International Journal of Pharmaceutics 470 (2014) 8–14

Contents lists available at ScienceDirect

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

In-line monitoring of pellet coating thickness growth by means of visual imaging Nika Oman Kadunc a, *, Rok Šibanc b , Rok Dreu b , Boštjan Likar a,c , Dejan Tomaževi9 c a,c a

Sensum, Computer Vision Systems, Tehnološki park 21, Ljubljana SI-1000, Slovenia Department of Pharmaceutical Technology, Faculty of Pharmacy, University of Ljubljana, Ašker9 ceva cesta 7, Ljubljana SI-1000, Slovenia c Laboratory of Imaging Technologies, Faculty of Electrical Engineering, University of Ljubljana, Tržaška cesta 25, Ljubljana SI-1000, Slovenia b

A R T I C L E I N F O

A B S T R A C T

Article history: Received 14 March 2014 Received in revised form 25 April 2014 Accepted 29 April 2014 Available online 2 May 2014

Coating thickness is the most important attribute of coated pharmaceutical pellets as it directly affects release profiles and stability of the drug. Quality control of the coating process of pharmaceutical pellets is thus of utmost importance for assuring the desired end product characteristics. A visual imaging technique is presented and examined as a process analytic technology (PAT) tool for noninvasive continuous in-line and real time monitoring of coating thickness of pharmaceutical pellets during the coating process. Images of pellets were acquired during the coating process through an observation window of a Wurster coating apparatus. Image analysis methods were developed for fast and accurate determination of pellets’ coating thickness during a coating process. The accuracy of the results for pellet coating thickness growth obtained in real time was evaluated through comparison with an off-line reference method and a good agreement was found. Information about the inter-pellet coating uniformity was gained from further statistical analysis of the measured pellet size distributions. Accuracy and performance analysis of the proposed method showed that visual imaging is feasible as a PAT tool for in-line and real time monitoring of the coating process of pharmaceutical pellets. ã 2014 Elsevier B.V. All rights reserved.

Keywords: Coating thickness Pellets PAT Fluid-bed coating Visual imaging Quality assurance

1. Introduction Coated pharmaceutical pellets that can be enclosed in capsules or compressed into tablets are increasingly used as controlledrelease systems in the production of solid dosage forms (MayoPedrosa et al., 2007; Muschert et al., 2009). In comparison to single-unit dosage forms, pellets offer the advantage of a more predictable gastric transit time and drug absorption, which improves the therapeutic effects of a medical treatment (Bechgaard and Nielsen, 1978). In addition to drug release control, various coatings are being applied in order to mask taste, improve stability of the drug or to physically separate incompatible components of a dosage form. The coating thickness is the most important attribute of coated pellets; with uncontrolled thickness, the end product would not meet the anticipated functionality for sustained release. Coatings that are too thick could result in delayed disintegration or dissolution in case of immediate release dosage forms with protective coating, whereas coatings that are too thin will not assure desired functionality in case of delayed release dosage forms (Knop and Kleinebudde, 2013). Accordingly,

* Corresponding author. Tel.: +386 40 792 333; fax: +386 1 620 3359. E-mail address: [email protected] (N. Oman Kadunc). http://dx.doi.org/10.1016/j.ijpharm.2014.04.066 0378-5173/ ã 2014 Elsevier B.V. All rights reserved.

coating thickness must be precisely controlled to ensure the quality of solid dosage form products. Following the process analytical technology (PAT) guidance for pharmaceutical industry (US Food and Drug Administration, 2004), issued by the U.S. Food and Drug Administration (FDA), quality assurance in drug manufacturing is achieved through a systematic approach to pharmaceutical innovation. Development of new in-line, on-line and at-line PAT tools facilitates the implementation of quality by design (QbD) process development introduced in the ICH Q8 guidance, which states that “quality cannot be tested into products; it should be built-in or should be by design” (ICH, 2008; Yu, 2008; Hinz, 2006). In-line monitoring is important, not only for the product quality but also for a better understanding of the manufacturing process. When enough information is collected, PAT can ultimately lead to real-time release approaches wherein end product analytical testing is no longer necessary to ensure product quality, as the real time information gathered already provides a guarantee that the product is acceptable (Vogt and Kord, 2011). The analytical techniques in use for direct evaluation of the average coating thickness are tests that require the use of spectroscopic methods, often together with separation technique, to determine the amount of coating material in the obtained solution (Joseph and Raghavan, 2013; Duerst, 2013; Ciurczak,

