Pharmaceutical Development and Technology

ISSN: 1083-7450 (Print) 1097-9867 (Online) Journal homepage: http://www.tandfonline.com/loi/iphd20

Developing a quality by design approach to model

tablet dissolution testing: an industrial case study

Ketsia Yekpe, Nicolas Abatzoglou, Bernard Bataille, Ryan Gosselin, Tahmer

Sharkawi, Jean-Sébastien Simard & Antoine Cournoyer

To cite this article: Ketsia Yekpe, Nicolas Abatzoglou, Bernard Bataille, Ryan Gosselin, Tahmer
Sharkawi, Jean-Sébastien Simard & Antoine Cournoyer (2017): Developing a quality by design
approach to model tablet dissolution testing: an industrial case study, Pharmaceutical Development
and Technology, DOI: 10.1080/10837450.2017.1392566

To link to this article: http://dx.doi.org/10.1080/10837450.2017.1392566

Published online: 02 Nov 2017.

Submit your article to this journal

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journalInformation?journalCode=iphd20

Download by: [Gothenburg University Library] Date: 03 November 2017, At: 02:52
 PHARMACEUTICAL DEVELOPMENT AND TECHNOLOGY, 2017
https://doi.org/10.1080/10837450.2017.1392566

RESEARCH ARTICLE

Developing a quality by design approach to model tablet dissolution testing: an

industrial case study
Ketsia Yekpea,b, Nicolas Abatzogloua, Bernard Batailleb, Ryan Gosselina, Tahmer Sharkawib,
Jean-S
ebastien Simarda,c and Antoine Cournoyera,c
a
Pfizer Industrial Research Chair on Process Analytical Technology in Pharmaceutical Engineering, Department of Chemical & Biotechnological
Engineering, Universit
e de Sherbrooke, Sherbrooke, Quebec, Canada; bFaculty of Pharmacy, University of Montpellier I, Montpellier, France;
c
Process Analytical Science Group, Pfizer, Saint-Laurent, Quebec, Canada

ABSTRACT ARTICLE HISTORY

This study applied the concept of Quality by Design (QbD) to tablet dissolution. Its goal was to propose a Received 4 May 2017
Downloaded by [Gothenburg University Library] at 02:52 03 November 2017

quality control strategy to model dissolution testing of solid oral dose products according to International Revised 10 October 2017
Conference on Harmonization guidelines. The methodology involved the following three steps: (1) a risk Accepted 10 October 2017
analysis to identify the material- and process-related parameters impacting the critical quality attributes of
dissolution testing, (2) an experimental design to evaluate the influence of design factors (attributes and KEYWORDS
parameters selected by risk analysis) on dissolution testing, and (3) an investigation of the relationship Dissolution testing; quality
between design factors and dissolution profiles. Results show that (a) in the case studied, the two parame- by design; design of
ters impacting dissolution kinetics are active pharmaceutical ingredient particle size distributions and tab- experiments; risk
let hardness and (b) these two parameters could be monitored with PAT tools to predict dissolution assessment; dissolution
profiles. Moreover, based on the results obtained, modeling dissolution is possible. The practicality and modeling; ICH Q8 guideline
effectiveness of the QbD approach were demonstrated through this industrial case study. Implementing
such an approach systematically in industrial pharmaceutical production would reduce the need for tablet
dissolution testing.

1. Introduction There has been a growing interest in the use of QbD tools for
understanding and prediction of drug release from tablets.
The concept of Quality by Design (QbD) was first introduced in
Hernandez et al. (2016) used NIR to predict in real time dissol-
1992 by Dr Joseph M. Juran (Juran 1992). The US Food and
ution of tablet exposed to different levels mechanical strength.
Drug Administration (FDA) encourages risk-based approaches
Yekpe et al. (2015) demonstrated that dissolution can be pre-
and the adoption of QbD principles within pharmaceutical indus-
dicted based on disintegrant distribution with NIR Chemical
try (Yu et al. 2014). QbD proposes that quality should be built
Imaging Technology. Basalious et al. (2011) applied QbD for
into initial product design rather than assessed at the end of tab- pharmaceutical development of felodipine solid mixture contain-
let manufacture (ICH Q8, R2 2009). Furthermore, the goal of QbD ing hydrophilic carriers and/or polymeric surfactants, for easier
is to ensure finished product quality, improve tablet manufactur- development of controlled-release tablets of felodipine. Dumarey
ing and curtail product costs. Thus, with the QbD initiative, rou- et al. (2015) studied the influence of granule and compression
tine quality control testing, such as dissolution testing (DT) variability introduced by a design of experiments on the entire
performed after tablet manufacturing, could be discontinued if dissolution profile with a multivariate tool. Cui et al. (2012) used a
influential parameters are controlled (Rathore et al. 2009). single time point to study and predict the drug release as a func-
DT aims at verifying active pharmaceutical ingredient (API) dis- tion of API physical properties.
solution in simulated gastrointestinal fluid (Dressman and Kramer Ultimately, if influential parameters are identified and con-
2005; Le Hir et al. 2009). Such pharmaceutical control testing has trolled a priori, efficient applications of QbD principles and risk-
now been applied for more than 30 years and is recognized as a assessment tools could reduce and perhaps eliminate the need
gold standard for in vivo dissolution correlation (Buri 1983; for tests such as “tablet DT” (ICH Q8, R2 2009; ICH Q9 2011a, ICH
Juenemann et al. 2011). Moreover, this control test is routinely Q10 2011b). Such evolution would be highly beneficial to the
employed by the pharmaceutical industry and regulatory agencies pharmaceutical industry in terms of process understanding, envir-
to improve monitoring of batch-to-batch reproducibility in manu- onmental impacts, and time to market.
facturing processes (Dressman and Kramer 2005). However, the The following methodology was adopted: after identifying
test consumes much time, material, equipment, and human drug product dissolution as a Critical Quality Attribute (CQA),
resources. The present work evaluates the possibility of modeling quality risk assessment ascertained potentially critical material and
DT for solid oral dosage products based on QbD principles and process parameters that may impact dissolution. Experiments
proposes a novel approach to ensure conformity with ICH were designed to further identify crucial process parameters and
guidelines. elucidate their impact on product dissolution profiles.

