European Journal of Pharmaceutics and Biopharmaceutics xxx (2014) xxx–xxx

Contents lists available at ScienceDirect

European Journal of Pharmaceutics and Biopharmaceutics journal homepage: www.elsevier.com/locate/ejpb

Research paper

In-line spatial filtering velocimetry for particle size and film thickness determination in fluidized-bed pellet coating processes Friederike Folttmann, Klaus Knop, Peter Kleinebudde, Miriam Pein ⇑ Heinrich-Heine-University Duesseldorf, Institute of Pharmaceutics and Biopharmaceutics, Duesseldorf, Germany

a r t i c l e

i n f o

Article history: Received 2 September 2014 Accepted in revised form 9 October 2014 Available online xxxx Keywords: Spatial filtering velocimetry In-line particle size determination Coating thickness Process analytical technology Pellet coating Wurster system Functional coating Real time monitoring

a b s t r a c t A spatial filtering velocimetry (SFV) probe was applied to monitor the increase in particle size during pellet Wurster coating processes in-line. Accuracy of the in-line obtained pellet sizes was proven by at-line performed digital image analysis (DIA). Regarding particle growth, high conformity between both analytical methods (SFV/DIA) was examined for different coating processes. The influence of ring buffer size and the process of filling the buffer were investigated. With buffer sizes of 30,000–50,000 particles best results were obtained in this study. Investigated process parameters, such as inlet air volume and spray rate, had different effects on the impact of the SFV probe. While the particle rate (the number of particles detected by the SVF probe per second) was highly dependent on the inlet air volume, different spray rates of up to ±1 g/min did not affect the detected particle growth. Artefacts and delays in SFV particle sizing appeared especially at the beginning of the coating processes. The slope of the particle growth during the final spraying period was therefore used to determine coating thickness. Ó 2014 Elsevier B.V. All rights reserved.

1. Introduction Multiple unit dosage forms often consist of spherical drug containing pellets filled into capsules or compressed into tablets. By coating pellets, enteric, sustained or controlled release, taste masking, improved stability or esthetic appearance [1] can be achieved. Pellet coating processes are usually performed in fluidized bed coaters [2,3]. Different types of fluid bed equipment exist for batch processes, such as top spray, bottom spray or rotor with tangential spray system. Most frequently used in pharmaceutical industry is the so called Wurster coater, which is an insert bottom spray coater. This system has been applied in the food industry [1,4], but was originally developed as pharmaceutical technique to coat powder particles, granules, tablets or capsules [5,6]. Due to the controlled particle circulation, which increases the drying rate and reduces undesirable agglomeration during coating, it is an efficient batch fluid bed coater [7]. The film thickness and the coating uniformity strongly affect the properties of coated pellets [8,9]. In different studies image analysis was applied to evaluate the film characteristics, such as film thickness or integrity [10–17]. The often applied dynamic ⇑ Corresponding author: Institute of Pharmaceutics and Biopharmaceutics, Heinrich-Heine-University, Universitaetsstrasse 1, 40225 Duesseldorf, Germany. Tel.: +49 211 811422; fax: +49 211 8114225. E-mail address: [email protected] (M. Pein).

image analysis measures the particle size of a high number of particles, which are detected as a particle flow passing through a measuring field [18–21]. It was observed that the DIA method can directly measure the increase in coating thickness down to 2% of added polymer coat weight [19]. The method was even discussed to serve as surrogate dissolution test for coated pellets [20]. In-line pellet sizing during fluid bed coating processes by image analysis based methods have recently been introduced [22,23]. However, a high optical contrast between the core and the coating material is necessary to successfully determine the film thickness by these in-line methods. The modified spatial filter velocimetry (SFV) is a technique that provides chord length values of moving particles by calculating their velocity and time of flight while passing through a laser array. To enable particle size determination in-line, this technique can be implemented into a stick probe of a manageable size. The modified SFV technique relies on the conventional velocimetry by a fiber optical spatial filtering velocimeter. Aizu and Asakura classified spatial filtering velocimeter based on their configuration in four typical groups: the transmission grating type, the detector type, the spectral grating type and the optical fiber type [24]. The optical fiber type was described to be beneficial due to the flexibility and stability of the optical and mechanical system. The SFV technique was further modified to determine the particle size by a technique called fiber optical spot scanning (FSS) [25,26]. The time of flight of a moving particle passing through a single optical fiber is observed

http://dx.doi.org/10.1016/j.ejpb.2014.10.004 0939-6411/Ó 2014 Elsevier B.V. All rights reserved.

