Because the wet granulation DoE was con-
ducted with drug substance having D90 of 68 mm, additional study
was necessary to demonstrate performance of the larger particle
size at the edge of the process design space. A risk-based approach
was used to achieve this purpose leveraging the knowledge
established from the wet granulation study which illustrated that
combination of the high levels of all 3 parameter values results in
the lowest dissolution rate. Hence, the worst-case combination of
granulation parameters which was identified from the DoE study
(high water level, high impeller speed and high wet massing time)
was evaluated in conjunction with large drug substance particle
size (D90 ¼ 106 mm) to ensure that the worst-case combination of
process parameters and drug substance particle size would result in
an acceptable dissolution rate. As shown in Figure 13, dissolution of
tablets manufactured with this worst-case combination of process
parameters and particle size showed acceptable dissolution profile
which met the acceptance criteria.
The other critical input material attribute (drug substance im-
purity content) was controlled by incoming material specifications,
with a limit of not more than 1.0% on the parent compound (BMS-
540215).
In-Process Attributes Control
Fluid-Bed Drying. Water content of the granules during drying was
tested and controlled by the NIR spectroscopy method to a limit of
178 S.I.F. Badawy et al. / Journal of Pharmaceutical Sciences 105 (2016) 168e181

Figure 13. Dissolution profile of tablets manufactured with large drug substance
particle size and high level of wet granulation parameters (water amount, impeller
speed, and wet massing time). Dissolution in 50-mM phosphate buffer at pH 6.8 with
1% Triton-X; USP Apparatus II; paddle speed of 75 rpm.

not more than 1.3% w/w. The limit was implemented to ensure that
water added during wet granulation is removed by drying in the
fluid-bed dryer and to account for moisture associated with the
extragranular excipients added downstream. In addition, the limit
was set to enable the coated tablet to meet water content accep-
tance limit in the final product.
The established NIR method was used to monitor granulation
moisture during fluid-bed drying of several development and
commercial-scale batches. Comparison between the NIR and LOD
results showed close agreement (Fig. 14), and it was determined
that moisture levels could be accurately measured between 0% and
3% with approximately 0.25% resolution. The NIR method suitably
monitored the drying process of commercial-scale batches at the
manufacturing site, enabling determination of the drying end
Figure 15. (a) NIR calibration curve for at-line measurement of core tablet potency.
point. Predicted tablet assay by NIR method, reference assay by HPLC. (b) Potency results for
full-scale brivanib tablet batches comparing composite HPLC (for coated tablets) versus
at-line NIR (for core tablets). “Minimum Tablet” and “Maximum Tablet” are for the NIR
Compression. As typically implemented in tablet compression, up- potency determination.
per and lower acceptance limits for core tablet weight (both indi-
vidual and mean tablet weight) were established to ensure
implemented to ensure that tablets could withstand the coating
acceptable assay and content uniformity of the core tablets. Simi-
process, packaging, and shipping.
larly, control of tablet hardness within the acceptance limits was

Film Coating. As mentioned previously, tablet-coating process was

Figure 14. Comparison of online NIR and at-line LOD for moisture monitoring during
fluid-bed drying of brivanib alaninate granulation.
designed to control the water level in the core tablets and to
remove any additional water picked up by the drug product during
film coating of the tablets. To achieve these objectives, the coating
process was designed to include 3 stages: (1) drying of core tablets,
(2) application of film coat (spray), and (3) drying of the coated
tablets. Water content of uncoated tablets (before the start of
coating) and coated tablets (after the end of coating) was deter-
mined by LOD method and controlled by an in-process specification
limit. Water content limit in the core tablet was established at not
more than 1.6% w/w. The control on the water level in the core
tablets ensured that any water picked up by the product during the
stages of manufacture after fluid-bed drying was removed before
coating.
A limit on the water level for the coated tablet was also set at not
more than 1.8% w/w to ensure that any water picked up by the
tablets during coating was removed. Effectiveness of this coating
approach to achieve acceptably low water level in the coated tab-
lets was established and previously reported.25 Water limit of the
coated tablets was based on the simulation of the effect of initial
water content on the rate of hydrolysis of brivanib alaninate. Using
 S.I.F. Badawy et al. / Journal of Pharmaceutical Sciences 105 (2016) 168e181 179