CONTACT Nicolas Abatzoglou Nicolas.Abatzoglou@USherbrooke.ca Department of Chemical & Biotechnological Engineering, Universit
e de Sherbrooke, 2500
boul. Universit
e, Sherbrooke, Quebec, J1K2R1, Canada
ß 2017 Informa UK Limited, trading as Taylor & Francis Group
 2 K. YEKPE ET AL.

2. Experimental methods According to the FMEA, parameters that impact dissolution

testing most are presented in Table 1 along with a justification of
2.1. Risk assessment
each RPN value.
Risk assessment is a valuable science-based process in quality risk 1. Tablet compression
management that can help identify which material attributes and 2. Coating
process parameters could potentially have an effect on product 3. Particle size of API
CQAs (ICH Q8, R2 2009; ICH Q9 2011a). A comprehensive risk 4. Ibuprofen properties (ibuprofen physical–chemical properties
assessment was undertaken to identify process parameters whose and quantity)
variability may influence DT of conventional, immediate-release 5. Storage humidity
tablets containing ibuprofen, a Biopharmaceutical Classification The first three parameters were expected to have the most
System (BCS) Class II API. Well-recognized risk assessment tools – important impact on tablet dissolution. Before carrying out a com-
Fishbone diagram and failure mode effects analysis – were used plete DoE, a preliminary study was performed on the other two
to classify process parameters and then quantify the associated parameters (ibuprofen properties and humidity) to determine their
degree of risk. Since DT is influenced by numerous variables, these impact on dissolution.
steps were instrumental in identifying the critical DT parameters. Wet granulation is a step of the product manufacturing and
Before carrying out a Design of Experiments (DoE), the parame- could eventually have an impact on dissolution. Moreover, it
ters that were initially thought to influence DT have been classi- should be noted that the product under study was already devel-
fied in five categories:
Downloaded by [Gothenburg University Library] at 02:52 03 November 2017

oped and properly validated using standard pharmaceutical

1. Tablet manufacturing and design of the drug product.
guidelines. Thus, the wet granulation process of the product
2. The properties of the materials (excipients) used to manufac-
under study is robust.
ture the final drug product.
Moreover, the wet granulation process of the product under
3. The properties of the drug product API: ibuprofen.
study was systematically controlled by means of:
4. The type/role of the excipients used in the drug product.
1. LOD measurement for measuring powder wetness;
5. The conditions under which the dissolution test is run.
2. Powder yield, which is measured after granulation. The pow-
Based on the above parameter classification, a Fishbone dia-
der yield is also an indicator of powder’s wetness. An
gram was developed (Figure 1).
Failure Mode and Effects Analysis (FMEA) was then applied to adequately granulated powder will result in a powder yield
identify the critical parameters that impact dissolution testing. The within specifications.
latter are used to build a DoE and establish a control strategy for Thus, thanks to these several stage gates, the risk of wet
the product under study. Each parameter presented in the granulation is mitigated.
Fishbone diagram was ranked based on its Occurrence (O), its
Impact (I), and its Detectability (D). A Risk Priority Number (RPN) 2.1.1. Ibuprofen properties
was the calculated according to: RPN ¼ (O)  (I)  (D). The hypothesis that tablet dissolution can be influenced by the
RPN was used to separate critical parameters from less critical physical properties and quantity of ibuprofen was tested in this
ones for the present system. Parameters that are characterized by study. Ibuprofen bought from three different suppliers was eval-
the highest RPN values were considered critical. These parameters uated: suppliers A, B, and C. Three properties – reported as factors
would strongly impact DT and would have to be first tested using in this study – were chosen: particles size distribution (more spe-
an appropriate DoE. cifically mean particle diameter), particle shape, and crystal form.

Figure 1. Fishbone Diagram of the factors thought to influence product CQAs.