Please cite this article in press as: F. Folttmann et al., In-line spatial filtering velocimetry for particle size and film thickness determination in fluidized-bed pellet coating processes, Eur. J. Pharm. Biopharm. (2014), http://dx.doi.org/10.1016/j.ejpb.2014.10.004

2

F. Folttmann et al. / European Journal of Pharmaceutics and Biopharmaceutics xxx (2014) xxx–xxx

by the FSS technique and the particle chord length is calculated from this single impulse time and the particle velocity. The ParsumÒ SFV probe is equipped with a disperser which encloses the measuring gap, dilutes the incoming particle stream and centers the particle flight path [27]. However, the particle trajectory can still be positioned randomly in relation to the fixed single fiber. The distribution obtained by the SFV probe is thus a chord length distribution (CLD) and the chord length is a random cut across the particle. Considering this fact, Petrak calculated the possible chord density distribution detected by the SFV method on narrow-sized glass beads and found a good agreement between the calculated and the measured distribution [25]. In another study the characteristics of the chord length were observed and it was pointed out that a CLD does not clearly reflect the underlying particle size distribution (PSD) [28]. Therefore, a mathematical approach to transform the SFV-CLD into a PSD was performed by Fischer et al. [29]. The individual values detected by the SFV probe are written continuously into the flexible storage system. It was mentioned in the literature that the selected size of this particle ring buffer influences the latency of the probe and observations on the optimal buffer size were made in preliminary investigations [29]. Plitzko and Dietrich focussed on the in-line SFV detection of pellet agglomeration during Wurster coating [30], while SFV has recently been established for in-line particle size monitoring in fluidized bed granulation [27,31–36]. The obtained SFV data improved the understanding of the impact of process variables by using model-based process control approaches, such as DoE, univariate and multivariate PLS. In these studies, the SFV probe was placed into a granule side-stream. Aim of the present work was to determine the increasing film thickness in-line by applying a SFV probe in fluidized bed coating processes. To prove reliability of the results (1) the accuracy of the in-line obtained particle sizes during the coating processes should be confirmed, (2) the influence of the particle buffer size on the results should be assessed and (3) effects on particle sizing based on process parameters, such as inlet air volume or spray rate should be evaluated. Finally, findings should be applied to present a suitable method to determine film thickness during coating (4).

2. Materials and methods 2.1. Core and coating materials HCT layered pellets (hydrochlorothiazide layered on CelletsÒ500 (IPC, Dresden, Germany) [37]) (Fig. 1c) black curve) or theophylline pellets (Temmler Ireland Ltd., Killorglin, Co Kerry, Ireland) (Fig. 1c) gray curve) were used as core materials. Basic butylated methacrylate copolymer (EudragitÒ EPO), Ammonio methacrylate copolymer type A (EudragitÒ RL 30D), Ammonio methacrylate copolymer type B (EudragitÒ RS 30D) and PlasACRYLÒ (a ready to use mixture of glyceryl monostearate, triethyl citrate, polysorbate 80 and water) were received by Evonik Industries AG (Darmstadt, Germany). Sodium lauryl sulfate and titanium dioxide were purchased from Caeser&Loretz GmbH (Hilden, Germany), stearic acid from Baerlocher (Lingen, Germany), triethyl citrate from Merck KGaA (Darmstadt, Germany) and talc from C.H. Erbsloeh (Krefeld, Germany). 2.2. Coating dispersions Coating dispersions for trial 1–4 were prepared based on the excipients summarized in Table 1. To prepare the dispersion of trial 1, water was heated up to 50 °C and 100 g less than the required amount was weight into the preparation vessel. Stearic acid and sodium lauryl sulfate were stirred into the warm water until the solution was clear. EudragitÒ EPO was added and the slurry was stirred, until a slightly yellow, light turbid solution emerged. Talc was added to the omitted 100 g of water (room temperature) and homogenized using an Ultra Turrax for 10 min. The talc water mixture was poured slowly to the turbid solution while stirring continuously. For trial 2, triethyl citrate and talc were homogenized in water for 10 min. The EudragitÒ RL30D and RS30D dispersions were combined at a four-to-six ratio and stirred. Both suspensions were combined by slowly pouring the first suspension into the EudragitÒ mixture while stirring gently. PlasACRYLÒ was shaken manually and stirred using a dissolver plate for the preparation of the dispersion for trial 3. The EudragitÒ RL30D and RS30D dispersions were combined and added to the PlasACRYLÒ suspension. Water

Fig. 1. Core material: (a) theophylline pellets, (b) HCT layered pellets (light microscopic pictures), (c) volume density distributions (q3 [%/lm]) measured by the dynamic image analysis. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Table 1 Composition of the coating dispersions, quantity based on the solid fraction [% (w/w)] is given.

Trial Trial Trial Trial

1 2 3 4

Core

EudragitÒ EPO

EudragitÒ RL 30D

EudragitÒ RS 30D

Sodium lauryl sulfate

Stearic acid

Triethylcitrate

PlasACRYLÒ

Talc

Titanium dioxide

Total solid content (%)

HCT layered cellets Theophyllin pellets Theophyllin pellets Theophyllin pellets

57.13 – – –

– 25 34.79 29.64

– 37.49 52.17 44.46

5.73 – – –

8.6 – – –

– 6.26 4.33 3.71

– – 8.67 7.37

28.53 31.25 – –

– – – 34.8

15 20 20 23.48

Please cite this article in press as: F. Folttmann et al., In-line spatial filtering velocimetry for particle size and film thickness determination in fluidized-bed pellet coating processes, Eur. J. Pharm. Biopharm. (2014), http://dx.doi.org/10.1016/j.ejpb.2014.10.004

3

F. Folttmann et al. / European Journal of Pharmaceutics and Biopharmaceutics xxx (2014) xxx–xxx Table 2 Process parameters ((1): process data run 1, (2): process data run 2, (3): process data run 3). Process time (min) Trial 1 Forerun Warm up Coating Drying Trial 2 Forerun Warm up Coating Break Curing Drying Trial 3 Forerun Warm up Coating Break Curing Drying Trial 4 Forerun Warm up Coating Break Curing Drying

5 10 135 10 5 10 190 10 30 3 5 10 190 10 30 30 5 10 190 10 30 30

Spray rate (g/ min)

Inlet air volume (m3/h)

Inlet air temperature (°C)

Product temperature (°C)

Exhaust air humidity (% r.h.)