Table 4
Control Strategy: Raw Materials, In Process Controls and Release Tests

Tablet Quality Attribute Raw Material In-Process Control At-line Release Test Off-line Release Test

Drug release Drug substance particle size Dissolution

Impurities Drug substance impurities
Water content Water content at end of fluid-bed
drying of granules (NIR, NMT 1.3%)
Water content of preheated core
tablets in coating pan (LOD, NMT 1.6%)
Water content of tablets at end of
coating (LOD, NMT 1.8%)
Potency Tablet weight during compression Core tablet potency by NIR
Content uniformity Tablet weight during compression Uniformity of dosage units by
weight variation on core tablets
Appearance Tablet hardness Physical imperfections AQL

AQL, acceptable quality limit of tablet defects; NMT, not more than.

the predictive model mentioned previously, the formation of
BMS-540215 as a function of the initial tablet water content was
examined, and results identified an acceptable water content
release limit of not more than 2.4% w/w for final coated tablets. The
1.8% in-process limit of the coated tablets is tighter than release
limit to account for the variability of the in-process LOD method
and the KarleFischer release test method and ensures that tablets
meet the specification limit for water content on release testing.25
Product Release Testing
At-Line Release Tests
Tablet Potency. An at-line release testing for tablet potency is
performed on the core tablets by the NIR method. Tablet samples
were collected throughout compression and used for the potency
analysis. Tablet potency by the NIR method must meet the potency
limit on the release specification. No potency release testing is
performed on the final coated tablets as the potency does not
change during film coating. Data from development and technol-
ogy transfer batches confirmed this requirement.
Figure 15a shows the brivanib alaninate tablet calibration curve
for the NIR method with 7 concentration levels of brivanib alani-
nate and 17 combinations of excipient concentration levels. The
calibration root mean squared error of approximately 1.6% w/w and
a correlation coefficient of 0.99 demonstrated strong correlations
between the HPLC measurements and NIR predictions. A compar-
ison between the HPLC composite potency results and the NIR
potency results was performed for production of full-scale batches
of brivanib alaninate tablets. The results, plotted in Figure 15b,
show excellent agreement between NIR and HPLC results,
demonstrating the capability of the NIR method to quantitate bri-
vanib alaninate level in the core tablets.
Tablet Content Uniformity. Tablet uniformity was also tested on
the core tablets using a weight variation method as allowed by the
compendia at the high drug loading (50% w/w).
Final Product Off-Line Release Tests. Off-line release tests are
implemented for dissolution and tablet identity (Raman spectros-
copy method) and must conform to the acceptance criteria of the
release specifications. In addition, a water content release test by a
KarleFischer method is implemented with a limit of not more than
2.4% w/w as described previously.
No final product release test was proposed for impurities. This is
based on product understanding and input material controls in
place. BMS-540215 is the only degradant of the drug product.
However, BMS-540215 increases only on shelf life and does not
change during drug product manufacturing, even at the worst-case
extremes of the granulation parameter design space. Therefore,
controlling the level of BMS-540215 in the drug substance ensures
acceptable degradant level in the drug product at release and
obviates the need for testing of impurities for tablet release. Im-
purities are tested on market life stability as BMS-540215 level
increases over-shelf life.
Discussion
An integrated and overarching application of QbD principles to
modern drug product development requires a holistic under-
standing of the effect of input variables which include (1) raw
material properties and their potential variability, (2) formulation
composition (qualitative and quantitative), and (3) manufacturing
process parameters. In this article, a drug product control strategy
was developed which encompasses both material property and
process parameter controls, with appropriate in-process and
finished product analytical tests and specification limits (Tables 3
and 4). These analytical tests encompass both conventional
analytical tests and process analytical technology. In the following
section, some of the key elements from the brivanib alaninate QbD
development are discussed further.
Building Process Robustness by Formulation Design