 PHARMACEUTICAL DEVELOPMENT AND TECHNOLOGY 3

Table 1. Failure Mode and Effects Analysis (FMEA) of parameters impacting the
dissolution tests – quantifying their Occurrence (O), Impact (I), and Detectability
(D).
Parameters I O D RPN Comments
Tablet 5 2 1 10 Tablets with inadequate hardness
compression will present unacceptable dis-
integration/dissolution profiles.
While its impact is important,
its presence is easily detect-
able and corrected by adjust-
ing compression setting.
Coating 5 2 1 10 Tablet coating is critical and
could have a very high impact
on dissolution profiles.
Inhomogeneous and/or under-
coated tablets can often easily
be detected.
API particle size 5 1 2 10 Particle size has a very important Figure 2. Ibuprofen PSD obtained from three suppliers: A – blue, B – red, and
impact on dissolution testing C – green.
as stated by the
Noyes–Whitney equation.
Downloaded by [Gothenburg University Library] at 02:52 03 November 2017

Ibuprofen physical– 5 1 2 10 Ibuprofen Physical–Chemical be rejected which indicates that the average particle sizes for the
chemical properties Properties have a significant three powders are statistically equal.
effect on dissolution. As this
study is for an existing prod-
uct, these parameters are not Factor 3: crystal identification. The third factor of this initial ibu-
included into the DoE. profen study was crystal identification with X-Ray Diffraction
Ibuprofen 5 2 1 10 Drug concentration has an (XRD). The purpose of this part of this article was to determine if
concentration important impact on dissol- the crystalline structure of the three ibuprofen powders is the
ution testing as stated by the
Noyes–Whitney equation.
same. As ibuprofen structure is crystalline, it was expected to
However, the ability to detect obtain sharp narrow diffraction peaks for the three powders.
a problem of drug concentra- The crystal identification was realized with an X’pert Pro MRD
tion is easily to detect during diffractometer from PANanalytical. The results are presented in
routine QC assay testing.
Figure 4. Peak position and intensity for the three ibuprofen pow-
Storage humidity 5 2 1 10 Storage Humidity has a signifi-
cant effect on physical and ders were compared using the internal library of the apparatus;
chemical properties of exci- the comparison shows that the three powders are identical.
pients and API contained in However, peak intensity for the three powders presents slight dif-
tablet. Thus, storage Humidity ference for some peak positions. This difference is explained by a
could impact API dissolution.
phenomenon called preferred orientation, defined as the non-ran-
dom distribution of the crystallites within a sample, and is the
likely cause of intensity variations in XRD powder diffractograms.
Thus, no polymorph of ibuprofen and no amorphous phase have
In the following sections, ibuprofen powders from supplier A, B, C been found on XRD spectra; consequently, the crystalline structure
will be referred to accordingly. of the three ibuprofen samples is the same.

Factor 1: particle size distribution. Particle size (PS) and particle Factor 4: Ibuprofen concentration. According to the FMEA, the
size distribution (PSD) are important variables in pharmaceutical impact of API concentration on dissolution testing is high (Smith
industry. Indeed, API and excipients PSD can affect final formula- 2015). It was decided to test tablets containing 90, 100, and 110%
tion (dissolution, appearance, and stability). Ibuprofen A, B, and C of ibuprofen claim in the tablets. Batches of 90 and 110% ibupro-
PS were measured via a light scattering method using a fen have the same formulation as regular ibuprofen tablets, but
Mastersizer 2000, Malvern instrument. The results are presented in excipient percentage within each formulation was modified in
Figure 2. They illustrate that the volumetric PS of all three pow- order to reach either 90 or 110% of API claim. Dissolution profiles
ders is unimodal, similar in shape with a relatively low proportion of 90% and 100% ibuprofen as well as those of 110 and 100%
of fines (mean particle size below 10 microns). ibuprofen were compared in order to estimate if the dissolution
profiles of the two formulated batches are similar.
Factor 2: particle shape. The shape and surface morphology of The dissolution testing was performed on six tablets of each
the particles were examined with a FEG-SEM S-4700, Hitachi batch, using USP apparatus 2 (Distek Inc., North Brunswick, NJ) in
instrument. The micrograph of the three powders (Figure 3) indi- 900 ml of dissolution media (phosphate buffer at pH 7.2) at
cates that particles for the three suppliers have similar shapes. 37 ± 0.5  C with UV detector as required in USP at 221 nm. Section
Moreover, the typical range of PS is confirmed: a wide range of 2.5 contains technical details concerning the dissolution test.
mean particle sizes (from 12 to 188 microns) can be seen for all Each dissolution profile was compared with dissolution profiles
three powders. Eight images were acquired for each of the three of 100% API Tablets. For all three cases, dissolution profiles show
powders. For each image, 10 particles were measured: ibuprofen that more than 85% of the drug is dissolved within 15 min, which
A, B and C present average particle sizes of 88.1(12.3), 82.4(8.7), indicates that the products dissolve rapidly. Indeed, in 6 min more
and 83.4(18.5) mm, respectively (values in parenthesis represent than 85% of ibuprofen is dissolved for all three batches.
standard deviation). An ANOVA was done to test the null hypoth- Coefficients of variation for the three batches were not more than
esis that the average particle sizes for the three powders are stat- 1% between 10 and 60 min indicating good data repeatability.
istically equal. With a p values of 0.68, the null hypothesis cannot USP dissolution testing specifications require that at least 85% of
 4 K. YEKPE ET AL.
Downloaded by [Gothenburg University Library] at 02:52 03 November 2017

Figure 3. SEM images obtained from 3 ibuprofen suppliers.

the labeled API amount should be dissolved within 60 min at pH

7.2.
Low coefficient of variation implies that API concentration does
not impact dissolution. As a consequence, all batches met USP
specification.

