Effects of baffle configuration and tank size on spherical agglomerates of dimethyl fumarate in a common stirred tank

International Journal of Pharmaceutics 495 (2015) 886–894

Contents lists available at ScienceDirect

International Journal of Pharmaceutics journal homepage: www.elsevier.com/locate/ijpharm

Effects of baffle configuration and tank size on spherical agglomerates of dimethyl fumarate in a common stirred tank Po Yen Lin, Hung Lin Lee, Chih Wei Chen, Tu Lee* Department of Chemical and Materials Engineering, National Central University, 300 Jhong-Da Road, Jhong-Li District, Taoyuan City 32001, Taiwan

A R T I C L E I N F O

A B S T R A C T

Article history: Received 9 June 2015 Received in revised form 31 August 2015 Accepted 18 September 2015 Available online 28 September 2015

To pave the way for technology transfer and scale up of the spherical agglomeration (SA) process for dimethyl fumarate, effects of the US, European and Kawashima type baffles and 0.5, 2.0 and 10 L-sized common stirred tank were studied. It was found that the particle size distribution varied significantly. However, the size-related properties such as dissolution profile and flowability of agglomerates from the same size cut after sieving could remain unchanged. The interior structure-related properties such as particle density and mechanical property of agglomerates upon baffle change and scale up from the same size cut were decayed and the agglomerates could become denser and stronger by prolonged maturation time. To maintain the same size distribution, agglomerates from any batch could have been separated and classified by sieving and then blended back together artificially by the desired weight% of each cut. ã 2015 Elsevier B.V. All rights reserved.

Keywords: Spherical agglomeration Dimethyl fumarate Baffle Scale up Dissolution profile Mechanical property

1. Introduction About 80% of all marketed drug products and more than 95% of the top selling drugs are solid oral dosage forms, which are convenient for transportation, packaging, storage, and highly acceptable to patients (Byn et al., 2005). To increase the dissolution rate and reach sufficient bioavailability of poorly water-soluble drugs, small micron-sized active pharmaceutical ingredient (API) crystals are often produced by crystallization and precipitation. However, the poor flow and mechanical properties of those fine crystals can make the downstream processing and dosage control difficult (Lee and Webb, 2003). The desired physical and mechanical properties for filtering, drying and handling, the decent flowability and packability for mixing, filling and tableting, and the uniform bulk density for predictable compressibility and dissolution, can often be obtained by agglomeration where the small crystals are assembled to larger agglomerates (Osborne et al., 1990). The conventional ways for agglomeration are by mixer granulation and fluidized-bed granulation (Faure et al., 2001), but granulation step is time-consuming and adds additional costs to the manufacturing (Lee et al., 2010a, 2010b). A better alternative is to produce the round agglomerates directly by spherical

* Corresponding author. Fax: +886 3 425 2296. E-mail address: [email protected] (T. Lee). http://dx.doi.org/10.1016/j.ijpharm.2015.09.056 0378-5173/ ã 2015 Elsevier B.V. All rights reserved.

crystallization in a common stirred tank at the crystallization step. If that worked out, granulation could be avoided, and direct tableting would become economically feasible. Consequently, less equipment and space, lower labor costs, less processing steps and lower energy consumption would be required (Lee et al., 2010a, 2010b). Spherical crystallization (Kova9 ci9 c et al., 2012) can be categorized by four different ways: (1) spherical agglomeration (SA), (2) quasi emulsion solvent diffusion (QESD), (3) ammonia diffusion system (ADS), and (4) neutralization. SA is important and popular (Kova9 ci9 c et al., 2012) because of its simplicity. However, it may not be applicable for complex problems. Normally, in the SA method, an antisolvent is firstly added to the nearly saturated solution of API in a good solvent for precipitating the drug crystals immediately. Under agitation, a third solvent called the bridging liquid (or the wetting agent) is fed, which is immiscible with the poor solvent and preferentially wet the precipitated crystals. As a result of the interfacial tension and capillary force, the bridging liquid acts to adhere the crystals to one another and turns them into agglomerates of larger sizes. For the ease of studying the influence of operating parameters for SA, ballmilled and sieved crystals were used to fix and standardize the size of the crystalline building blocks of SA at the starting point. The bridging liquid was fed directly to the crystal suspension made of a poor solvent. Therefore, only two solvents were used instead of three (Chow and Leung, 1996). If water is being used as one of the solvents, it will be possible to reduce the production cost for many systems.

P.Y. Lin et al. / International Journal of Pharmaceutics 495 (2015) 886–894

In general, solvents and solvent composition (Osborne et al., 1990; Gordon and Chowhan, 1990), temperature (Osborne et al., 1990), amount of bridging liquid (Blandin et al., 2003; AmaroGonzález and Biscans, 2002), feeding rate of bridging liquid (Kawashima et al., 1982), feeding rate of suspension (Kawashima et al., 1982), initial particle size (Kawashima et al., 1981a), solid loading (Blandin et al., 2000), stirring rate (Maghsoodi, 2011) and maturation time (Blandin et al., 2000; Thati and Rasmuson, 2012) are the many important operating parameters which influence the success of spherical crystallization. These parameters affect not only the productivity but also the particle size distribution (Subero-Couroyera et al., 2006), morphology (Amaro-González and Biscans, 2002; Thati and Rasmuson, 2012), strength (Blandin et al., 2000) and dissolution rate (Maghsoodi, 2011; Varshosaz et al., 2011) of the agglomerates. The effects of above operating parameters had been explored earlier by using different materials as model compounds. For instances, (1) the increase in the amount of bridging liquid would increase the agglomerate size (Blandin et al., 2003). Adding too much bridging liquid would make the agglomerates soft and pasty (Amaro-González and Biscans, 2002), (2) at high feeding rate of bridging liquid, the product particle size became larger, whereas at high feeding rate of suspension, the product agglomerate size became lower (Kawashima et al., 1982), (3) the agglomerate size increased as either the initial particle size decreased (Kawashima et al., 1981a) or the solid loading increased (Blandin et al., 2000), (4) a higher stirring speed also led to faster particle dissolution rate which might be related to particle size reduction of the agglomerates (Maghsoodi, 2011), (5) with increasing the stirring speed and the amount of bridging liquid, the particle size distribution tended to shift toward the larger particle size range (Subero-Couroyera et al., 2006), (6) the choice of the bridging liquid had an influence on the rate of agglomeration and the strength of the agglomerates (Kawashima et al., 1981b), (7) as the maturation time increased, the particle size increased. The particles became more spherical with higher strength (Thati and Rasmuson, 2012), (8) various solvent compositions would influence the particle size distribution, morphology and mechanical strength differently (Amaro-González and Biscans, 2002; Thati and Rasmuson, 2012), and (9) lower temperature led to larger and stronger agglomerates (Thati and Rasmuson, 2012). To scale up the process for SA properly, special attention must also be paid to the effects of baffle configuration and tank size. However, systematic studies of those parameters are rare. For instances, (1) the agitator torque was found to be a function of the speed of the agitator and the number of baffles in the system (Kawashima and Capes, 1974), (2) the agglomeration rate constant was seen to increase exponentially with the shear force applied to the system as measured by the agitator torque, the total number of agglomerates per unit volume of suspension (i.e., population density) increased as the mean agglomerate size decreased (Kawashima and Capes, 1974), and (3) spherical crystallization was carried out in 2.5 L-sized (Blandin et al., 2005), 30 L-sized (Kawashima et al., 1994), and 100 L-sized (Bos and Zuiderweg,