It should be recognized that building manufacturing process
robustness starts with appropriate formulation design. Traditional
product development efforts focus on sequential development of
formulation composition, manufacturing process, and drug product
control strategy. Because excipients affect process performance,
selection of formulation composition without consideration of the
process can limit process robustness. Process development efforts
aimed at establishing parameter space where process performance
is minimally impacted by sources of variability would be signifi-
cantly hampered by suboptimal formulation composition. In the
integrated development exemplified in this article, the assessment
of robustness was studied across the unit operations at the
formulation design stage through fewer, mechanistically driven
experiments.
As an example of how process robustness can be built into
excipient selection and formulation design, we studied the effect of
the disintegrant, CCS versus CPVP, on the end point of the wet
granulation process. Performance of the wet granulation process
with both disintegrants was evaluated and CCS was selected as the
intragranular disintegrant on account of the greater robustness of
the CCS formulation with respect to granulation end point.
Another example was provided by the formulation optimization
to enhance its performance in the tablet compression process with
respect to sticking to compression tooling. As a key risk to process
robustness, the effect of formulation composition on sticking po-
tential was studied in the formulation-ruggedness DoE. Findings
180 S.I.F. Badawy et al. / Journal of Pharmaceutical Sciences 105 (2016) 168e181
related to the role of binder and lubricant and their interaction on
sticking were the basis for the optimization of the concentration of
HPC and magnesium stearate in the final formulation. The opti-
mized formulation demonstrated excellent robustness and showed
no sticking in the process development studies across the full range
of parameters tested.
Design Space Strategy
An integrated design space that incorporates a range of
permissible or expected variability in material properties and
process variables presents significant logistic challenge. These
challenges stem from the wide range of input variables and re-
sponses that must be evaluated. The following 2 strategies can be
adopted to address these challenges: (1) generate an extended
experimental space using multifactorial statistical DoE study or (2)
adopt a selective and sequential study designs of which parameters
and material attributes are evaluated based on mechanistic un-
derstanding of their effect on product attributes.
A single DoE that incorporates variables across material prop-
erties and process parameters collectively, although feasible, be-
comes prohibitive in material, time, and effort requirements.27 To
overcome the limitations of a single comprehensive study, we
highlight in this article the alternative strategy of using a sequential
and selective DoE. Using this approach, studies do not necessarily
use exhaustive statistical designs but are based on sound scientific
and mechanistic understanding of underlying causes of product
responses. For example, the effects of wet granulation process pa-
rameters and material properties on the dissolution rate of brivanib
alaninate tablets were evaluated in separate studies. CPPs, their
design space, and the mechanism of their effect on tablet dissolu-
tion were established in the process DoE. Material properties study
identified drug substance particle size as the critical property for
dissolution and established an acceptable particle size range.
Because the wet granulation study used target drug substance
particle size, while the particle size study was conducted at target
wet granulation parameters, combination of the 2 acceptable
ranges from the 2 studies required further justification. Product and
process understanding established in the respective studies were
leveraged to generate experimental space that captures the worst-
case combination of wet granulation parameters and drug sub-
stance particle size. Additional experimentation of this worst-case
space was conducted to confirm the multivariate acceptability of
the combination of acceptable ranges from the process and mate-
rial property studies. The application of such strategy, as demon-
strated in this article, rationally reduces the extent of
experimentation required by leveraging the scientific and mecha-
nistic understanding of variables that impact product quality and
enables holistic approach to the development of a control strategy
for assuring product quality.
