International Journal of Pharmaceutics 490 (2015) 47–54
Contents lists available at ScienceDirect
International Journal of Pharmaceutics journal homepage: www.elsevier.com/locate/ijpharm
Understanding the impact of microcrystalline cellulose physicochemical properties on tabletability Gregory Thoorens a, * , Fabrice Krier a , Eric Rozet b , Brian Carlin c, Brigitte Evrard a a b c
Laboratory of Pharmaceutical Technology, Department of Pharmacy, C.I.R.M., University of Liège, 4000 Liège, Belgium Laboratory of Analytical Chemistry, Department of Pharmacy, C.I.R.M., University of Liège, 4000 Liège, Belgium FMC Health and Nutrition, Ewing, NJ, USA
A R T I C L E I N F O
A B S T R A C T
Article history: Received 3 April 2015 Received in revised form 7 May 2015 Accepted 9 May 2015 Available online 14 May 2015
The quality by design (QbD) initiative is promoting a better understanding of excipient performance and the identification of critical material attributes (CMAs). Despite microcrystalline cellulose (MCC) being one of the most popular direct compression binders, only a few studies attempted identifying its CMAs. These studies were based either on a limited number of samples or on MCC produced on a small scale and/or in conditions that deviate from those normally encountered in production. The present work utilizes multivariate analyses first to describe a large database of MCCs produced on a commercial scale, including an overview of their physicochemical properties, and secondly to correlate the most significant material attributes with tabletability. Particle size and moisture content are often considered as the most common if not the sole CMAs with regard to MCC performance in direct compression. The evaluation of more than 80 neat MCCs and the performance of selected samples in a model formulation revealed the importance of other potential critical attributes such as tapped density. Drug product developers and excipient suppliers should work together to identify these CMAs, which may not always be captured by the certificate of analysis. ã 2015 Elsevier B.V. All rights reserved.
Chemical compounds studied in this article: Microcrystalline cellulose (PubChem CID: 14055602) Ascorbic acid (PubChem CID: 54670067) Magnesium stearate (PubChem CID: 11177) Keywords: Microcrystalline cellulose Direct compression Tabletability Multivariate analysis Quality by design Critical material attributes
1. Introduction Microcrystalline cellulose (MCC) has been considered for the last fifty years as the diluent having the best binding properties and is recognized as one of the preferred direct compression (DC) binders (Bolhuis and Armstrong, 2006; Carlin, 2008; Patel et al., 2006; Saigal et al., 2009). The reasons for this preference include compactibility, tabletability, tradition, supply, handling, and physiological inertness (Bolhuis and Chowhan, 1996). MCC is a purified, partially depolymerized cellulose prepared by treating with mineral acids alpha cellulose (type Ib), obtained as a pulp from fibrous plant material, mostly from wood (Albers et al., 2006; Shlieout et al., 2002). The rate of hydrolysis slows to a certain level-off degree of polymerization (LODP). The LODP is a characteristic of a particular pulp and is typically found in the 200–300 range (Doelker, 1993).
* Corresponding author at: CHU, Tower 4, 2nd Floor, Laboratory of Pharmaceutical Technology, Department of Pharmacy, University of Liège, Avenue de l'hôpital, 1, 4000 Liège, Belgium. Tel.: +32 43664301; fax: +32 43664302. E-mail address: [email protected] (G. Thoorens). http://dx.doi.org/10.1016/j.ijpharm.2015.05.026 0378-5173/ ã 2015 Elsevier B.V. All rights reserved.
MCC is commonly manufactured by spray drying the neutralized aqueous slurry resulting from the hydrolysis of cellulose. Most commercial grades are formed by varying and controlling the spray drying conditions in order to manipulate the degree of agglomeration (particle size distribution) and moisture content (loss on drying) (Reier, 2000). Other drying techniques may be used (Christiansen and Sardo, 2001), which may require additional screening steps post drying to control particle size distribution. Several studies have compared microcrystalline cellulose from various sources, including different manufacturers and different sites (Albers et al., 2006; Doelker, 1993; Landín et al., 1993a,b,c; Williams et al., 1997). It was generally recognized that batch-tobatch variability from a sole manufacturing site was less important than differences observed between multiple sources. However, these conclusions were based only on single samples from two to three batches. Since MCC is manufactured by continuous production, a batch is defined as a certain period of time and could represent two days up to one week from a larger production campaign. It could therefore be argued that one sample (few hundred grams up to few kilograms) is not representative of the variability of a high volume continuously produced material.
48
G. Thoorens et al. / International Journal of Pharmaceutics 490 (2015) 47–54
Table 1 Frequency table, number of MCC samples classified by manufacturing site (origin) and date of manufacture. Origin
Month-Year
Total sample
Total %
May-2011
Jul-2011
Aug-2011
Oct-2011
Jan-2012
Feb-2012
Mar-2012
Cork Newark
– 14
18 –
6 –
6 –
6 12
– 16
– 6
36 48
43 57
Total
14
18
6
6
18
16
6
84
100
Only a few studies have tried to correlate the manufacturing conditions of microcrystalline cellulose with its physicochemical properties and its performance in tableting applications (Dybowski, 1997; Shlieout et al., 2002; Wu et al., 2001a,b). This information could be highly valuable to control or even to optimize the performance of MCC. However, these MCC samples were prepared on a small scale and/or in conditions that deviate from those normally targeted in production. Any conclusion drawn at small scale might not correlate with large scale operating conditions. There is an opportunity for a new and systematic study to gauge the variability encountered in a large manufacturing scale and to identify the physicochemical parameters of microcrystalline cellulose, i.e., its critical material attributes (CMAs), that may impact its tableting performance. The present study describes the evaluation of a substantial series of commercial samples, which allowed the implementation of proven multivariate analysis methods and the determination of statistically sound correlations (Haware et al., 2009; Kushner, 2013; Kushner et al., 2014; Souihi et al., 2013; Tho and Bauer-Brandl, 2011, 2012). 2. Material and methods 2.1. Material In order to capture the variability of microcrystalline cellulose type 102 as manufactured by FMC Health and Nutrition, samples were randomly collected from two manufacturing plants, Cork, Ireland and Newark, DE, USA. This type of MCC has a median particle size of about 100 mm, a bulk density close to 0.3 g/cm3, and is commonly used in direct compression. As summarized in Table 1, a total of 84 samples were obtained from 6 batches from each plant. Each batch was represented by at least 6 samples. Ascorbic acid (Hebei Welcome Pharmaceutical Co., Ltd., China) and magnesium stearate (code 2257, Mallinckrodt Pharmaceuticals, USA) were also used to assess the impact of MCC on the tabletability of the model formulation described in Section 2.4.2. 2.2. MCC manufacture MCC type 102 was produced under normal manufacturing conditions. The key steps of a typical manufacturing process are illustrated in Fig. 1.
