Upload
katsuhide
View
223
Download
6
Embed Size (px)
Citation preview
Cgm
Ta
b
c
a
ARR1AA
KADHMS
1
pvttfbtb(tpsi
T
0h
International Journal of Pharmaceutics 461 (2014) 495– 504
Contents lists available at ScienceDirect
International Journal of Pharmaceutics
j o ur nal ho me page: www.elsev ier .com/ locate / i jpharm
larifying the mechanism of aggregation of particles in high-shearranulation based on their surface properties by usingicro-spectroscopy
akeo Kanoa,b,∗, Yasuo Yoshihashib, Etsuo Yonemochic, Katsuhide Teradab
Chugai Pharmaceutical Co., Ltd., 5-5-1 Ukima, Kita-ku, Tokyo 115-8543, JapanFaculty of Pharmaceutical Sciences, Toho University, 2-2-1 Miyama, Funabashi, Chiba 274-8510, JapanInstitute of Medicinal Chemistry, Hoshi University, 2-4-41 Ebara, Shinagawa, Tokyo 148-8501, Japan
r t i c l e i n f o
rticle history:eceived 21 June 2013eceived in revised form1 November 2013ccepted 14 December 2013vailable online 22 December 2013
eywords:ggregationistribution map
a b s t r a c t
The present study aimed to clarify, by means of micro-spectroscopy, the mechanism of aggregation ofparticles into granules during high-shear granulation. We used two types of pharmaceutical granulesprepared by high-shear granulator, one containing mefenamic acid and the other containing flavox-ate hydrochloride as poorly soluble active pharmaceutical ingredients (APIs). Lactose, cornstarch, andmicrocrystalline cellulose were used as excipients; and hydroxypropyl cellulose (HPC) was used as thebinding agent. The distributions of components in granules were visualized by mapping cross-sections ofindividual granules with techniques utilizing mid-infrared spectroscopy at the SPring-8 synchrotronradiation facility and micro-Raman spectroscopy. In the distribution maps of mefenamic acid gran-ules, distributions of mefenamic acid, cornstarch, and microcrystalline cellulose overlapped; in flavoxate
igh-shear granulationicro-spectroscopy
urface free energy
hydrochloride granules, on the other hand, distributions of flavoxate hydrochloride and lactose over-lapped. Assessment of the surface free energy of each component found that ingredients with overlappingdistribution had similar surface properties. Therefore, it was revealed that in high-shear granulation, inaddition to the granulator operating conditions and general properties of the formulation itself (such asthe solubility and particle size of each ingredient), the surface properties of the ingredients and theirinterrelationships were also factors that determined the aggregation behavior of the particles.
. Introduction
In the pharmaceutical industry, granulation is a generally usedrocess in the manufacture of tablets and capsules. Among theariety of granulation methods, high-shear granulation is one ofhe most widely used. In high-shear granulation, powders are agi-ated by an impeller in a vessel and sprayed with a binder solutionrom the top, while, simultaneously, coarse particles are crushedy a chopper. As droplets of the binder solution are dispersed intohe powder, liquid bridges form between primary particles thatecome nuclei around which granule size subsequently increasesFaure et al., 2001). Granulation is carried out in order to con-rol dust of the active pharmaceutical ingredients (APIs) and to
repare granules of high flowability, which facilitates subsequentteps in the manufacturing process (e.g., tableting). Also, produc-ng homogeneous granules is an important objective in ensuring∗ Corresponding author at: Chugai Pharmaceutical Co., Ltd., 5-5-1 Ukima, Kita-ku,okyo 115-8543, Japan. Tel.: +81 3 3968 4806; fax: +81 3 3968 4810.
E-mail address: [email protected] (T. Kano).
378-5173/$ – see front matter © 2013 Elsevier B.V. All rights reserved.ttp://dx.doi.org/10.1016/j.ijpharm.2013.12.013
© 2013 Elsevier B.V. All rights reserved.
uniformity of content. Therefore, granulation is considered oneof the most important processes in the manufacturing process ofpharmaceutical products, and a lot of research has been directedat understanding the complex mechanisms of granule formation.Studies of granulation have been conducted under various manu-facturing conditions to examine the effects that process parametersand formulation variables (such as the rotational speeds of theimpeller and chopper, granulation time, proportion of water, pro-portions of feed ingredients, and viscosity of the binder solution)have on physical properties of the granules (e.g., particle size, parti-cle size distribution, structure, and strength) (Belohlav et al., 2007;Bouwman et al., 2005; Cavinato et al., 2011; Nguyen et al., 2010;Rahmanian et al., 2011; Smirani-Khayati et al., 2009; Tu et al., 2009).Previous research has shown that the formation of granules is con-trolled by a combination of 3 major mechanisms:
(1) Wetting and nucleation: droplets of binder solution are dis-
persed into powder and liquid bridges are formed betweenprimary particles; these form the nuclei of granules.(2) Consolidation and aggregation: nuclear particles coalesce witheach other, and particles enlarge and become stronger.
4 l of Pharmaceutics 461 (2014) 495– 504
(
dVNtErprpi(ite2
mgusdsL2tuFib(iata2
pioupmcttstomsag
2
2
flo(
Table 1Formulations of the two types of granules experiments.
Materials Percentage (%)
Mefenamic acidgranules
Flavoxatehydrochloride granules
Mefenamic acid 9.80 –Flavoxate hydrochloride – 9.80Lactose 52.90 52.90Cornstarch 26.40 26.40Microcrystalline cellulose 8.80 8.80Hydroxypropyl cellulose 2.10 2.10
96 T. Kano et al. / International Journa
3) Breakage and growth: particles are crushed by impeller bladesand chopper.
Many researchers have tried to explain these mechanisms inetail (Benali et al., 2009; Iveson et al., 2001a; Van den Dries andromans, 2004, 2009; Van den Dries et al., 2003; Vonk et al., 1997).ucleation occurs at the beginning of granulation, and aggrega-
ion and fragmentation is repeated as particles continue to grow.ventually, granules reach maximum strength and equilibrium iseached and further growth and breakage does not occur, at whichoint granulation is complete. Other studies have evaluated theelationships between the physical properties of the granules (e.g.,article size, particle size distribution, flowability, compressibil-
ty, and uniformity) and the properties of the APIs and excipientse.g., particle size, specific surface area, solubility, and wettabil-ty). For example, it has been found that highly soluble materialsend to grow into granules rapidly (Iveson et al., 2001b; Saleht al., 2005; Van den Dries and Vromans, 2002; Vemavarapu et al.,009).