The manufacturing scale used to establish design space presents
another challenge. Ideally, design space would be established at the
commercial production scale. However, this strategy presents sig-
nificant challenges associated with cost and availability of drug
substance. Limited availability of drug substance during drug
product development stage makes this approach impractical in
many cases. ICH Q8 permits development of design space at pilot
scale and extrapolation to the production scale if properly justified.
However, the ICH guideline does not prescribe an approach for
justifying extrapolation of design space from pilot scale. In this
article, an approach for the extrapolation of wet granulation design
space is used which consists of 3 components: (1) wet granulation
parameters at the pilot scale are expressed in scale-independent
terms (impeller tip speed, ratio of water to solids and wet mass-
ing time); (2) scale independence of these parameters was
demonstrated by manufacturing product at 3 different scales (form
pilot to production scales) using target parameters; key product
attributes were shown to be similar across the 3 scales; (3) pro-
duction scale batches were manufactured at 2 points in the process
parameter space which represent worse case parameter combina-
tions with respect to CQAs. Both batches showed acceptable attri-
butes. The approach used was hence based on process
understanding established from the development studies and was
successful in minimizing number of batches manufactured at the
production scale.
Model-Based Specification Limits
Final product and in-process specification limits are key com-
ponents of the control strategy. Specifications need to be linked to 1
or more CQA with limits established within which product per-
formance is acceptable and should be established based on un-
derstanding of the factors affecting the CQA. Traditionally,
specification limits are based on data of key batches (e.g., long-term
stability and clinical batches used in pivotal trials) and perceived
process capability. Such an approach is not very effective because
there are usually other factors affecting a given CQA, in addition to
the attribute for which the specification limit is being set. Estab-
lished limits using the traditional approach do not account for the
multivariate effect of these factors on the CQA. Mathematical
modeling is a useful tool to quantify the effect of these factors on
CQAs and provides the basis for product and in-process specifica-
tions. For brivanib alaninate tablets, degradant formation depends
on initial water content of tablets and on packaging factors such as
package permeability and presence of desiccant. A model-based
approach was used to predict the effect of these different factors
on rate of degradant formation and to rationally establish a speci-
fication limit on initial tablet water content. The model quantita-
tively predicted the effect of the different variables on degradation
rate. Hence, the model allowed for the initial tablet water content
limit to be set based on an in silico study of the variable space
instead of relying on limited set of batch data.
Conclusions
Identification of CQAs of brivanib alaninate tablets, risk assess-
ment, and focused development studies were instrumental to the
identification of an effective control strategy based on QbD prin-
ciples. Controls were implemented on input and in-process mate-
rials and on manufacturing process parameters, which ensure that
final product consistently meets criteria established for CQAs. This
demonstration of QbD principles to integrated and holistic drug
product development highlights the value of mechanistically
driven experiments in building product and process robustness
while addressing key risks to the patient, manufacturing plant, and
product development.
References
1. Yu LX. Pharmaceutical quality by design: product and process development,
understanding, and control. Pharm Res. 2008;25(4):781-791.
2. Yu LX, Amidon G, Khan MA, et al. Understanding pharmaceutical quality by
design. AAPS J. 2014;16(4):771-783.
3. Awotwe-Otoo D, Agarabi C, Wu GK, et al. Quality by design: impact of
formulation variables and their interactions on quality attributes of a lyophi-
lized monoclonal antibody. Int J Pharm. 2012;438(1-2):167-175.
4. Kona R, Fahmy RM, Claycamp G, Polli JE, Martinez M, Hoag SW. Quality-by-
design III: application of near-infrared spectroscopy to monitor roller
compaction in-process and product quality attributes of immediate release
tablets. AAPS PharmSciTech. 2014;16(1):202-216.
5. Merritt JM, Viswanath SK, Stephenson GA. Implementing quality by design in
pharmaceutical salt selection: a modeling approach to understanding dispro-