Pulp (raw material)
Filtration Hydrolysis
(washing)
After depolymerization with mineral acids, the soluble components of cellulose are washed out and the insoluble MCC is dried to obtain the well-known white, odorless, tasteless, direct compression binder (Guy, 2009). MCC was sampled immediately after the drying step and did not go through subsequent processing steps such as cyclones, screening and packaging. 2.3. MCC characterization 2.3.1. Moisture content Moisture content, loss on drying (LOD), was determined with a halogen moisture analyzer (Mettler Toledo HR73, Switzerland) prior to any bulk density and tableting evaluation. In the case of tableting, moisture content was also measured once after the trial in order to calculate a mean value. The ‘standard’ drying program was selected. About 3 g (10%) of MCC was exposed to 105 C until the mean weight loss was less than 1 mg during 50 s. 2.3.2. Particle size Particle size distribution was obtained by laser diffraction (Malvern Mastersizer 2000 equipped with the Sirocco Dry Powder Feeder, UK). One aliquot of about 2 g (a tablespoon) of MCC powder was fed to the measurement cell using a vibration feed rate of 75% and a dispersive air pressure of 3 bars. The refractive index was set to 1.45 and the desired obscuration was about 5%. Volume weighted particle size distributions were described by the 10, the 50 (median) and the 90 percentiles. 2.3.3. Bulk density A Scott Volumeter (Paul N. Gardner Company, Inc., USA) was used to measure bulk density according to the method described in the MCC monograph and to the Method II of the General Chapter (USP37-NF32, 2014a). The volumeter is composed of a funnel with a 10 mesh screen, a chute with glass baffles to minimize packing, and a 25 ml brass cup. MCC powder is poured through the assembly into the brass cup until powder overflows. The excess powder is then scored off with a spatula. The cup is tapped and moved to a two-decimal place balance (Mettler Toledo PM4800, Switzerland). Density is calculated based on sample weight and its known volume. The measurement is repeated three times in order to calculate a mean value.
Spray dryer Reslurry tank
Packaging
Sampling Fig. 1. Microcrystalline cellulose manufacturing overview.
G. Thoorens et al. / International Journal of Pharmaceutics 490 (2015) 47–54
2.3.4. Tapped density Tapped density was determined according to the Method II of the General Chapter (USP37-NF32, 2014a). An apparent volume tester (Pharma Test PT-TD, Germany) equipped with a 250 ml graduated cylinder was used. About 60 g of MCC powder was introduced into the cylinder using a vibrator (Electromagnetic Sieve Vibrator EMS 755, Topas GmbH, Germany) set to an amplitude of 7 (maximum of 10). The ensemble was set up in such a way that the transmission of vibrations from the vibrator to the cylinder was minimized. The actual sample weight was measured with a two-decimal place balance (Mettler Toledo PM4800, Switzerland). Tapped volume was recorded after 2500 taps, at a rate of about 250 taps per minute from a height of about 3 mm. 2.3.5. True density A helium pycnometer (Micromeritics AccuPyc 1330, USA) was used to measure true density (density at zero porosity). About 4 g of MCC was poured into the sample cup. The weights of the empty sample cup and of the sample were precisely measured with an analytical balance (Mettler Toledo ML104, Switzerland). Following 10 purge cycles, MCC true volume was determined by injecting helium in the measuring chamber up to 19.5 psig. Volume and density were averaged from 10 consecutive runs. 2.3.6. Compressibility index The compressibility index, also called Carr’s index, was calculated from bulk density (rbulk) and tapped density (rtapped) values, as defined previously, as shown in Eq. (1). Compressibilityindexð%Þ ¼
100 ðrtapped rbulk Þ
rtapped
(1)
The compressibility index has been proposed as an indirect measure of bulk density, size and shape, surface area, moisture content, and cohesiveness of materials because all of these can influence the observed index. Low indices, e.g., below 20%, are typical of free-flowing materials. Cohesive powders might be characterized by indices superior to 40% (USP37-NF32, 2014b). 2.3.7. Degree of polymerization Degree of polymerization of hydrolyzed cellulose is based on ASTM Method No. D1795 (ASTM International, 2013), which describes the determination of the intrinsic viscosity of purified celluloses. In a second step the intrinsic viscosity is correlated with degree of polymerization according to Eq. (2) (USP37-NF32, 2014a): DP ¼
ð95 ½hc Þ ðWs ½ð100 %LODÞ=100Þ
(2)
[h]c = intrinsic viscosity Ws = weight of the microcrystalline cellulose taken (g) %LOD = value obtained from the test for loss on drying The intrinsic viscosity ([h]c) of MCC was determined by interpolation of the relative viscosity (hrel), which is calculated by dividing (KV)1, the kinematic viscosity of an MCC solution (2.6% MCC), by (KV)2, the kinematic viscosity of 0.5 M cupriethylenediamine hydroxide solution (Cuene). The kinematic viscosity of the MCC solution was measured with a number 150 Cannon-Fenske viscometer, whereas a number 100 Cannon-Fenske viscometer was used to determine the kinematic viscosity of the blank solution (CFRC series, Cannon Instrument Company, USA). This test was most commonly performed on a composite sample, which was prepared by blending smaller samples taken at regular intervals from a larger batch.