The many studies mentioned above were carried out at theacroscopic level. However, it is also important to evaluate
ranules from a more microscopic perspective to more clearlynderstand the mechanisms of granulation (Le et al., 2011). Micro-copic imaging techniques are used in many studies, and it is amplyemonstrated that they are suitable for assessing extremely micro-copic regions in formulations of various types (Gendrin et al., 2007;i et al., 2008; Maurer and Leuenberger, 2009; Yonemochi et al.,008). For example, the distribution of the API and additives athe surface of tablets has been assessed by microscopic imagingsing NIR (near-infrared) spectroscopy (Amigo and Ravn, 2009;ranch-Lage et al., 2011), and the distribution of API and additivesn ointments produced by different methods has been evaluatedy using attenuated total reflection infrared (ATR-IR) spectroscopyYamamoto et al., 2012). An investigation using techniques combin-ng micro-Raman mapping, scanning electron microscopy (SEM),nd transmission electron microscopy (TEM) has been conductedo differentiate between the amorphous molecular level dispersionnd nano dispersions of API in a solid dispersion (Karavas et al.,007).
By using micro-spectroscopy, it is possible to assess the dis-ersion state of each component in individual granules, and this
nformation could be very useful for explaining the mechanismf granulation. In this study to describe the mechanism of gran-lation, the dispersion states of each component in granulesrepared using different granulation times were examined usingicro-spectroscopy. We confirmed that the dispersions of each
omponent in the granules changed with the progress of granula-ion. In addition, all ingredients except the binder were consideredo be almost insoluble under the granulation conditions in thistudy. In such cases, it is considered that it is the surface proper-ies of the powdered ingredients that most affects the distributionf each component in the granules during granulation. By usingicro-spectroscopic mapping techniques and assessment of the
urface properties of ingredients, we attempted to clarify the mech-nism of aggregation of components that occurs in high-shearranulation.
. Materials and methods
.1. Materials
Mefenamic acid (Sanchemipha Co., Ltd, Miyagi, Japan) andavoxate hydrochloride (Sanchemipha) were used as modelsf poorly water soluble APIs: solubility of mefenamic acid0.00174 mg/mL at 20 ◦C) (PMRJ, 2002a) was lower than that
Total 100.00 100.00
of flavoxate hydrochloride (17.2 mg/mL at 20 ◦C) (PMRJ, 2002b).Lactose (DMV-Fonterra Excipients GmbH & Co, Goch, Germany),cornstarch (Nihon Shokuhin Kako Co., Ltd., Inc., Tokyo, Japan),and microcrystalline cellulose (Asahi Kasei Chemicals Corporation,Tokyo, Japan) were used as excipients, and hydroxypropyl cellulose(HPC; Nippon Soda Co., Ltd., Tokyo, Japan) was used as the bindingagent.
2.2. Granulation conditions
Two types of granules were prepared, each with a different typeof API. The percentage of API in each type of granule was 9.8%,and the percentages of excipients were the same (Table 1). Gran-ules were prepared in a high-shear granulator (VG-25; PowrexInc., Hyogo, Japan) with impeller speed of 180 rpm and chop-per speed of 3000 rpm. The volume of the granulator was 25 Land the charge amount was set at approximately 7.5 kg. After2 min of premixing, a 7% (w/v) aqueous solution of HPC wasadded all at once. Granules were sampled from three fixed pointspositioned equally around the circumference of the powder bed,with the samples taken at specified times from the time thatthe HPC was added (10 s, 30 s, 1 min, 1.5 min, 2 min, 3 min, 4 min,5 min, 7 min, 10 min, 15 min), and then dried on a shelf dryer at80 ◦C.
2.3. Measurement of mean particle size
The average particle size of the granules was measured by sieveanalysis. Samples of about 100 g were sieved through 12-, 32-,and 48-mesh (1.40, 0.50, and 0.30 mm) sieves. The proportion ofsample retained on each sieve was used to obtain the mean par-ticle size, and granules of the greatest weight fraction remainingon the sieve at each granulation time were analyzed by micro-spectrophotometer.
2.4. Mid-infrared spectroscopy (MIR)
2.4.1. Determination of standard spectra by MIRThe spectrum of each ingredient was determined with the BL43
IR beamline of the SPring-8 synchrotron radiation facility (Hyogo,Japan) and a mercury cadmium tellurium (MCT) detector. Thespectra were collected in transmission mode and measured at awavenumber resolution of 4 cm−1 with 32 scans in the mid-infraredrange (4000–400 cm−1).
2.4.2. Sample preparation for MIRIn order to map cross-sections of the granules in transmission
mode, granules were embedded in acrylic resin (LR white resin;Okenshoji Co., Ltd, Tokyo, Japan) and then sliced to a thickness of5 �m using a microtome. At least three granules were evaluated foreach of the specified sampling times.
l of Ph
2
mtt2mum(tnpluasoPfmdpchaycZmictptu
2
2
RJptc
2
uSnWie
2
usocutec
T. Kano et al. / International Journa
.4.3. 2D mapping measurement by MIR2D mapping was performed under the same conditions used to
easure standard spectra at a step increment of 10 �m in bothhe x- and y-directions over a range of more than a quarter ofhe cross-section of the granule. In the case of measurements of a00 �m × 200 �m area using 10 �m × 10 �m steps, the distributionap consists of 441 spectra (=pixels). Mapping data was computed
sing ISys chemical imaging software (version 5.0; Malvern Instru-ents, Ltd., Malvern UK), and partial least squares regression 2
PLS2) analysis—which is a standard tool for modeling linear rela-ions between multivariate measurements—was performed afterormalization processing. The PLS score of each component in eachixel was calculated based on the ratio to the spectrum in the
ibrary. The library is composed of the spectra of standard prod-cts (pure components). For example, the PLS score of the API in
pixel is calculated by comparing the spectrum to the standardpectrum of pure API in the library, taking the standard spectrumf the pure API as equivalent to a PLS sore of 1, and calculating aLS score of between 0 and 1. These calculations were carried outor all components in all pixels (Koide et al., 2013). Using chemo-
etric analysis, the components present and their densities wereetermined for each pixel in the distribution map, allowing us toroduce distribution maps showing the dispersion states of eachomponent in the granule. In the distribution maps, the areas ofigh PLS scores indicated areas of high concentration, whereasreas of low PLS scores indicated areas of low concentration. Anal-ses of mefenamic acid, lactose, cornstarch, and acrylic resin wereonfirmed to be valid by loading spectra (Johansson et al., 2005;hang et al., 2005). However, validity was not confirmed for eithericrocrystalline cellulose or HPC: for microcrystalline cellulose,
t was considered that this was because the spectrum of micro-rystalline cellulose was similar to that of other additives and thathe amount of microcrystalline cellulose was less than other com-onents, and for HPC it was considered that HPC was difficulto detect under this study’s conditions due to the small amountsed.
