A practical approach for the scale-up of roller compaction process

Accepted Manuscript A practical approach for the scale up of roller compaction process Weixian Shi, Omar L. Sprockel PII: DOI: Reference:

S0939-6411(16)30007-8 http://dx.doi.org/10.1016/j.ejpb.2016.02.005 EJPB 12118

To appear in:

European Journal of Pharmaceutics and Biopharmaceutics

Received Date: Revised Date: Accepted Date:

25 August 2015 7 January 2016 9 February 2016

Please cite this article as: W. Shi, O.L. Sprockel, A practical approach for the scale up of roller compaction process, European Journal of Pharmaceutics and Biopharmaceutics (2016), doi: http://dx.doi.org/10.1016/j.ejpb. 2016.02.005

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

A PRACTICAL APPROACH FOR THE SCALE UP OF ROLLER COMPACTION PROCESS Weixian Shi & Omar L. Sprockel Drug Product Science and Technology, Bristol-Myers Squibb, 1 Squibb Drive New Brunswick, NJ 08901 USA [email protected]

ABSTRACT An alternative approach for the scale up of ribbon formation during roller compaction was investigated, which required only one batch at the commercial scale to set the operational conditions. The scale up of ribbon formation was based on a probability method. It was sufficient in describing the mechanism of ribbon formation at both scales. In this method, a statistical relationship between roller compaction parameters and ribbon attributes (thickness and density) was first defined with DoE using a pilot Alexanderwerk WP120 roller compactor. While the milling speed was included in the design, it has no practical effect on granule properties within the study range despite its statistical significance. The statistical relationship was then adapted to a commercial Alexanderwerk WP200 roller compactor with one experimental run. The experimental run served as a calibration of the statistical model parameters. The proposed transfer method was then confirmed by conducting a mapping study on the Alexanderwerk WP200 using a factorial DoE, which showed a match between the predictions and the verification experiments. The study demonstrates the applicability of the roller compaction transfer method using the statistical model from the development scale calibrated with one experiment point at the commercial scale. KEYWORDS Roller Compaction; Scale Up; Technology Transfer; Statistical Model; DoE; Granulation 1. INTRODUCTION Roller compaction is a process typically used in the pharmaceutical and food industry to increase the flowability of a powder by forming granules of greater size and density [1]. In a roller compaction process, a powder blend containing active pharmaceutical ingredient (API) and excipients passes through two counter-rotating rolls that compact the blend into ribbons. The ribbons subsequently are milled into granules of a desired size distribution that are mixed with extragranular materials prior to tabletting or capsule filling. While milling can be a mechanically separate unit operation, some roller compactors have combined ribbon formation and milling mechanism, making it a continuous operation. In this study, we focus on the scale up of such roller compactor. As the granules are denser and larger compared to the pre-roller-compacted blend, the granules flow freely and result in consistent feeding on a tablet press or capsule filling machine. However, the trade-off of the roller compaction process is the reduction in compactability as a result of the powder being stressed, which may cause chipping or capping in tablets. Therefore, a balance between increase in flow and reduction in compactability is a key design consideration in the development of a roller compaction process for tablet products. The flow and compactability are directly linked to ribbon density [2]. While there is no evidence that ribbon thickness could affect

followability or compactability, the ribbon thickness is used as a process stability index. Its variation is related to functionality of the feeding control system, i.e. the responsiveness of the control logic to the input material or other variations [2]. Consequently, maintaining appropriate ribbon attributes (both ribbon density and thickness) during process scale-up [3] is crucial in the development of solid dosage forms (tablet or capsule). Roller compaction as a unit operation has been studied extensively. Based on the Johanson model which describes the principle of ribbon formation, various methods [4-7] with the emphasis of matching normal stress betweens scales have been derived and utilized for scale up. However, since measuring the material properties required in the models is challenging and there are variations in mechanical conditions among roller compactors, design of experiment (DoE) at the commercial scale is still conducted to identify proper conditions for operating roller compaction. Such tedious experiments at commercial scales consumes valuable API due to the large batch size 567