49
2.3.8. pH The pH of MCC samples was measured on dispersions containing 15% solids (52 g of MCC in 281 ml of deionized water) prepared using a Waring blender (Waring Commercial Model 700G, USA). The pH meter (Mettler Toledo FiveEasy PlusTM, Switzerland) was calibrated with buffers solutions of pH 4.0 and 7.0. Measurements were conducted at ambient temperature. 2.3.9. Conductivity The conductivity of MCC samples was measured on dispersions containing 15% solids (52 g of MCC in 281 ml of deionized water). The conductivity meter (Jenway Model 4510, UK) was run at ambient temperature. 2.3.10. Water soluble substances As described in the MCC monograph (USP37-NF32, 2014a), water soluble substances were quantified by shaking 5 g of MCC in 80 ml of water for 10 min. The dispersion was then filtered twice through a pre-moistened Whatman No. 42 ashless filter paper (GE Healthcare Life Sciences, USA). The filtrate was dried at 105 C overnight. The weight of residues in a blank was subtracted from the weight of residues in the MCC sample, as determined with an analytical balance (Mettler Toledo AE200, Switzerland). The resulting residues were then reported in mg/5 g of MCC. This test was performed on a composite sample representative of the entire batch. 2.3.11. Ether soluble substances According to the MCC monograph (USP37-NF32, 2014a), the amount of ether soluble substances was determined by placing 10 g of MCC in a chromatographic column having an internal diameter of about 20 mm, filled with a plug of glass wool and a porous disc, and by pouring 50 ml of diethyl ether into the column. After 2 min, the eluting solution was collected in a glass beaker, then evaporated in a fume hood, and finally dried at 105 C for 30 min. The weight of residues in a blank was subtracted from the weight of residues in the MCC sample, as determined with an analytical balance (Mettler Toledo AE200, Switzerland). The resulting residues were then reported in mg/10 g of MCC. The test was performed on a composite sample. 2.4. Tableting 2.4.1. Tablet preparation MCC tablets were prepared on an ESH Compaction Simulator (Phoenix Calibration & Services Ltd., UK), using a sine wave profile for the upper punch displacement and maintaining the lower punch stationary during compression. High compression speeds were obtained by selecting a cycle duration of 0.12 s, which resulted in a mean vertical speed of the upper punch of 300 mm/s and in dwell times of about 6 ms. A consistent tablet weight close to 500 mg was obtained by weighing every powder sample manually on an analytical balance (Mettler Toledo AG104, Switzerland) and by pouring this powder manually into the die cavity. Punches were flat-faced, round, with a 13 mm diameter. Die walls and punch tips were lubricated every three tablets by applying a dispersion of magnesium stearate in acetone with a cotton swab. Five tableting forces were selected between 4 and 15 kN in order to obtain tablet solid fractions between 55 and 80%. For each tableting force, three tablets were produced to assess reproducibility. The temperature and the relative humidity (RH) of the compaction simulator room were maintained to 20 1 C and 40 2% RH, respectively. MCC samples were tested as received and their moisture contents were not equilibrated. Moisture pick-up of the tablets was also minimized by placing the freshly produced tablets in two
50
G. Thoorens et al. / International Journal of Pharmaceutics 490 (2015) 47–54
consecutive sealed plastic bags and storing the tablets in the compaction simulator room during 24 h. Tablet properties, i.e., weight (Mettler Toledo ML104, Switzerland), dimensions (Mitutoyo Digimatic Thickness Gauge No. 547-301 and Caliper No. 500-321, Japan) and crushing force (Erweka TBH 30 MD, Germany), were measured after this 24-hour period. The tensile strength of cylindrical tablets was calculated based on their crushing force and dimensions as follows (Eq. (3)) (Fell and Newton, 1970): TS ¼
ð2 FÞ ðp t DÞ
Table 2 Composition of the ascorbic acid model formulation. Component
Function
%
Ascorbic acid Microcrystalline cellulose Magnesium stearate
Active pharmaceutical ingredient Binder Lubricant
60.0 39.5 0.5
Total
–
100.0
(3)
where TS is the tensile strength in MPa, F is the crushing force in N, t is the thickness and D the diameter, both in mm. Tabletability, the capacity of a powdered material to be transformed into a tablet of specified strength under the effect of compaction pressure (Joiris et al., 1998; Sun, 2008), was determined by plotting the tensile strength of fifteen tablets versus compaction pressure. In this study tabletability was represented by a unique descriptor, i.e., TS at 85 MPa compaction pressure, which was calculated from the linear regression of the TS versus pressure plot. 2.4.2. Model formulation A model formulation was also considered to assess the tabletability of MCC when combined with a poorly tabletable active pharmaceutical ingredient (API) and a hydrophobic lubricant. Ascorbic acid (60%) and MCC (39.5%) were initially screened through a 710 mm sieve then blended for 10 min with a Turbula T2F (Willy A. Bachofen AG—Maschinenfabrik, Switzerland) set to a speed of 49 revolutions per minute. Magnesium stearate (0.5%) was added to a part of the pre-blend, screened trough the 710 mm sieve, added to the remainder of the pre-blend and finally blended for an additional 2 min with the same blender. The composition of this model formulation is detailed in Table 2. The formulation was compressed on the compaction simulator previously described; using the same compression speed and the same toolings, and targeting the same tablet weight of 500 mg. Six tableting forces were selected between 5 and 30 kN. For each tableting force, two tablets were produced to assess reproducibility. 2.5. Statistical analysis Principal component analysis (PCA) is a data visualization technique that describes large quantities of data in a reduced number of principal components (PCs) that maximize explained variance in the data on each successive component under the constraint of being orthogonal to the previous PCs (Rajalahti and Kvalheim, 2011). The application of PCA provides means for excellent overviews and the detection of trends, groupings and