.5. Raman spectroscopy
.5.1. Determination of standard spectra by Raman spectroscopyThe spectrum of each ingredient was determined with a
aman spectrophotometer (Cornes Technologies Limited, Tokyo,apan) equipped with a 532 nm laser. Spectra collection waserformed at room temperature under the following condi-ions: 50× microscope objective, 6 s exposure time, and 20ounts.
.5.2. Sample preparation for Raman spectroscopyTo prepare cross-sections of the granules for mapping, gran-
les were first embedded in embedding agent (O.C.T. Compound;akura Finetek Japan Co., Ltd., Tokyo, Japan) and sliced to a thick-ess of 5 �m in a cryostat (CM1850; Leica Microsystems GmbH,etzlar, Germany). This preparation was similar to the preprocess-
ng for MIR (Section 2.4.2). At least five granules were evaluated forach of the specified sampling times.
.5.3. 2D mapping by Raman spectroscopy2D mapping of the cross-section of a granule was performed
nder the same conditions used for measuring standard Ramanpectra at a step increment of 5 �m in both the x- and y-directionsver a 200 �m × 200 �m range. In this case, the distribution maponsists of 1681 spectra (=pixels). Mapping data was computed
sing ISys software, with PLS2 analysis performed after 2nd deriva-ive processing. The components present and their densities inach pixel in the distribution map were determined by usinghemometric analysis in a manner similar to that described abovearmaceutics 461 (2014) 495– 504 497
for MIR mapping (Section 2.4.3). Analyses of mefenamic acid,flavoxate hydrochloride, lactose, cornstarch, and microcrystallinecellulose were confirmed to be valid by loading spectra. How-ever, validity was not confirmed for HPC, which is thought tobe because HPC was difficult to detect due to the small amountused.
2.6. Determination of cluster size of APIs in granules
If dispersion of the API during the granulation process is poor,there is a possibility that the API will agglomerate and be segre-gated in the granules. Therefore, the sizes of clusters of API werecalculated. From over the whole of the API concentration distribu-tion map, the 9.8% of pixels with the highest API score (which isthe same percentage as the amount of API charged) were selected;specifically, 165 pixels corresponding to 9.8% of the 1681 pixelsthat constituted the distribution map were selected. The size ofeach resultant API cluster was calculated on the basis of the diam-eter of a circle with an area equivalent to that of the selected pixels(Clarke, 2004).
Fig. 1 shows in detail the method of determining cluster size.An example distribution map of mefenamic acid composed of441 pixels is shown in Fig. 1a. First, the 9.8% of pixels with thehighest API score are selected and a binary image created. In thebinary image, these selected pixels are displayed in red and theother pixels that are not selected are displayed in blue (Fig. 1b).Finally, the cluster size is calculated on the basis of the diame-ter of a circle of equivalent area as the selected pixels. In thiscase, the presence of four clusters was identified (the four domainsindicated by the red circles in Fig. 1b). When detecting agglom-erates of APIs, however, the selected clusters include a numberof large clusters and a lot of fine clusters. Therefore, the aggre-gation behavior of the API during granulation was evaluated bycalculating the size of the largest cluster of API in a granule; inthe case of Fig. 1b, the selected largest cluster is indicated by thearrow.
2.7. Evaluation of score in each pixel
Within each pixel in the image, single or multiple componentscan be present. A score indicating the concentration of each compo-nent in each pixel was calculated (scaled from 0 to 1). By evaluatingthe score of each additive in the region covered by the largest clus-ter of API in the granule (which was selected by evaluation of clustersize) (e.g., the cluster indicated by the arrow in Fig. 1b), the ratios ofthe score of each additive (lactose, cornstarch, and microcrystallinecellulose) could be calculated and the miscibility of API and othercomponents could be evaluated.
2.8. Measurement of surface free energy
Surface free energies of each component were measured usinga surface tensiometer (Processor Tensiometer K12; Krüss GmbH,Hamburg, Germany) at 25 ◦C. A stainless steel tube packed witheach component in powder form was attached to the deviceand lowered into liquid, and then the time and the weight ofthe liquid which penetrated into the sample powder bed wererecorded when penetration of the liquid into the powder hadstarted. Hexane, tetrachloromethane, 1-octanol, benzene, toluene,formamide, dichloromethane, and N,N-dimethylformamide (allpurchased from Sigma–Aldrich Japan K.K., Tokyo, Japan) were used
as probe solvents of different polarity. The surface free energieswere calculated from the Owens–Wendt–Rabel–Kaelble equationusing the contact angle between liquid and powder (Chung et al.,2003).498 T. Kano et al. / International Journal of Pharmaceutics 461 (2014) 495– 504
50
100
0
150
200
50 10 00 150 200
(a) Distribut ion maps of Me fenamic aci d
Lengt h X (μm)
Len
gth
Y (μ
m)
PLS sc ore
High
Low
50
100
0
150
200
50 1000 150 200
(b) Binar y image maps of Mefena mic aci d
Length X (μm)
Len
gth
Y (μ
m)
The largest cluster of API in this granule
Fig. 1. Method of determining the size of clusters of API in a granule. (a) Distribution map of mefenamic acid; (b) binary image map of mefenamic acid. Red indicates selectedpixels (the 9.8% of pixels with the highest API score) and blue indicates unselected pixels. (For interpretation of the references to color in this figure legend, the reader isr
3
3
Thaaotbspupaottiobemto2g
eferred to the web version of this article.)