– –

100 100

0 40

22(1), 27(2), 22(3) 33

40(1), 25(2), 34(3) 33(1), 24(2), 29(3)

7.9 –

100 100

40 40

27–31 37

35(1), 30(2), 32(3) 20(1), 19(2), 18(3)

– –

100 100

0 40

23 31

44(1), 39(2), 42(3) 39(1), 33(2), 35(3)

7.9. – 7.9 –

100 100 100 100

40 67 67 0

28–31 up to 49 45–50 50

36(1), 19(1), 18(1), 11(1),

– –

100 100

0 45

20(1), 23(2), 22(3) 34

33(1), 33(2), 39(3) 27(1), 27(2), 35(3)

4.8 – 4.8 –

100 100 100 50–25

45 67 67 0

33–35 Up to 49 45–50 31

18(1), 22(2), 26(3) 11(1), 15(2), 17(3) 10(1), 16(2), 18(3) 9(1), 9(2), 12(3)

– –

100 100

0 45

21(1), 27(2), 24(3) 33

32(1), 19(2), 46(3) 28(1), 20(2), 42(3)

4.8 – 4.8. –

100 100 100 100

45 68 68 0

33–35 Up to 49 45–50 31

22(1), 13(1), 14(1), 16(1),

and triethyl citrate were poured into the EudragitÒ mixture while stirring gently. The dispersion of trial 4 was prepared equally to trial 3 with additional titanium dioxide. All spray suspensions were passed through a 0.5 mm sieve before spraying.

35(2), 21(2), 20(2), 14(2),

20(2), 14(2), 15(2), 18(2),

36(3) 17(3) 19(3) 13(3)

31(3) 21(3) 21(3) 24(3)

exposure to air. One additional sample was taken after the preheating period in trial 3 and trial 4. The samples weighed 18 g each in trial 1 (about 50,500 pellets), 16 g in trial 2 (about 16,900 pellets) and 12 g in trial 3 and trial 4 (about 12,600 pellets).

2.3. Coating and sampling procedure Per trial, three batches of 1000 g core material were coated under the same conditions (Table 2) and labeled as run 1–3. The corresponding coating dispersions were sprayed for 135 min (trial 1) or 190 min (trial 2–4), respectively. By adjusting a spray rate of approximately 7.9 g/min (trial 1,2) and 4.8 g/min (trial 3,4), this led to a theoretical weight gain of 15.9% (w/w) in trial 1, 30.0% (w/w) in trial 2, 18.2% (w/w) in trial 3 and 21.4% (w/w) in trial 4. At the time of sampling during the spraying period, the sprayed amount of dispersion was weighed and the actual spray rate was calculated. All coating runs were performed in a laboratory-scale fluid bed Wurster coater (GPCG1, Glatt, Binzen, Germany). The Wurster height was adjusted to 30 mm and the C 1-122-00088-3 bottom plate was applied. It was sprayed with 2 bar atomizing air pressure using a 0.8 mm (trial 1) or 1.2 mm (trial 2–4) nozzle. A peristaltic pump (Isomatec, IDEX Health&Science GmbH, Wertheim, Germany) and a silicone tube with 2 mm internal tube diameter were used for the liquid transport. Five samples were consecutively sampled every 27 (trial 1) or 38 min (trial 2–4), respectively, during the spraying period. In trial 1, the sixth sample was taken after the in process drying phase lasting 10 min and the last sample was taken after removing the pellets from the coater and 2 h of drying at 40 °C in a circulating air oven and afterward 10 h of storage with exposure to air. Within trial 2–4 the sixth sample was taken after the in-process curing phase lasting 30 min (10–20% r.h.) and one last sample after removing the pellets from the coater and 12 h of storage with

Fig. 2. Position of the SFV probe in a Glatt GPCG1 fluid bed equipment. Highlighted excerpt: Dispersing unit and measuring gap. Graphic: Autodesk Inventor, support by A. Madani and P. Regier, University Duisburg/Essen. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Please cite this article in press as: F. Folttmann et al., In-line spatial filtering velocimetry for particle size and film thickness determination in fluidized-bed pellet coating processes, Eur. J. Pharm. Biopharm. (2014), http://dx.doi.org/10.1016/j.ejpb.2014.10.004

4

F. Folttmann et al. / European Journal of Pharmaceutics and Biopharmaceutics xxx (2014) xxx–xxx

Fig. 3. (a and b) Volume density distributions (q3 [%/lm]) measured at-line (DIA) and in-line (SFV) during the coating processes of run 2 (trial 1, trial 2) and run 1 (trial 3, trial 4) (black curves: coating start, light gray curves: coating end). (c) Particle size (D50 [lm]) during coating process SFV vs DIA (black: DIA, gray: SFV, circles: run 1, quadrates: run 2, triangles: run 3) (buffer size 50,000 particles).