outliers (Gabrielsson et al., 2002). Objects clustered in groups on the same side of a PC have similar features, in opposition to objects situated further apart (Klevan et al., 2010). Descriptive statistics, plots and PCAs were generated in Minitab 16 (Minitab Inc., USA). PCs were calculated using the correlation matrix. JMP 11 (SAS Institute Inc., USA) was used in a first step to perform stepwise regressions. The stopping rule of the stepwise regressions was based on a p-value threshold of 0.10, both for forward selection and backward elimination from the model. In a second step, after having identified the most statistically significant variables, multiple linear regression models were created using the least-squares method. 3. Results and discussion 3.1. Description of the database For each of the 84 MCC samples, ten physicochemical properties were recorded. Eight were certificate of analysis parameters: degree of polymerization (DP), pH, conductivity, water soluble substances, ether soluble substances, moisture content (loss on drying, LOD), median particle size (d50), and bulk density. The additional two properties were the tapped and true densities. The resulting database is described with the help of a principal component analysis (PCA), as shown in Table 3 and Figs. 2 and 3. The PCA model captures about 86% of the total variability with five principal components. The first principal component (PC1), which captures the largest variation in the data set, i.e., 29% (R2 = 0.290) of the total variability, shows two clusters in Fig. 2. These clusters correlate with the site of origin, which is not a variable in the PCA analysis. PC1 is mostly influenced by the degree of polymerization, the pH and the conductivity. The latter two parameters may relate to the residual acid post washing and filtration (Fig. 1). This apparent site-to-site difference is restricted to these three parameters. The score plot for the third and fourth principal components shown in Fig. 3 suggests that there is good intermixing of the data from the two sites for these principal components. There is no site-to-site variability in the properties captured by the second to the fifth principal component (variables listed in Table 3).
Table 3 Principal component analysis, main variables for the first five principal components. Principal component 1
2 3
4 5
Variable captured pH Degree of polymerization, DP Conductivity (mS/cm) Water soluble substances (mg/5g) True density (g/cm3) Bulk density (g/cm3) Tapped density (g/cm3) Ether soluble substances (mg/10g) Moisture content, loss on drying (%) Median particle size, d50 (mm)
Eigen value
R2 (x)
R2 (x) cum
2.9
0.290
0.290
2.0 1.7
0.199 0.166
0.488 0.654
1.3
0.127
0.781
0.8
0.080
0.862
G. Thoorens et al. / International Journal of Pharmaceutics 490 (2015) 47–54
51
Fig. 2. Principal component analysis, score plot for the first two principal components.
Fig. 3. Principal component analysis, score plot for the third and the fourth principal components.
3.2. Tabletability distribution
the manufacturing site (origin, Cork or Newark) was also considered as an independent variable. The dependent variable representing tabletability was the tablet tensile strength at a compaction pressure of 85 MPa. As highlighted in Table 5, four attributes were statistically significant (p-value < 0.05): moisture content, conductivity, tapped density and pH. The multiple linear regression model (Eq. (4)) based on these four independent variables is statistically sound with an overall p-value less than 0.001. The model also explains close to 70% of the variability in tabletability (R2 = 0.70).
In a first phase, the 84 neat MCC samples were compressed on a compaction simulator to assess tabletability, i.e., the strength a tablet attains under a given pressure. In a second phase, detailed in Section 3.3.2, 8 MCC samples representative of the larger set were incorporated in an ascorbic acid model formulation and compressed into tablets. As illustrated in Fig. 4 and Table 4, the tensile strength of neat MCC at a compaction pressure of 85 MPa ranged around a mean value of 4.5 MPa. As expected the addition of a poorly tabletable API and of a hydrophobic lubricant to MCC resulted in much lower tensile strengths and in a tighter spread around the mean value of 0.6 MPa. The QbD initiative encourages the identification of the CMAs that influence the tableting performance of MCC, as these CMAs may in turn impact the critical quality attributes (CQAs) of finished products. 3.3. Factors influencing tabletability 3.3.1. Physicochemical properties of neat MCC The impact of 84 neat MCC physicochemical parameters on tabletability was analyzed by means of a stepwise regression (Table 5). In addition to the ten parameters described in Section 3.1,
TS@85MPa ¼ 4:5085 þ 0:3787 LOD 0:0089 Conductivity 5:9276 Tappeddensity þ 0:2722 pH þ e
(4)
where e, the error term or residuals, follows a normal distribution with constant variance. Following the hydrolysis step, the cellulose slurry is filtered and washed to remove the amorphous and solubilized cellulose. In the model (Eq. (4)), the tabletability of neat MCC is negatively influenced by a decrease of pH and an increase of conductivity, both of which could relate to the propensity of the cellulose slurry to be washed. An inverse correlation of tapped density with tabletability was also identified. High tapped density may be the resultant of
Fig. 4. Histogram of TS at 85 MPa, 84 neat MCCs (gray) and 8 MCCs in a model formulation (black).