Aiming to streamline transfers to commercial scales without impacting the reliability of the process, we have developed the probability based method that maximizes the usage of development scale data, and only requires one batch at commercial scale to define the acceptable range of operational parameters. 2. EXPERIMENTAL METHOD Materials and preparation A low drug loading formulation containing API X at 5% w/w was used in the study for the proposed transfer method. Other components in the formulation included microcrystalline cellulose (68.5%, FMC Biopolymer), lactose anhydrous (20%, Kerry Bio-science), crospovidone (4%, Ashland), silicon dioxide (1.5%, Evonik), and magnesium stearate (1%, Covidien). All ingredients except magnesium strearate were mixed in a bin blender for 108 revolutions, comilled through a Quadro Comil (with 0.6 mm screen opening) and mixed again in the bin blender for 120 revolutions with magnesium strearate. The milling operation was used for enhanced local mixing, rather than size reduction as previously reported [8]. Since lactose anhydrous, the largest ingredient in the formulation, had a particle size of about 200 µm, the particle size of blend was not expected to change after milling through the 600 µm opening. As API X had a small particle size (100% less than 30 µm) and was a cohesive material with a low bulk density of about 0.1 g/ml, the resulted powder had a poor mass flow at 5% drug loading and required roller compaction prior to tabletting. A drug loading higher than 5% increased risk of sticking, arching or rate holing during process. Roller compactor Alexanderwerk WP120 (pilot scale) and WP200 (commercial scale) roller compactors were used in the study. The major differences between the two roller compactors are roller width (120 mm vs 200 mm) and roller depth (40 mm vs 75 mm). While the WP120 uses a horizontal single screw feeder to deliver the powder from the hopper to the roller units, the WP200 uses a horizontal double screw feeder to accommodate the wider rollers. As a result, the WP200 can have a throughput up to 400 kg/hr compared to 40 kg/hr for the WP120. Pilot scale roller compaction development The process was developed on an Alexanderwerk WP120 roller compactor which has an incorporated milling operation (fine rotor granulators). A 3-factor central composite design or CCD (JMP 7.0.1, SAS Institute Inc.) as shown in Figure 1 was used. The inscribed option was checked, which limited the experiment points within the minimum and maximum of design parameters. The design was rotatable, a desirable property of response surface design. In the design, roll pressure

(hydraulic pressure, P, 50-70 bar), gap setting between the rolls (G, 2.0-2.6 mm), and milling speed (MS, 40-80 rpm) were studied. The response in this design is granules properties (flow, bulk density, and particle size distribution) as well as ribbon attributes (ribbon density and thickness). The development work was used to determine parameter ranges yielding acceptable granules or ribbons, i.e., flowability, compactability and tablet dissolution rate. As milling of ribbons proceeded ribbon formation, the 3-factor design could be used to study ribbon formation only, i.e., using ribbon density and thickness as the responses and only considering roll pressure and gap setting as the factors without milling speed. This transforms the design to a two-factor central composite design with factorial points repeated once and the center point repeated 6 times as shown in Figure 2. From each run, ribbon and powder samples were taken. Ribbon thickness was measured with a digital thickness gauge. Ribbon density was determined using the weight and volume of 1-inch ribbon disks cut from the ribbon samples [9]. Powders were characterized for particle size distribution with a sonic sifter. Bulk density was measured by the weight and volume of the power filled in a 100 ml glass cylinder. The flowability was determined by the mass flow rate of the powder going through an agitated tapered steel cylinder with an 8 mm opening at the bottom. The compatibility of granules was determined during tabletting, using the ratio between tablet hardness and the corresponding compression force. The tableting operation was set at a fixed target hardness of 14 Strong-Cobbs unit (SCU), thus compression force was adjusted based on the compactability of powder to yield the target hardness. A smaller compression force indicates a higher compactability of the granules. Once the development work concluded that ribbons made in the entire DoE study were acceptable based on the flow and compactability of the resulting materials and milling had no practical effect on the powder properties, the target and acceptable ranges of ribbon density and thickness were determined. Ribbon density and thickness were then expressed as a function of hydraulic pressure and the gap on the Alexanderwerk WP120 roller compactor using a statistical model.

Figure1. Experimental Design on Alexanderwerk WP120. A total of 18 experiments were conducted with 6 star points (star), 8 factorial points (diamond), and 4 replicates of the centre point (cross).