N. Oman Kadunc et al. / International Journal of Pharmaceutics 470 (2014) 8–14

2013; Oberdier, 2013). The selection of the specific spectroscopic method is substance-dependent, so that different coating materials require different analytical techniques. The initial pellet core size information is still required and is usually obtained by manually measuring a pellet sample using the microscope or an off-line visual imaging method. The requirement for substance-dependent multivariate calibration methods in spectroscopic analytical techniques further contributes to the complexity of the design and implementation of a quantitative analysis. Consequently, the analytical techniques are timeconsuming and the results of the analyses are obtained after the coating process is completed. A number of analytical tools are being investigated with the aim of improving the accuracy of coating characteristics measurements of pharmaceutical pellets (Andersson et al., 2000a,b; Kennedy and Niebergall, 1997; Kucheryavski et al., 2010; Laksmana et al., 2009; Larsen et al., 2003; Lee et al., 2011; Liew et al., 2010; Možina et al., 2009; Možina et al., 2010; Naelapää et al., 2007; Närvänen et al., 2008; Podczeck et al., 1999), among them, several studies of visual inspection methods (Kennedy and Niebergall, 1997; Kucheryavski et al., 2010; Laksmana et al., 2009; Larsen et al., 2003; Liew et al., 2010; Možina et al., 2009; Možina et al., 2010; Närvänen et al., 2008; Podczeck et al., 1999; Heinicke and Schwartz, 2005). These visual inspection techniques, however, are not designed to give results in real time and under in-process conditions. Under the quality assurance and PAT premise that quality should be “designed into” processes through systematic use of control strategies to continuously ensure quality of the product, the interest for developing PAT tools for in-line, on-line and at-line monitoring of pellet coating process has increased over the last few years (Knop and Kleinebudde, 2013). Studies of in-line monitoring of pellet coating process have been conducted using near-infrared (NIR) spectroscopy (Andersson et al., 2000a,b; Lee et al., 2011) and a combination of NIR and Raman spectroscopy (Bogomolov et al., 2010). The development of new visual inspection techniques for inline and real time monitoring of the coating process could present many advantages as they offer means of non-invasive, nondestructive, fast and continuous automatic measurements of pellets’ coating thickness. In this study, a visual imaging system was designed and examined for automatic in-line and real time monitoring of pellet coating thickness during the coating process. Image analysis methods were developed for determination of pellets’ coating thickness in the in-process visibility conditions. Images of pellets were acquired in-line, in the course of a pellet coating process, through an observation window of a laboratory Wurster coating apparatus. The results for pellet coating thickness growth were obtained in real time and their accuracy was evaluated through comparison with a reference spectrophotometric analytical method. Information about the inter-pellet coating uniformity was gained from further statistical analysis of the measured pellet size distributions. The visual imaging system was also examined in terms of performance and computational efficiency to assess the feasibility of the system for real time monitoring of pellet coating process. 2. Materials and methods 2.1. Materials A narrow size fraction of microcrystalline cellulose pellets (800–1000 mm, Cellets1 700, HARKE, Germany) was obtained using test sieves (800 mm and 1000 mm) and sieve shaker (Retsch AS 200 basic, Germany). Pellets were coated with an aqueous solution consisting of 8.08% hydroxypropyl methylcellulose (Pharmacoat 606, Shin-Etsu,

9

Japan), 1.01% polyethylene glycol (PEG 6000, Fluka, Switzerland), and 1.094% tartrazine coloring agent (Sigma–Aldrich, Germany). Tartrazine has a proven stability within the slightly acidic to neutral pH range (Wade and Weller, 1994). Coating solution was prepared by adding the polyethylene glycol and hydroxypropyl methylcellulose to preheated water at 70  C and then stirred for 30 min. After leaving the solution to cool down to room temperature, coloring agent was admixed and the portion of evaporated water was replaced. In the experiment, 1473.7 g of the coating solution was applied to 1300 g of starting cores (approximately 2,500,000 pellets).

2.2. Coating of pellets For coating of the pellets, bottom spray fluidized bed coater with a draft tube (GPCG-1, Glatt GmbH, Dresden, Germany, Wurster insert) was used. The coating solution was sprayed using a binary nozzle with tip diameter of 0.8 mm and cap opening diameter of 2.50 mm. Coating process parameters were kept constant throughout the process: inlet airflow rate 130 m3/h; inlet air temperature 55  C; spray rate 10.5 g/min; atomizing air pressure 2.0 bar; gap between distribution plate and the Wurster insert bottom edge 20 mm. The spraying-coating process was performed for 140 min. For the purpose of off-line spectrophotometric measurements, periodic samplings of 2 g of pellets were made every 10 min.

