OR WAIT 15 SECS
Thermal effusivity and power consumption may help predict granulation end point in high-shear granulators. In this study, power consumption was monitored and compared with percent relative standard deviation (RSD) on thermal effusivity measured at-line. Lactose monohydrate, microcrystalline cellulose, and magnesium oxide were granulated, and the effect of load size on granule growth in a fixed-volume granulator was evaluated using three load levels. Load size, liquid addition rate, and impeller speed were measured, and the correlation among RSD on effusivity, power consumption, mean granule specific surface area, and granule compressibility index were determined.
Process analytical technology (PAT) involves the application of process tools (e.g., in-process monitoring devices), chemometrics, and process control techniques in the development and manufacturing environments. Collaboration between regulatory agencies such as the US Food and Drug Administration, the European Medicines Agency, Health Canada, the International Conference on Harmonization (ICH), and the industry has initiated the application of analytical tools, including automation in pharmaceutical process monitoring. The PAT framework and the ICH Q8 document have provided a basis for risk mitigation to achieve quality by design and process understanding. PAT also affords the industry a system for designing, analyzing, and controlling manufacturing through timely measurements during the processing of critical quality and performance attributes of raw and in-process materials and other processes with the goal of ensuring final product quality (1–3).
This study used a 33-1 fractional factorial design to elucidate the use of thermal effusivity and power consumption comparatively in predicting granulation end point in high-shear mixers. The factors included in the design were load size, liquid-addition rate, and impeller speed. The response variables were relative standard deviation (RSDs) on effusivity, power consumption (kW), mean granule specific surface area (SSAm), and Carr's index. These factors and response variables were selected on the basis of the available amount of research supporting the high level of sensitivity of these parameters during wet granulation (4–7). This study also would be a contribution to the ongoing scientific discussion about PAT in the pharmaceutical industry, where in-line, at-line, on-line, or in-process determination of process or unit-operation characterization is needed.
Wet granulation using high-shear mixers has become a mainstay operation in the pharmaceutical and food industries. This form of granulation enables intimate mixing of drug components with the help of granulating liquids such as water, starch paste, or a polymeric solution, in most cases. The mechanism of such intimate mixing involves the generation of liquid bridges between the powdery materials and the liquid through electrostatic forces or hydrogen bonding. As the liquid is dispersed, granule growth is affected, resulting in the formation of agglomerates. This process of granule growth ultimately approaches a state of dynamic equilibrium in which the granulating liquid becomes evenly distributed throughout the entire product mass. The continuous addition of the liquid with concurrent granulation would result in the collapse of the liquid bridges, a wet mass, and finally to overgranulation. It is therefore pertinent to end such a granulating event during the optimum massing period or equilibrium phase. Depending on the granulation type and the prevailing process conditions, the massing "equilibrium" phase could be as long as 5 min. During this massing phase, the agglomerates propagate and constantly improve their shapes and sizes. Hence, the granulation process may be terminated once the desired attributes (e.g., particle size) are obtained. The determination of the granulation end point during the massing equilibrium phase typically is conducted with the help of power consumption (or direct impeller torque), transducer measurements, or personal experience. A recent publication has indicated the application of other PAT tools such as acoustics emissions in granulation monitoring (8). A search of the literature revealed no published complete work that directly compared power consumption with any other currently available PAT tool.
Numerous publications have reviewed the use of power consumption (or direct impeller torque), transducers, or physical inspection for the determination of granulation end point (9–15). In most cases, reliance on these predictors depends on one's experience with such instruments and the ability to develop an acceptable level of confidence. Hence the establishment of an empirical value for the granulation end point is in some cases unattainable because of a lack of differentiation in the peak signals of power consumption (16). This study introduces thermal effusivity as an alternative end point predictor for wet-granulation processes. Such optimal granulation is hypothesized not only to possess desired granular growth, but also homogeneity of mix. Mathis et al. previously used thermal effusivity as a predictor of blend uniformity (17).