887

1987) stirred vessels separately in several unrelated studies where no scale up guidelines were provided at all. Therefore, the aim of this paper is to study the effects of baffle configuration and tank size on the relatively simple SA method systematically in addition to two other operating parameters: (1) the amount of bridging liquid added (BSR: bridging liquid volume to solid ratio), and (2) the maturation time. A newly approved re-positioned drug in the year of 2013 for treating relapsing forms of multiple sclerosis, dimethyl fumarate (Jarvis, 2014), was chosen as our model API because it is commercially available, inexpensive and its poor solvent is water. Toluene was selected as the bridging liquid for the SA method because of its immiscibility with water (0.50 mg of toluene/mL of water at 20
C) (Murov, 1997) and the relatively good solubility of dimethyl fumarate in toluene. The micromeritic properties of the commercial dimethyl fumarate, and the ball-milled and sieved dimethyl fumarate feed standardized for all trials were shown in Figs. S1 and S2, respectively. The loading of dimethyl fumarate solids in the aqueous suspension was kept at 0.025 g/mL. The chemical identity, polymorphism, morphology, size distribution, mechanical strength, flowability and dissolution rate of all agglomerates were characterized and checked. 2. Materials and methods 2.1. Chemicals and solvents Dimethyl fumarate (C6H8O4, 99% purity, MW 144.13, mp 102–105
C, true density 1.37 g/cm3, Lot 10183993), white crystalline platelets were purchased from Alfa Aesar (Ward Hill, MA, USA) (Fig. S1(a)). Toluene (C6H5CH3, ACS grade, 99.5% purity, MW 92.14 bp, 111
C, Lot ETA140403) was received from Echo Chemical (Miaoli, Taiwan). Reversible osmosis (RO) water was clarified by a water purification system (model Milli-RO Plus) bought from Millipore (Billerica, MA). Potassium phosphate buffer concentrate pH 6.8 (pH 6.78–6.82, Lot BCBN3286V) was purchased from Fluka (Ireland). 2.2. Experimental methods 2.2.1. Ball milling 100 g of commercial platelet-shaped dimethyl fumarate powders were ball-milled (MUBM-236-RTD, Shin Kwang Machinery, Taiwan, ROC) to produce the standardized dimethyl fumarate as starting materials. The volume of the ball jar was 1.5 L, and the mass ratio of ceramic-ball-to-powder was fixed to 6:1. The ball milling process was taken with a rotation speed of 500 rpm for 5 h, and the ball-milled powders are characterized in Fig. S2. 2.2.2. Spherical agglomeration 2.2.2.1. Apparatus. The temperature for all experiments was kept at 25
C. Cylindrical glass vessels of three different sizes of 0.5, 2 and 10 L equipped with four vertical baffles of either European,

Table 1 Cylindrical tank sizes and vertical baffle configurations. Tank size (L) 0.5

2 10

Tank diameter, T (cm) 8

13 20

Impeller diameter, D (cm)

Baffle type

Baffle width, B (cm)

Stirring rate, n (rpm)

3.6

US European Kawashima

0.7 0.9 1.1

600

5 7

Kawashima Kawashima

1.6 2.5

450 350

888

P.Y. Lin et al. / International Journal of Pharmaceutics 495 (2015) 886–894

US or Kawashima type having the baffle-width-to-tank-diameter ratio of 1/10,1/12 and 1/8, respectively (Bates et al.,1966; Kawashima et al., 2003) were used for the SA process. Their combinations were summarized in Table 1. If D = impeller diameter (cm), C = clearance of the impeller from the tank bottom (cm), T = tank diameter (cm), H = liquid height (cm), and B = baffle width (cm), D¼C¼

T 3

ð1Þ

H=T

B¼

(2)

T T T ðUS typeÞ or ðEuropean typeÞ or ðKawashima typeÞ 12 10 8 ð3Þ

All vessels were agitated by a pitched blade down flow turbine (PTD-45
, four blades). The pitched blade turbine impeller is especially efficacious in flow sensitive mixing operations such as solid suspension (Cui et al., 1999). To keep particles suspended, the impeller speed, N (rev s1), was pre-determined by the Zwietering equation (Zweitering, 1958): N / SD0.85

(4)

where S = a constant and D = the impeller diameter (m). The impeller speeds for 0.5 L-, 2 L- and 10 L-sized tanks were calculated to be 600, 450 and 350 rpm, respectively. 2.2.2.2. Effects of baffle configuration. 10 g of ball-milled and sieved dimethyl fumarate powders were suspended in 0.4 L of water in a 0.5 L-sized tank in each trial. Since BSR values of 0.4, 0.5 and 0.6, the maturation times of 1, 2 and 3 h, and the baffle types of the US, European or Kawashima were chosen as variables, there were a total of 3  3  3 = 27 combinations of experiments. Therefore, for each baffle type, there was a set of 3  3 = 9 operating conditions based on the coupling of different BSRs and the maturation times. BSR values of 0.4, 0.5 and 0.6 were equivalent to 2.91, 3.64 and 4.37 mL of toluene. All agglomerates grown were filtered at the end of the maturation time, oven dried at 40
C for 12 h, and then sieved. For BSR of 0.5 and the maturation time of 2 h, the 710–1000 mm dried agglomerates were characterized for their chemical identity, mechanical properties, and dissolution rates. The 710–1000 mm sized agglomerates were characterized selectively because this size range had fallen between the 250 mm size where agglomerates had started to become round and the 2 mm size where the agglomerates in the marketed dimethyl fumarate capsules were found. In a separate SA experiment, agglomerates grown from a stirred tank with the Kawashima type baffle at BSR of 0.5 and maturation time of 2 h were filtered at the end of the maturation time and sieved directly. The 710–1000 mm wet agglomerates were characterized for their mechanical properties. The morphology of all the 0–1000 mm dried and wet agglomerates harvested from those stirred tanks were observed by optical microscopy (OM). 2.2.2.3. Effects of tank size. The solid loadings of dimethyl fumarate and the amounts of toluene used for the various tank