2.3. Determination of coating thickness by spectrophotometric measurement of pellet samples Ten groups of ten pellets were randomly sampled from each pellet sample (2 g) taken during coating. Determination of the amount of coloring agent and corresponding amount of coating material deposited on the pellet cores was conducted via spectrophotometric measurements (UV spectrophotometer HP 8453, Hewlett-Packard, USA) at 425 nm. Groups of ten coated pellets were dispersed in a 6 mL of dihydrogen phosphate buffer with pH 6.5 in order to dissolve coating. After 4 h and prior to UV analysis, each test solution was subjected to filtration (Minisart RC 25 filters, Sartorius AG, Göttingen, Germany). Prior to the sample analysis, a calibration line was constructed (c = 002064*A), where coated pellets without tartrazine served as a blank (c – tartrazine concentration (mg/mL), A – absorbance). By knowing the coating composition and coating density, determined by helium pycnometer (1.272 g/cm3), the volume of the applied coating was calculated from the amount of the coloring agent. Calculated coating volumes for ten pellet groups, consisting of ten pellets, were averaged and expressed as per pellet value. Value of the average per particle coating volume was used in a sphere model, a pellet shape approximation, in order to calculate the time specific coating thickness and pellets diameter change. For the pellet diameter estimation, the equivalent spherical diameter (ESD) was used. ESD is calculated from the projected area of a segmented pellet (A) (Jennings and Parslow, 1988): sffiffiffiffiffiffiffiffiffiffiffiffi   A ESD ¼ 2 : (1)

p

Average ESD value for uncoated pellets, determined by an off-line reference visual imaging method (Možina et al., 2009), was used as initial input parameter (882.8 mm). Values of 10,000 pellet ESDs were used to determine the relative standard deviation (RSD) of the surface area for groups of 100 pellets. Obtained surface area RSD value (2.79%) was used to estimate the standard deviation of the coating thickness,

10

N. Oman Kadunc et al. / International Journal of Pharmaceutics 470 (2014) 8–14

Fig. 1. Schematic illustration of the experimental setup.

determined by spectrophotometric measurement, as one can assume that volume of the coating equals product of particle surface and coating thickness for small coating thickness values. 2.4. In-line visual imaging method 2.4.1. Image acquisition The in-line visual imaging system consisted of a monochrome (black and white) CMOS camera (MV1-D1312, Photonfocus, Switzerland); 50 mm lens, F-mount, 1:1.4D (Nikon Japan); a circular (150 mm in diameter) stroboscopic light source consisting of LEDs distributed equidistantly on a cylinder carrier; a personal computer for image processing. White light LEDs were used with emission in the visible light spectrum (wavelengths between 400 nm and 700 nm) and average luminous intensity of 1344 mcd. The visual inspection system was calibrated with a measurement of a caliber of known dimensions to determine the pixel size Spix. A value Spix = 22.8 mm/pix was obtained. Image acquisition was performed through an observation window of the Wurster coating apparatus. The imaging system was mounted on the window frame where pellets in a free fall movement toward the pellets bed (Christensen and Bertelsen, 1997) can be observed. Images of pellets were obtained continuously throughout the coating process with acquisition frame rate being 20 frames per second and exposure time 100 ms for each image. A schematic illustration of the experimental setup is depicted in Fig. 1.

Fig. 2. Illustration of the pellet segmentation procedure. (a) The lines (rays) that intersect the pellet’s border (white) and regression points (white dots) are plotted. (b) Plot of the regression points (white dots) and the resulting regression curve (white curve) representing the pellet’s border.

2.4.2. Image analysis In obtaining accurate results for coating thickness estimation, proper segmentation of the pellets is the most important part of the visual inspection method. Because of the characteristics of pellets’ visual appearance during coating, simple threshold-based or gradient-based methods that are commonly used for object segmentation are not appropriate for segmentation of pellets in the in-process conditions. Considering the pellet visibility characteristics, the appropriate segmentation procedure was developed. First, the central areas of the pellets were segmented using a threshold method and pellets’ center points were estimated. The segmented border points served as starting points for the second part of the segmentation. The second part of the segmentation was done by a procedure based on gradient edge detection method to find candidate points for pellet border (Davies, 2005). For each pellet, the radial gradient component gr was calculated along the lines connecting the center point and each starting point (Fig. 2(a)). The maximum of the radial gradient component gr on each ray was found and marked as a candidate for a pellet’s border point. In Fig. 2(a), the white lines represent the rays, where gr was calculated, and white dots the maximal gr found on each ray. Next, the result of the segmentation procedure was obtained for each pellet in form of a curve of m-th order that best fitted the border candidate points obtained in the previous step of the segmentation procedure. The curve representing a pellet’s border is written in polar form as a Fourier series: rð’Þ ¼ a0 þ

m X ðaj cosðj’Þ þ bj sinðj’ÞÞ;