Thermal effusivity is a material property that combines thermal conductivity, density, and heat capacity. Hence, it can differentiate between solids, liquids, and powder components in a system based on heat-transfer properties (e.g., in a formulation) (18). As the system undergoes dynamic process (e.g., blending or mixing), the various components distribute and mix until a homogenous state is attained. In most mixing systems, overmixing would result in segregation or demixing (19). Determining the extent of mixing or blending using effusivity is predicated on how close or comparable the measured values are to each other (e.g., low RSDs). Hwang et al. demonstrated that a poorly mixed or unmixed blend would have a high RSD, and a homogenous blend would have a low RSD (20).
Lactose monohydrate, microcrystalline cellulose, and magnesium oxide all were of pharmaceutical grade and were generously supplied by Par Pharmaceutical, Inc. (New York, NY) Purified water, USP, was used as the granulating liquid. All pieces of equipment used were in acceptable calibrated states.
Table I: Formulation ingredients and composition.
All ingredients and their respective percentages in the placebo formulations are listed in Table I. The factorial design included nine batches that were processed in a high-shear granulator: three granulations at a 240-kg batch size, three granulations at a 210-kg batch size, and three granulations at a 180-kg batch size (see Table II). Two levels of water addition were used: 3.5 and 4.6 kg/min. Dry-powder mixing (premix) and wet massing were carried out using a 600-L high-shear mixer (Zanchetta Roto G, Romaco, Pompton Plains, NJ). Approximately two thirds of the total vessel volume was designated for material charge, and the remaining one third of the volume was void space to ensure proper movement of the material. The granulator had a two-speed, bottom-driven impeller (75 and 150 rpm) and a single-speed chopper (1200 rpm). On the basis of the study design for the nine runs, the impeller speed was operated at either low or high speed during wet massing. The chopper was operated in the "OFF" position during premix, but switched to the "ON" position during wet massing. Each bowl charge of materials was initially premixed in dry form at low speed and with the chopper off. The granulating liquid was then metered into the granulator through the liquid-addition port using a pressurized spray tank fitted with a full-jet nozzle. Liquid addition was carried out with concurrent granulation for 6 min. At the end of the water addition, the wet mass was granulated further for 3–6 min with the chopper on (see Tables I and II).
Table II: Fractional factorial design for the wet-granulation process.
Regression analysis. During the granulating phase, the design involved dependent variables y (namely, RSD on effusivity, power consumption, SSAm, or Carr's index for compressibility) and three independent or controlled variables x1, x2, and x3 (namely, load, liquid-addition rate, and impeller speed, respectively). The regression can be expressed as y = f (x1, x2, ... xn) for each dependent variable. The contribution of the process variables was compared using analysis of variance at p < 0.005. The main effects and interactions among the independent variables in the design were analyzed to elucidate the correlation between the power consumption and effusivity using statistical software (Minitab, State College, PA).
Power consumption. During each phase of the granulation, power consumption was monitored with an in-line instrument (Poly Power model 3030A, Valhalla Scientific, San Diego, CA). The instrument used three- to four-phase connections to determine the true power consumption based on the product load as well as physical changes that occurred in the material during granulation (see Figure 1). At the end of the granulation process, the data were downloaded onto a laptop computer for evaluation.
Effusivity. At each phase of the granulation, effusivity analysis was conducted at-line using representative samples taken from multiple locations in the granulator bowl. The effusivity test was conducted using the TC Probe (Mathis Instruments, Fredericton, NB, Canada). The instrument detected interfacial heat flow from materials and displayed the results on a laptop computer screen. The rate of heat transfer is a function of energy transfer according to the following relationship: effusivity = (kρα)1/2 in which k is the thermal conductivity (W/m × K), ρ is the density (kg/m3 ), and α is the heat capacity (J/kg × K). Based on the principle that various materials have different effusivities, a mixing event with several components can, at some periods in the process, show homogeneity (low RSDs). Before sample analysis, the instrument was calibrated using high-density polyethylene standard material. Each calibration was conducted in triplicate, and the sampling and testing for RSD on effusivity was conducted in quadruplicate. The value of RSD on the effusivity was indicative of the product's homogeneity or nonhomogeneity.