2.4. Modified spatial filtering velocimetry (SFV) The pellet size was measured in-line using a spatial filtering velocimetry probe (ParsumÒ IPP 70, Parsum GmbH, Chemnitz, Germany). The SFV probe should be installed at a location, where every particle size is present during the process. Therefore, the SFV probe was positioned in the lower part of the expansion chamber but still above the particle level with the fluidization turned off (at a height of 9.5 cm and a depth of 2 cm) (Fig. 3). The SFV probe was not removed from the coater between the coating runs. On the rod tip the stick probe is equipped with a disperser, which is removable attached to the stick and encloses the 10 mm wide measuring gap. The pellets pass through a small column inside the disperser (Fig. 2), and thus the maximum size of the incoming pellets is limited to 4 mm by the column´s diameter. The disperser contains a system of air-slides, which is connected to two main air channels located inside the stick. Pressurized air is passed through these air channels into the disperser to dilute the incoming particle stream, to center the particle flight path and to keep the sapphire windows clean from dust and coating liquid. The incoming air stream is reduced previously by an air supply unit to 3 l/min (external main channel) and 20 l/min (internal main channel). The SFV probe provides a set of light obscuration data produced by the particle passing through a laser array, which is formed from several optical fibers. The measured shading signals are transferred to the computer and based on these data a chord length value is

calculated. The results are stored in a circular storage system. The capacity of this ring buffer can be adjusted from approximately 1000 to 2 million individual particles. By the time the ring buffer is filled, the total number of classified particles maintained constant by storing new data starting at the beginning of the buffer and overwriting the old (FIFO-buffer; first in–first out). Thus, a moving, constantly updated chord length distribution (number size distribution Q0) is obtained. It is possible to monitor a fluidized bed process by four individual SFV probes using only one ParsumÒ software package as control module. The single probe, which was installed into the coater in the current study, was controlled by three out of four possible control channels. Therefore, three different probe settings could be defined for the single probe. Buffer sizes of 50,000, 100,000 and 200,000 particles (trial 1, trial 2) and 5000, 30,000, 50,000 particles (trial 3, trial 4) were examined using three different probe settings for each coating run. The buffer size was the only varied parameter in the three probe settings. Chord length distributions are displayed from the ring buffer automatically every second and saved every 10 s (data log interval). From the primary distributions (Q0 distribution), volume size distributions were calculated. In addition to the chord length the probe measures the velocity of every single pellet and a velocity distribution is calculated. By using the disperser, particles will be accelerated by the air stream inside the measurement channel. The width of the entire pellet velocity distribution should be smaller than the specified velocity

Please cite this article in press as: F. Folttmann et al., In-line spatial filtering velocimetry for particle size and film thickness determination in fluidized-bed pellet coating processes, Eur. J. Pharm. Biopharm. (2014), http://dx.doi.org/10.1016/j.ejpb.2014.10.004

F. Folttmann et al. / European Journal of Pharmaceutics and Biopharmaceutics xxx (2014) xxx–xxx

5

Fig. 4. Particle size (D50 [lm]) during process using different buffer sizes (black curves: run 1, gray curves: run 2, light gray curves: run 3).

measurement range. Otherwise pellets with a higher or lower velocity than the defined maximum or minimum velocity, will not be recorded. In this study the velocity measurement range was adjusted to 0–50 m/s. The probe technique determines continuously the relative particle concentration in the measuring volume (loading [%]). To compensate the overestimated particle sizes, which are due to coincidences between the pellets, a limiting value (maximum loading [%]) can be defined. By exceeding the limiting value, the current date will not be saved into the ring buffer. In this study the maximum loading was adjusted to 30% because especially in pulsating process stages, e.g. at the beginning of the preheating or spraying period, the reduced loading value is a mathematical tool to avoid errors in particle sizing which are due to particle coincidences. Another opportunity to keep the pellets size distribution free from clearly overestimated particles, is the limitation of the pellet size range (size measurement range [lm]). The size range was limited to 10–2200 lm in the current study. Ten sieve sizes were chosen for calculation in trial 1 (0 lm, 400 lm, 600 lm, 800 lm, 900 lm, 1000 lm, 1100 lm, 1200 lm, 1300 lm, 1400 lm, 1600 lm) and in trial 2–4 (0 lm, 200 lm, 400 lm, 600 lm, 650 lm, 700 lm, 750 lm, 800 lm, 850 lm, 900 lm, 1000 lm). 2.5. Dynamic image analysis (DIA) The obtained samples were analyzed at-line using a DIA system (n = 1). Image analysis was conducted on a CamsizerXTÒ (RetschÒ Technology GmbH, Haan, Germany) using the X-Fall module (free fall mode). The CamsizerXTÒ uses two LED stroboscope light

sources, which illuminate the dispersed particles to cast sharp shadows toward two cameras on the opposite side. The first camera with a high magnification captures small particles with a resolution of 1 lm/pixel. The second camera captures larger particles with a lower resolution but wider field of view with a resolution of 15 lm/pixel. The determined chord length (xcmin) is defined as width of the particle projection. It is the shortest of maximum chord length measured in 32 directions. The measuring directions are defined by the evaluation of distances between edge pixels of the particle shadow image. Thereby every edge pixel is connected with every edge pixel resulting in a large number of distances. They are summarized and classified into 32 main groups by the RetschÒsoftware. These groups are named ‘‘directions of measuring.’’ The volume density distribution calculated on the xcmin chord length was used for further evaluation.