2.1.2. Stability study

According to ICH Q1 stability testing provides evidence on “how
the quality of a drug substance or drug product varies over time
under the influence of a variety of environmental factors such as
temperature, humidity, and light ( … )” (ICH Q1A (R2) 2003).
In order to evaluate the impact of humidity on dissolution, ibu-
profen tablet stability data of five batches were analyzed and
compared to ibuprofen tablets dissolution testing USP specifica-
tions. Stability storage conditions for these tablets are 25  C and
60% relative humidity.
Figure 5 shows stability results for dissolution testing to stor-
age conditions 25  C/60%. The USP specifications for dissolution
testing results are: Q (amount of dissolved API) should not be
lower than 85% in 60 min. Specifications are met: the dissolution
percentage for all tested batches is above this limit. Moreover, no
trend is observed for the results over 36 months of storage. Based
on one-way ANOVA, as p values is higher than the significance
level (p values ¼0.521), all means at 60 min are equal at a signifi-
Figure 4. XRD spectra for ibuprofen A, B, and C. cance level of 0.05. Thus, these results prove that these stability
conditions have no influence on ibuprofen dissolution.
 PHARMACEUTICAL DEVELOPMENT AND TECHNOLOGY 5

Table 3. List of samples.

Hardness Coating weight
Group Batch (% of the nominal) (% of the nominal) API PS
1 1 100 86 Nominal
2 100 86 Nominal
3 100 86 Nominal
4 70 75 Nominal
5 70 100 Nominal
6 140 75 Nominal
7 140 100 Nominal
2 8 70 75 Reduced
9 70 100 Reduced
10 100 86 Reduced
11 100 86 Reduced
12 140 75 Reduced
13 140 100 Reduced
Figure 5. Stability dissolution testing data of ibuprofen tablets at 25  C and 60%
RH.

Table 2. Factors and levels of the DoE.

Downloaded by [Gothenburg University Library] at 02:52 03 November 2017

Tablet Hardness Coating weight

(% of the nominal) (% of the nominal) API PS (lm)
Low level 70% 75% 90
Central Point 100% 86% N/A
High level 140% 100% 0.40-190
Comments:
Nominal PS is 0.40–190 lm
Reduced PS is 90 lm

2.1.3. Conclusion on preliminary study

This preliminary study showed that:
 The ibuprofen from different suppliers exhibited the same
physical properties (based on suppliers A, B, and C).
Consequently, it is expected they do not affect ibuprofen
tablet dissolution results; Figure 6. Dissolution profiles of all 13 batches at medium pH 6.4, 6.8, and 7.2.
Red line represents USP dissolution specifications.
 Ibuprofen concentration (90 and 110% API) has no influ-
ence on ibuprofen tablet dissolution results;
 Storage humidity (conditions: 25  C/60%) has no influence the DoE are 75 and 100% with an intermediate point at 86%.
on ibuprofen tablet dissolution results. Coating percentage of 75, 86, and 100% indicate that batches
were coated respectively with 75, 86, and 100% of actual coating
2.2. Design of experiments solution quantity resulting in different coating thicknesses.
More over a lower coating thickness would affect the product
The risk assessment identified API PS, tablet hardness and coating appearance as the coating would appear not homogeneous; such
thickness as parameters that may significantly affect the dissol- a defect would be easily detected by In-Process Checks.
ution of the product under investigation. To evaluate the effects Experimental design and analysis were conducted with Design
of these variables, a DoE was carried out. As the study must be as Expert version 8.0.6 (Stat-ease Inc., Minneapolis, MN). Table 2 sum-
representative as possible of current manufacturing processes, we
marizes the adopted DoE. As the ibuprofen used in this study is
selected factor levels mirroring specifications for API particle size
ionic and its solubility depends on pH, we performed the DoE at
(PS), tablet hardness, and tablet coating thickness (or tablet coat-
different pH levels.
ing weight).
The experimental design space consisted of 13 experiments
(two factors at three levels and one factor at two levels). The fac- 2.3. Tablet manufacture
tors at three levels were: tablet hardness (70, 100, and 140% of
the nominal value) and tablet coating weight (75, 86, and 100% Small amounts of ibuprofen powder were withdrawn from com-
w/w of the nominal value) and the factor at two levels was API mercial batches after wet granulation, drying, sieving, and final
mean particle size (0.40–190 lm and lower than or equal to blending steps to produce the samples belonging to Group 1
90 lm). Our experiments were designed to evaluate the effects of (Table 3). Tablet compression and coating steps were carried out
these three factors on responses, i.e. ibuprofen dissolution percen- after the final blending step.
tages over time. To produce Group 2 samples containing smaller API particles,
Coating percentage of 100% represents the nominal coating the API commercial powder: 0.40–190 lm (“Nominal” in Table 3)
according to the manufacturing instructions. The nominal value is was separated with a mesh sieve in order to obtain API particles
the current coating value validated for this product. This value sizes lower than or equal to 90 lm (“Reduced” in Table 3). After
was selected as the high level in the DoE since it would not likely mixing the API with other formulation ingredients, wet granula-
be possible to add additional coating solution as per manufactur- tion, drying, and sieving steps were carried out. After the sieving
ing instructions. Thus, variables for coating percentage level for step, the powder was compressed at the typical tablet hardness
 6 K. YEKPE ET AL.