. Results and discussion
.1. Change in mean particle size with progression of granulation
The average particle size at each sampling time is shown in Fig. 2.he two types of granules (mefenamic acid granules and flavoxateydrochloride granules) each showed similar behavior. Immedi-tely after the start of granulation, the particle size was very largend there were a lot of large lumps due to the poor dispersionf the binder into the powder. With the progression of granula-ion, particle size was reduced by dispersion of the binder andy crushing by impeller blade and chopper; the average particleize was approximately 500 �m at 5 min. Thereafter, the averagearticle diameter gradually increased with the growth of the gran-les. Particles stopped growing at around 10 min, and the averagearticle diameter became constant (approx. 700 �m). There was
difference between the two types of granules in the behaviorf particle size decrease in the early stage of granulation. It washought that because the flavoxate hydrochloride is more solublehan mefenamic acid, dispersion of the binder occurs more quicklyn flavoxate hydrochloride granules, resulting in quicker growthf granules. However, the particle size was almost the same foroth types of granule at the end of granulation. This was consid-red to be because additives other than the API, which account forore than 90% of the formulation, were exactly the same in the
wo types of granules. In the mapping measurements, distributionsf components in the granules were evaluated in the early (10 s,
min), middle (5 min, 7 min), and late (10 min, 15 min) stages ofranulation.
300
500
700
900
1100
151050time (min)Granulation
Mea
n p
arti
cle
dia
met
er (
μm
)
acid(a) Mefenamic
Fig. 2. Mean particle size of granules. (a) Mefe
3.2. Change in size of API clusters during granulation
The state of aggregation of the API and sizes of API clustersduring granulation as revealed by distribution maps created bymid-infrared spectroscopy (MIR) were compared with those cre-ated by Raman spectroscopy. Fig. 3 shows the distribution maps ofcomponents of the mefenamic acid granules at 10 min as measuredby mid-infrared spectroscopy (Fig. 3a) and by Raman spectroscopy(Fig. 3b). Areas of white in the MIR distribution map (Fig. 3a) werepixels in which only the acrylic resin was present. Average API clus-ter size at 10 min as measured by MIR (29 ± 12.7 �m) was almostthe same as that measured by Raman spectroscopy (26 ± 15.7 �m).Cluster size as measured by MIR was slightly larger than that byRaman spectroscopy because the pixel size in the MIR distributionmap was greater than in the Raman spectroscopy distribution map.Thus, it was confirmed that for mefenamic acid granules, MIR anal-ysis of the API cluster size obtained results equal to those of Ramanspectroscopic analysis. However, analysis of microcrystalline cellu-lose by MIR spectroscopic analysis was uncertain because the MIRspectrum of microcrystalline cellulose was similar to that of lac-tose and cornstarch, whereas in the case of Raman spectroscopicanalysis, differences were detected between the spectrum of themicrocrystalline cellulose and that of other additives. Therefore, thefollowing discussion shows detailed results of Raman spectroscopyonly.
Fig. 4 shows the relationship between the API cluster size,which was the largest cluster size in a granule as measured byRaman analysis, and the granulation time. The average largestcluster size at 10 s was approximately 20 ± 4.8 �m in mefenamic
300
500
700
900
1100
151050
hydrochloride(b) Flavoxate
time (min)Granulation
Mea
n p
arti
cle
dia
met
er (
μm
)
namic acid; (b) flavoxate hydrochloride.
T. Kano et al. / International Journal of Pharmaceutics 461 (2014) 495– 504 499
Lactose CornstarchacidMefenamic
100 3002000
100
200
0
50
150
250
100 3002000
100
200
0
50
150
250
100 3002000
100
200
0
50
150
250
Raman spectroscopy(b)
(a) Mid-inf rared spe ctrosc opy
Microcrystalli ne cellul oseCornstarchLactose
50
100
0
150
200
100500 150 200
50
100
0
150
200
100500 150 200
50
100
0
150
200
100500 150 200
50
100
150
200
100500 150 200
acidMefenamic
Length X (μm)
Len
gth
Y (μ
m)
PLS scor e
High
Low
F Mid-b on of to
autt4rliiiwaAlamdus
Fa(
ig. 3. Distribution maps of components of mefenamic acid granules at 10 min. (a)lue indicates pixels of low score, and white indicates acrylic resin. (For interpretatif this article.)
cid granules and 22 ± 0.8 �m in flavoxate hydrochloride gran-les, roughly the same size. Cluster sizes of both APIs increasedhrough the middle stage of granulation, and at 5 min of granula-ion, the average largest cluster size in mefenamic acid (approx.9.2 ± 9.2 �m) was markedly greater than in flavoxate hydrochlo-ide (approx. 36.3 ± 3.6 �m). Cluster size then decreased in theate stage, with the average largest cluster size at 15 min approx-mately 24 ± 8.3 �m in mefenamic acid granules and 22 ± 2.8 �mn flavoxate hydrochloride granules, roughly the same size. Thus,n both types of granule, the cluster size of the API increased
ith the progress of granulation through the early to middle stagend decreased in the late stage. This suggests that aggregates ofPI were formed from the early to middle stage, and then col-
apsed along with growth of the particles. Moreover, there was difference in the behavior of the cluster size decrease from theiddle to late stage. In mefenamic acid granules, the cluster size
ecreased from the late stage; in flavoxate hydrochloride gran-
les, on the other hand, the cluster size decreased from the middletage.0
10
20
30
40
50
60
151050
Clu
ster
siz
e (μ
m)
Granul ati on ti me (min )
hydrochloride(b) Flavoxateacid(a) Mefenamic
0
10
20
30
40
50
60
151050
Clu
ster
siz
e (μ
m)
Granul ati on ti me (min )
ig. 4. Relationship between the average largest cluster size of the API in a granules measured by Raman analysis and granulation time. (a) Mefenamic acid granules;b) flavoxate hydrochloride granules.
infrared spectroscopy; (b) Raman spectroscopy. Red indicates pixels of high score,he references to color in this figure legend, the reader is referred to the web version
3.3. Analysis of distribution maps
The change in dispersion states of the API and additives withchange in cluster size of API was determined by analysis of distri-bution maps of the early (10 s, 2 min), middle (5 min, 7 min), andlate (10 min, 15 min) stages of granulation. Overlaps and separa-tion in dispersion states among components in the granules wereobserved, and mechanisms of aggregation that occur in granulationare discussed.
