Investigation into the inherent variability of pharmaceutical excipients

The International Pharmaceutical Excipients Council (IPEC) defines excipients as, Substances, other than the Active Pharmaceutical Ingredient (API) in finished dosage form, which have been appropriately evaluated for safety and are included in a drug delivery system to either aid the processing or to aid manufacture, protect, support, enhance stability, bioavailability or patient acceptability, assist in product identification, or enhance any other attributes of the overall safety and effectiveness of the drug delivery system during storage or use. Although they do not produce medical effect, excipients have been proven to be essential in both biopharmaceutical and technical aspects. Due to their important roles, interchangeability and uniformity of excipients are necessary in order to achieve consistent quality in the finished products. All raw materials, including excipients, must be standardized to the acceptance of the regulatory authorities. However, there are examples of excipients meeting the monograph standard but performing differently during processing and in the final dosage form1. Many factors can contribute to the batch-to-batch variability within a source or between sources, such as differences in raw material, manufacturing process, storage conditions and transportation. Despite the multiple sources of variability, it has been suggested that the pharmacopoeia is primarily based on the verification of identity, purity and chemical stability with only limited testing on particle and powder physical properties, which can affect an excipients functionality. As a result, certificate of analysis of excipients may not provide sufficient confidence of equivalency between vendors or batches. In fact, excipients functionality may not only depend on their intrinsic properties but also their applications and formulation details. Therefore, it is controversial to include functionality or physical testing related to functional properties in the monograph due to the many different ways that an excipient can be used23. Nevertheless, a better understanding of the properties of excipients and their relationships to the functionalities can help formulators to select appropriate excipient and validate manufacturing process accordingly, hence improve process control and move towards more controlled products with consistent quality in line with the FDAs Quality by Design initiative. In fact, powders processibility and functionality can be influenced by many variables including physical and mechanical properties and environmental effects. It is well known that particle size, shape, and size distribution can impact on flow and compaction. Moisture content and electrostatic effects can also influence flow depending on the material and the environment, including humidity and storage conditions. Mechanical properties, for example elasticity,

plasticity, hardness, brittleness, can also have a profound effect on powders tabletting properties4. The aim of this paper is to investigate batch-to-batch, vendor-to-vendor variations of excipients and their impact on processing by studying the physical and function-related properties of the excipients, including particle size, specific surface area, surface energy, relative flowability, compressibility, compactibility and disintegration. Three commonly used excipients received from the major suppliers were chosen for characterisation, including lactose anhydrous, microcrystalline cellulose (MCC) and croscarmellose sodium (CCS). The three Avicel grades of MCC, PH101, PH102 and PH200, which were noted as having different particle sizes and processing behaviours, were also characterised and functionally compared.

Lactose anhydrous Lactose is a naturally occurring disaccharide, which is composed of one galactose and one glucose moiety bonded through a glycosidic link. Due to the presence of an asymmetric carbon, lactose can exist in two isomeric forms, or , which have different properties. Anhydrous lactose typically contains 70-80% anhydrous - lactose and 20-30% anhydrous -lactose. It occurs in white crystalline form with no water of hydration and is usually produced by roller drying a lactose solution, which is then milled and sieved to the desired size. It is widely used in direct compression tabletting process and employed as a capsule or tablet filler and binder, especially in moisture sensitive compound where low moisture content is desirable5. The popularity of lactose as an excipient can be contributed by its cost, availability, bland taste, low hygroscopicity, good compatibility with other ingredients, excellent stability and water solubility6. Microcrystalline cellulose Microcrystalline cellulose is derived from a naturally occurring polymer, -cellulose, obtained as a pulp from fibrous plant materials. It is comprised of linear chains formed by glucose units joining together by a 1-4 glycosidic bonds. It is widely used as tablet filler/binder in both granulation and direction compression. In addition to its low chemical reactivity, high plasticity and ease of availability, it also possesses some lubricant and disintegrant properties. MCC is manufactured by spray drying the aqueous slurry resulting from the acid-catalysed hydrolysis of cellulose and the commercial grades are produced by manipulating the spray drying process to vary the moisture content and the degree of agglomeration, hence particle size5. MCC is available in many different grades with differing particle size, shape and moisture content to meet the needs of the formulators. For example, higher PH grades, which are reported as having larger nominal particle sizes, have been developed to provide better flow properties than the conventional grades. Today, MCC products can be obtained from a range of suppliers in different countries. However, differences in the wood used as raw materials and manufacturing processes can attribute significantly to the source and batch variations. Manufacturer dependent and/or batch-wise variations have been reported in a number of molecular-related properties7,8 (such as crystallinity, lignin content, etc.), powder properties9,10 (such as particle size, surface area, etc.) and tabletting behaviour9,11 (such as crushing strength, friability, etc.). Moreover,