of commercial product family. The tablets obtained after this step Table 4. Dissolution models under study.
have been coated. Model Equation
Zero-order (Costa and Sousa Lobo 2001) %disso ¼ kt (1)
First-order (Polli et al. 1997) %disso ¼ 1  ekt (2)
2.4. Tablet properties Higuchi (1963) %disso ¼ kt1=2 (3)
Korsmeyer–Peppas (Korsmeyer et al. 1983) %disso ¼ ktn (4)
Tablet properties (weight uniformity, length, thickness, and hard- 3
Hixson–Crowell (Hixson and Crowell 1931) %disso ¼ ð1  ktÞ b (5)
ness) were measured in randomly selected tablets from each Weibull (Dokoumetzidis et al. 2006) %disso ¼ 1  ðe ðatÞ
) (6)
batch. Five (5) tablets were tested for weight uniformity, length,
thickness, and hardness. Tablet weight uniformity was established
by AT460 Delta Range electronic balance (Mettler Toledo,
Columbus, OH). Length and thickness were quantified with a CD-
6" CX digital caliper (Mitutoyo, Aurora, IL). Hardness was analyzed
with a Type HT500 tablet hardness tester (Key International, Inc.,
New-York, NY). Relative standard deviation (RSD) values were cal-
culated for each batch.

2.5. DT of manufactured tablets

Downloaded by [Gothenburg University Library] at 02:52 03 November 2017

Dissolution studies were performed using a Distek 2100 A USP dis-

solution apparatus 2 (Distek Inc., North Brunswick, NJ). Six tablets
from each batch were placed separately, each in 900 ml of dissol-
ution medium (potassium phosphate buffer at pH 6.4, 6.8, or 7.2:
0.05 M). Temperature was fixed at 37.0 ± 0.5  C and paddle stirring
was set at 50 rpm. Samples were analyzed by a continuous flow Figure 7. Coefficients of determination (R2) of kinetic models in all 13 batches at
system attached to an Agilent 8453 UV/VIS spectrophotometer pH 6.4, 6.8, and 7.2.
(Agilent Technologies, Inc., Wilmington, NC). Wavelength was set
as recommended by the USP monograph on API, which is 221 nm.
Absorbance was monitored every 2 min over a 60 min period
(totalizing of 30 times points). Data were acquired by Chemstation USP dissolution testing specifications require that at least 85%
software (Agilent Technologies, Inc., Wilmington, NJ). Values were of the labeled API amount should be dissolved within 60 min at
compared against a calibration curve, and percentages of dis- pH 7.2 (the red threshold limit bar in Figure 6 represents USP
solved drug were plotted for time profiles. According to USP, spe- specifications). Every tested tablet, from every batch, met USP
cification results must be higher than or equal to 85% of API specifications. Thus, despite different medium pH values and
dissolution at 60 min (FDA 1997). Dissolution percentages over changes in tablet properties, USP specifications were still met
time served as response variables for DoE analysis (13 DoE runs at within 60 min. As indicated in Figure 6, dissolution results for
three different pH values). batches tested at pH 7.2 were slightly above 100%. The observed
drug level variability was expected and acceptable. It may be a
reflection of DT variability (i.e. systematic errors) or analytical
2.6. Statistical analysis artifacts.
Three variables were studied: API PS (A), tablet hardness (B), and
coating weight (C). Analysis Of Variance (ANOVA) was used to 3.3. Impact of dissolution medium pH on API release kinetics
understand relationships between variables (i.e. A, B, and C) and
dissolution percentages over test time. API release kinetics were studied. The data were compared
through the applicability of models available in the literature:
zero-order, first-order, Higuchi, Korsmeyer–Peppas, Hixson–Crowell,
3. Results and discussion and Weibull equations. These models are considered as the most
relevant and commonly used for dissolution time profiles and are
3.1. Tablet properties
presented in Table 4 (Costa and Sousa Lobo 2001). Figure 7
Mean RSD values for all batches were 0.77, 0.16, 0.07, and 6.96% presents average R2 values for the 13 batches at the pH 6.4, 6.8,
for tablet weight uniformity, thickness, length, and hardness, and 7.2 for all kinetics models.
respectively. Thus, the results varied insignificantly for all tablet The Weibull and first order models give the highest R2 values;
properties studied, confirming good process reproducibility. thus, they were chosen for further study. In order to select the
most appropriate prediction model, the values of a and b parame-
ters of Weibull model and the constant k of the first-order equa-
3.2. Dissolution profile in medium pH study
tion were studied. Table 5 presents the means and standard
As described earlier, DoE established how design factors – API PS, deviations associated with these values for the 13 lots to pH 6.4,
tablet hardness, and coating weight – affected responses (API dis- 6.8, and 7.2. It shows that the standard deviations of a and b
solution percentage over time with test medium pH). Figure 6 parameters are very high for the pH 7.2 tests. Statistical analysis
shows the dissolution profiles of all 13 DoE runs at three different using the Student t-test (a ¼ 0.05) indicates that the parameter k
pH values. Dissolution at each time point represented an average of the first order model is highly significant in all cases, while the
of six tablets. For each time point, the dissolution RSD was less parameters a and b of the Weibull model at pH 7.2 are not.
than 5% and, therefore, for clarity, the error bars are not shown in Moreover, the Weibull equation overfits individual profiles.
Figure 6. It can be observed that dissolution varied with pH Consequently, the use of the first order model is considered more
values. appropriate. The computed k values for this model applied to all
 PHARMACEUTICAL DEVELOPMENT AND TECHNOLOGY 7