3. Results and discussion 3.1. Accuracy of SFV particle size data Results imply that the SFV probe provides highly accurate in-line measured particle size. In Fig. 3 (column a and b), the volume density distributions (q3) measured at-line (DIA) and in-line (SFV) during the coating process are displayed based on one run of each trial. A broad particle size distribution with a tendency toward smaller particle sizes becomes apparent for the theophylline

Please cite this article in press as: F. Folttmann et al., In-line spatial filtering velocimetry for particle size and film thickness determination in fluidized-bed pellet coating processes, Eur. J. Pharm. Biopharm. (2014), http://dx.doi.org/10.1016/j.ejpb.2014.10.004

6

F. Folttmann et al. / European Journal of Pharmaceutics and Biopharmaceutics xxx (2014) xxx–xxx

of the consecutively sampled coated pellets was observed in-line and at-line. This observation was in good agreement with off-line obtained results by Heinicke et al. [18]. In this study the increasing pellet size during Wurster coating was determined by at-line dynamic image analysis (DIA) and a displacement of the D50 value of the uncoated core from the best fit straight line through the D50s of the consecutively sampled coated pellets was observed. The filling of valleys or pores in the core surface by the initial coat and the dissolving or subsequent smoothing of the surface material were considered as possible reasons [18]. Since the observed discrepancy between the results of the analytical methods in run 3, trial 2 might be due to a differing particle rate in the final spraying period, this was evaluated and will be discussed in Section 3.3. A good repeatability between run 1, 2 and 3 was achieved in all trials. During the 10 min drying period in trial 1, the size of the coated particles did not increase or decrease much. Detected size differences in the final period of trial 3 might be due to the differing exhaust air humidity during curing between the runs (Table 2, trial 3, curing period) and are less pronounced for the SFV data. 3.2. The influence of the circular buffer size

Fig. 5. (a) Inlet air volume [m3/h] and (b) particle rate [1/s] during process time [min] (black curves: run 1, gray curves: run 2, light gray curves: run 3).

pellets used in trial 2–4. In contrast, the applied HCT layered pellets in trial 1 are allocated almost optimal in a narrow distribution. In all trials the distributions shift to larger particle sizes due to particle growth during spraying. This is clearly visible in those trials, in which the higher spray rate of 7.9 g/min was adjusted (trial 1, trial 2). Both analytical methods showed a high agreement on these results. However, the volume distribution based on the in-line SFV-measurements in trial 1 shows a relatively high frequency of particles at particle sizes of 500 lm. The PSDs determined by the SFV probe are generally broader than those detected by the DIA method. For reasons of clarity, the median (D50) of the volume size distribution was taken for further evaluation and defined as particle size. The at-line measured DIA-particle sizes of the samples were compared to their corresponding SFV-values obtained at the same time in-line (Fig. 3c). As expected, due to the differing chord length evaluation methods, no absolute match of the particle sizes measured by the two analytical methods was proven. However, results displayed in Fig. 3c) clearly illustrate comparable slopes during the coating process. In trial 1, there is a higher probability to detect comparable particle sizes by both methods due to the spherical, narrow sized starter cores. In run 1, trial 1 a lag, in which no particle growth was recorded by the SFV probe, was observed in the beginning of the spraying period (Fig. 4, trial 1, see e.g. 50,000 buffer size). This observation could not be confirmed at-line (Fig. 3c), trial 1). Therefore it was assumed to be either an effect of the SFV-technique or due to temporary effects inside the pellet bed such as electrostatic interactions between particles. A lag time during the initial spraying period occurred as well in granulation trials [27]. In trial 3 and trial 4, an additional sample was therefore taken after the preheating period, and analyzed by DIA. A significant reduction of the D50 values of the uncoated cores to those