tablet characteristics, including drug dissolution rate12 and drug absorption13, were also found to be variable in formulations produced of MCC from different sources. Croscarmellose sodium Croscarmellose sodium is a cross- linked polymer made up of carboxymethyl cellulose sodium. It is insoluble in water but highly absorbent and therefore, it possesses good water wicking and swelling properties without losing its fibrous integrity. These enable its role working as a superdisintegrant at concentrations up to 5% in the production of oral formulations by direct compression or wet granulation. Multiple sources of croscarmellose sodium are now available, however unlike MCC neither batch-to-batch nor vendor-to-vendor variation has been studied extensively in the reason of its utility in only small fraction (normally 2-3%) within a formulation.

Materials The following batches of excipients were used in this study:

Vendor DMV- Fonterra Excipients GmbH & Co. Borculo Domo Lactose anhydrous Origin Grade Germany Netherlands SuperTab 21AN Lactopress 250 Lactose anhydrous NF DT Batch number 10421277 10428173 10444667 632751 633379 631678 1320010016 1320010017 1320010021 1320019937

Kerry Bioscience

USA

Vendor

Microcrystalline cellulose Origin Grade Avicel PH-101

FMC

Ireland

Avicel PH- 102

Avicel PH-200

JRS Pharma

Germany

Vivapur 102

Batch number 60811C 60727C 70745C 70709C 70728C XN07818924 P208819026 M0913C M0728C 5610291512 5610291009 5610290506 5610290203 5610288045 5610291612 5610285531

Vendor DMV

Croscarmellose sodium Origin Grade Netherlands Primellose

JRS Pharma

Germany

Vivasol

FMC

Ireland

Ac-Di-Sol

Batch number FN00023 FN00005 10417292 3201083113 3201083167 3201082080 3201082082 3201083104 3201084219 3201091027 T0937C T0731C T0850C

Methods Specific surface area Micromeritics Gemini 2390A surface area analyser was employed to measure the SSAs of the samples. The saturation pressure was needed to calculate the surface area and so was measured at the beginning of the tests each day. Prior to the surface area measurement, the samples were weighed accurately into sample tubes and outgassed overnight at 1000C to remove any gases and vapours that might have become physically absorbed onto the sample surfaces. The weight of lactose, MCC and CCS tested were 3g, 1g and 1.5g respectively. After degassing, the samples were allowed to cool to room temperature with flowing nitrogen and were then weighed prior to analysis. The specific surface area (m2/g) was determined by the physical absorption resulting from the van der waals forces between nitrogen and the absorbent surface of test samples using Brunauer, Emmett and Teller (BET) theory. The samples were evacuated at a rate of 500mmHg/min for 5 minutes and equilibrated for 5 minutes. Multipoint measurements (5 points) over the range of 0.06-0.02 were performed and triplicate results were generated for each sample. Particle size Samples were tested on the Malvern Morphologi G3 image based particle size analyser. They were dispersed using compressed air and allowed to settle on the glass plate which was then analysed with appropriate magnification lens. The settings for each class of excipient tested are detailed in table. Both geometric (volume-based) and arithmetic (number-based) particle size distribution were produced and triplicate results were generated for each sample. Some filtering settings, by means of convexity or circularity, were employed to remove any partially imaged and/or overlapping particles. An intensity filter, whereas the primary particles could be isolated through inversion of the intensity cut-off setting for the agglomerates, was also employed for the study of agglomeration in MCC samples.

Material

Dry air pressure (bar) 1.0

Magnification lens

Morphological filters applied (limits)

Lactose anhydrous

10X and 5X (3.5-420 micron resolution range) 10X (3.5-210 micron resolution range) 10X (3.5-210 micron resolution range)

Convexity (<0.9)

MCC

4.0

Whole population: Circularity (<0.45) Primary particles: Convexity (<0.8) + Intensity mean (>40) Agglomerates: Intensity mean (<40) Convexity (<0.85)