sizes were listed in Table 2. The SA experiments were carried out in 0.5, 2 and 10 L-sized tanks equipped with the Kawashima baffle type and the stirring rates were kept at 600, 450 and 350 rpm, respectively. The maturation time was fixed at 2 h for all experiments. Maturation time was also prolonged to 6 h. However, it was implemented to the SA experiments merely in 2 and 10 L-sized tanks. All the agglomerates were filtered at the end of the maturation time, oven dried at 40
C for 12 h, and then sieved. The 710– 1000 mm dried agglomerates were characterized for their chemical identity, mechanical properties, and dissolution rates, and the 500–710 mm dried agglomerates grown from 0.5 and 10 L-sized stirred tanks were characterized only for dissolution rates. 2.2.3. Solubility The solubility value of dimethyl fumarate in toluene at 25
C was determined by the gravimetric method (Lee et al., 2012). About 50 mg of dimethyl fumarate were weighted in a 20 mL scintillation vial, and toluene was titrated carefully by a micropipette into the vial with an intermittent shaking in a 25
C water bath until all the solids were just dissolved as determined by eye. The solubility value was calculated as the weight of dimethyl fumarate divided by the total volume of toluene added. Even though the gravimetric method had been reported and verified having an inherent inaccuracy of about 10–20% (Lee and Lee, 2015), its advantages were its robustness, simplicity, without the need of performing any calibration and concerning for the formation of polymorph, hydrate or solvate, and sublimation. 2.3. Analytical methods 2.3.1. Sieving The particle size distributions of commercial dimethyl fumarate powders, ball-milled dimethyl fumarate powders and dimethyl fumarate spherical agglomerates were determined through a stack of metal sieve plates from the largest aperture to the finest in the order of 1410, 1000, 710, 500, 355, 250, 177, 125 and 88 mm (Kuang Yang, Taiwan). The diameter of metal sieve plate was 9 cm. To eliminate the unwanted mesh plugging and to minimize particle breakage and aggregation on the mesh, 5–10 g of samples were placed at the center of the 1410 mm-sized sieve plate in the beginning. Vibration was then generated by holding the 1410 mmsized sieve plate with one hand and tapping the sieve plate sideways with a spatula by another until no more powders (or agglomerates) on the 1410 mm-sized sieve plate passed through by eye. Agglomerates, which passed through the 1410 mm-sized sieve plate, were collected on the surface of the 1000 mm-sized sieve plate. The same shaking method was then repeated successively for the other sieve plates with smaller-sized openings. For this method, there was no need to worry about the effects of sample loading, the speed of shaking, and the measurement time on the data obtained for analyzing the particle size distributions (Lee and Hsu, 2007). The weight of the powders (or agglomerates) retained on the surface of each sieve plate was divided by the total sample weight of the powders (or agglomerates) to obtain the corresponding weight% oversize for each sieve fraction. The particle size distribution was presented as a cumulative oversize distribution curve.

Table 2 The solid loading conditions of dimethyl fumarate and volume of bridging liquid added for the various tank sizes Tank size (L)

Weight of dimethyl fumarate (g)

Volume of water (mL)

BSR

Volume of toluene (mL)

0.5 2 10

10 43 157

400 1725 6280

0.5 0.5–0.7 0.7

3.6 15.7–22.0 80.2

P.Y. Lin et al. / International Journal of Pharmaceutics 495 (2015) 886–894

889

2.4. Particle morphology The exterior particle morphologies of the agglomerates from 355, 500 and 710 mm-sized cuts were separately imaged by optical microscopy. The interior porous structures of those agglomerates were characterized separately by scanning electron microscopy (SEM). 2.4.1. Circularity The circularity of dimethyl fumarate agglomerates from the 710 mm size cuts were measured through the image software analysis called “ImageJ” (Helmy and Azim, 2012). ! Ap ð5Þ Circularity ¼ 4p Pp 2 where Ap = area of the particle outline (cm2), Pp = perimeter of particle outline (cm). 2.4.2. Particle strength The fracture forces of agglomerates were determined separately by compression under a punch with a water reservoir. Two to four agglomerates from the same size cut were subjected together to a constant movement of the punch due to the increasing weight of the water reservoir by gravimetric titration. Water titration would continue until all the agglomerates were fractured. The force applied was the weight of water, and the average fracture force exerted on each agglomerate was the total force applied divided by the number of agglomerates. Furthermore, the mechanical strength of each agglomerate was calculated by dividing the average fracture force by the circular cross-sectional area of each spherical agglomerate. 2.4.3. Flowability The rheological properties of the ball-milled 188–250 mm dimethyl fumarate powders and the 710–1000 mm dimethyl fumarate agglomerates were measured by calculating the Carr’s index, CI (Lee and Hsu, 2007): ðrt  rp Þ  100% CI ¼

rt

ð6Þ

where the poured density (g/cm3), rp = the mass of sample divided by the undisturbed volume in a 5 mL graduated cylinder after filling, and the tapped density (g/cm3), rt = the mass of sample divided by the disturbed volume after tapping until no change in the volume was seen. The number of times for tapping was about 50. 2.4.4. Particle density 20 agglomerates were isolated from the sieved fraction of the 710 mm size cut. Their particle density was determined by counting and weighting the agglomerates according to Eq. (7) (Lee et al., 2010b): M 3 ¼ rv K v d v N

Fig. 1. 27 cumulative oversize distribution curves for dimethyl fumarate agglomerates grown from (a) US baffle type, (b) European baffle type, and (c) Kawashima baffle type, with BSR from 0.4 to 0.6 and maturation time 1–3 h. X-axis denotes the sieve size and Y-axis denotes the weight%.