3.3.1. Overlapping distribution of components in granules—earlystage
The early stage was immediately after starting granulation, atwhich time there were many coarse particles and dispersion of thebinder was unsatisfactory. Fig. 5 shows the distribution maps ofeach component at 10 s. Because this point is immediately after thestart, granules sampled after 10 s granulation time are consideredto reflect the state of the pre-mix. There was almost no overlapin the distribution of components in the early stage. In regionswhere concentration of mefenamic acid was high (e.g., the domainindicated by the red circles in Fig. 5a), the concentration of lactosewas low. Therefore it was suggested that mefenamic acid and lac-tose were difficult to blend. The same phenomenon was confirmedbetween lactose and cornstarch in both mefenamic acid granulesand flavoxate hydrochloride granules.
3.3.2. Overlapping distribution of components ingranules—middle stage
The middle stage was the beginning of growth of granules. Fig. 6shows the distribution maps at 5 min. In the case of mefenamic acidgranules, in regions where the API was high (e.g., the domain indi-cated by the red circles in Fig. 6a), the concentrations of cornstarch
and microcrystalline cellulose were also high, and the distribu-tions of these three components overlapped. However, in the samearea, the concentration of lactose was low, and was the reverseof the distribution of the other three components. In the case of500 T. Kano et al. / International Journal of Pharmaceutics 461 (2014) 495– 504
Length X (μm)
Len
gth
Y (μ
m)
50
100
0
150
200
100500 150 200
50
100
0
150
200
100500 150 200
50
100
0
150
200
100500 150 200
50
100
0
150
200
100500 150 200
50
100
0
150
200
100500 150 200
50
100
0
150
200
100500 150 200
50
100
0
150
200
100500 150 200
50
100
0
150
200
100500 150 200
hydrochloride granules(b) Flavoxate
hydrochlorideFlavoxate
granulesacid(a) Mefenamic
Microcrystalline celluloseCornstarchLactoseacidMefenamic
Microcrystalline celluloseCornstarchLactose
PLS score
High
Low
F . (a) Mo ces to
flhttc
Fpa
ig. 5. Distribution maps of each component in the early stage of granulation (10 s)f high score and blue indicates pixels of low score. (For interpretation of the referen
avoxate hydrochloride granules, concentrations of lactose wereigh in the areas where concentration of the API was high (e.g.,
he domain indicated by the red circles in Fig. 6b), and their dis-ributions were found to overlap. However, in the same area, theoncentrations of cornstarch and microcrystalline cellulose wereLength X (μm
Len
gth
Y (μ
m)
50
100
0
150
200
100500 150 200
50
100
0
150
200
100500 150 200
50
100
0
150
200
0
50
100
0
150
200
100500 150 200
50
100
0
150
200
100500 150 200
50
100
0
150
200
0
hydrochlor(b) Flavoxate
gacid(a) Mefenamic
LactoseacidMefenamic
hydrochlorideFlavoxate Lactose
ig. 6. Distribution maps of each component in the middle stage of granulation (5 min)ixels of high score and blue indicates pixels of low score. (For interpretation of the referrticle.)
efenamic acid granules; (b) flavoxate hydrochloride granules. Red indicates pixels color in this figure legend, the reader is referred to the web version of this article.)
low, and were the reverse of the distribution of the other twocomponents.
Although such a distribution was confirmed in granules at aplurality of positions, places that showed different characteris-tics were also observed. In the case of mefenamic acid granules,
)
10050 150 200
50
100
0
150
200
100500 150 200
10050 150 200
50
100
0
150
200
100500 150 200
ide granules
ranules
Microcrystalli ne cell uloseCornstarch
Microcrystalli necelluloseCornstarch
PLS score
High
Low
. (a) Mefenamic acid granules; (b) flavoxate hydrochloride granules. Red indicatesences to color in this figure legend, the reader is referred to the web version of this
T. Kano et al. / International Journal of Pharmaceutics 461 (2014) 495– 504 501
Length X (μm)
Len
gth
Y (μ
m)
50
100
0
150
200
100500 150 200
50
100
0
150
200
100500 150 200
50
100
0
150
200
100500 150 200
50
100
0
150
200
100500 150 200
50
100
0
150
200
100500 150 200
50
100
0
150
200
100500 150 200
50
100
0
150
200
100500 150 200
50
100
0
150
200
100500 150 200
hydrochloride granules(b) Flavoxate
granulesacid(a) Mefenamic
hydrochlorideFlavoxateMicrocrystalline
cell uloseCornstarchLactose
Microcrystalli ne celluloseCornstarchLactoseacidMefenamic
PLS score
High
Low
F ). (a) Mo ces to
waaemhtc
odbcdchlmdl
3s
pdtwcwhcdsd
ig. 7. Distribution maps of each component in the late stage of granulation (15 minf high score and blue indicates pixels of low score. (For interpretation of the referen
e observed regions where the concentration of only mefenamiccid was high and regions where the concentrations of mefenamiccid and cornstarch or microcrystalline cellulose were high; how-ver, we identified no regions where the concentration of bothefenamic acid and lactose was high. In the case of flavoxate
ydrochloride granules, we observed areas with high concentra-ion of only the API, but did not observe any overlapping of API andornstarch or microcrystalline cellulose.
Such phenomena were considered to reflect the aggregationf the API in the granules. In the middle stage, overlappingistributions among the components were confirmed, but the com-inations in mefenamic acid granules were different from theombinations in flavoxate hydrochloride granules. For example,istributions of mefenamic acid, cornstarch and microcrystallineellulose overlapped in mefenamic acid granules; on the otherand, distributions of flavoxate hydrochloride and lactose over-
apped in flavoxate hydrochloride granules. In addition, in bothefenamic acid granules and flavoxate hydrochloride granules, the
istribution of cornstarch and microcrystalline cellulose were over-apping, and were the reverse of the distribution of lactose.