2.7

Roll Gap (mm)

2.5

2.3

2.1

Center point Factorial point Star point

1.9 45

50

55

60 65 Roll Pressure (Bar)

70

75

Figure 2. Experimental Design on Alexanderwerk WP120 for Ribbon Formation. Experimental design (ribbon formation) showing roll pressure and gap setting as the factors since milling occurs after ribbon formation and has no impact on ribbon formation.

Commercial scale roller compaction process transfer A calibration batch on a commercial Alexanderwerk WP200 roller compactor was manufactured at the target ribbon density and thickness. The corresponding roll pressure/gap from the Alexanderwerk WP200 was recorded. The data set from the calibration batch on the Alexanderwerk WP200 was used to identify the shift in the regression parameters of the statistical model derived from the DoE on the Alexanderwerk WP120. The redefined model for the Alexanderwerk WP200 was applied to project the acceptable operation ranges of roll pressure and roll gap on the roller compactor. Mapping experiments using the projected values were conducted to verify the accuracy of prediction from the redefined model on the Alexanderwerk WP200. This validated the one-point calibration method for process transfer between the two roller compactors. 3. RESULTS AND DISCUSSION Statistical analysis of granule properties from development experiment Statistical analysis of the DoE on the Alexanderwerk WP120 is shown in Table 1 with granule properties as the responses. The average particle sizes, percentage of coarse and fines were all derived from the particle size distribution test on the sonic sifter. The average particle size was arithmetic means of particles retained from each sieve. The coarse was the mass fraction retained on the largest (top) sieve (>840 µm, 20 mesh) while the fine was the mass fraction passing through the smallest (bottom) sieve (<53 µm, 270 mesh). The p values, representing the likelihood of the particular factor being statistically meaningful for a specific response, were obtained by fitting each response with a quadratic regression model in JMP. The blank cells indicate an insignificant effect associated with the factor-response pair (p value>0.1). Although some significant effects were observed between some factor-attribute pairs, there was no practical impact for the downstream process. In another word, the parameter ranges investigated were acceptable. It also suggested that the ribbon density and thickness generated in the study were acceptable, i.e., the resulted granules had flowability and compactability that were suitable for continuous tabletting operation and there were no impact on the dissolution behaviour of the resulted tablets.

Granule Properties

P

Bulk Density Flowability

G 0.0411

0.0005 0.0847

Coarse %

0.0733 0.0145

P*G

R*MS

G*G

MS*MS

0.0498 0.0580

Average Particle Size

Fines %

MS

0.0041

0.0705

0.0835

0.0746 0.0226

0.0044

Table 1. Statistical analysis (p values) of granule properties vs. roller compaction parameters.

Development of one-point transfer model for scaling up ribbon formation To transfer the ribbon formation process, the DoE was reanalyzed based on the design shown in Figure 2 using ribbon density and thickness as the responses. Although the quadratic and interaction terms were included in the original response surface model, it was found that these terms were statistically insignificant (p>0.10) and is thus removed from the final statistical model. The p values in the table represent a simple linear regression model without interaction. As expected, the ribbon thickness is dominantly controlled by gap setting and the ribbon density is primarily impacted by roll pressure.

Ribbon Properties

P

G

T

0.0599

<.0001

ρ.

<.0001

0.1722

Table 2. Statistical analysis (p values) of ribbon properties vs. roller compaction parameters.

The model fit for ribbon density and thickness is shown in Figures 3 and 4, respectively. The statistical equations are defined in Eq. 1 and Eq. 2, respectively. The unit of each parameter is shown in parenthesis. Only factors showing an effect with p<0.1 are considered in fitting the models for ribbon density and thickness.

Eq. (1) Eq. (2)

1.25

Ribbon Density (g/ml)

1.2

1.15

1.1

1.05

1 45

Circle: Experimental data points Line: Linear regression

50

55

60

65

70

75

Roll Pressure (Bar)

Figure 3. Effect of roller compaction on ribbon density (Alexanderwerk WP120). The ribbon density vs. roll pressure follows the typical linear relationship.