ð7Þ

where M = the cumulative mass (g), N = the cumulative number, rv = the particle density (g/cm3), Kv = the shape factor, and dv = the volume mean diameter of the upper and the lower sieve aperture of the sieve fraction (cm). dv = [(d13 + d23)/2]1/3 where d1 and d2 were the lower and upper Feret’s statistical diameter of the specific sieve fraction. To simplify the calculation of dv, the average of the upper and lower sieve aperture were used instead of the mean diameter of the agglomerates by image analysis. The particles were assumed to be spherical by which Kv = p/6 (Thati and Rasmuson, 2011).

Weight of dimethyl fumarate (mg)

890

P.Y. Lin et al. / International Journal of Pharmaceutics 495 (2015) 886–894

2.4.7. Instrumentations 120

2.4.7.1. Fourier-transform infrared spectroscopy (IR). FTIR spectroscopy was used to identify the samples. IR spectra were recorded on a PerkinElmer Spectrum One spectrometer (PerkinElmer Instruments LLC, Shelton, CT, USA). The KBr sample disk was scanned with a scan number of 8 from 4000 to 400 cm1 having a resolution of 2 cm1.

100 80 60

t50 = 19 min t50 = 24 min

40

US baffle type European baffle type Kawashima baffle type Raw materials

20 0

0

10

20

30

40

50

60

70

80

90

100

Time (min) Fig. 2. Dissolution profiles of 710–1000 mm sized dimethyl fumarate agglomerates grown from 0.5 L-sized tanks installed with the US, European and Kawashima baffle types at impeller speed = 600 rpm, BSR = 0.5, and maturation time = 2 h. Dissolution profile of commercial dimethyl fumarate platelets was also included for comparison.

2.4.5. Dissolution rate A dissolution test station (SR6, Hanson Research Corporation, Chatsworth, California, USA) Type II (paddle method) at a rotation speed of 100 rpm was used for testing the in vitro dissolution of dimethyl fumarate (Chikhale, 2012) agglomerates. Dissolution was carried out on an equivalent of 120 mg of dimethyl fumarate agglomerates. Potassium phosphate buffer of pH 6.8 was used as the dissolution medium. The volume and temperature of the dissolution medium were 500 mL and 37.0  0.2
C, respectively. Samples of 1–2 mL were withdrawn at 0, 1, 2.5, 5, 6.5, 8, 10, 12.5, 15, 20, 30, 40, 60 and 90 min by a plastic syringe near the stirring paddle. Each sample was filtered by a 0.22 mm syringe filter (Millex-GV, Millipore, Massachusetts, USA), diluted 40 times in a 20 mL scintillation vial with RO water and then assayed for the absorbance of dimethyl fumarate at l = 210 nm by a UV/Vis spectrophotometer. 2.4.6. Specific energy The specific energy (kJ/kg) was calculated by the power consumption, P (kW) (McCabe et al., 2005) and the total mass in the vessel, M (kg). For a turbine impeller, P = 0.00127 n 3D5r. Therefore, the specific energy: Pt 0:00127 n3 D5 rt ¼ M M

ð8Þ

where t = the maturation time (s), n = the rotation speed (rev/s), D = impeller diameter (m), r = the density of water = 1000 (kg/m3). Since the 710–1000 mm agglomerates were large and uniform, and much easy to measure their mechanical properties and dissolution tests, they had been chosen for the analysis. All the analysis including solubility measurement, particle strength, flowability, particle density and dissolution rate were repeated three times, and the standard deviations were provided.

2.4.7.2. Ultraviolet and visible spectrophotometry (UV–vis). The concentration of dimethyl fumarate was monitored over the dissolution test. About 1 mL of solution was withdrawn from the dissolution tester at different time points, filtered by a 0.22 mm syringe filter (Millex-GV, Millipore, Massachusetts, USA), and diluted for 40 times with water for dissolution in a 20 mL scintillation vial. The clear solution was assayed based on the characteristic UV absorbance peak at 210 nm by an UV–vis spectrophotometer (Lambda 25, PerkinElmer, Norwalk, CT, USA.). The concentration of dimethyl fumarate in the filtered solution for each run was converted from the absorbance value, A, to the corresponding concentration, C, by a linear calibration line: A = 0.008 + 104.2  C (mg/mL), according to Beer’ Law established from four standard aqueous solutions of acetaminophen with known concentrations (Fig. S3). 2.4.7.3. Differential scanning calorimetry (DSC). Thermal analytical data of 3–5 mg of samples in perforated aluminum sample pans (60 mL) were collected on a Perkin Elmer DSC-7 calorimeter (PerkinElmer Instruments LLC, Shelton, CT, USA) with a heating rate of 10
C/min from 40
C to 130
C under a constant nitrogen 99.99% purge. This instrument was calibrated with indium and zinc 99.999% having reference temperatures of 156.6
C and 419.47
C, respectively (PerkinElmer Instruments LLC, Shelton, CT, USA). 2.4.7.4. Powder X-ray diffraction (PXRD). PXRD patterns were detected by Bruker D8 Advance (Germany). The source of PXRD was CuKa (l = 1.5418 Å) and the diffractometer was operated at 40 kV and 40 mA. The X-ray was passed through a 1 mm slit and the signal a 1 mm slit, a nickel filter, and another 0.1 mm slit. The detector type was a scintillation counter. The scanning rate was set at 0.03
2u/s ranging from 5
to 35
. PXRD patterns provided another piece of information for the identification and crystallinity of dimethyl fumarate powders and agglomerates. 2.4.7.5. Optical microscopy (OM). The optical microscope (Olympus SZII, Tokyo, Japan) with a charge couple device (CCD) camera (SONY, model: SSCDC50A, Tokyo, Japan) was used to observe the morphology of dimethyl fumarate powders and agglomerates. 2.4.7.6. Scanning electron microscopy (SEM). Scanning electron microscope (Hitachi S-3500N, Tokyo, Japan) was used to observe the morphology and the cross-section of the agglomerates. Both secondary electron imaging (SEI) and backscattered electron imaging (BEI) were used for the SEM detector and the magnification was 15–300,000-fold. The operating pressure was 105 Pa vacuum and the voltage was 15.0 keV. For cross-section,

Table 3 Particle densities, flowability and mechanical properties of 710–1000 mm sized agglomerates grown from 0.5 L-sized tank installed with the various baffle types for 2 h. Baffle type

Carr’s Index

Particle density (g/cm3)

Fracture force (N)

Particle strength (MPa)