Each coating process was analyzed using three different probe settings with buffer sizes of 200,000, 100,000 and 50,000 particles in trial 1 and trial 2 and 50,000, 30,000 and 5 000 particles in trial 3 and trial 4 (Fig. 4). All other parameters (e.g. size range, sieve list, data log interval) were kept constant. The particle sizes increased almost linearly during the spraying period. The curing phase could clearly be distinguished from the spraying period in trial 2, using the probe setup with a buffer size of 50,000 particles. By adjusting higher buffer capacities, the process phases were not easily identifiable due to the delayed response. To evaluate the influence of the filling level of the particle buffer prior to the coating process, trials were conducted firstly without previously filled particle size buffer (trial 3, trial 4), where the buffer was filled with particles during the preheating phase. Secondly, in trial 1, run 2 and trial 2, run 2, 3, the buffer was kept filled with the last coated particles that have been estimated in the final stage of the previous run. Between the runs in trial 1 and trial 2 the SFV probe was not removed from the coater and it was not turned off after the preceding run. Thirdly, the buffer was filled up with uncoated pellets before the probe was installed into the coating chamber (trial 1, run 1 and trial 2, run 1). Therefore, particles were fed over a vibrating hopper into the measuring gap, comparable to the off-line setup by Närvänen et al. [38]. At last, the buffer was filled with particles in-line during an extended lead time of 53 min (trial 1, run 3). Least fluctuating particle growth during the beginning of the process was monitored, when the buffer was filled off-line with uncoated pellets before the probe was installed into the coating chamber. However, final results are comparable and thus, independent of the filling procedure. Oscillations, which could be due to fluctuations in the particle rate, were most pronounced, when using the smallest buffer size. The time to fill the particle buffer is dependent on the particle rate and the particle rate is highly dependent on the inlet air volume (see Section 3.4). By adjusting an inlet air volume of 100 m3/h and using 1 kg pellets, a particle buffer of 50,000 particles is completely filled during the preheating period, lasting 15 min. Consequently, the buffer size should be adapted to each individual process with regard to the inlet air volume, the particle loading and the preheating time. 3.3. The influence of the inlet air volume Results, which are summarized in Fig. 5, clearly illustrate that the number of particles measured per second by the SFV probe

Please cite this article in press as: F. Folttmann et al., In-line spatial filtering velocimetry for particle size and film thickness determination in fluidized-bed pellet coating processes, Eur. J. Pharm. Biopharm. (2014), http://dx.doi.org/10.1016/j.ejpb.2014.10.004

7

F. Folttmann et al. / European Journal of Pharmaceutics and Biopharmaceutics xxx (2014) xxx–xxx Table 3 Results of the film thickness determination on SFV and DIA particle size data. Run

Initial particle size (lm) (extrapolized intercept)

Final particle size (lm) (end of spraying period)

Estimated film thickness (lm)

SFV

SFV

SFV

DIA

Trial 1

1 2 3

746 750 750

DIA 746 746 750

785 786 787

793 793 792

20 18 19

24 24 21

Trial 2

1 2 3

1037 1049 1059

1023 1020 1023

1133 1133 1131

1110 1114 1127

48 42 36

43 47 52

Trial 3

1 2 3

1023 1020 1006

1037 1021 1029

1064 1058 1060

1080 1082 1084

21 19 27

22 31 27

Trial 4

1 2 3

983 993 993

1024 1025 1030

1047 1043 1041

1088 1087 1080

32 25 24

32 31 25

(particle rate) is highly dependent on the inlet air volume: the higher the fluctuation in the inlet air volume, the higher the fluctuation in the particle rate (see trial 1 vs trial 4). Minor differences in the particle rate (Fig. 5) between the runs of one trial did not result in differing final particle sizes (Fig. 4); for example visible in trial 2, where the particle rate of runs 3 shifted during the spraying period but the final particle size did not differ. In order to prove, if greater variations in the particle rate led to varying in-line particle sizes, a progressive filter blockage was induced in trial 3 run 2,3. Consequently, the inlet air volume and the particle rate decreased over process time (Fig. 5, trial 3). Final particle sizes (Fig. 4, trial 3, buffer size 50,000) in these runs were only slightly smaller (1060 lm) compared to those measured in run 1 (1064 lm). Accordingly, final particle sizes of trial 4, performed with an inlet air volume of 70 m3/h, were significantly smaller (1045 lm) compared to those of trial 3 despite the same applied spray rate of 4.8 g/min.

3.4. Film thickness determination In all trials the increase in particle size was measured with high conformity between the coating runs in the final spraying phase. This fact was taken into account, when particle growth of the last two-fifth of the spraying periods of each run was linearized and extrapolated to determine the film thickness. The according extrapolated intercepts provided the initial pellet sizes of the uncoated pellets and the coating thickness was then calculated based on the measured particle size data as follows:

Film thickness ¼

ðd coated pellets  d uncoated pelletsÞ 2

ð1Þ

where d is the particle size (D50). Results are summarized in Table 3. The film thicknesses generally agreed between the analytical methods, with the exception of trial 2, run 3 and trial 3, run 2. However a difference between the SFV and DIA film thicknesses of approximately 4 lm could be determined. Although in trial 2–4 the same cores were used, the estimated initial SFV particle sizes differ significantly, which might be due to the decreased particle rate and was already discussed in Section 3.4. The differing film thicknesses between the trials were due to differing solid contents and spray rates within the trials. The proportions of the obtained film thickness were in good agreement with the theoretical weight gain (15.9% trial 1, 30% trial 2, 18.2% trial 3, 21.4% trial 4).