.3.3. Overlapping distribution of components in granules—latetage
In the case of mefenamic acid granules, distribution of com-onents in the late stage of granulation (Fig. 7) was similar toistribution in the middle stage. Areas of overlapping distribu-ion of cornstarch, microcrystalline cellulose, and mefenamic acidere confirmed. However, in some areas, it was found that con-
entrations of cornstarch and microcrystalline cellulose were lowhere concentration of the API was high. In the case of flavoxateydrochloride granules, on the other hand, none of the distribution
haracteristics seen in the middle stage, for example, overlappingistribution of API and lactose, were seen in the late stage. In the lasttage, when the granules had reached equilibrium, the overlappingistribution of components that was observed in the middle stageefenamic acid granules; (b) flavoxate hydrochloride granules. Red indicates pixels color in this figure legend, the reader is referred to the web version of this article.)
had changed in both the mefenamic acid and flavoxate hydrochlo-ride granules.
3.4. Miscibility of the API and additives during granulation
The percentage of each additive present in the areas of highAPI concentration was calculated from the proportion of the valuesof the scores of lactose, cornstarch, and microcrystalline cellulosepresent in the largest cluster of API in the distribution map (whichwas selected by evaluation of cluster size in Section 3.2) that com-prised the highest concentrations of API (Fig. 8).
In mefenamic acid granules, the percentage of lactose in theareas of high API concentration was only 15% in the early stage,declining further to 1% in the middle stage, and remaining a lowpercentage (5%) in the later stage. On the other hand, percentagesof cornstarch and microcrystalline cellulose were constantly high. Itwas considered that aggregates in the late stage came from those inmiddle stage because the ratios of additives in the regions of highAPI concentration in the late stage were close to those in middlestage.
In flavoxate hydrochloride granules, the ratio of lactoseincreased in the middle stage of granulation, but in the early andlate stages the proportions of additives were close to the ini-tial charge. Flavoxate hydrochloride was uniformly mixed withadditives in the premixing stage; the miscibility between API andlactose then increased in the middle stage, but it became uniformlymixed again in the late stage.
3.5. Evaluation of affinity of each component
3.5.1. Surface free energy of each ingredientThe state of dispersion of each component in the granules could
be visualized from the distribution maps. Clumps of API were con-firmed in the middle stages of granulation in both mefenamic acidgranules and flavoxate hydrochloride granules; it was thus consid-ered that segregation in the granules had occurred. To understand
502 T. Kano et al. / International Journal of Pharmaceutics 461 (2014) 495– 504
(b) Flavoxate hydrochloride granules(a) Mefe namic acid granu les
0
20
40
60
80
0 5 10 15
Ex
iste
nce
rat
io (
%)
Granula tion ti me ( min)
Lac tose
Cornstarch
Microcrystalline
cellulose 0
20
40
60
80
0 5 10 15
Ex
iste
nce
rat
io (
%)
Granulatio n time (min)
Lactose
Cornstarc h
Microcrystalline
cellulose
Fig. 8. The percentages of each component present in the areas of high API concentration based on scores of each additive. (a) Mefenamic acid granules; (b) flavoxatehydrochloride granules.
Table 2Surface free energy of each ingredient: (�p) polar part, (�d) dispersive part.
Ingredients �d (mJ/m2) �p (mJ/m2)
Mefenamic acid 13.9 9.9Flavoxate hydrochloride 10 27.2Lactose 16.3 27Cornstarch 17.9 6.9Microcrystalline cellulose 14.3 12.6Hydroxypropyl cellulose 17.8 16.5
10
14
18
22
3026221814106
γ d(m
J/m
2)
γp (mJ/m2)
Mefenamic acid
Flavoxate hy drochloride
Lactose
Cornstarch
Mic roc rystalli ne
cellulose
Fet
taslptta
3d
tptie(
D
an
c
Table 3Distance of surface free energy (DSFE) for each combination of components.
Combination of ingredients Distance of surfacefree energy (DSFE)
Mefenamic acid—lactose 17.3Mefenamic acid—cornstarch 5.0Mefenamic acid—microcrystalline cellulose 2.7Flavoxate hydrochloride—lactose 6.3Flavoxate hydrochloride—cornstarch 21.8Flavoxate hydrochloride—microcrystalline cellulose 15.2Lactose—cornstarch 20.2
ig. 9. Surface free energy of each ingredient. �p: polar part of the surface freenergy, �d: dispersive part. The shorter the distance between two points, the higherhe affinity between the two components.
he mechanisms underlying this segregation we examined factorsffecting the miscibility of each component. Table 2 shows theurface free energies of mefenamic acid, flavoxate hydrochloride,actose, cornstarch, microcrystalline cellulose, and HPC. The polarart (�p) of the surface free energy for mefenamic acid was lowerhan that for flavoxate hydrochloride. In addition, the polar part ofhe surface free energy for lactose was by far the highest amongdditives.
.5.2. Relationship between miscibility and affinity based onistance of surface free energy (DSFE)
For each component ingredient of the formulations, we plottedhe polar part (�p) of the surface free energy against the dispersiveart (�d) (Fig. 9). For any two component ingredients plotted inhe figure, a short distance between the two represents high affin-ty and a long distance represents low affinity. Distance betweenach component was defined as the distance of surface free energyDSFE), calculated as follows:
SFE =√
{[(�d Component1) − (�d Component2)]2 + [(�p Component1) − (�p Component2)]2}
nd a small value of DSFE indicated a high affinity between compo-ents.
In the combination of mefenamic acid, cornstarch, and micro-rystalline cellulose, the miscibilities of which were considered
Lactose—microcrystalline cellulose 14.5Corn starch—microcrystalline cellulose 6.7
high in middle stage of granulation, the DSFE of each combina-tion of them was less than the DSFE of other combinations thatwere separated in the middle stage (e.g., lactose versus mefenamicacid, cornstarch, and microcrystalline cellulose; and flavoxatehydrochloride versus cornstarch and microcrystalline cellulose)(Table 3). These results suggested that high affinity between com-ponents was the factor underlying formation of aggregates in themiddle stage of granulation. Thus, if affinity between componentswas high, their miscibility was high and they clumped together ingranules. Conversely, if the affinity between components was low,miscibility was low and there was a tendency to separate in thegranules.
3.6. Mechanism of granule formation in high-shear granulation
From this micro-spectroscopic study with evaluation of affin-ity between ingredients, we propose the following mechanism ofgranule formation in high-shear granulation. It was confirmed thatthe mechanism of aggregation of components in high-shear gran-ulation was composed of three stages.