US European Kawashima

9.08  1.89 8.09  1.64 9.74  2.71

0.75  0.09 0.83  0.10 0.98  0.06

0.057  0.003 0.072  0.003 0.136  0.008

0.102  0.006 0.128  0.005 0.239  0.013

P.Y. Lin et al. / International Journal of Pharmaceutics 495 (2015) 886–894

3. Results and discussion

Cumulative percent of mass larger (%)

Toluene was selected as the bridging liquid for the SA method because of the relatively good solubility of dimethyl fumarate in toluene near 51.5  1.5 mg of dimethyl fumarate/mL of toluene at 25
C. The US and European baffle types are more common than the Kawashima baffle type especially for the 4000 and 10,000 L-sized tanks in manufacturing. One of our concerns was whether spherical crystallization was robust enough to withstand the variations of baffle configuration and tank size during technology transfer and process scale-up. IR spectra, DSC scans and PXRD patterns in Figs. S4–S6 all indicated the chemical identity, polymorphism and crystallinity of the dimethyl fumarate, respectively, would not be altered by the solvents used in the SA process as compared with the data of the commercial standard. The commercial dimethyl fumarate powders could hardly flow, and the ball-milled powders also displayed a poor flowability with a Carr’s index of 38.6  7.5. And yet, the flowability of dimethyl fumarate had been significantly improved by the SA process. Sieving the just-filtered, wet agglomerates could reveal the original morphology of agglomerates in the suspension (Fig. S7). However, their particle size distributions could not be obtained by weight due to the presence of residual solvents. This problem could be overcome if agglomerates were completely oven dried and their morphology was assumed to remain unchanged after the drying process (Fig. S7). In a stirred tank, the added bridging liquid was floating on the surface of the suspension initially and then being dragged into the suspension by turbulence. For all the baffle types at maturation time of 1 h, the agglomerates grown from the 0.5 L-sized tank gradually shifted toward to larger particle size range as the BSR increased from 0.4 to 0.6 (Fig. S8). For examples, the highest weight fraction was moved from the size cut of <250 mm to the size cuts of 710–1000 mm and 1000–1400 mm in Fig. S8a, and to the size cuts of 500–710 mm and 710–1000 mm in Fig. S8b. In Fig. S8c, the weight

100

0.5 L 2L 10 L

90 80 70 60 50 40 30 20 10 0

>2

50

>3

55

>5

00

>7

10

>1

000

>1

400

Particle size range ( m) Fig. 3. Cumulative oversize distribution curves for dimethyl fumarate agglomerates grown from (a) 0.5 L-sized, (b) 2 L-sized, and (c) 10 L-sized tanks installed with the Kawashima baffle type at BSR = 0.5, and maturation time = 2 h.

fraction of the size cut of <250 mm decreased from 31% to 8% and the ones for the size cuts of >710 mm were increased to more than 50% all together. As a result, the BSR had shown a direct correlation with the agglomerate size (Blandin et al., 2000). For all the baffle types at the same BSR of 0.4 in Figs. S8a–c, more than one half of the weight fractions were still smaller than 500 mm even with the increase of maturation time of 1–3 h. This suggested that the maturation time had been mainly associated with the level of completion in spherical agglomeration (Blandin et al., 2000). 3.1. Effects of baffle configuration The SA process was the interplay of events of particle-toparticle adhesion due the presence of bridging liquid, particle-toparticle collision, particle-to-impeller collision, particle-to-baffle collision, and the balance between cohesion and disruption (Blandin et al., 2000; Thati and Rasmuson, 2011, 2012). At the BSR of 0.5, the agglomerates grown from the 0.5 L-sized tank installed with the US baffle type were mainly distributed from 500 to 1000 mm, 355 to 1400 mm and 710 to 1400 mm in Fig. S8a when the maturation was 1, 2 and 3 h, respectively. As for the Kawashima baffle type in Fig. S8c, it mostly produced 355–710 mm, 500–1000 mm and 710–1400 mm-sized agglomerates at the maturation time of 1, 2 and 3 h, respectively. Apparently, the wider baffle width of Kawashima type could provide a higher frequency for collision and disruption resulting in smaller particle size distributions compared to the ones of the US type. Therefore, the Kawashima type baffle gave the relatively loosely-held agglomerates of smaller sizes when the BSR value was 0.6 and the maturation time was only 1 h in Fig. S8c. Noticeably, for the European baffle type and the BSR value of 0.5, the particle size distributions of the agglomerates did not shift too much and mostly centered around 355–710 mm even with the change in the maturation time. The constant distributions might be due to the balance achieved between cohesion and disruption at the earlier maturation time of 1 h. However, Fig. 1 clearly illustrated that the 27 cumulative oversize distribution curves of the agglomerates were different from each other. No two cumulative oversize distribution curves looked identical. Therefore, two important points can be made: (1) when BSR and

Weight of dimethyl fumarate (mg)

samples were prepared by cutting the 710–1000 mm agglomerate in half by a razor blade. All the samples were mounted on a carbon conductive tape (Prod. No. 16073, TED Pella Inc., California) and then sputter-coated with gold (Hitachi E-1010 ion spotter, Tokyo, Japan) with a thickness of about 6 nm. The discharge current used was about 0–30 mA and the vacuum was around 10 Pa.

891

120 100 80 60 40

0

0.5 L 2L 10 L

t50 = 19 min

20

0

10

20

30

40

50

60

70

80

90

100

Time (min) Fig. 4. Dissolution profiles of 710–1000 mm sized dimethyl fumarate agglomerates grown from 0.5, 2, and 10 L-sized tanks installed with the Kawashima baffle type at BSR = 0.5 and maturation time = 2 h.