portionation. Pharm Res. 2013;30(1):203-217.
 S.I.F. Badawy et al. / Journal of Pharmaceutical Sciences 105 (2016) 168e181
6. Saurí J, Mill ~e
an D, Sun -Negre JM, et al. Quality by design approach to under-
stand the physicochemical phenomena involved in controlled release of
captopril SR matrix tablets. Int J Pharm. 2014;477(1-2):431-441.
7. Visser JC, Dohmen WMC, Hinrichs WLJ, Breitkreutz J, Frijlink HW,
Woerdenbag HJ. Quality by design approach for optimizing the formulation
and physical properties of extemporaneously prepared orodispersible films. Int
J Pharm. 2015;485(1-2):70-76.
8. Bohlin M, Jones H, Black S. New developments in scale-up and QbD to ensure
control over product quality. Am Pharm Rev. 2009;12(7):52-57.
9. Kayrak-Talay D, Dale S, Wassgren C, Litster J. Quality by design for wet gran-
ulation in pharmaceutical processing: assessing models for a priori design and
scaling. Powder Technol. 2013;240:7-18.
10. Kushner J, Langdon BA, Hicks I, et al. A quality-by-design study for an
immediate-release tablet platform: examining the relative impact of active
pharmaceutical ingredient properties, processing methods, and excipient vari-
ability on drug product quality attributes. J Pharm Sci. 2014;103(2):527-538.
11. Prpich A, Am Ende MT, Katzschner T, Lubczyk V, Weyhers H, Bernhard G. AIChE
Annual Meeting, Conference Proceedings. 2008.
12. Huynh H, Ngo VC, Fargnoli J, et al. Brivanib alaninate, a dual inhibitor of
vascular endothelial growth factor receptor and fibroblast growth factor re-
ceptor tyrosine kinases, induces growth inhibition in mouse models of human
hepatocellular carcinoma. Clin Cancer Res. 2008;14(19):6146-6153.
13. Narang AS, Badawy S, Ye Q, et al. Role of self-association and supersaturation in
oral absorption of a poorly soluble weakly basic drug. Pharm Res. 2015. Article
in press; doi:10.1007/s11095-015-1645-y.
14. Narang AS, Badawy SIF, Subramanian GA, LaMarche KR, Bindra DS, Rao VM.
High Drug Load Tablet Formulation of Brivanib Alaninate. 2012. WO/2012/
097222.
15. Narang AS, Yamniuk AP, Zhang L, et al. Reversible and pH-dependent weak
drug-excipient binding does not affect oral bioavailability of high dose drugs.
J Pharm Pharmacol. 2012;64(4):553-565.
16. Badawy SI, Narang AS, Lamarche K, Subramanian G, Varia SA. Mechanistic basis
for the effects of process parameters on quality attributes in high shear wet
granulation. Int J Pharm. 2012;439(1-2):324-333.
181
17. Wolfe S, Pafiakis S, Tang D, et al. Effect of Tablet Pore Size Distribution on
Dissolution upon Accelerated Storage Conditions. 2013. AAPS Annual Meeting,
San Antonio, TX.
18. Narang AS, Sheverev VA, Stepaniuk V, et al. Real-time assessment of granule
densification in high shear wet granulation and application to scale-up of a pla-
cebo and a brivanib alaninate formulation. J Pharm Sci. 2015;104(3):1019-1034.
19. Pandey P, Sinko PD, Bindra DS, Hamey R, Gour S, Vema-Varapu C. Processing
challenges with solid dosage formulations containing vitamin E TPGS. Pharm
Dev Technol. 2013;18(1):296-304.
20. Pandey P, Tao J, Chaudhury A, Ramachandran R, Gao JZ, Bindra DS. A combined
experimental and modeling approach to study the effects of high-shear wet
granulation process parameters on granule characteristics. Pharm Dev Technol.
2013;18(1):210-224.
21. Lerk CF, Bolhuis GK, Smallenbroek AJ, Zuurman K. Interaction of tablet dis-
integrants and magnesium stearate during mixing. II. Effect on dissolution rate.
Pharm Acta Helv. 1982;57(10-11):282-286.
22. Pingali KC, Mendez R. Nanosmearing due to process sheardinfluence on
powder and tablet properties. Adv Powder Technol. 2014;25(3):952-959.
23. Sheskey PJ, Robb RT, Moore RD, Boyce BM. Effects of lubricant level, method of
mixing, and duration of mixing on a controlled-release matrix tablet con-
taining hydroxypropyl methyl cellulose. Drug Dev Ind Pharm. 1995;21(19):
2151-2165.
24. Zhao F, Derbin G, Miller S, Badawy S, Hussain M. Acid-catalyzed hydrolysis of
BMS-582664: degradation product identification and mechanism elucidation.
J Pharm Sci. 2012;101(9):3526-3530.
25. Badawy SI, Lin J, Gokhale M, et al. Quality by design development of brivanib
alaninate tablets: degradant and moisture control strategy. Int J Pharm.
2014;469(1):111-120.
26. International Conference on Harmonization of Technical Requirements for
Registration of Pharmaceuticals for Human Use. ICH Harmonized Tripartite
Guideline, Pharmaceutical Development Q8(R2), Current Step 4 Version.
27. Narang A. Addressing excipient variability in formulation design and drug
development. In: Narang A, Boddu S, eds. Excipient Applications in Formulation
Design and Drug Delivery. New York, NY: Springer; 2015.