Table 5. Means and standard deviations associated with k, a, and b values for the 13 lots to pH 6.4. 6.8, and 7.2.
k Constant (1st order) a Parameter (Weibull) b Parameter (Weibull)
1 1
pH value Average (min ) Standard-deviation (min ) Average (min) Standard-deviation (min) Average (adimensional) Standard-deviation (adimensional)
6.4 0.13 0.03 1.33 0.27 0.30 0.05
6.8 0.22 0.05 0.78 0.13 0.20 0.04
7.2 0.40 0.12 16.58 34.88 1.51 0.95

Table 6. Significant models in ANOVA with a ¼ 0.05.

pH 6.4 pH 6.8 pH 7.2
Between 6–12 min No Yes Yes
Between 14–60 min No No No

Table 7. ANOVA at pH 6.8 and 7.2.

pH 6.8 pH 7.2
Source F value Probability (p-value) F value Probability (p-value)
Downloaded by [Gothenburg University Library] at 02:52 03 November 2017

Model 6.84 0.0251 6.54 0.0276

A 33.14 0.0022 35.90 0.0019
B 7.54 0.0405 7.20 0.0437
C 0.26 0.6326 0.35 0.5801
AB 0.044 0.8418 0.70 0.4410
AC 2.73 0.1597 0.20 0.6765
BC 3.10 0.1387 0.013 0.9142
Figure 8. First-order dissolution rate constant (k) values for all 13 batches at pH ABC 1.08 0.3466 1.39 0.2909
6.4, 6.8, and 7.2. Lack of fit 0.35 0.7329 1.30 0.3920

batches are presented in Figure 8. It is shown that, as expected,

that k values vary with batch and increase with pH. As expected,
drug release performance is improved in more alkaline media.
Table 8. Selected limits for this modified control strategy.
Factors Tablet hardness Coating weight API PS
3.4. Statistical analysis
limit (% of the nominal) (% of the nominal) (lm)
Design factor effects on dissolution were studied quantitatively by Low threshold value 70% 75% 90
ANOVA. According to Table 6, with 13 runs, ANOVA indicates that High threshold value 140% 100% 0.40–190
models were significant for dissolution results with testing time
between 6 and 12 min at pH 6.8 and 7.2. Between 14 and 60 min,
none of the models and none of the effects were proven signifi- Y ¼ ð100:31 6 0:49Þ–ð2:28 6 0:38ÞAPI PS þ ð1:30 6 0:48Þ
cant (p < 0.05). Indeed, API particle size, tablet hardness, and coat- (8)
Tablet Hardness for pH 7:2
ing weight showed no significant effects on DT at 60 min.
At the end of steady state of the dissolution process, no critical
process variables impacted the dissolution results. This suggests Good correlations were obtained between observed and predicted
that these variables do not contribute to API dissolution anymore. values, as indicated by the high R2 values with both models: 0.905
At 10 min, modeling is possible only at pH 6.8 and 7.2. ANOVA is and 0.902 for pH 6.8 and 7.2, respectively.
presented in Table 7 for pH 6.8 and 7.2, respectively. Based on
the ANOVA results, two main effects were significant at pH 6.8 3.5. Discussion
and 7.2 (p < 0.05): API PS and tablet hardness.
These variables have a greater impact on API dissolution at an Risk analysis is an important step in QbD strategy application. For
earlier dissolution time than at the end of the dissolution process. this reason, in the first part of our study, risk analysis was per-
This phenomenon can be explained by the fact that dissolution is formed and three factors were identified as potential impact fac-
dependent on disintegration, which is mostly influenced by tablet tors. They included API PS, tablet hardness, and coating weight.
hardness. DoE involves relatively few runs (13 runs) and can estimate the
Thus, at earlier time points, tablets are still disintegrating and significance of main effects with high efficiency and accuracy. As
API particles are being released from them. The rate of API dissol- mentioned, at USP specifications of 60 min, no parameters were
ution, therefore, depends on API PS. Moreover, during the dissol- found to statistically affect dissolution results of the tablets, under
ution process, a visual inspection proved that coating was the ranges studied.
systematically removed in less than 1 min from tablets of all Two factors proved to be statistically significant, namely, API
batches, confirming qualitatively that this parameter has no sig- PS and tablet hardness. Of these factors, low API PS had a positive
nificant impact on dissolution results. effect on DT at earlier time points, emphasizing that small API par-
The final equations in terms of coded factors were determined. ticle size enhanced dissolution performance because of increased
The statistical models were: surface/volume ratio (Noyes and Whitney 1897).
First order models successfully described the dissolution rate,
Y ¼ ð86:95 6 0:49Þð2:82 6 0:49ÞAPI PS þ ð1:71 6 0:62Þ but further validation is required to evaluate their general predict-
(7)
Tablet Hardness for pH 6:7 ive capability.
 8 K. YEKPE ET AL.