892

P.Y. Lin et al. / International Journal of Pharmaceutics 495 (2015) 886–894

Table 4 Particle densities, flowability and mechanical properties of 710–1000 mm sized agglomerates grown from 0.5, 2 and 10 L-sized tanks installed with the Kawashima baffle type for 2 h. Tank size (L)

Carr’s Index

Particle density (g/cm3)

Fracture force (N)

Particle strength (MPa)

0.5 2 10

9.74  2.71 10.08  2.24 11.77  1.96

0.98  0.06 0.73  0.04 0.72  0.06

0.136  0.008 0.057  0.006 0.055  0.005

0.239  0.013 0.099  0.011 0.097  0.008

maturation time were constant, the particle size distribution was strongly dependent on the baffle type (Figs. 1 and S8), and conversely, (2) once the particle size distribution was altered by the change of a baffle type, the particle size distribution could not be reversed by simply adjusting BSR and the maturation time. The corresponding optical micrographs of agglomerates grown from the various baffle configurations of different size cuts in Fig. 1 were tabulated in Fig. S9. Fig. S9 showed that 355–500 mm sized agglomerates, 500–710 mm sized agglomerates and 710–1000 mm sized agglomerates were spherical and uniform in size. Interestingly, the other way to look at the cumulative oversize distribution curves is to image that the weight% of a given size cut of agglomerates was actually the productivity of that particular size cut of agglomerates in the batch. That particular agglomerates size was governed by the growth events of the agglomerate which was related to a family of specific traveling pathways in the tank. Those pathways were directed by the streamlines and the Stokes number for collisions (Snow et al., 1997). The number of possible and successful pathways from a certain hydrodynamics pattern could be affected by the baffle configuration (Kumaresan et al., 2005) when all other operating parameters were held constant. Noticeably, the dissolution profiles and the Carr’s indices of the same size cut of agglomerates grown from SA processes by using different baffle types did not change much as clearly shown in Fig. 2 with a 50% drug release time, t50, of 19 min, and Table 3 with around 9.0, respectively. This was because both dissolution occurred from the agglomerate surface (Maghsoodi, 2011) and flowability originated from the agglomerate Euclidean geometry were the functions of particle size (Fig. S9 and Table S1). However, the same size cut of agglomerates grown from SA processes displayed an increase in both fracture force from 0.057 to 0.072 to 0.136 N and particle strength from 0.102 to 0.128 to 0.239 MPa as the baffle configuration was changed from the US type to the European type to the Kawashima type, respectively, as indicated in Table 3. The particle mechanical properties depended less on the agglomerate size, but more so, on the internal structure of the agglomerates as reflected by particle density (Table 3). There was a direct correlation between density and fracture force or particle strength (Thati and Rasmuson, 2011). Installation of the baffles effectively destroyed the circular liquid pattern, and axial flows became much stronger as a result. With an increase in the baffle width from the US type to the European type to the Kawashima type (Table 1), the shear rate values and the turbulent dissipation rate would be increased giving a higher collision rate and more energetic collisions (Kawashima and Capes, 1974; Kumar, 2010), and the kneading action accompanied by collisions led to an increase in particle density from 0.75 to 0.98 g/cm3 and a rise in particle strength from 0.102 to 0.239 MPa, respectively, as shown in Table 3 (Thati and Rasmuson, 2011). It is necessary to take caution that the particle strength of the agglomerates was quite weak during the separation and drying process. For instance, the mechanical strength of the wet 710–1000 mm agglomerates prepared in the stirred tank installed with Kawashima baffle type, BSR of 0.5 and the maturation time of 2 h was measured to 0.07 N which was one-half the strength of the dried counterparts of 0.136 N. Therefore, a thick wet cake of agglomerates should be avoided.

3.2. Effects of tank size The significant shift in the particle size distributions of the agglomerates grown from 0.5, 2 and 10 L-sized tanks installed with the Kawashima baffle types were observed in Figs. 3 and S10. Plenty of ball-milled dimethyl fumarate powders did not assemble to form spherical agglomerates under the same operating conditions in 2 and 10 L-tanks. Fig. 3 shows that the three cumulative oversize distribution curves agglomerates grown from 0.5, 2 and 10 L-sized vessels installed with the same Kawashima baffle type looked very different given with all the other operating parameters were the same. The optical micrographs of agglomerates grown from the three various tank sizes of different size cuts in Fig. 3 are tabulated in Fig. S12. Fig. S12 shows that 355–500 mm, 500–710 mm and 710–1000 mm agglomerates were uniform in size. The degree of circularity (i.e., for a perfect sphere, circularity = 1.0) of a specific size cut, for instance the 710–1000 mm agglomerates, began to decrease from 0.84 to 0.74 to 0.57 as the tank size became larger as it went from 0.5 to 2 to 10 L. Fig. S13 displayed the SEM images of the ball-milled powders and the surface morphology and cross-section of dimethyl fumarate agglomerates prepared in 0.5, 2 and 10 L-sized vessels. The very fine milled dimethyl fumarate powders were irregular platelets (Fig. S13(a)) and the dimethyl fumarate agglomerates were made up of those irregular building blocks (Fig. S13(e)–(g)). The crosssections of those agglomerates showed that they were not hollow and their inner structures were filled with irregular platelets (Fig. S13(h)–(j)). Therefore, the spherical agglomerates were formed by the assembly of many fine milled dimethyl fumarate crystals in the suspension. The dissolution profiles and Carr’s indices of agglomerates from a given size cut, such as 710– 1000 mm agglomerates, did not change much upon scale up from 0.5 to 2 to 10 L as shown in Fig. 4 with a 50% drug release time, t50,

Fig. 5. Cumulative oversize distribution curves for dimethyl fumarate agglomerates grown from 2 and 10 L-sized tanks installed with the Kawashima baffle type at BSR = 0.5 and maturation time = 2 and 6 h.

P.Y. Lin et al. / International Journal of Pharmaceutics 495 (2015) 886–894

893

Table 5 Particle densities, flowability and mechanical properties of 710–1000 mm sized agglomerates grown from 2 L- and 10 L-sized tanks installed with the Kawashima baffle type at the BSR = 0.5 and the maturation time = 2 and 6 h. Tank size (L)

Maturation time (h)

Carr’s Index

Particle density (g/cm3)

Fracture force (N)

Particle strength (MPa)

2 2 10 10

2 6 2 6

10.08  2.24 11.34  1.53 11.77  1.96 5.90  1.43

0.73  0.04 0.84  0.05 0.72  0.06 0.76  0.02

0.057  0.006 0.069  0.002 0.055  0.005 0.089  0.001

0.099  0.011 0.107  0.003 0.097  0.008 0.137  0.002

Table 6 Relationships among tank size, impeller diameter, agitator speed and power. Tank size (L)

Impeller diameter (cm)

Stirring rate (rpm)

Power (hp)

Stirring rate (rpm)

Power (hp)