(2)

j¼1

A robust iterative method of linear regression was employed to calculate the curves’ coefficients, where the border candidate points were used as regression points. The fitting was done using the iteratively reweighted least squares method (Holland and Welsch, 1977). The weight for each regression point was calculated in each iterative step with two contributing factors: the radial gradient component magnitude calculated in the regression point gr (weight value proportional to the gradient magnitude value) and the regression point’s residual distance dres – the Euclidean distance from the regression point to the current regression curve (Holland and Welsch, 1977). In Fig. 2(b), the white dots are the border candidate points and the white curve is the curve, obtained as the result of the robust iterative procedure. The curve represents the pellet border and is the result of the segmentation procedure for a single pellet. The segmentation procedure was done for every pellet found on the image by the initial threshold method. From all segmented pellets on an image, only pellets in focus were automatically selected and added to the final segmented pellet set using a squared gradient focus measure (Santos et al., 1997) along the rays calculated in the segmentation procedure. Also, pellets that lay on the border of the image were automatically eliminated from the segmented pellet set. The segmentation parameters’ values were set before the beginning of coating to give good results in terms of speed and accuracy of the segmented pellet borders. Threshold value of 0.6 (for equalized image histogram on the interval (0, 1)) was used in the first part of segmentation procedure, values nit = 3 for the number of iterations for the iterative procedure and m = 2 for the final order of the fitted curves were used. 2.4.3. Coating thickness determination Coating thickness can be calculated from the obtained segmentation results through determination of parameters

N. Oman Kadunc et al. / International Journal of Pharmaceutics 470 (2014) 8–14

11

describing pellets’ size. In our evaluation, we used the parameter ESD, obtained according to Eq. (1), for pellet size determination. For a set of N segmented pellets, the mean ESD value is then calculated as: ! N 1X ESD ¼ ESDk  s ; (3) ESD N k¼1

~ , the is related to s where standard error of the mean s ESD corresponding estimated standard deviation of the pellet set,pand ffiffiffiffi ~~= N. to N, the number of pellets measured in the set: s ¼s ESD With continuous measurement, pellets in a pellet set i are measured over a period of time Dt; we define measurement time ti of the pellet set i as the mean time from beginning to the end of acquisition of pellets in the set: ti=Dt/2. The estimated increase of coating thickness between two pellet sets i and j (CTEij) can be evaluated through calculating the mean of ESD values of the two sets of pellets, measured at different times of the coating process; the mean ESD of set i (measured at time ti) ESD i and the mean ESD of set j (measured at time tj) ESD j , where ti > tj. CTEij is then determined as (Možina et al., 2010):   (4) CTEij ¼ ð1=2Þ ESD i Þ  ESD j  2~~ij : The estimated standard error 2~~ij of the coating thickness estimation is calculated as: sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi    2  2 1 ~ s þ s : (5) 2~ij ¼ ESD i ESD j 2 For each segmented pellet, ESD was calculated from the segmentation curve. A pellet set was obtained every 3 min (Dt = 3 min) and the mean value of pellets’ ESDs was calculated from the acquired segmented pellet set. Finally, the CTEi0 was calculated for each set to give the results for the coating thickness growth through consecutive pellet sets, i.e., through the process of coating of pellets.

Fig. 4. Coating thickness estimation (CTE) as a function of coating process time: inline visual imaging measurements (black circles); off-line spectrophotometric measurements of pellet samples (gray squares).

3. Results and discussion 3.1. Image acquisition and computational efficiency In the in-line visual imaging measurements of the pellets, 1200 images of pellets in coating process were obtained each minute. Image analysis was performed in real time. With image acquisition frame rate of 20 fps, on average, approximately 90 pellets per second (5400 pellets per minute) were segmented and measured for size. An example of an acquired image together with the results of the segmentation procedure (the segmentation curves) is shown in Fig. 3. Computational efficiency of the image analysis was assessed after the coating process; together with image preprocessing and the elimination of out-of-focus and border pellets, the average computational time for each obtained pellet was 3.72 ms. This result shows that with a higher image acquisition frame rate, many more pellets could be measured in real time with this image analysis method. 3.2. Pellet coating thickness evaluation The results for coating thickness were calculated from the segmentation results and were observed in three minute time intervals (Dt = 3 min). The average number of pellets in a set was 16,545. Over the course of the coating process, 48 sets of pellets

Fig. 3. Example of an acquired image during the coating process together with the results of the segmentation. The white curves are the pellet borders obtained by the segmentation procedure.