Regarding ICH Q8/Q9/Q10 (R2) which explains the role of mod- between formulation and manufacturing process variables and
els in QbD (ICH Q8/Q9/Q10 (R2) 2011c), the models presented in BCS Class II API DT to improve our fundamental knowledge of dis-
this article are categorized as high-impact models. solution testing (DT).
An interesting finding is that the last tested factor, coating Risk analysis, identification of critical formulation/process varia-
weight, had no effect on drug dissolution. This can be explained bles, understanding the effects of these critical variables and inter-
by the fact that the tablets under study were film-coated and actions on key product quality attributes are essential steps to
principally composed of very soluble excipients that dissolved rap- achieve enhanced product and process awareness and offer
idly upon wetting, thus not interfering in tablet disintegration or opportunities to develop control strategies while ensuring final
API dissolution. product quality. Indeed, as DT results at 60 min were all similar, it
According to QbD principles, a specific method is required to was not possible to model dissolution with operating conditions
specify the design space (ICH Q8, R2 2009). Design space gener- specified by USP. Preliminary results have shown that modeling at
ation begins with CQA identification, risk assessment, DoE, model DT time of 10 min can be performed allowing us to understand
building, and design space visualization (Yu 2008; Little, 2013). and predict DT. In these situations, opportunities exist to develop
According to ICH Q8 (R2), a design space can be described in
more flexible regulatory approaches, for example, to facilitate real-
terms of ranges of material attributes and process parameters, or
time quality control, leading to reduction of end-product release
through more complex mathematical relationships (ICH Q8, R2
testing.
2009).
This case study exemplified the application of QbD principles
Thus, the relationship between the process inputs (material
Downloaded by [Gothenburg University Library] at 02:52 03 November 2017

and tools to drug product and process development. It demon-

attributes and process parameters) and the critical quality attrib-
strated that DoE effects/response surface analysis represents a
utes can be described thanks to the design space. Inside the
design space, product quality is assured. powerful tool to investigate the effects of selected factors (API PS,
The design space is proposed by the applicant and is subject tablet hardness, and coating weight) on response–dissolution per-
to regulatory assessment and approval. Therefore, in order to be centages, showing product and process robustness. The following
approved, the model underlying the design space requires fulfill- points comprise the main new information brought forward by
ing specifications. our study:
The design space is this article can be explained mathematic-  The two parameters impacting dissolution kinetics at
ally through the equation describing relationships between quality 10 min are API PS and tablet hardness. They can easily be
attributes and process parameters and/or Material Attributes: controlled during and after the manufacturing process
Y(Dissolution percentage at a specific time) ¼ F(API PS and tablet with standard equipment such as an online laser-granul-
hardness). ometer and hardness tester, this information is highly use-
Moreover, according to the results of the current study, the ful for formulation improvement.
edge of failure of the design space is beyond the selected limits  Models can be used to predict DT if appropriate validation
for this modified control strategy. The selected limits for this was to be performed.
modified control strategy are presented in Table 8.  This QbD approach demonstrates product and process
Thus, the edge of failure of the design space was beyond the robustness.
limits selected. As stated in ICH Q8, R2 and confirmed in The QbD concept can completely eliminate the need of dissol-
“Questions and Answers on Design Space Verification” published ution testing or eventually either decreases required dissolution
on 24 October 2013 (FDA and EMEA 2013), the determination or testing time as the finale dissolution percentage can be predicted
demonstration of edge of failure is not an essential part of the knowing the first minute dissolution percentage the first minutes’
definition of design space; the edge of failure of this design space dissolution percentage. Thus, dissolution test can be reduced, or
was found to be outside the selected parameters studied. even eliminated, if confirmed by a validation procedure.
Moreover, our study proves that the examined manufacturing pro- In conclusion, the methodology presented in this article (FMEA,
cess was robust. DoE, and Design space definition) could be replicated and per-
For the formulation under study, which has a relatively simple formed in other dosage forms like controlled release tablets, emul-
dissolution behavior, the first order model worked satisfactorily. sions, suspensions, etc.
However, more complex formulation where drug release depends
on amount of other ingredients like controlled-release polymers
and pH modifiers might be determined by other models (Costa Disclosure statement
and Sousa Lobo 2001). In this case, the advantages and disadvan-
tages of model dependent and model independent methods The author declares that she has no relevant or material financial
should be examined to predict dissolution. Once the ability to interests that relate to the research described in this paper.
predict dissolution with a model is well demonstrated, meeting
the remaining requirement for implantation in pharmaceutical
control quality strategy like method validation based on current References
pharmaceutical guidelines will be necessary.
Basalious EB, El-Sebaie W, El-Gazayerly O. 2011. Application of
Pharmaceutical QbD for enhancement of the solubility and dis-
4. Conclusion solution of a Class II BCS drug using Polymeric Surfactants and
QbD tools made it possible to identify critical parameters, trans- Crystallization Inhibitors: development of controlled-release tab-
late them to attributes that the drug product should possess, and lets. AAPS PharmSciTech. 12:799–810.
establish how critical process parameters can be varied to consist- Buri P. 1983. Mesure de la vitesse de dissolution pour
ently produce drugs with desired characteristics. QbD principles l'am
elioration et le contro ^le de la qualit

e des m
edicaments
and tools presented in ICH Q8, ICH Q9, and ICH Q10 guidelines [Measurement of dissolution rate for drug quality improvement
were considered in the current case study to explore relationships and control]. J Pharm Biomed Anal. 1:393–402.
 PHARMACEUTICAL DEVELOPMENT AND TECHNOLOGY 9