0.5 2 10 4000

3.6 5 7 110

600 450 350 30

1.03  104 2.25  104 5.68  104 0.34

600 600 600 600

1.03  104 5.32  104 2.87  103 2742.89

of 19 min, and Table 4 with around 10, respectively. However, Table 4 also indicated the decay in particle density from 0.98 to 0.73 to 0.72 g/cm3 and the deterioration of fracture force from 0.136 to 0.057 to 0.055 N, and particle strength from 0.239 to 0.099 to 0.097 MPa, respectively, of 710–1000 mm agglomerates as the tank size increased. All those results implied that as the tank size increased, the number of energetic collisions among agglomerates was decreased. This was probably due to the drop of the specific energy input into the tank according to Eq. (7), which decreased from 1.35 to 0.68 to 0.47 kJ/kg as the tank size increased from 0.5 to 2 to 10 L. Moreover, the growth events were dependent on the hydrodynamic pattern and the traveling distance. Upon scale up, even hydrodynamic pattern was nearly the same because of similar geometry. However, the traveling distance would increase significantly (Smith et al., 1990). This would drastically decrease the number of successful collisions for particle consolidation providing with the same amount of maturation time. The interior structure-related properties such as density and mechanical properties, and the completion of agglomerates governed by the number of successful collisions, might only be controlled in our current situation by prolonging the maturation time from 2 to 6 h as verified by Fig. 5 and Table 5. The weight% of spherical agglomerates larger than 250 mm increased as the maturation time increased from 2 to 6 h. The particle density, fracture force and particle strength of 710–1000 mm agglomerates prepared in a 2-L sized tank were increased from 0.73 to 0.84 g/cm3, from 0.057 to 0.069 N and from 0.099 to 0.107 MPa as the maturation time increased from 2 to 6 h, respectively. Similar trends were observed in a 10-L sized tank. The particle density, fracture force and particle strength of 710–1000 mm agglomerates were increased from 0.73 to 0.76 g/cm3, from 0.036 to 0.089 N and from 0.054 to 0.137 MPa, when the maturation time increased from 2 to 6 h, respectively. Obviously, increasing the maturation time was along the right track if a common stirred tank was used. A much higher density and stronger mechanical force might also be achieved by a faster impeller speed. However, the ultimate power consumption, P (hp) (McCabe et al., 2005): P¼

1:27 n3 D5 r 745:69

ð9Þ

where n = rotational speed (rev/s), D = impeller diameter (m),

r = water density = 1000 (kg/m3), would put a limit on the rate as

the tank size increased (Table 6). For a 4000 L-sized vessel, an agitator speed of 600 rpm would require a power of 2743 hp. Therefore, our future study will focus on the optimization of the

maturation time and the impeller speed for producing the same granular density and mechanical properties from the same size cut by sieving upon scale up. 4. Conclusions It was impossible to reproduce a desired particle size distribution for agglomerates in the SA process upon either baffle change or scale up in a common stirred tank simply by holding all other parameters such as solid loading, bridging liquid type, BSR and maturation time constant. However, the size-related dissolution behavior and flowability for the agglomerates of the same size cut could more or less remain unchanged. If that so, sieving after the SA process would be necessary. Agglomerates from different size cuts would be separated, classified and stored up for future use. The additional advantage of this post-treatment is that if agglomerates from any of the size cut does not meet a desired dissolution rate, this situation may be coped with by blending the agglomerates from two size cuts afterwards based on the mixing rules (Lee et al., 2008) instead of re-designing the entire SA process. The unwanted decay in the granular densities and mechanical properties of the same size cut upon baffle change or scale up could be simply remedied by prolonging the maturation time of the SA process. The maturation time and the impeller speed in a common stirred tank will be optimized in the future. To maintain the same particle size distribution upon scale up, agglomerates from a common stirred tank could have been separated and classified first by sieving and then blended together artificially according to the desired weight% of each cut in the particle size distribution. Acknowledgments This research was supported by the grants from the Ministry of Science and Technology of Taiwan, ROC (MOST 104-2221-E-008070-MY3). We thank Ms. Jui-Mei Huang for assistance with DSC, Ms. Shew-Jen Weng for PXRD and Ms. Ching-Tien Lin for SEM. All with the Precision Instrument Center at National Central University are gratefully acknowledged. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j. ijpharm.2015.09.056.

894

P.Y. Lin et al. / International Journal of Pharmaceutics 495 (2015) 886–894

References Amaro-González, D., Biscans, B., 2002. Spherical agglomeration during crystallization of an active pharmaceutical ingredient. Powder Technol. 128, 188–194. Bates, R.L., Fondy, P.L., Fenic, J.G., 1966. Impeller characteristics and power. In: Uhl, V. W., Gray, J.B. (Eds.), Mixing, Theory and Practice. Academic Press, New York. Blandin, A.-F., Rivoire, A., Mangin, D., Klein, J.-P., Bossoutrot, J.-M., 2000. Using in situ image analysis to study the kinetics of agglomeration in suspension. Part. Part. Syst. Charact. 17, 16–20. Blandin, A.F., Mangin, D., Rivoire, A., Klein, J.P., Bossoutrot, J.M., 2003. Agglomeration in suspension of salicylic acid fine particles: influence of some process parameters on kinetics and agglomerate final size. Powder Technol. 130, 316– 323. Blandin, A.F., Mangin, D., Subero-Couroyer, C., Rivoire, A., Klein, J.P., Bossoutrot, J.M., 2005. Modelling of agglomeration in suspension: application to salicylic acid microparticles. Powder Technol. 156, 19–33. Bos, A.S., Zuiderweg, F.J., 1987. Size of agglomerates in batch wise suspension agglomeration. Chem. Eng. Res. Des. 65, 187–194. Byn, S., Morris, K., Comella, S., 2005. Reducing time to market with a science-based management strategy. Pharm. Technol. 46–56. Chikhale, E., 2012. Clinical Pharmacology and Biopharmaceutics ReviewCenter for Drug Evaluation and Research Food and Drug Administration. . February http:// www.accessdata.fda.gov/drugsatfda_docs/nda/2013/ 204063Orig1s000ClinPharmR.pdf. Chow, A.H.L., Leung, M.W.M., 1996. A study of the mechanisms of wet spherical agglomeration of pharmaceutical powders. Drug Dev. Ind. Pharm. 22, 357–371. Cui, L.-C., Wu, Y., Shen, X.-R., Xu, X.-Z., 1999. A LDA study of the turbulent flow field in a baffled vessel generated by a PTD and a PTU. J. Hydrodyn. B 4, 52–59. Faure, P., York, P., Rowe, R.C., 2001. Process control and scale-up of pharmaceutical wet granulation processes: a review. Eur. J. Pharm. Biopharm. 52, 269–277. Gordon, M.S., Chowhan, Z.T., 1990. Manipulation of naproxen particle morphology via the spherical crystallization technique to achieve a directly compressible raw material. Drug Dev. Ind. Pharm. 16, 1279–1290. Helmy, I.M., Azim, A.M.A., 2012. Efficacy of ImageJ in the assessment of apoptosis. Diagn. Pathol. 7, 15–21. Jarvis, L.M., 2014. The year in new drugs. C&EN 92, 10–13. Kawashima, Y., Capes, C.E., 1974. An experimental study of the kinetics of spherical agglomeration in a stirred vessel. Powder Technol. 10, 85–92. Kawashima, Y., Furukawa, K., Takenaka, H., 1981a. The physicochemical parameters determining the size of agglomerate prepared by the wet spherical agglomeration technique. Powder Technol. 30, 211–216. Kawashima, Y., Takagi, H., Takenaka, H., 1981b. Wet spherical agglomeration of binary mixtures. II. Mechanism and kinetics of agglomeration and the crushing strength of agglomerates. Chem. Pharm. Bull. 29, 1403–1409. Kawashima, Y., Kurachi, Y., Takenaka, H., 1982. Preparation of spherical wax matrices of sulfamethoxazole by wet spherical agglomeration technique using a CMSMPR agglomerator. Powder Technol. 32, 155–161. Kawashima, Y., Cui, F., Takeuchi, H., Niwa, T., Hino, T., Kiuchi, K., 1994. Improvements in flowability and compressibility of pharmaceutical crystals for direct tabletting by spherical crystallization with a two-solvent system. Powder Technol. 78, 151–157. Kawashima, Y., Imai, M., Takeuchi, H., Yamamoto, H., Kamiya, K., Hino, T., 2003. Improved flowability and compactibility of spherically agglomerated crystals of ascorbic acid for direct tableting designed by spherical crystallization process. Powder Technol. 130, 283–289.