The slope in Eq. 1 suggests an interpretation of the compressibility (0.0046 g/ml/bar) of the powder blend feeding into the roller compactor. The two slopes in Eq. 2 suggests an interpretation of the plastic deformation (-0.0031 mm/bar) and elastic relaxation (1.18 mm/mm or 18% relaxation) of the powder blend, respectively. It is understood that powder undergoes both plastic and elastic deformation with an applied force during roller compaction [10]. While plastic deformation reduces ribbon thickness (the negative slope in Eq. 2), elastic deformation is recoverable and increase ribbon thickness via relaxation upon removal of the applied load (the positive slope in Eq. 2). All three slopes reflect powder blend properties, which are constant between the development scale and commercial scale batches, i.e., roller compactor independent. Such assumption of constant powder properties is valid, given that the API specification is in place and all other formulation components are widely used in the industry. In both equations, the intercept is assumed to be equipment specific, i.e., roller compactor dependent. The intercepts depend on the specific condition of the roller compactor, such as difference due to equipment model, routine maintenance and calibration. As the intercepts are the sole parameters that changes from equipment to equipment, one calibration experiment at the commercial scale is able to determine the shift in intercepts from the Alexanderwerk WP120 to the Alexanderwerk WP200.

ness (mm) Ribbon Thick

3.2 3.0 2.8 2.6 2.4

p Ga

2.2 2.5 2.4 2.3 2.2 2.1

m) (m

g ttin Se

2.0

45

55

60

65

70

75

Bar) ( e r su Pres l l o R

50

Figure 4. Effect of roller compaction on ribbon thickness (Alexanderwerk WP120). Ribbon thickness increases linearly with gap setting and decrease linearly with roll pressure.

The calibration batch was conducted on the Alexanderwerk WP200 at a ribbon density of 1.13 g/ml and thickness of 2.55 mm. The corresponding roll pressure and gap were 71 bar and 2.2 mm, respectively. The equations for the Alexanderwerk WP200 with adjusted intercepts are shown in Eq. 3 and Eq. 4.

Eq. (3) Eq. (4) Confirmation of the one-point transfer methodology Eq. 3 and Eq. 4 were used to predict the parameters (roll pressure and roll gap) for a parameter mapping study consisting of two batches. The first batch had 7 runs and maintained the 5% API loading. There were 5 runs in the second batch and a 2.5% drug loading was used. Given the low percentage of the API, such difference between the two batches were expected to have no impact on ribbons. These runs were designed to yield combinations of low (L), medium (M) and high (H) levels of ribbon density and thickness. The ribbon density and thickness were experimentally measured and compared to the predicted values using the actual roll pressure and gap observed at each process parameter set. The differences between the predicted and the obtained ribbon density and the thickness values also are included in Table 3.

Design

p

G

(Bar)

(mm)

Ribbon Density (g/ml) Obtained

Predicted

Ribbon Thickness (mm)

% Obtained Difference

Predicted

% Difference

Batch 1 LL

61

2.0

1.08

1.08

0.0%

2.27

2.4

-5.4%

LH

60

2.6

1.07

1.08

-0.9%

2.97

3.0

-1.0%

HH

80

2.6

1.15

1.18

-2.5%

2.95

3.0

-1.7%

HL

82

2.0

1.13

1.18

-4.2%

2.38

2.4

-0.8%

LM

61

2.3

1.04

1.08

-3.7%

2.68

2.7

-0.7%

HM

82

2.3

1.12

1.18

-5.1%

2.63

2.7

-2.6%

MM

73

2.3

1.09

1.13

-3.5%

2.69

2.7

-0.4%

Batch 2 LL

60

2.0

1.05

1.08

-2.8%

2.38

2.4

-0.8%

LH

59

2.6

1.08

1.08

0.0%

2.96

3.0

-1.3%

HH

82

2.6

1.12

1.18

-5.1%

2.93

3.0

-2.3%

HL

84

2.0

1.19

1.18

0.8%

2.39

2.4

-0.4%

MM

71

2.3

1.11

1.13

-1.8%

2.71

2.7

0.4%

Table 3. Model validation using process mapping (Alexanderwerk WP200).

As shown in Table 3, the obtained ribbon attributes agree with the predicted values. The R 2 from the linear regression of the predicted vs. the obtained values of ribbon density or ribbon thickness are 0.74 and 0.98, respectively. Figure 5 shows the linear plots of the predicted values vs. the obtained values.

Figure 5. Ribbon attributes predicted vs. obtained (Alexanderwerk WP200). The model predicts ribbon density (circle and solid line) and ribbon thickness (diamond and dashed line) with a maximum of 5.4% difference.