Costa P, Sousa Lobo JM. 2001. Modeling and comparison of dissol- Use. 2011b. Pharmaceutical Quality System, Q10. Rockville, MD:
ution profiles. Eur J Pharm Sci. 13:123–133. Guidance for Industry.
Cui Y, Song X, Reynolds M, Chuang K, Xie M. 2012. International Conference on Harmonisation Quality
Interdependence of drug substance physical properties and Implementation Working Group Points To Consider (R2). 2011c.
corresponding quality control strategy. J Pharm Sci. 101: International Conference on Harmonisation – Endorsed Guide
312–321. for ICH Q8/Q9/Q10 Implementation. Rockville, MD: Guidance for
Dokoumetzidis A, Papadopoulou V, Macheras P. 2006. Analysis Industry.
of dissolution data using modified versions of Noyes–Whitney Juenemann D, Jantratid E, Wagner C, Reppas C, Vertzoni M,
equation and the Weibull function. Pharm Res. 23:256–261. Dressman JB. 2011. Biorelevant in vitro dissolution testing of
Dressman J, Kramer J. (2005). Pharmaceutical Dissolution testing products containing micronized or nanosized fenofibrate with a
(Vol. 140 of Drugs and Pharmaceutical Sciences Series). New view to predicting plasma profiles. Eur J Pharm Biopharm.
York: Taylor and Francis. 77:257–264.
Dumarey M, Goodwin DJ, Davison C. 2015. Multivariate modelling Juran JM. (1992). Juran on quality by design: the new steps for
to study the effect of the manufacturing process on the com- planning quality into goods and services. New York (NY): The
plete tablet dissolution profile. Int J Pharm. 486:112–120. Free Press.
FDA. 1997. Dissolution Testing of Immediate Release Solid Oral Korsmeyer RW, Gurny R, Doelker E, Buri P, Peppas NA. 1983.
Dosage Forms [Internet]. Rockville, MD: Guidance for industry. Mechanisms of potassium chloride release from compressed,
[cited 216 Dec 31]. Available from http://www.fda.gov/down-
Downloaded by [Gothenburg University Library] at 02:52 03 November 2017

hydrophilic, polymeric matrices: effect of entrapped air. J Pharm

loads/Drugs/GuidanceComplianceRegulatoryInformation/
Sci. 72:1189–1191.
Guidances/ucm070237.pdf
Le Hir A, Chaumeil J-C, Brossard D. (2009). Pharmacie Gal
enique:
FDA and EMEA. 2013. Questions and Answers on Design Space
Bonnes pratiques de fabrication des m
edicaments. Paris:
Verification [Internet]. [cited 216 Dec 31]. Available from: http://
Elsevier Masson.
www.ema.europa.eu/docs/en_GB/document_library/Other/2013/
Little T. 2013. Essentials in establishing and using design space.
11/WC500153784.pdf
BioPharm Int. 26:50–54. Available from http://www.biopharmin-
Hernandez E, Pawar P, Keyvan G, Wang Y, Velez N, Callegari G,
~ach RJ. 2016. ternational.com/essentials-establishing-and-using-design-space
Cuitino A, Michniak-Kohn B, Muzzio FJ, Roman
Noyes AA, Whitney WR. 1897. The rate of solution of solid sub-
Prediction of dissolution profiles by non-destructive near infra-
stances in their own solutions. J Am Chem Soc. 19:930–934.
red spectroscopy in tablets subjected to 148 different levels of
Polli JE, Rekhi GS, Augsburger LL, Shah VP. 1997. Methods to com-
strain. J Pharm Biomed Anal. 117:568–565.
Higuchi, T. 1963. Mechanism of sustained-action medication theor- pare dissolution profiles and a rationale for wide dissolution
etical analysis of rate of release of solid drugs dispersed in solid specifications for metoprolol tartrate tablets. J Pharm Sci.
matrices. J Pharm Sci. 52:1145–1149. 86:690–700.
Hixson AW, Crowell JH. 1931. Dependence of reaction velocity Rathore A, Saleki-Gerhardt A, Montgomery S, Tyler S. 2009. Quality
upon surface and agitation. Ind Eng Chem. 23:923–931. by design: industrial case studies on defining and implementing
International Conference on Harmonisation of Technical design space for pharmaceutical processes part 2. Biopharm Int.
Requirements for Registration of Pharmaceuticals for Human 22:41–47.
Use. 2003 Stability Testing of New Drug Substance and Smith J. 2015. Research Works. Remington Education. Phys Pharm.
Products Q1A(R2). Rockville, MD: Guidance for Industry. 86:31–50.
International Conference on Harmonisation of Technical Yekpe K, Abatzoglou N, Bataille B, Gosselin R, Sharkawi T, Simard
Requirements for Registration of Pharmaceuticals for Human J-S, Cournoyer A. 2015. Predicting the dissolution behavior of
Use. 2009. Pharmaceutical Development Q8 (R2). Rockville, MD: pharmaceutical tablets with NIR chemical imaging. Int J Pharm.
Guidance for Industry. 486:242–251.
International Conference on Harmonisation of Technical Yu LX. 2008. Pharmaceutical quality by design: product and pro-
Requirements for Registration of Pharmaceuticals for Human cess development, understanding, and control. Pharm Res.
Use. 2011a. Quality Risk Management, Q9. Rockville, MD: 25:781–791.
Guidance for Industry. Yu LX, Amidon G, Khan MA, Hoag SW, Polli J, Raju GK,
International Conference on Harmonisation of Technical Woodcock J. 2014. Understanding pharmaceutical quality by
Requirements for Registration of Pharmaceuticals for Human design. AAPS J. 16:771–783.