3.6.1. Early to middle stage: process of formation of aggregates ingranules (Sections 3.3.1, 3.3.2, 3.5.2)
In the early stage of granulation, aggregates of the API couldnot be confirmed in either mefenamic acid granules or flavoxatehydrochloride granules. This indicates that the API was distributeduniformly immediately after premixing. In the middle stage of gran-ulation, the API formed aggregates with high-affinity additives toform coarse particles that were decreased in size by distribution ofthe binder and by destruction by the chopper.
3.6.2. Middle stage: aggregates based on affinity (Sections 3.2,3.3.2, 3.4, 3.5.2)
Granules in the middle stage of granulation contained API clus-ters which were based on affinity as determined by the surface free
l of Ph
eaplahsausopsm
3a
gwtocotogd
4
gagsdacAbImucmfrtwwg
mcosstpittob
T. Kano et al. / International Journa
nergy. In mefenamic acid granules, mefenamic acid, cornstarch,nd microcrystalline cellulose, each of which had similar surfaceroperties, formed agglomerations together and were isolated from
actose whose surface property was much less similar. In flavox-te hydrochloride granules, aggregations consisted of flavoxateydrochloride and lactose, which were separated from both corn-tarch and microcrystalline cellulose; the relationships of affinitymong the components were the same as in mefenamic acid gran-les. That is to say, if the surface properties of components wereimilar, they were easy to mix in the granulation process; on thether hand, if the surface properties were not similar the com-onents were hard to mix. It was considered that componentegregation occurred in the granules in middle stage due to thisechanism.
.6.3. Middle stage—late stage: process of disintegration ofgglomerates (Sections 3.2, 3.3.3)
In both mefenamic acid granules and flavoxate hydrochlorideranules, the size of API clusters was reduced in the late stage. Itas considered that the aggregates that formed in the granules in
he middle stage collapsed and were dispersed along with progressf granulation. In the middle stage of granulation, mefenamic acidlusters were larger than clusters of flavoxate hydrochloride. More-ver, in mefenamic acid granules, the cluster size decreased fromhe late stage, whereas, in flavoxate hydrochloride granules, on thether hand, the cluster size decreased from the middle stage. Whenranules enlarged and reached equilibrium, aggregates of API wereestroyed and were dispersed in the granules.
. Conclusions
In this study using granules containing mefenamic acid andranules containing flavoxate hydrochloride, we were able to visu-lize the dispersion state of the components in the 2 types ofranule by using micro-spectroscopic techniques (mid-infraredpectroscopy [MIR] and Raman spectroscopy). MIR analysis toetermine the size mefenamic acid clusters obtained results equiv-lent to those of the Raman analysis. Micro-Raman mapping ofross-sections of the granules confirmed the distributions of thePI, lactose, cornstarch, and microcrystalline cellulose. The distri-utions of these components changed as granulation proceeded.n addition, by using chemometrics to analyze the distribution
aps, we could explain that API formed agglomerates in gran-les from the early to middle stage of granulation, and that theselusters collapsed in the late stage. Moreover, in order to discussechanisms of granulation, we evaluated the distances of sur-
ace free energy (DSFE) between each ingredient. Examining DSFEevealed that, in the case of high-shear granulation, API and addi-ives which had high affinity toward each other were easy to blendhile those with low affinity were hard to blend. This phenomenonas considered to be the cause of segregation of components in
ranules.By using a combination of experimental techniques, namely
icro-spectroscopy and evaluating surface free energy, we suc-essfully explained granulation mechanisms such as the formationf agglomerates. The present study demonstrated that micro-pectroscopic techniques and evaluation of affinities based onurface free energy between components is a valuable techniqueo clarify mechanisms of granulation at the level of componentarticles. We consider this method will also be useful in elucidat-
ng the mechanisms of granulation in granulation methods other
han high-shear granulation, such as fluidized bed granulation. Fur-hermore, the techniques could be important to the elucidation ofther mechanisms of general manufacturing processes, such as thelending process.armaceutics 461 (2014) 495– 504 503
Acknowledgements
We express our appreciation to Powrex Inc. (Hyogo, Japan) formanufacturing the pharmaceutical granules for us.
The synchrotron radiation experiments were performed at theBL43 of SPring-8 with the approval of the Japan SynchrotronRadiation Research Institute (JASRI) (Proposal Nos. 2007A1249,2007B1231, and 2008A1311).
References
Amigo, J.M., Ravn, C., 2009. Direct quantification and distribution assessment ofmajor and minor components in pharmaceutical tablets by NIR-chemical imag-ing. Eur. J. Pharm. Sci. 37, 76–82.
Belohlav, Z., Brenkova, L., Hanika, J., Durdil, P., Rapek, P., Tomasek, V., 2007. Effectof drug active substance particles on wet granulation process. Chem. Eng. Res.Des. 85, 974–980.
Benali, M., Gerbaud, V., Hemati, M., 2009. Effect of operating conditions and physico-chemical properties on the wet granulation kinetics in high shear mixer. PowderTechnol. 190, 160–169.
Bouwman, A.M., Henstra, M.J., Westerman, D., Chung, J.T., Zhang, Z., Ingram, A.,Seville, J.P.K., Frijlink, H.W., 2005. The effect of the amount of binder liquid onthe granulation mechanisms and structure of microcrystalline cellulose granulesprepared by high shear granulation. Int. J. Pharm. 290, 129–136.
Cavinato, C., Andreato, E., Bresciani, M., Pignatone, I., Bellazzi, G., Franceschinis, E.,Realdon, N., Canu, P., Santomaso, A.C., 2011. Combining formulation and processaspects for optimizing the high-shear wet granulation of common drugs. Int. J.Pharm. 416, 229–241.
Chung, H.Y., Yonemochi, E., Saitoh, T., Terada, K., Tozuka, Y., Oguchi, T., Yamamoto,K., Chung, H.Y., Choi, W.S., 2003. Factors affecting the apparent solubility ofursodeoxycholic acid in the grinding process. Int. J. Pharm. 255, 49–56.
Clarke, F., 2004. Extracting process-related information from pharmaceutical dosageforms using near-infrared microscopy. Vib. Spectrosc. 34, 25–35.