DIA

4. Conclusion Spatial filtering velocimetry (SFV) could be proven to give reliable results when monitoring a coating process. Within different trials, a high conformity in the determination of the particle growth was given between SFV and the reference method. To monitor the different stages of a coating process – preheating, spraying, drying/curing – a particle buffer size of not more than 50,000 particles should be chosen, dependent on the inlet air capacity, the material loading and the preheating time. Smoothest results for the initial process were observed when the ring buffer was completely filled with particles during the preheating period. However, final results were independent of the initial filling. While differences in the applied spray rate had minor influence on the detected increase in particle size, it was highly effected by differing inlet air volume, which resulted in differing particle rates in the SFV-probe. However, further studies will evaluate, whether this effect was maybe due to segregation inside the coating chamber. Most reproducible results in the particle size increase were detected in the last two-fifth of spraying time of each coating run. By linearizing and extrapolating this part, it was possible to calculate the applied film thickness. Conflict of interest Authors declare no conflict of interest. Acknowledgements We thank Stefan Dietrich (Parsum GmbH Chemnitz) for providing the Parsum IPP 70 probe used in this study. He gave useful information about the handling and data evaluation. Furthermore, we acknowledge Abdelkader Madani and Peter Regier (University Duisburg/Essen) for their support using the Autodesk Inventor graphic software. We also would like to thank the Evonik Industries for kindly donating the coating polymers. References [1] K. Dewettinck, A. Huyghebaert, Fluidized bed coating in food technology, Trends Food Sci. Technol. 10 (1999) 163–168. [2] D. Jones, Air suspension coating for multiparticulates, Drug Dev. Ind. Pharm. 20 (1994) 3175–3206. [3] M. Wesseling, R. Bodmeier, Influence of plasticization time, curing conditions, storage time, and core properties on the drug release from aquacoat-coated pellets, Pharm. Dev. Technol. 6 (2001) 325–331. [4] S.R.L. Werner, J.R. Jones, A.H.J. Paterson, R.H. Archer, D.L. Pearce, Airsuspension particle coating in the food industry: Part I – state of the art, Powder Technol. 171 (2007) 25–33.

Please cite this article in press as: F. Folttmann et al., In-line spatial filtering velocimetry for particle size and film thickness determination in fluidized-bed pellet coating processes, Eur. J. Pharm. Biopharm. (2014), http://dx.doi.org/10.1016/j.ejpb.2014.10.004

8

F. Folttmann et al. / European Journal of Pharmaceutics and Biopharmaceutics xxx (2014) xxx–xxx

[5] F.N. Christensen, P. Bertelsen, Qualitative description of the Wurster-based fluid-bed coating process, Drug Dev. Ind. Pharm. 23 (1997) 451–463. [6] H.S.P. Hall, R.E. Pondell, The Wurster process, in: Controlled Release Technologie: Methods, Theory and Applications, CRC Press Inc., Boca Raton, Florida, USA, 1980, pp. 133–154. [7] E. Teunou, D. Poncelet, Batch and continuous fluid bed coating – review and state of the art, J. Food Eng. 53 (2002) 325–340. [8] N. Haddish-Berhane, C. Nyquist, K. Haghighi, C. Corvalan, A. Keshavarzian, O. Campanella, J. Rickus, A. Farhadi, A multi-scale stochastic drug release model for polymer-coated targeted drug delivery systems, J. Control. Release 110 (2006) 314–322. [9] L. Ho, Y. Cuppok, S. Muschert, K.C. Gordon, M. Pepper, Y. Shen, F. Siepmann, J. Siepmann, P.F. Taday, T. Rades, Effects of film coating thickness and drug layer uniformity on in vitro drug release from sustained-release coated pellets: a case study using terahertz pulsed imaging, Int. J. Pharm. 382 (2009) 151–159. [10] M. Andersson, B. Holmquist, J. Lindquist, O. Nilsson, K.-G. Wahlund, Analysis of film coating thickness and surface area of pharmaceutical pellets using fluorescence microscopy and image analysis, J. Pharm. Biomed. Anal. 22 (2000) 325–339. [11] P. Borgquist, G. Zackrisson, B. Nilsson, A. Axelsson, Simulation and parametric study of a film-coated controlled-release pharmaceutical, J. Controll. Release 80 (2002) 229–245. [12] S. Kucheryavski, K.H. Esbensen, A. Bogomolov, Monitoring of pellet coating process with image analysis-a feasibility study, J. Chemom. 24 (2010) 472– 480. [13] N. Laitinen, O. Antikainen, J. Yliruusi, Characterization of particle sizes in bulk pharmaceutical solids using digital image information, AAPS PharmSciTech. 4 (2003) [electronic resource]. [14] F.L. Laksmana, L.J. Van Vliet, P.J.A. Hartman Kok, H. Vromans, H.W. Frijlink, K. Van Der Voort Maarschalk, Quantitative image analysis for evaluating the coating thickness and pore distribution in coated small particles, Pharm. Res. 26 (2009) 965–976. [15] C.C. Larsen, J.M. Sonnergaard, P. Bertelsen, P. Holm, Validation of an image analysis method for estimating coating thickness on pellets, Eur. J. Pharm. Sci. 18 (2003) 191–196. [16] T. Nakano, H. Yuasa, Y. Kanaya, Suppression of agglomeration in fluidized bed coating. III. Hofmeister series in suppression of particle agglomeration, Pharm. Res. 16 (1999) 1616–1620. [17] M.A. Taylor, Quantitative measures for shape and size of particles, Powder Technol. 124 (2002) 94–100. [18] G. Heinicke, F. Matthews, J.B. Schwartz, The effects of substrate size, surface area, and density on coat thickness of multi-particulate dosage forms, Pharm. Dev. Technol. 10 (2005) 85–96. [19] G. Heinicke, J.B. Schwartz, Particle size distributions of inert spheres and pelletized pharmaceutical products by image analysis, Pharm. Dev. Technol. 9 (2004) 359–367. [20] G. Heinicke, J.B. Schwartz, Assessment of dynamic image analysis as a surrogate dissolution test for a coated multiparticulate product, Pharm. Dev. Technol. 11 (2006) 403–408. [21] F. Priese, B. Wolf, Development of high drug loaded pellets by Design of Experiment and population balance model calculation, Powder Technol. 241 (2013) 149–157.