Fluid-bed drying. Each granulation was immediately transferred to a 700-L fluid-bed dryer (Aeromatic, GEA/Niro, Columbia, MD). The granulation was dried to a predetermined moisture content using product temperature as a guide. The product's temperature was measured with a portable datalogger (model OM-3000, Omega Engineering, Stamford, CT). The loss-on-drying was measured at 95 °C for 10 min using a moisture analyzer (Ohaus, Pine Brook, NJ). The dried granules were subsequently milled (CoMil model 194, Quadro, Waterloo, ON, Canada, with screen 2C039R03125 and a round impeller at 700 rpm).
Compressibility (Carr's) index. The bulk and tapped densities of the dried granules were determined using graduated cylinders and a mechanical tapping device. The Carr's indices for the experiments, which provided an indication of the flow characteristics of the powder, were determined as:
Mean specific surface area of the granules or particles. Particle-size analysis and specific surface area computation were conducted as previously described (2, 3). Attainment of the desired granulate particle size is indicative of an optimal granulation process. Particle-size analysis was conducted on 100-g samples using Gilsonic sieves (Gilson Co., Lewis Center, OH). Each sample was shaken for 11 min using 425-, 250-, 180-, 150-, and 75-μm standard sieves. Mean specific surface area was calculated as:
in which xi is the weight of a sieve portion in grams and pi is the mean geometrical diameter of the granules of a given sieve portion, which is equal to p1 × p2 where p1 and p2 are the opening sizes of the two sieves sandwiching a given portion.
Results and discussion
Power consumption. Researchers have discussed using power consumption for predicting wet granulation end point (10–13). Accordingly, this experiment aimed to determine wet granulation end point by measuring true-power consumption (kW). Such determination would be compared with the results of effusivity from the samples collected at intervals during the granulation process. Temperature change during dry to wet granulation was monitored to evaluate its effect on the entire process.
Figure 1 shows the power-consumption profile of the impeller during the dry-mix, liquid addition, postaddition granulation, and additional granulation stages to effect overgranulation. For all the granulations at the three load sizes, the dry-mix phase generally had fairly constant true-power use at a nominal value of 12.5 kW. The fluctuation in kilowatts was attributed to the presence of lumps in the materials, which was observed visually during bowl charging.
Figure 1: Power consumption profile during granulation at the 210-kg batch size. The x-axis represents the time of granulation; the y-axis represents the true power consumed during granulation; A is the dry mix phase; BâC is liquid addition with initial massing phase; CâD is massing phase with peak granulation point; and DâE is over-granulation or breakdown of agglomeration.
The initial high-kilowatt value reflected the work done by the impeller blades to break up these lumps (Stage A). The second block showed the fusion of the liquid-addition and massing phases (Stages B–C). This phase initially showed slow increment in true-power consumption, followed by a significant rise in true-power value. This change in the value of true power is attributed to the progressive increase in wetting through the formation of liquid and solid bridges (by means of electrostatic, van der Waals forces and hydrogen bonding) between the powder materials and the granulating liquid. This change in the value of true power also is believed to reflect the distribution of the granulating fluid across the powder universe. Once this occurred, the powders continued to mass (Stages B–C) until a peak value of approximately 20.0 kW was reached. This stage signified the area of optimal massing or densification (Stages C–D). Additional granulation at the postequilibrium phase resulted in a decrease of the power-consumption values, demonstrating a breakdown of the already formed granules with subsequent overwetting (Stages D–E). Further granulation beyond this stage did not yield any further rise in true-power consumption. Comparable patterns were observed with respect to power consumption for all granulations used in the study.
Effusivity. The individual effusivities of all components used in the granulation process are listed in Table III. Water had the highest effusivity value (1600 Ws1/2 /m2 K), and air had the lowest value (5.5 Ws1/2 /m2 K). The computed average effusivity value for the components (523 Ws1/2 /m2 K), clearly, was different from the experimental effusivity value from a wet granulation (454.7 Ws1/2 /m2 K) (see Figure 2). In a typical mixing or granulation process, the ingredients rarely are mixed in equal quantities. Hence, this difference was not unexpected and was speculated to be the result of numerous factors, including the physical changes in the particle size resulting from agglomeration, liquid imbibition, formation of liquid and solid bridges, changes in particle–particle interfacial temperature, densification, and the presence of electrostatic forces and hydrogen bonding.