Faure, A., York, P., Rowe, R.C., 2001. Process control and scale-up of pharmaceuticalwet granulation processes: a review. Eur. J. Pharm. Biopharm. 52, 269–277.
Franch-Lage, F., Amigo, J.M., Skibsted, E., Maspoch, S., Coello, J., 2011. Fastassessment of the surface distribution of API and excipients in tablets usingNIR-hyperspectral imaging. Int. J. Pharm. 411, 27–35.
Gendrin, C., Roggo, Y., Collet, C., 2007. Content uniformity of pharmaceutical soliddosage forms by near infrared hyperspectral imaging: a feasibility study. Talanta73, 733–741.
Iveson, S.M., Litster, J.D., Hapgood, K., Ennis, B.J., 2001a. Nucleation growth andbreakage phenomena in agitated wet granulation processes: a review. PowderTechnol. 117, 3–39.
Iveson, S.M., Wauters, P.A.L., Forrest, S., Litster, J.D., Meesters, G.M.H., Scarlett, B.,2001b. Growth regime map for liquid-bound granules: further developmentand experimental validation. Powder Technol. 117, 83–97.
Johansson, J., Pettersson, S., Folestad, S., 2005. Characterization of different laserirradiation methods for quantitative Raman tablet assessment. J. Pharm. Biomed.Anal. 39, 510–516.
Karavas, E., Georgarakis, M., Docoslis, A., Bikiaris, D., 2007. Combining SEM TEM, andmicro-Raman techniques to differentiate between the amorphous molecularlevel dispersions and nanodispersions of a poorly water-soluble drug within apolymer matrix. Int. J. Pharm. 340, 76–83.
Koide, T., Nagato, T., Kanou, Y., Matsui, K., Natsuyama, S., Kawanishi, T., Hiyama, Y.,2013. Detection of component segregation in granules manufactured by highshear granulation with over-granulation conditions using near-infrared chem-ical imaging. Int. J. Pharm. 441, 135–145.
Le, P.K., Avontuur, P., Hounslow, M.J., Salman, A.D., 2011. A microscopic study ofgranulation mechanisms and their effect on granule properties. Powder Technol.206, 18–24.
Li, W., Woldu, A., Kelly, R., McCool, J., Bruce, R., Rasmussen, H., Cunningham, J.,Winstead, D., 2008. Measurement of drug agglomerates in powder blending sim-ulation samples by near infrared chemical imaging. Int. J. Pharm. 350, 369–373.
Maurer, L., Leuenberger, H., 2009. Terahertz pulsed imaging and near infrared imag-ing to monitor the coating process of pharmaceutical tablets. Int. J. Pharm. 370,8–16.
Nguyen, T.H., Shen, W., Hapgood, K., 2010. Effect of formulation hydrophobicity ondrug distribution in wet granulation. Chem. Eng. J. 164, 330–339.
Pharmaceutical and Medical Device Regulatory Science Society of Japan (PMRJ),2002a. Iryoyo Iyakuhin Hinshitsu Johoshu No. 11. Yakuji Nippo Limited, Tokyo.
Pharmaceutical and Medical Device Regulatory Science Society of Japan (PMRJ),2002b. Iryoyo Iyakuhin Hinshitsu Johoshu No. 13. Yakuji Nippo Limited, Tokyo.
Rahmanian, N., Naji, A., Ghadiri, M., 2011. Effects of process parameters on gran-ules properties produced in a high shear granulator. Chem. Eng. Res. Des. 89,512–518.
Saleh, K., Vialatte, L., Guigon, P., 2005. Wet granulation in a batch high shear mixer.Chem. Eng. Sci. 60, 3763–3775.
Smirani-Khayati, N., Falk, V., Bardin-Monnier, N., Marchal-Heussler, L., 2009. Binderliquid distribution during granulation process and its relationship to granulesize distribution. Powder Technol. 195, 105–112.
Tu, W.-D., Ingram, A., Seville, J., Hsiau, S.-S., 2009. Exploring the regime map forhigh-shear mixer granulation. Chem. Eng. J. 145, 505–513.
5 l of Ph
V
V
V
V
V
pharmaceutical granules using IR-to-THz imaging. Infrared Phys. Technol. 51,450–453.
04 T. Kano et al. / International Journa
an den Dries, K., Vromans, H., 2002. Relationship between inhomogeneityphenomena and granule growth mechanisms in a high-shear mixer. Int. J. Pharm.247, 167–177.
an den Dries, K., Vromans, H., 2004. Qualitative proof of liquid dispersion andpenetration-involved granule formation in a high shear mixer. Eur. J. Pharm.Biopharm. 58, 551–559.
an den Dries, K., Vromans, H., 2009. Quantitative proof of liquid penetration-involved granule formation in a high shear mixer. Powder Technol. 189,165–171.
an den Dries, K., De Vegt, O.M., Girard, V., Vomans, H., 2003. Granule breakage
phenomena in a high shear mixer; influence of process and formulation variablesand consequences on granule homogeneity. Powder Technol. 133, 228–236.emavarapu, C., Surapaneni, M., Hussain, M., Badawy, S., 2009. Role of drug sub-stance material properties in the processibility and performance of a wetgranulated product. Int. J. Pharm. 374, 96–105.
armaceutics 461 (2014) 495– 504
Vonk, P., Guillaume, C.P.F., Ramaker, J.S., Vromans, H., Kossen, N.W.F., 1997. Growthmechanisms of high-shear pelletisation. Int. J. Pharm. 157, 93–102.
Yamamoto, Y., Fukami, T., Koide, T., Suzuki, T., Hiyama, Y., Tomono, K., 2012.Pharmaceutical evaluation of steroidal ointments by ATR-IR chemical imaging:distribution of active and inactive pharmaceutical ingredients. Int. J. Pharm. 426,54–60.
Yonemochi, E., Furuyama, N., Bunko, M., Moriwaki, T., Ikemoto, Y., Terada, K., 2008.Evaluation of dispersion state of the two racemic compounds of troglitazone in
Zhang, L., Henson, M.J., Sekulic, S.S., 2005. Multivariate data analysis forRaman imaging of a model pharmaceutical tablet. Anal. Chim. Acta 545,262–278.