52
G. Thoorens et al. / International Journal of Pharmaceutics 490 (2015) 47–54
Table 4 Descriptive statistics, tensile strength at 85 MPa compaction pressure. Statistics
Mean Range St. Dev
TS of 84 MCCs (MPa)
TS of 8 MCCs (MPa)
Neat
Neat
Formulation
4.5 1.6 0.3
4.7 1.6 0.5
0.6 0.3 0.1
Table 5 Stepwise regression, physicochemical parameters influencing the tabletability of neat MCC. Independent variable
p-Value
Moisture content, LOD (%) Conductivity (mS/cm) Tapped density (g/cm3) pH Origin (Cork or Newark) Ether soluble substances (mg/10g) True density (g/cm3) Water soluble substances (mg/5g) Bulk density (g/cm3) Median particle size, d50 (mm) Degree of polymerization, DP
Contents lists available at ScienceDirect
International Journal of Pharmaceutics journal homepage: www.elsevier.com/locate/ijpharm
Understanding the impact of microcrystalline cellulose physicochemical properties on tabletability Gregory Thoorens a, * , Fabrice Krier a , Eric Rozet b , Brian Carlin c, Brigitte Evrard a a b c
Laboratory of Pharmaceutical Technology, Department of Pharmacy, C.I.R.M., University of Liège, 4000 Liège, Belgium Laboratory of Analytical Chemistry, Department of Pharmacy, C.I.R.M., University of Liège, 4000 Liège, Belgium FMC Health and Nutrition, Ewing, NJ, USA
A R T I C L E I N F O
A B S T R A C T
Article history: Received 3 April 2015 Received in revised form 7 May 2015 Accepted 9 May 2015 Available online 14 May 2015
The quality by design (QbD) initiative is promoting a better understanding of excipient performance and the identification of critical material attributes (CMAs). Despite microcrystalline cellulose (MCC) being one of the most popular direct compression binders, only a few studies attempted identifying its CMAs. These studies were based either on a limited number of samples or on MCC produced on a small scale and/or in conditions that deviate from those normally encountered in production. The present work utilizes multivariate analyses first to describe a large database of MCCs produced on a commercial scale, including an overview of their physicochemical properties, and secondly to correlate the most significant material attributes with tabletability. Particle size and moisture content are often considered as the most common if not the sole CMAs with regard to MCC performance in direct compression. The evaluation of more than 80 neat MCCs and the performance of selected samples in a model formulation revealed the importance of other potential critical attributes such as tapped density. Drug product developers and excipient suppliers should work together to identify these CMAs, which may not always be captured by the certificate of analysis. ã 2015 Elsevier B.V. All rights reserved.
Chemical compounds studied in this article: Microcrystalline cellulose (PubChem CID: 14055602) Ascorbic acid (PubChem CID: 54670067) Magnesium stearate (PubChem CID: 11177) Keywords: Microcrystalline cellulose Direct compression Tabletability Multivariate analysis Quality by design Critical material attributes
1. Introduction Microcrystalline cellulose (MCC) has been considered for the last fifty years as the diluent having the best binding properties and is recognized as one of the preferred direct compression (DC) binders (Bolhuis and Armstrong, 2006; Carlin, 2008; Patel et al., 2006; Saigal et al., 2009). The reasons for this preference include compactibility, tabletability, tradition, supply, handling, and physiological inertness (Bolhuis and Chowhan, 1996). MCC is a purified, partially depolymerized cellulose prepared by treating with mineral acids alpha cellulose (type Ib), obtained as a pulp from fibrous plant material, mostly from wood (Albers et al., 2006; Shlieout et al., 2002). The rate of hydrolysis slows to a certain level-off degree of polymerization (LODP). The LODP is a characteristic of a particular pulp and is typically found in the 200–300 range (Doelker, 1993).
* Corresponding author at: CHU, Tower 4, 2nd Floor, Laboratory of Pharmaceutical Technology, Department of Pharmacy, University of Liège, Avenue de l'hôpital, 1, 4000 Liège, Belgium. Tel.: +32 43664301; fax: +32 43664302. E-mail address: [email protected] (G. Thoorens). http://dx.doi.org/10.1016/j.ijpharm.2015.05.026 0378-5173/ ã 2015 Elsevier B.V. All rights reserved.
MCC is commonly manufactured by spray drying the neutralized aqueous slurry resulting from the hydrolysis of cellulose. Most commercial grades are formed by varying and controlling the spray drying conditions in order to manipulate the degree of agglomeration (particle size distribution) and moisture content (loss on drying) (Reier, 2000). Other drying techniques may be used (Christiansen and Sardo, 2001), which may require additional screening steps post drying to control particle size distribution. Several studies have compared microcrystalline cellulose from various sources, including different manufacturers and different sites (Albers et al., 2006; Doelker, 1993; Landín et al., 1993a,b,c; Williams et al., 1997). It was generally recognized that batch-tobatch variability from a sole manufacturing site was less important than differences observed between multiple sources. However, these conclusions were based only on single samples from two to three batches. Since MCC is manufactured by continuous production, a batch is defined as a certain period of time and could represent two days up to one week from a larger production campaign. It could therefore be argued that one sample (few hundred grams up to few kilograms) is not representative of the variability of a high volume continuously produced material.
48
G. Thoorens et al. / International Journal of Pharmaceutics 490 (2015) 47–54
Table 1 Frequency table, number of MCC samples classified by manufacturing site (origin) and date of manufacture. Origin
Month-Year
Total sample
Total %
May-2011
Jul-2011
Aug-2011
Oct-2011
Jan-2012
Feb-2012
Mar-2012
Cork Newark
– 14
18 –
6 –
6 –
6 12
– 16
– 6
36 48
43 57
Total
14
18
6
6
18
16
6
84
100
Only a few studies have tried to correlate the manufacturing conditions of microcrystalline cellulose with its physicochemical properties and its performance in tableting applications (Dybowski, 1997; Shlieout et al., 2002; Wu et al., 2001a,b). This information could be highly valuable to control or even to optimize the performance of MCC. However, these MCC samples were prepared on a small scale and/or in conditions that deviate from those normally targeted in production. Any conclusion drawn at small scale might not correlate with large scale operating conditions. There is an opportunity for a new and systematic study to gauge the variability encountered in a large manufacturing scale and to identify the physicochemical parameters of microcrystalline cellulose, i.e., its critical material attributes (CMAs), that may impact its tableting performance. The present study describes the evaluation of a substantial series of commercial samples, which allowed the implementation of proven multivariate analysis methods and the determination of statistically sound correlations (Haware et al., 2009; Kushner, 2013; Kushner et al., 2014; Souihi et al., 2013; Tho and Bauer-Brandl, 2011, 2012). 2. Material and methods 2.1. Material In order to capture the variability of microcrystalline cellulose type 102 as manufactured by FMC Health and Nutrition, samples were randomly collected from two manufacturing plants, Cork, Ireland and Newark, DE, USA. This type of MCC has a median particle size of about 100 mm, a bulk density close to 0.3 g/cm3, and is commonly used in direct compression. As summarized in Table 1, a total of 84 samples were obtained from 6 batches from each plant. Each batch was represented by at least 6 samples. Ascorbic acid (Hebei Welcome Pharmaceutical Co., Ltd., China) and magnesium stearate (code 2257, Mallinckrodt Pharmaceuticals, USA) were also used to assess the impact of MCC on the tabletability of the model formulation described in Section 2.4.2. 2.2. MCC manufacture MCC type 102 was produced under normal manufacturing conditions. The key steps of a typical manufacturing process are illustrated in Fig. 1.