[22] D. Markl, M. Zettl, G. Hannesschläger, S. Sacher, M. Leitner, A. Buchsbaum, J.G. Khinast, Calibration-free in-line monitoring of pellet coating processes via optical coherence tomography, Chem. Eng. Sci. (2014), http://dx.doi.org/ 10.1016/j.ces.2014.05.049. [23] N. Oman Kadunc, R. Šibanc, R. Dreu, B. Likar, D. Tomazˇevicˇ, In-line monitoring of pellet coating thickness growth by means of visual imaging, Int. J. Pharm. 470 (2014) 8–14. [24] Y. Aizu, T. Asakura, Principles and development of spatial filtering velocimetry, Appl. Phys. B Photophys. Laser Chem. 43 (1987) 209–224. [25] D. Petrak, Simultaneous measurement of particle size and particle velocity by the spatial filtering technique, Part. Part. Syst. Char. 19 (2002) 391–400. [26] D. Petrak, S. Dietrich, G. Eckardt, M. Köhler, In-line particle sizing for real-time process control by fibre-optical spatial filtering technique (SFT), Adv. Powder Technol. 22 (2011) 203–208. [27] S. Schmidt-Lehr, H.U. Moritz, K.C. Jürgens, Online-control of the particle size during fluid-bed granulation/evaluation of a novel laser probe for a better control of particle size in fluid-bed granulation, Pharm. Ind. 69 (2007) 478–484. [28] M. Li, D. Wilkinson, K. Patchigolla, Determination of non-spherical particle size distribution from chord length measurements. Part 2: Experimental validation, Chem. Eng. Sci. 60 (2005) 4992–5003. [29] C. Fischer, M. Peglow, E. Tsotsas, Restoration of particle size distributions from fiber-optical in-line measurements in fluidized bed processes, Chem. Eng. Sci. 66 (2011) 2842–2852. [30] M. Plitzko, S. Dietrich, Optimale Prozessführung im Wurster-Coating, CAV Chemie, Anlagen + Verfahren 5 (2010) 16–18. [31] A. Burggraeve, T. Van Den Kerkhof, M. Hellings, J.P. Remon, C. Vervaet, T. De Beer, Evaluation of in-line spatial filter velocimetry as PAT monitoring tool for particle growth during fluid bed granulation, Eur. J. Pharm. Biopharm. 76 (2010) 138–146. [32] A. Burggraeve, T. Van Den Kerkhof, M. Hellings, J.P. Remon, C. Vervaet, T. De Beer, Batch statistical process control of a fluid bed granulation process using in-line spatial filter velocimetry and product temperature measurements, Eur. J. Pharm. Sci. 42 (2011) 584–592. [33] H. Ehlers, A. Liu, H. Räikkönen, J. Hatara, O. Antikainen, S. Airaksinen, J. Heinämäki, H. Lou, J. Yliruusi, Granule size control and targeting in pulsed spray fluid bed granulation, Int. J. Pharm. 377 (2009) 9–15. [34] J. Huang, C. Goolcharran, J. Utz, P. Hernandez-Abad, K. Ghosh, A. Nagi, A PAT approach to enhance process understanding of fluid bed granulation using inline particle size characterization and multivariate analysis, J. Pharm. Innov. 5 (2010) 58–68. [35] T. Lipsanen, T. Närvänen, H. Räikkönen, O. Antikainen, J. Yliruusi, Particle size, moisture, and fluidization variations described by indirect in-line physical measurements of fluid bed granulation, AAPS PharmSciTech. 9 (2008) 1070– 1077. [36] T. Närvänen, T. Lipsanen, O. Antikainen, H. Räikkönen, J. Heinämäki, J. Yliruusi, Gaining fluid bed process understanding by in-line particle size analysis, J. Pharm. Sci. 98 (2009) 1110–1117. [37] L. Suhrenbrock, G. Radtke, K. Knop, P. Kleinebudde, Suspension pellet layering using PVA–PEG graft copolymer as a new binder, Int. J. Pharm. 412 (2011) 28– 36. [38] T. Närvänen, T. Lipsanen, O. Antikainen, H. Räikkönen, J. Yliruusi, Controlling granule size by granulation liquid feed pulsing, Int. J. Pharm. 357 (2008) 132–138.

Please cite this article in press as: F. Folttmann et al., In-line spatial filtering velocimetry for particle size and film thickness determination in fluidized-bed pellet coating processes, Eur. J. Pharm. Biopharm. (2014), http://dx.doi.org/10.1016/j.ejpb.2014.10.004