Figure 2: Computed and experimental values of average effusivity.
Effusivity sensors identify an ingredient on the basis of that material's unique properties. As such, in an unmixed state, a computed average effusivity value tends to be different from the experimental value from granulation. A possible explanation for this difference could be that the ingredient particles tend to mix in various proportions as they attain homogeneity of mix. Another explanation for this difference could be that it is a result of the particle attrition and aeration that occur during processing events such as blending and mixing. This is similar to the progression from divergence toward convergence or uniformity (see Table III and Figure 2).
Table III: Effusivity values of ingredients and components.
RSD on effusivity was generally low (<2.0%) at the end of the 3-min dry mixing (see Figures 3–5). For all batch sizes (180, 210, and 240 kg), the high RSD on effusivity correlated with the high fluctuation in power consumption as recorded at the onset of the wet-granulation process (i.e., liquid addition). These variances in the values of effusivity and power consumption appeared to reflect a nonhomogenous mix. It is believed that the granulating liquid (in this case, water with an effusivity value of 1600) increased both the average effusivity value and the variance. As more liquid was imbibed by the powdery mass, the authors observed the resultant formation of liquid and solid bridges as well as a lower RSD (<5.0%). Although further granulation beyond the massing phase (Stages D–E) failed to elicit a rise in true power, the RSD on effusivity was significantly lower for the 240-kg batch size. This low value for RSD on effusivity at the end of the additional granulation phase for this batch size was attributed to the extended granulation time required to compensate for the high fill volume that exceeded the manufacturer-recommended working capacity for the granulator. The extended granulation time was necessary to achieve homogeneity of the mix.
Figure 3: Plots of power consumption (kw) and relative standard deviation (RSD) on effusivity at 240 kg.
Power consumption versus RSD on effusivity. Comparative analysis of the power-consumption and effusivity data gave the best correlation at the 210-kg batch size based on a curvilinear response (p < 0.05). Figures 3–6 show the correlation of power-consumption values and RSDs on effusivity at the 180-, 210-, and 240-kg batch sizes.
Figure 4: Plots of power consumption (kW) and relative standard deviation (RSD) on effusivity at 210 kg.
For the 210-kg batch size, low values of power consumption at the dry-mix stage corresponded to lower values of RSD on effusivity. In general, the RSD during the liquid addition and granulation (wet-massing stage) showed a sharp rise in their values, comparable with the increases in power consumption that were typically observed for this stage. Nonetheless, a significant drop in the RSD on effusivity was observed with additional granulating time (see Figure 3). This period of low RSD on effusivity was perceived to be indicative of homogeneity within granulation.
Figure 5: Plots of power consumption (kW) and relative standard deviation (RSD) on effusivity at 180 kg.
Regression analysis. The use of power consumption, torque, or transducer to predict wet-granulation end point is based on the assumption that the load size as well as the physical changes that occur in the material during granulation would elicit some measurable energy-related response. In addition, the yardstick measuring how efficiently a wet-granulation process was conducted would depend on the acceptable granular growth or particle size. Based on these principles, general linear and quadratic models were generated for each response parameter (RSD on effusivity, kW, SSAm, and Carr's index). Each response parameter listed in Table IV was fitted to both linear and quadratic models. The regression coefficients for each term in the model are summarized in Table V, together with the r2 correlation coefficient of the regression model. The SSAm value of the granules was most responsive for a linear model (see Table V). The authors concluded that the combination of desired granular growth and homogeneous mix appeared to be key indicators for a successful granulation event (see Table IV).
Figure 6: Regression plot for relative standard deviation (RSD) versus power consumption for the 210-kg batch size.