Pulp (raw material)
Filtration Hydrolysis
(washing)
After depolymerization with mineral acids, the soluble components of cellulose are washed out and the insoluble MCC is dried to obtain the well-known white, odorless, tasteless, direct compression binder (Guy, 2009). MCC was sampled immediately after the drying step and did not go through subsequent processing steps such as cyclones, screening and packaging. 2.3. MCC characterization 2.3.1. Moisture content Moisture content, loss on drying (LOD), was determined with a halogen moisture analyzer (Mettler Toledo HR73, Switzerland) prior to any bulk density and tableting evaluation. In the case of tableting, moisture content was also measured once after the trial in order to calculate a mean value. The ‘standard’ drying program was selected. About 3 g (10%) of MCC was exposed to 105 C until the mean weight loss was less than 1 mg during 50 s. 2.3.2. Particle size Particle size distribution was obtained by laser diffraction (Malvern Mastersizer 2000 equipped with the Sirocco Dry Powder Feeder, UK). One aliquot of about 2 g (a tablespoon) of MCC powder was fed to the measurement cell using a vibration feed rate of 75% and a dispersive air pressure of 3 bars. The refractive index was set to 1.45 and the desired obscuration was about 5%. Volume weighted particle size distributions were described by the 10, the 50 (median) and the 90 percentiles. 2.3.3. Bulk density A Scott Volumeter (Paul N. Gardner Company, Inc., USA) was used to measure bulk density according to the method described in the MCC monograph and to the Method II of the General Chapter (USP37-NF32, 2014a). The volumeter is composed of a funnel with a 10 mesh screen, a chute with glass baffles to minimize packing, and a 25 ml brass cup. MCC powder is poured through the assembly into the brass cup until powder overflows. The excess powder is then scored off with a spatula. The cup is tapped and moved to a two-decimal place balance (Mettler Toledo PM4800, Switzerland). Density is calculated based on sample weight and its known volume. The measurement is repeated three times in order to calculate a mean value.
Spray dryer Reslurry tank
Packaging
Sampling Fig. 1. Microcrystalline cellulose manufacturing overview.
G. Thoorens et al. / International Journal of Pharmaceutics 490 (2015) 47–54
2.3.4. Tapped density Tapped density was determined according to the Method II of the General Chapter (USP37-NF32, 2014a). An apparent volume tester (Pharma Test PT-TD, Germany) equipped with a 250 ml graduated cylinder was used. About 60 g of MCC powder was introduced into the cylinder using a vibrator (Electromagnetic Sieve Vibrator EMS 755, Topas GmbH, Germany) set to an amplitude of 7 (maximum of 10). The ensemble was set up in such a way that the transmission of vibrations from the vibrator to the cylinder was minimized. The actual sample weight was measured with a two-decimal place balance (Mettler Toledo PM4800, Switzerland). Tapped volume was recorded after 2500 taps, at a rate of about 250 taps per minute from a height of about 3 mm. 2.3.5. True density A helium pycnometer (Micromeritics AccuPyc 1330, USA) was used to measure true density (density at zero porosity). About 4 g of MCC was poured into the sample cup. The weights of the empty sample cup and of the sample were precisely measured with an analytical balance (Mettler Toledo ML104, Switzerland). Following 10 purge cycles, MCC true volume was determined by injecting helium in the measuring chamber up to 19.5 psig. Volume and density were averaged from 10 consecutive runs. 2.3.6. Compressibility index The compressibility index, also called Carr’s index, was calculated from bulk density (rbulk) and tapped density (rtapped) values, as defined previously, as shown in Eq. (1). Compressibilityindexð%Þ ¼
100 ðrtapped rbulk Þ
rtapped
(1)
The compressibility index has been proposed as an indirect measure of bulk density, size and shape, surface area, moisture content, and cohesiveness of materials because all of these can influence the observed index. Low indices, e.g., below 20%, are typical of free-flowing materials. Cohesive powders might be characterized by indices superior to 40% (USP37-NF32, 2014b). 2.3.7. Degree of polymerization Degree of polymerization of hydrolyzed cellulose is based on ASTM Method No. D1795 (ASTM International, 2013), which describes the determination of the intrinsic viscosity of purified celluloses. In a second step the intrinsic viscosity is correlated with degree of polymerization according to Eq. (2) (USP37-NF32, 2014a): DP ¼
ð95 ½hc Þ ðWs ½ð100 %LODÞ=100Þ
(2)
[h]c = intrinsic viscosity Ws = weight of the microcrystalline cellulose taken (g) %LOD = value obtained from the test for loss on drying The intrinsic viscosity ([h]c) of MCC was determined by interpolation of the relative viscosity (hrel), which is calculated by dividing (KV)1, the kinematic viscosity of an MCC solution (2.6% MCC), by (KV)2, the kinematic viscosity of 0.5 M cupriethylenediamine hydroxide solution (Cuene). The kinematic viscosity of the MCC solution was measured with a number 150 Cannon-Fenske viscometer, whereas a number 100 Cannon-Fenske viscometer was used to determine the kinematic viscosity of the blank solution (CFRC series, Cannon Instrument Company, USA). This test was most commonly performed on a composite sample, which was prepared by blending smaller samples taken at regular intervals from a larger batch.