Fig. 5. Relationship between the measured values for coating thickness obtained by off-line spectrophotometric measurements and in-line visual imaging along with the linear regression line.

12

N. Oman Kadunc et al. / International Journal of Pharmaceutics 470 (2014) 8–14

Fig. 6. Estimated probability density functions of measured ESD values for the three pellet sets measured in the coating process (uncoated pellets – at time t = 0, at time t = 70.5 min and at time t = 140 min).

were measured and CTE for each set i was calculated to give the result for the coating thickness at times ti. The results of the in-line visual imaging measurement of the coating thickness together with results of the off-line spectrophotometric measurements of pellet samples are presented in Fig. 4. The estimated standard errors for measurements of both methods are provided with the measured coating thickness estimation values in Fig. 4. To evaluate the obtained results for coating thickness growth, the measurements of the visual imaging method were compared with the reference spectrophotometric method. From Fig. 4, we see a good agreement between the in-line visual imaging results for coating thickness and off-line spectrophotometric measurements of collected pellet samples. Although the process times of measurement do not coincide completely (maximal difference is 1.5 min at t = 120 min), an assessment of the accuracy of the presented visual imaging method can be made by comparing the in-line results that lie closest to the times of pellet sampling. A comparison of values for CTE at process time t = 120 min, where maximal discrepancy between them is found, gives the difference of 0.64 mm. Comparing this to previously presented results for offline (Larsen et al., 2003) and on-line (Heinicke and Schwartz, 2005) visual imaging measuring techniques for pellet coating thickness determination, this result shows an excellent performance of the presented in-line visual imaging method. The estimated standard errors of the visual imaging method, obtained according to Eq. (5), only describe the statistical errors of the measured pellet sets and do not account for possible other sources of error or short-term process variability. The underlying

Fig. 8. (a) Estimated probability density functions of coating thickness distributions of pellet sets at t = 70.5 min and at t = 140 min; (b) relative standard deviation of coating thickness distributions for all pellet sets obtained in the coating process.

causes of the observed discrepancies between in-line visual imaging results for CTE and off-line spectrophotometric measurements, although small, could include not totally spherical shape of the pellet cores, which affects the accuracy of pellet size evaluation (Podczeck et al., 1999), or possible changes in film density through the process, which could affect the performance of the reference method, especially with larger coating thicknesses. The relationship between the measured values for coating thickness obtained by the two methods along with a linear regression line are shown in Fig. 5, where results for in-line imaging measured at times closest to the times of sampling of pellets were taken for comparison. The measure of agreement between the methods was assessed by evaluating their coefficient of determination R2; a strong positive correlation is observed and confirmed by the R2 value: 0.998. 3.3. Analysis of pellet size distributions

Fig. 7. Coating thickness gain for size percentiles for eight pellet sets obtained in the coating process.

The advantage of automatic digital visual imaging is that large number of measurements can be obtained in a relatively short time, yielding good statistical estimations of pellets in a coating process. The statistical variation in the measured pellet coating thickness can be observed through the distribution of measured ESD values of individual pellets. Kernel density estimations of measured ESD values were calculated (with Gaussian kernel and bandwidth value of 8 mm) to estimate the probability density functions of pellet size distributions. The estimated probability density functions for three pellet sets measured in the coating process (uncoated pellets – at time t = 0, at time t = 70.5 min and at time t = 140 min) are shown in Fig. 6. For the purpose of assessing the changes in the shape of the measured pellet size distributions through the course of the coating