Although the Carr's indices indicated good compressibility, the fitted regression had poor coefficient of determination r2 and adjusted r2 values (r2 = 50.7%, adjusted r2 = 21.1%). The r2 is a measure of the linearity of the regression equation, and the adjusted r2 signifies how well the linearity fits within some degree of freedom. In addition, analysis of the response parameters using the quadratic models indicated that the model was inappropriate for the design. On the other hand, the linear regression model using SSAm gave good r2 and adjusted r2 values (r2 = 88.3%, adjusted r2 = 81.3%). Hence, the regression equation is expressed as:
SSAm = 0.064 + 0.013 load - 0.003 impeller speed - 0.002 liquid addition rate
Regression analysis showed that the load size had the highest impact on particle size (p < 0.005) and both the impeller speed and liquid-addition rate had less impact (see Table V). The analysis of variance in Table VI showed that the regression model selected for this design was appropriate (p = 0.009) and that the equation was definitive.
Table IV: Factors and responses for the regression models.
In addition, Pearson's correlation coefficient, which is a measure of linear relationship, was used to further elucidate the relationship between the power consumption and the percent RSD on effusivity. The 240-kg batch size gave a poor correlation coefficient of 0.227, while the 210-kg and 180-kg batch sizes gave excellent values of 0.995 and 0.989, respectively. Results further indicated that the batch size (load) influenced process performance (see Tables V and VI).
Table V: Regression analysis: SSAm versus load, liquid rate, and impeller speed.*
The relationship between power consumption and effusivity was evident from the analysis of the main effects and interactions within the experimental factors (i.e., load size, liquid-addition rate, and impeller speed). Based on the main effects of the factors, a correlation was seen between load size and RSD on effusivity (see Figure 7). Increases in load size resulted in higher RSD on effusivity. Similar increases in liquid-addition rate and impeller speed, however, produced lower RSD on effusivity. This effect was attributed to the fact that higher loads required longer time to attain the same level of homogeneity as smaller loads. In addition, faster addition rate and higher speed resulted in quicker miscibility. Interaction plots showed further evidence of correlation between power consumption and RSD on effusivity (see Figures 7–9).
Table VI: Analysis of variance.
A faster liquid-addition rate produced low RSD on effusivity for all load sizes. The power consumption barely changed for the 240- and 180-kg batch sizes. On the other hand, the 210-kg batch size showed a dramatic rise in power consumption, hence better mixing action. This effect could be explained by the fact that this load size may be more appropriate for the granulator size.
Figure 7: Main effect of the factors, with relative standard deviation (RSD) on effusivity as the response.
End point prediction during wet granulation in a high-shear granulator with various load sizes was attempted using comparative responses of true-power consumption and relative standard deviation on effusivity. Colinearity was established between the observation from power consumption and RSD on effusivity. This was based on the interpretation of the regression analysis and the interactions among the load size, liquid-addition rate, and impeller speed. Load size was the most significant factor (p < 0.05) under a fixed granulator volume. Depending on the load, granular growth—and hence particle size of the final product—was influenced. The mean granule specific surface area (SSAm) was the most responsive variable (p = 0.009) and confirmed the desired results obtained from the particle-size analysis.
Figure 8: Interaction between the factors, with relative standard deviation (RSD) on effusivity as the response.
Although this study was based on placebo formulations, the authors believe that understanding the factors that affect granulation process and the interpretation, when using process analytical technology tools such as power consumption and thermal effusivity, would guide future studies involving active pharmaceutical ingredients.
Figure 9: Interaction between the factors, with power consumption as the response.
The authors express gratitude to S. Matz and N. Ragunathan, PhD, for their valuable reviews and comments.
Greg Fariss is a validation scientist and Roger Keintz is a manager of technical service, both at Par Pharmaceutical (Spring Valley, NY). Patrick Okoye* is a global PAT field application specialist at Mathis Instruments Ltd., 21 Alison Blvd., Fredericton, NB, Canada, tel. 845.566.6061 or 845.598.4906, fax 425.660.6061, email@example.com.
*To whom all correspondence should be addressed.