4. CONCLUSION The current investigation demonstrates the successful application of the one-point method for process transfer between different roller compactors. The study shows that the statistical model from development studies on the Alexanderwerk WP 120 can be used to define operation parameters on the commercial scale Alexanderwerk WP 200 with only one calibration run at the scale. The method streamlines process transfer by maximizing the understanding from development scale and can result in significant savings as DoE at commercial scale is not needed. The method is an efficient way to transfer the process between different roller compactors. There are other potential uses of the one-point transfer method. Modern instrument, like the roller compactor, encompass various electronics that measure and transfer data for hardware or software processing. As these electronic components age, drifting of the measured or transferred data is unavoidable. Such drifting results in discrepancy in the actual and displayed values of parameters. Meanwhile, the mechanic components can also wear over the time, another element for drifting. Such drifting results in machines that may operate at different parameters ranges even if the machines are the exact same model and made by the same manufacture. While severe drifting warrants recalibration of the instrument, very often slight drifting is overlooked since it is likely within the calibration tolerance. The current methodology can be used to set up a product specific baseline using the two intercepts in equation (3) and (4). Significant deviations in the actual ribbon density or thickness from the theoretical predictions suggest a change in equipment conditions.

LIST OF SYMBOLS ρ

Ribbon Density

[g/ml]

P

Roll Pressure

[bar]

G

Roll Gap

[mm]

T

Ribbon Thickness

[mm]

MS

Milling Speed

[RPM]

REFERENCES [1] P. Kleinebudde, Roll Compaction/dry granulation: pharmaceutical applications, Eur. J. Pharm. Biopharm. 58 (2004) 317-326. [2] G.E. Peck, J.L.P. Soh, K.R. Morris, Dry Granulation, : in L.L. Augsburger, S.W. Hoag (Eds.), Pharmaceutical Dosage Forms-Tablets, Third Edition; CRC Press, Boca Raton, 2008, 303-336 [3] T.J. Smith, G. Sackett, P. Sheskey, L. Liu, Development, scale-up, and optimization of process parameter: Roller compaction, in: Y. Qiu, Y Chen, G.Z. Zhang, L. Liu, L. Yu, V. Rao, W. Porter (Eds.), Developing solid oral dosage forms: pharmaceutical theory & practice development.; Academic Press, Burlington, 2009, 715-724. [4] R.T. Dec, A. Zavaliangos, J.C. Cunningham, Comparison of various modeling methods for analysis of powder compaction in roller press, Powder Technol. 130 (2003) 265-271. [5] V.V. Nesarikar, N. Vatsaraj, C. Patel, W. Early, P. Pandey, O. Sprockel, Z. Gao, R. Jerzewski, R. Miller, M. Levin, Instrumented roll technology for the design space development of roller compaction process, Int. J. Pharm. 426 (2012) 116-131. [6] A. V. Zinchuk, M.P. Mullarney, B.C. Hancock Simulation of roller compaction using a laboratory scale compaction simulator, Int. J. Pharm. 269 (2004) 403-415. [7] G. Reynolds, R. Ingale, R. Roberts, S. Kothan, B. Gururajan, Practical application of roller compaction process modeling, Comput. Chem. Eng., 34(2010), 1049-1057 [8] W. Shi, E. Galella, O. Sprockel, Macro- and micro-mixing of a cohesive pharmaceutical powder during scale up, Powder Technlo. 274 (2015) 319-323 [9] A.M. Miguélez-Morán, C.Y. Wu, H. Dong, J.P.K, Seville, Characterization of density distributions in roller-compacted ribbons using micro-indentation and X-ray micro-computed tomography, Eur. J. Pharm. Biopharm. 72 (2009) 173-182. [10] R.W. Miller, Roller compaction in: D.M. Parikh (Eds.), Handbook of pharmaceutical granulation technology, third edition; CRC Press, Boca Raton, 2010, 163-182

Graphical abstract

Obtained Thickness (mm) 2.0

2.3

2.6

2.9

3.2

1.3

Predicted Density (g/ml)

y = 0.9835x + 0.0843 2

R = 0.9794

2.7

1.2

y = 0.9718x + 0.0581 R2 = 0.7415

1.1

Ribbon density

2.3

1.9

Ribbon thickness 1 1.00

1.5 1.05

1.10 1.15 Obtained Density (g/ml)

1.20

Predicted Thickness (mm)

3.1