49
2.3.8. pH The pH of MCC samples was measured on dispersions containing 15% solids (52 g of MCC in 281 ml of deionized water) prepared using a Waring blender (Waring Commercial Model 700G, USA). The pH meter (Mettler Toledo FiveEasy PlusTM, Switzerland) was calibrated with buffers solutions of pH 4.0 and 7.0. Measurements were conducted at ambient temperature. 2.3.9. Conductivity The conductivity of MCC samples was measured on dispersions containing 15% solids (52 g of MCC in 281 ml of deionized water). The conductivity meter (Jenway Model 4510, UK) was run at ambient temperature. 2.3.10. Water soluble substances As described in the MCC monograph (USP37-NF32, 2014a), water soluble substances were quantified by shaking 5 g of MCC in 80 ml of water for 10 min. The dispersion was then filtered twice through a pre-moistened Whatman No. 42 ashless filter paper (GE Healthcare Life Sciences, USA). The filtrate was dried at 105 C overnight. The weight of residues in a blank was subtracted from the weight of residues in the MCC sample, as determined with an analytical balance (Mettler Toledo AE200, Switzerland). The resulting residues were then reported in mg/5 g of MCC. This test was performed on a composite sample representative of the entire batch. 2.3.11. Ether soluble substances According to the MCC monograph (USP37-NF32, 2014a), the amount of ether soluble substances was determined by placing 10 g of MCC in a chromatographic column having an internal diameter of about 20 mm, filled with a plug of glass wool and a porous disc, and by pouring 50 ml of diethyl ether into the column. After 2 min, the eluting solution was collected in a glass beaker, then evaporated in a fume hood, and finally dried at 105 C for 30 min. The weight of residues in a blank was subtracted from the weight of residues in the MCC sample, as determined with an analytical balance (Mettler Toledo AE200, Switzerland). The resulting residues were then reported in mg/10 g of MCC. The test was performed on a composite sample. 2.4. Tableting 2.4.1. Tablet preparation MCC tablets were prepared on an ESH Compaction Simulator (Phoenix Calibration & Services Ltd., UK), using a sine wave profile for the upper punch displacement and maintaining the lower punch stationary during compression. High compression speeds were obtained by selecting a cycle duration of 0.12 s, which resulted in a mean vertical speed of the upper punch of 300 mm/s and in dwell times of about 6 ms. A consistent tablet weight close to 500 mg was obtained by weighing every powder sample manually on an analytical balance (Mettler Toledo AG104, Switzerland) and by pouring this powder manually into the die cavity. Punches were flat-faced, round, with a 13 mm diameter. Die walls and punch tips were lubricated every three tablets by applying a dispersion of magnesium stearate in acetone with a cotton swab. Five tableting forces were selected between 4 and 15 kN in order to obtain tablet solid fractions between 55 and 80%. For each tableting force, three tablets were produced to assess reproducibility. The temperature and the relative humidity (RH) of the compaction simulator room were maintained to 20 1 C and 40 2% RH, respectively. MCC samples were tested as received and their moisture contents were not equilibrated. Moisture pick-up of the tablets was also minimized by placing the freshly produced tablets in two
50
G. Thoorens et al. / International Journal of Pharmaceutics 490 (2015) 47–54
consecutive sealed plastic bags and storing the tablets in the compaction simulator room during 24 h. Tablet properties, i.e., weight (Mettler Toledo ML104, Switzerland), dimensions (Mitutoyo Digimatic Thickness Gauge No. 547-301 and Caliper No. 500-321, Japan) and crushing force (Erweka TBH 30 MD, Germany), were measured after this 24-hour period. The tensile strength of cylindrical tablets was calculated based on their crushing force and dimensions as follows (Eq. (3)) (Fell and Newton, 1970): TS ¼
ð2 FÞ ðp t DÞ
Table 2 Composition of the ascorbic acid model formulation. Component
Function
%
Ascorbic acid Microcrystalline cellulose Magnesium stearate
Active pharmaceutical ingredient Binder Lubricant
60.0 39.5 0.5
Total
–
100.0
(3)
where TS is the tensile strength in MPa, F is the crushing force in N, t is the thickness and D the diameter, both in mm. Tabletability, the capacity of a powdered material to be transformed into a tablet of specified strength under the effect of compaction pressure (Joiris et al., 1998; Sun, 2008), was determined by plotting the tensile strength of fifteen tablets versus compaction pressure. In this study tabletability was represented by a unique descriptor, i.e., TS at 85 MPa compaction pressure, which was calculated from the linear regression of the TS versus pressure plot. 2.4.2. Model formulation A model formulation was also considered to assess the tabletability of MCC when combined with a poorly tabletable active pharmaceutical ingredient (API) and a hydrophobic lubricant. Ascorbic acid (60%) and MCC (39.5%) were initially screened through a 710 mm sieve then blended for 10 min with a Turbula T2F (Willy A. Bachofen AG—Maschinenfabrik, Switzerland) set to a speed of 49 revolutions per minute. Magnesium stearate (0.5%) was added to a part of the pre-blend, screened trough the 710 mm sieve, added to the remainder of the pre-blend and finally blended for an additional 2 min with the same blender. The composition of this model formulation is detailed in Table 2. The formulation was compressed on the compaction simulator previously described; using the same compression speed and the same toolings, and targeting the same tablet weight of 500 mg. Six tableting forces were selected between 5 and 30 kN. For each tableting force, two tablets were produced to assess reproducibility. 2.5. Statistical analysis Principal component analysis (PCA) is a data visualization technique that describes large quantities of data in a reduced number of principal components (PCs) that maximize explained variance in the data on each successive component under the constraint of being orthogonal to the previous PCs (Rajalahti and Kvalheim, 2011). The application of PCA provides means for excellent overviews and the detection of trends, groupings and