Submitted: Sept. 9, 2005. Accepted: Feb. 27, 2006
Keywords: wet granulation, high-shear granulation, thermal effusivity, power consumption, process analytical technology, manufacturing
1. US Food and Drug Administration, Guidance for Industry: PAT— A Framework for Innovative Pharmaceutical Development, Manufacturing, and Quality Assurance (Rockville, MD, Sept. 2004), http://www.fda.gov/cder/guidance/index.htm, accessed May 5, 2006.
2. FDA, Innovation and Continuous Improvement in Pharmaceutical Manufacturing: Pharmaceutical CGMPs for the 21st Century (Rockville, MD, Nov. 2004), http://www.fda.gov/cder/guidance/index.htm, accessed May 5, 2006.
3. J. Berridge, Q8 Pharmaceutical Development, International Conference on Harmonization draft no. 4.3: Step 2, Nov. 2004, http://www.fda.gov/cder/guidance/index.htm, accessed May 5, 2006.
4. T. Schaefer et al., "Granulation in Different Types of High-Speed Mixers, Part I: Effects of Process Variables and Up-Scaling," Pharm. Ind. 48 (9), 1083–1089 (1986).
5. M. Meshali, H. El-Banna, and H. El-Sabbagh, "Use of a Fractional Factorial Design to Evaluate Granulations Prepared in a Fluidized Bed," Pharmazie 38 (5), 323–325 (1983).
6. H. Lieberman, L. Lachman, and J. Schwartz, Pharmaceutical Dosage Forms: Tablets Vol. 2 (Marcel Dekker, New York NY, 2nd ed., 1990), pp. 245–348.
7. T. Chirkot, "Scale-Up and End Point Issues of Pharmaceutical Wet Granulation in a V-Type Low-Shear Granulator," Drug Dev. Ind. Pharm. 28 (7), 871–888 (2002).
8. S. Badawy, T. Lee, and M. Menning, "Effects of Drug Substance Particle Size on the Characteristics of Granulation Manufactured in a High-Shear Mixer," AAPS PharmSciTech 1 (4), article 33 (2000).
9. New PMA-Advanced High Shear Mixer–Granulator With New GMP and PAT Features,www.pharmaceuticalonline.com/content/news/article.asp , accessed June 10, 2005.
10. P. Holm, T. Schaefer, and C. Larsen, "End-Point Detection in a Wet Granulation Process," Pharm. Dev. Technol. 6 (2), 181–192 (2001).
11. M. Kopcha, G. Roland, and W. Vading, "Monitoring the Granulation Process in a High-Shear Mixer-Granulator: An Evaluation of Three Approaches to Instrumentation," Drug Dev. Ind. Pharm. 18 (18), 1945–1968 (1992).
12. L.L. Augsburger et al., "A New Approach to Scale-Up of a High-Shear Granulation Process," Pharm. Technol. (supplement), 2–10 (Oct. 1996).
13. K. Terashita et al., "Analysis of End Point with Power Consumption in a High-Speed Mixer," Chem. Pharm. Bull. 38 (7), 1977–1982 (1990).
14. Comparison of Wet-Granulation Processes Using Top and Bottom-Drive Granulators,www.vectorcorporation.com/papers, accessed May 5, 2006.
15. A. Royce et al., "Process Control and Scale-Up of High-Shear Wet Granulation," J. Proc. Anal. Technol. 2 (2), 8–16 (2005).
16. T. Hlinak, "Granulation and Scale-Up Issues in Solid Dosage Form Development," Am. Pharm. Rev. 3 (4), 33–36 (2000).
17. N. Mathis et al., "Monitoring Blend Uniformity with Effusivity," Pharm. Technol. 26 (4), 80–84 (2002).
18. N. Mathis, "Effusivity as a Blend Uniformity Technique," retrieved from www.mathisInstruments.com, Aug. 2002.
19. H. Vromans, H. Poels-Janssen, and H. Egermann, "Effects of High-Shear Granulation on Granulate Homogeneity," Pharm. Dev. Technol. 4 (3), 297–303 (1999).
20. R. Hwang, Y. Huang, and G. Peck, "Evaluation of Blend Sampling Errors: A Statistical Approach," Pharm. Technol. 23 (6), 56–65 (1999).