N. Oman Kadunc et al. / International Journal of Pharmaceutics 470 (2014) 8–14

process, the mean shift has been eliminated by aligning the mean values to zero. This way, changes in the shape of the presented probability densities can be observed. In an ideal coating process, the change in the underlying pellet size distribution should only be in the mean shift and no skewness or modality change in the distribution’s shape should appear with the pellets’ coating thickness growth. Comparing the pellets size distributions in Fig. 6, some differences in the overall shape are observed. The differences in pellet size distributions may indicate a dependence of coating deposited on pellet size, which is often the case with coating of pellets with wide pellet size distributions (Luštrik et al., 2012). This dependence affects the inter-pellet coating uniformity, which is desired to be as small as possible (KuShaari et al., 2006; Knop and Kleinebudde, 2013; Turton, 2008). The differences in pellet size distributions were further examined by additional analysis of the distributions. Consecutive pellets’ measurements, divided into percentiles according to pellet size, were used to calculate the estimated coating thickness gain from the initial distribution of uncoated pellets for each size percentile separately. The coating thickness gain for each size percentile is shown in Fig. 7 for eight pellet sets obtained in the coating process. Pellets’ measurements were divided into 100-percentiles and sizes below 5-percentile and above 95-percentile were omitted because of the low accuracy of quantile (percentile) analysis in the regions of low frequency values of pellet size distributions. In an ideal case, where all pellets (regardless of initial size) gain the same amount of coating, the curves representing the pellet coating thicknesses as assessed through distribution percentiles would be straight horizontal lines. In Fig. 7, we see that the presented curves show a slight dependence of coating growth on pellets size; the changes are most pronounced between 70percentile and 80-percentile of the pellet size distributions and the effect becomes more apparent towards the end of the process. Inter-pellet coating variability of a pellet set can be assessed through the variability of values for coating thickness gain per percentiles. Values for coating thicknesses for a pellet set (represented as a curve in Fig. 7) can be gathered into a relative frequency histogram. The variability of coating thickness gain is directly related to the width of the histogram, i.e., its estimated standard deviation. Probability densities obtained through kernel density estimations are presented in Fig. 8(a) for CTE distributions of pellet sets at t = 70.5 min and at t = 140 min. Gaussian kernel and bandwidth of value 0.5 mm were used for the estimations. Comparing the presented probability densities in Fig. 8(a), we see that the coating thickness distribution is wider (coating thickness variability is larger) for the end point coating thickness distribution (at t = 140 min). The relative standard deviation (RSD) was calculated for each measured pellet set to evaluate the pellet coating variability through the course of the coating process. RSD was calculated as: RSD = s CTE/CTE, where s CTE is the standard deviation of the estimated coating thickness distributions for size percentiles and CTE is the coating thickness estimation of a pellet set obtained according to Eq. (4). In Fig. 8(b), the values obtained for the RSD of coating thickness estimations for all pellet sets obtained by the visual imaging method in the course of the coating process are presented. We see in Fig 8(b) that the RSD value drops to the plateau level of around 12%. This estimation of inter-pellet coating variability represents the minimal coating variability of a pellet batch, as possible transitions of pellets through different percentiles of the distribution are not accounted for in the results. Under the assumption, however, that there is no correlation between pellet size and other pellet parameters that might influence the coating speed of pellets in a given coating process, the analysis per quantiles (percentiles) gives a good assessment of inter-pellet coating variability.

13

4. Conclusion A visual imaging technique was designed and examined for automatic, in-line and real time monitoring of coating thickness of pharmaceutical pellets during the coating process. An image analysis method was developed that allows fast and accurate determination of pellets’ coating thickness in the in-process visibility conditions. Images of pellets were acquired during the coating process through an observation window of a laboratory Wurster coating apparatus and the results for coating thickness growth were obtained in real time. Their accuracy was evaluated through comparison with a reference spectrophotometric measurement of pellet samples obtained in the coating process and a good agreement between the two methods was found with R2 value of 0.998. The results obtained during the coating process were further examined and inter-pellet coating uniformity was assessed through statistical analysis of pellet size distributions. The accuracy of the obtained results together with computational efficiency of image analysis methods proves visual imaging feasible for in-line and real time monitoring of the pellet coating process. Further evaluation of the robustness and performance of the presented visual imaging method is subject of future work, including studies with different pellets, pellet size distributions and coating materials. The development of new visual inspection techniques for inline and real time monitoring of the coating process could present many advantages over the existing methods for determining coating thickness of pellets, as they offer means of non-invasive, non-destructive, fast and continuous automatic measurements of pellets’ coating thickness, which allows us the determination of the process end point. With large number of pellets measured, the information about pellet size distribution and inter-pellet coating uniformity can be obtained in real-time and the behavior observed throughout the process, allowing possible necessary changes of process parameters during the coating process. Another advantage of visual imaging methods is that the calibration of the measurement system is independent of the coating substance and other coating process parameters. The benefits of visual imaging methods for monitoring the coating process of pharmaceutical pellets offer new possibilities for optimizing and a better understanding of coating processes and, subsequently, of the end product quality assurance, which are the main goals of PAT guidance. Acknowledgments This work was supported by the Slovenian Research Agency under grant L2–4072, by Sensum, Computer Vision Systems, and by the European Union, European Social Fund. References Andersson, M., Folestad, S., Gottfries, J., Johansson, M.O., Josefson, M., Wahlund, K.G., 2000a. Quantitative analysis of film coating in a fluidized bed process by in-line NIR spectrometry and multivariate batch calibration. Anal. Chem. 72, 2099– 2108. doi:10.1021/ac990256r. Andersson, M., Holmquist, B., Lindquist, J., Nilsson, O., Wahlund, K.G., 2000b. Analysis of film coating thickness and surface area of pharmaceutical pellets using fluorescence microscopy and image analysis. J. Pharm. Biomed. Anal. 22, 325–339. doi:10.1016/s0731-7085(99)00289-7. Bechgaard, H., Nielsen, G.H., 1978. Controlled-release multiple-units and single-unit doses a literature review. Drug Dev. Ind. Pharm. 4, 53–67. doi:10.3109/ 03639047809055639. Bogomolov, A., Engler, M., Melichar, M., Wigmore, A., 2010. In-line analysis of a fluid bed pellet coating process using a combination of near infrared and Raman spectroscopy. J. Chemometr. 24, 544–557. doi:10.1002/cem.1329. Christensen, F.N., Bertelsen, P., 1997. Qualitative description of the Wurster-based fluid-bed coating process. Drug Dev. Ind. Pharm. 23, 451–463. doi:10.3109/ 03639049709148494.