Kova9 ci9c, B., Vre9 cer, F., Planinšek, O., 2012. Spherical crystallization of drugs. Acta Pharm. 62, 1–14. Kumar, B., 2010. Energy dissipation and shear rate with geometry of baffled surface aerator. Chem. Eng. Res. Bull. 14, 92–96. Kumaresan, T., Nere, N.K., Joshi, J.B., 2005. Effect of internals on the flow pattern and mixing in stirred tanks. Ind. Eng. Chem. Res. 44, 9951–9961. Lee, T., Hsu, F.B., 2007. A cross-performance relationship between Carr’s index and dissolution rate constant: the study of acetaminophen batches. Drug Dev. Ind. Pharm. 33, 1273–1284. Lee, H.L., Lee, T., 2015. Direct co-crystal assembly from synthesis to cocrystallization. CrystEngComm doi:http://dx.doi.org/10.1039/c5ce01205h. Lee, D.C., Webb, M.L., 2003. Pharmaceutical Analysis. In: Warman, M., Hammond, S. (Eds.), Process Analysis in the Pharmaceutical Industry. Blackwell Publishing Ltd., Oxford, pp. 324–356. Lee, T., Hou, H.J., Hsieh, H.Y., Su, Y.C., Wang, Y.W., Hsu, F.B., 2008. The prediction of the dissolution rate constant by mixing rules: the study of acetaminophen batches. Drug Dev. Ind. Pharm. 34, 522–535. Lee, T., Su, Y.C., Hou, H.J., Hsieh, H.Y., 2010a. Spherical crystallization for lean soliddose manufacturing (Part I). Pharm. Technol. 34, 72–75. Lee, T., Su, Y.C., Hou, H.J., Hsieh, H.Y., 2010b. Spherical crystallization for lean soliddosage manufacturing (Part II). Pharm. Technol. 34, 88–103. Lee, T., Chen, H.R., Lin, H.Y., Lee, H.L., 2012. Continuous co-crystallization as a separation technology: the study of 1:2 co-crystals of phenazine-vanillin. Cryst. Growth Des. 12, 5897–5907. Maghsoodi, M., 2011. Effect of process variable on physicomechanical properties of the agglomertaes obtained by spherical crystallization technique. Pharm. Dev. Technol. 16, 474–482. McCabe, W.L., Smith, J.C., Harriott, P., 2005. Agitation and mixing of liquid. In: McCabe, W.L., Smith, J.C., Harriott, P. (Eds.), Unit Operations of Chemical Engineering. McGraw-Hill Inc., New York, pp. 244–293. Murov, S., 1997. MSDS compilations (Section II-B-4 of the Chemistry Webercises Directory). Available: http://murov.info/orgsolvents.htm. Osborne, J.D., Sochon, R.P.J., Cartwright, J.J., Doughty, D.G., Sano, A., Kuriki, T., Kawashima, Y., Takeuchi, H., Hino, T., Niwa, T., 1990. Particle design of tolbutamide by the spherical crystallization technique. III. Micromeritic properties and dissolution rate of tolbutamide spherical agglomerates prepared by the Quasi-Emulsion Solvent Diffusion Method and the solvent exchange method. Chem. Pharm. Bull. 38, 733–739. Smith, G.W., Tavlarides, L.L., Placek, J., 1990. Turbulent flow in stirred tanks: scale-up computations for vessel hydrodynamics. Chem. Eng. Commun. 93, 49–73. Snow, R.H., Allen, T., Ennis, B.J., Litster, J.D., 1997. Size reduction and size enlargement. In: Ennis, B.J., Litster, J.D. (Eds.), Perry’s Chemical Engineers’ Handbook. McGraw-Hill, New York, pp. 1823–1912. Subero-Couroyera, C., Mangina, D., Rivoireb, A., Blandinc, A.F., Kleina, J.P., 2006. Agglomeration in suspension of salicylic acid fine particles: analysis of the wetting period and effect of the binder injection mode on the final agglomerate size. Powder Technol. 161, 98–109. Thati, J., Rasmuson, Å.C., 2011. On the mechanisms of formation of spherical agglomerates. Eur. J. Pharm. Sci. 42, 365–379. Thati, J., Rasmuson, Å.C., 2012. Particle engineering of benzoic acid by spherical agglomeration. Eur. J. Pharm. Sci. 45, 657–667. Varshosaz, J., Tavakoli, N., Salamat, F.A., 2011. Enhanced dissolution rate of simvastatin using spherical crystallization technique. Pharm. Dev. Technol. 16, 529–535. Zweitering, Th.N., 1958. Suspension of solid particles in liquids by agitators. Chem. Eng. Sci. 8, 244–253.