outliers (Gabrielsson et al., 2002). Objects clustered in groups on the same side of a PC have similar features, in opposition to objects situated further apart (Klevan et al., 2010). Descriptive statistics, plots and PCAs were generated in Minitab 16 (Minitab Inc., USA). PCs were calculated using the correlation matrix. JMP 11 (SAS Institute Inc., USA) was used in a first step to perform stepwise regressions. The stopping rule of the stepwise regressions was based on a p-value threshold of 0.10, both for forward selection and backward elimination from the model. In a second step, after having identified the most statistically significant variables, multiple linear regression models were created using the least-squares method. 3. Results and discussion 3.1. Description of the database For each of the 84 MCC samples, ten physicochemical properties were recorded. Eight were certificate of analysis parameters: degree of polymerization (DP), pH, conductivity, water soluble substances, ether soluble substances, moisture content (loss on drying, LOD), median particle size (d50), and bulk density. The additional two properties were the tapped and true densities. The resulting database is described with the help of a principal component analysis (PCA), as shown in Table 3 and Figs. 2 and 3. The PCA model captures about 86% of the total variability with five principal components. The first principal component (PC1), which captures the largest variation in the data set, i.e., 29% (R2 = 0.290) of the total variability, shows two clusters in Fig. 2. These clusters correlate with the site of origin, which is not a variable in the PCA analysis. PC1 is mostly influenced by the degree of polymerization, the pH and the conductivity. The latter two parameters may relate to the residual acid post washing and filtration (Fig. 1). This apparent site-to-site difference is restricted to these three parameters. The score plot for the third and fourth principal components shown in Fig. 3 suggests that there is good intermixing of the data from the two sites for these principal components. There is no site-to-site variability in the properties captured by the second to the fifth principal component (variables listed in Table 3).
Table 3 Principal component analysis, main variables for the first five principal components. Principal component 1
2 3
4 5
Variable captured pH Degree of polymerization, DP Conductivity (mS/cm) Water soluble substances (mg/5g) True density (g/cm3) Bulk density (g/cm3) Tapped density (g/cm3) Ether soluble substances (mg/10g) Moisture content, loss on drying (%) Median particle size, d50 (mm)
Eigen value
R2 (x)
R2 (x) cum
2.9
0.290
0.290
2.0 1.7
0.199 0.166
0.488 0.654
1.3
0.127
0.781
0.8
0.080
0.862
G. Thoorens et al. / International Journal of Pharmaceutics 490 (2015) 47–54
51
Fig. 2. Principal component analysis, score plot for the first two principal components.
Fig. 3. Principal component analysis, score plot for the third and the fourth principal components.
3.2. Tabletability distribution
the manufacturing site (origin, Cork or Newark) was also considered as an independent variable. The dependent variable representing tabletability was the tablet tensile strength at a compaction pressure of 85 MPa. As highlighted in Table 5, four attributes were statistically significant (p-value < 0.05): moisture content, conductivity, tapped density and pH. The multiple linear regression model (Eq. (4)) based on these four independent variables is statistically sound with an overall p-value less than 0.001. The model also explains close to 70% of the variability in tabletability (R2 = 0.70).
In a first phase, the 84 neat MCC samples were compressed on a compaction simulator to assess tabletability, i.e., the strength a tablet attains under a given pressure. In a second phase, detailed in Section 3.3.2, 8 MCC samples representative of the larger set were incorporated in an ascorbic acid model formulation and compressed into tablets. As illustrated in Fig. 4 and Table 4, the tensile strength of neat MCC at a compaction pressure of 85 MPa ranged around a mean value of 4.5 MPa. As expected the addition of a poorly tabletable API and of a hydrophobic lubricant to MCC resulted in much lower tensile strengths and in a tighter spread around the mean value of 0.6 MPa. The QbD initiative encourages the identification of the CMAs that influence the tableting performance of MCC, as these CMAs may in turn impact the critical quality attributes (CQAs) of finished products. 3.3. Factors influencing tabletability 3.3.1. Physicochemical properties of neat MCC The impact of 84 neat MCC physicochemical parameters on tabletability was analyzed by means of a stepwise regression (Table 5). In addition to the ten parameters described in Section 3.1,
TS@85MPa ¼ 4:5085 þ 0:3787 LOD 0:0089 Conductivity 5:9276 Tappeddensity þ 0:2722 pH þ e
(4)
where e, the error term or residuals, follows a normal distribution with constant variance. Following the hydrolysis step, the cellulose slurry is filtered and washed to remove the amorphous and solubilized cellulose. In the model (Eq. (4)), the tabletability of neat MCC is negatively influenced by a decrease of pH and an increase of conductivity, both of which could relate to the propensity of the cellulose slurry to be washed. An inverse correlation of tapped density with tabletability was also identified. High tapped density may be the resultant of
Fig. 4. Histogram of TS at 85 MPa, 84 neat MCCs (gray) and 8 MCCs in a model formulation (black).
52
G. Thoorens et al. / International Journal of Pharmaceutics 490 (2015) 47–54
Table 4 Descriptive statistics, tensile strength at 85 MPa compaction pressure. Statistics
Mean Range St. Dev
TS of 84 MCCs (MPa)
TS of 8 MCCs (MPa)
Neat
Neat
Formulation
4.5 1.6 0.3
4.7 1.6 0.5
0.6 0.3 0.1
Table 5 Stepwise regression, physicochemical parameters influencing the tabletability of neat MCC. Independent variable
p-Value
Moisture content, LOD (%) Conductivity (mS/cm) Tapped density (g/cm3) pH Origin (Cork or Newark) Ether soluble substances (mg/10g) True density (g/cm3) Water soluble substances (mg/5g) Bulk density (g/cm3) Median particle size, d50 (mm) Degree of polymerization, DP