14

N. Oman Kadunc et al. / International Journal of Pharmaceutics 470 (2014) 8–14

Ciurczak, E.W., 2013. Spectroscopic methods of analysis: near-infrared spectrometry, Encyclopedia of Pharmaceutical Technology. third ed. Taylor & Francis, pp. 3434–3439. doi:10.1081/E-EPT3-100200013. Davies, E.R., 2005. Machine Vision, Electronics & Electrical. Morgan Kaufmann. Duerst, M.D., 2013. Spectroscopic methods of analysis: near-infrared spectrometry, Encyclopedia of Pharmaceutical Technology. third ed. Taylor & Francis, pp. 3405–3418. doi:10.1081/E-EPT3-100200020. Heinicke, G., Schwartz, J.B., 2005. Particle size distributions of inert spheres and pelletized pharmaceutical products by image analysis. Pharm. Dev. Technol. 9, 359–367. doi:10.1081/PDT-200032996. Hinz, D.C., 2006. Process analytical technologies in the pharmaceutical industry: the Fda’s pat initiative. Anal. Bioanal. Chem. 384, 1036–1042. doi:10.1007/s00216005-3394-y. Holland, P.W., Welsch, R.E., 1977. Robust regression using iteratively reweighted least-squares. Commun. Statist. – Theor. Meth. 6, 813–827. doi:10.1080/ 03610927708827533. ICH, 2008. "International Conference on Harmonization of Technical Requirements for Registration of Pharmaceuticals for Human Use, Ich Harmonized Tripartite Guideline, Draft Step 4." Pharmaceutical Development Q8(R1). Jennings, B.R., Parslow, K., 1988. Particle size measurement – the equivalent spherical diameter. Proc. Royal Soc. London Series A 419, 1856. doi:10.1098/ rspa.1988.0100. Joseph, J.C., Raghavan, R., 2013. Spectroscopic methods of analysis: ultraviolet and visible spectrophotometry, Encyclopedia of Pharmaceutical Technology. third ed. Taylor & Francis, pp. 3460–3475. doi:10.1081/E-EPT3100000436. Kennedy, J.P., Niebergall, P.J., 1997. Preliminary assessment of an image analysis method for the evaluation of pharmaceutical coatings. Pharm. Dev. Technol. 2, 205–212. doi:10.3109/10837459709031440. Knop, K., Kleinebudde, P., 2013. Pat-tools for process control in pharmaceutical film coating applications. Int. J. Pharm. 457, 527–536 http://dx.doi.org/10.1016/j. ijpharm.2013.01.062. Kucheryavski, S., Esbensen, K.H., Bogomolov, A., 2010. Monitoring of pellet coating process with image analysis—a feasibility study. J. Chemometr. 24, 472–480. doi:10.1002/cem.1292. KuShaari, K.Z., Pandey, P., Song, Y., Turton, R., 2006. Monte carlo simulations to determine coating uniformity in a Wurster fluidized bed coating process. Powder Technol. 166, 81–90 http://dx.doi.org/10.1016/j.powtec.2006.05.001. Laksmana, F.L., Van Vliet, L., Kok, P.H., Vromans, H., Frijlink, H.W., Van der Voort Maarschalk, K., 2009. Quantitative image analysis for evaluating the coating thickness and pore distribution in coated small particles. Pharm. Res. 26, 965– 976. doi:10.1007/s11095-008-9805-y. Larsen, C.C., Sonnergaard, J.M., Bertelsen, P., Holm, P., 2003. Validation of an image analysis method for estimating coating thickness on pellets. Eur. J. Pharm. Sci. 18, 191–196. doi:10.1016/s0928-0987(02)00260-9. Lee, M.-J., Seo, D.-Y., Lee, H.-E., Wang, I.-C., Kim, W.-S., Jeong, M.-Y., Choi, G.J., 2011. In line nir quantification of film thickness on pharmaceutical pellets during a fluid bed coating process. Int. J. Pharm. 403, 66–72. doi:10.1016/j. ijpharm.2010.10.022.

Liew, C.V., Wang, L.K., Heng, P.W.S., 2010. Development of a visiometric process analyzer for real-time monitoring of bottom spray fluid-bed coating. J. Pharm. Sci. 99, 346–356. doi:10.1002/jps.21820. Luštrik, M., Dreu, R., Šibanc, R., Sr9ci9 c, S., 2012. Comparative study of the uniformity of coating thickness of pellets coated with a conventional Wurster chamber and a swirl generator-equipped Wurster chamber. Pharm. Dev. Technol. 17, 268–276. doi:10.3109/10837450.2010.531740. Mayo-Pedrosa, M., Alvarez-Lorenzo, C., Lacík, I., Martinez-Pacheco, R., Concheiro, A., 2007. Sustained release pellets based on poly(N-isopropyl acrylamide): matrix and in situ photopolymerization-coated systems. J. Pharm. Sci. 96, 93–105. doi:10.1002/jps.20708. Možina, M., Bürmen, M., Tomaževi9 c, D., Pernuš, F., Likar, B., 2009. Automatic visual inspection of pharmaceutical pellets in coating process. Paper Presented at the Industrial Technology ICIT 2009IEEE International Conference on, 10–13 Feb. 2009. doi:10.1109/ICIT.2009.4939618. Možina, M., Tomaževi9 c, D., Leben, S., Pernuš, F., Likar, B., 2010. Digital imaging as a process analytical technology tool for fluid-bed pellet coating process. Eur. J. Pharm. Sci. 41, 156–162. doi:10.1016/j.ejps.2010.06.001. Muschert, S., Siepmann, F., Leclercq, B., Carlin, B., Siepmann, J., 2009. Prediction of drug release from ethylcellulose coated pellets. J. Control. Release 135, 71–79. doi:10.1016/j.jconrel.2008.12.003. Naelapää, K., Veski, P., Pedersen, J.G., Anov, D., Jørgensen, P., Kristensen, H.G., Bertelsen, P., 2007. Acoustic monitoring of a fluidized bed coating process. Int. J. Pharm. 332, 1–2. doi:10.1016/j.ijpharm.2006.09.036. Närvänen, T., Kari, S., Osmo, A., Jouko, Y., 2008. A new rapid on-line imaging method to determine particle size distribution of granules. AAPS PharmSciTech 9 1, 228– 287. doi:10.1208/s12249-008-9043-y. Oberdier, J.P., 2013. Spectroscopic methods of analysis: atomic absorption and emission spectrophotometry, Encyclopedia of Pharmaceutical Technology. third ed. Taylor & Francis, pp. 3367–3374. doi:10.1081/E-EPT3-100000433. Podczeck, F., Rahman, S.R., Newton, J.M., 1999. Evaluation of a standardised procedure to assess the shape of pellets using image analysis. Int. J. Pharm. 192, 123–138. doi:10.1016/s0378-5173(99)00302-6. Santos, A., Ortiz De SolÓRzano, C., Vaquero, J.J., PeÑA, J.M., Malpica, N., Del Pozo, F., 1997. Evaluation of autofocus functions in molecular cytogenetic analysis. J. Microsc. 188, 264–272. doi:10.1046/j.1365-2818.1997.2630819.x. Turton, R., 2008. Challenges in the modeling and prediction of coating of pharmaceutical dosage forms. Powder Technol. 181, 186–194 http://dx.doi. org/10.1016/j.powtec.2006.12.006. US Food and Drug Administration. 2004. Guidance for industry: pat – a framework for innovative pharmaceutical development. Manufacturing, and Quality Assurance. Vogt, F.G., Kord, A.S., 2011. Development of quality-by-design analytical methods. J. Pharm. Sci. 100, 797–812. doi:10.1002/jps.22325. Wade A., P.J. Weller, 1994. Handbook of Pharmaceutical Excipients, second ed. The Pharmaceutical press, London: American Pharmaceutical Association, Washington. Yu, L., 2008. Pharmaceutical quality by design: product and process development, understanding, and control. Pharm. Res. 25, 781–791. doi:10.1007/s11095-0079511-1.

In-line monitoring of pellet coating thickness growth by means of visual imaging.

Coating thickness is the most important attribute of coated pharmaceutical pellets as it directly affects release profiles and stability of the drug. ...
1MB Sizes 0 Downloads 4 Views