CCS

0.8

Surface energy Inverse gas chromatography (IGC) was employed to characterise the surface energy, which is often divided into polar ( P) and dispersive ( D) components. The samples were packed into 30cm (3cm inside diameter) silonised glass columns, plugged at either end by silonised glass wool. They were conditioned at 30oC (303K), 0% RH, 10sccm for three hours prior to analysis. The dispersive surface energy analysis was conducted by injecting a range of hydrocarbon probes; decane, nonane, octane, heptane, and hexane at 0.04p/po. The specific polar energy analysis was conducted by injecting a range of polar probes; acetone, acetonitrile, ethyl acetate, and ethanol at 0.04p/po. The column dead time was determined using an inert probe (methane at 0.04p/po). The acidic (ka) and basic (Kd) polar interaction were also determined for each sample. All samples were analysed in triplicate. Flow behaviours Angle of Repose Samples were tested on the Geldart Mark4 Angle of Repose tester. Due to its free flowing and stable behaviour, spray-dried mannitol was employed for calibration each day prior to the test. Temperature and relative humidity were also recorded. 100g of each sample was poured slowly onto the upper part of the chute using a small metal scoop. A motor was employed to generate a minimum degree of vibration required to aid the samples to slide down the upper chute and

feed into the metal hopper. As the powder left the lower chute and reached the L-shaped plastic base unit, a semi-cone with well-defined, sharp apex was formed. An electrostatic eliminator was also employed when the sample showed signs of static-like behaviour such as sticking to the lower chute and blocking the passage to flow. The height (h) of the semi-cone was read off and the average radius (r) of the base was determined by taking readings at 7 positions. (Fig) The samples were recovered after the test and triplicate results were generated for each of them. The AOR value was calculated from the following equation: AOR = tan-1(h/r)

Compressibility and compatibility The compression study was conducted on the Stylcam compression simulator 100R serial no. 0507104 (Medelpharm, France) which was fitted with 11.28mm round flat face plan punches using coupling plate no. 7. The tablets were made in one tablet mode using hand filling of the die by individually weighing each sample to the desired value (table). Before filling the die, the punches were lubricated using magnesium state slurry (2% magnesium stearate in acetone) which was painted on by hand. Each sample was compressed to a target solid fraction of 0.85. Literature true density values were used to determine the target tablet thickness (table). Six compacts were produced for each sample and once ejected, the accurate weight and thickness of each tablet was measured using an analytical balance and digital calipers. Tablet hardness was measured using a Schleuniger hardness tester HT1. Heckel analysis using an in-die method was employed to obtain the yield pressure (Py) of the particles. The strength of the compact, i.e. tensile strength, was calculated using the following equation: Tensile strength (MPa) = 2P/ (Dt) Where P= Hardness (N)

t= tablet thickness (mm) D= diameter of tablet (mm)

Literature true density (g/cm3) 1.589 1.59 1.543 2.389 Desired tablet thickness (mm) 2.96 2.96 3.05 1.97 Tablet Weight (mg) 400 300 400 400

Material Lactose anhydrous MCC CCS Dicalcium phosphate

Disintegration Tablets produced of pure dicalcium phosphate, 5% and 0.1% croscarmellose sodium were tested on disintegration immediately after the compression using the Stylcam compression simulator. Disintegration was conducted according to the USP method without disc at 37 1oC. Beaker filled with approximately 800ml distilled water was placed into a disintegration bath ZT502 (Copley, Copley Scientific Limited, Nottingham, UK). Six compacts were placed into the basket and moved in and out of the beaker at constant vertical motion until all of them had disintegrated. Disintegration time for each of the tablet was recorded. Statistical tests All statistical analyses were conducted using Minitab software (Minitab 15, Minitab Corp., PA). One-way analysis of variance (ANOCA) was carried out on the results. Interval plots were constructed to show differences between and/or within vendors. A p-value of less than 0.05 was considered to be significant.

Results and Discussion Lactose anhydrous Physical properties Specific surface area The specific surface area (SSA) results appear to indicate notable differences between the three vendors, with the highest SSA results found for the four Kerry batches and the lowest for the three DMV batches. Vendor batch-to-batch variability for both DMV and Borculo Domo materials was observed to be low whilst Kerry batches were noticeably more variable, with two lots (batch ref: 1320010021, 1320009937) having higher SSAs and the other two (batch ref: 1320010016, 1320009937) being lower (Figure X). Particle size The particle size data (Table X) was observed to show similar trends in variability between the batches and / or vendors as the SSA data, however no direct correlation between particle size and SSA was noted. This is probably due to the wide particle size distribution of lactose anhydrous. As the magnification lens used for particle size analysis had a resolution range of 3.5 to 400 microns, any particles small than 3.5 micron were outside the scope of analysis. However, the BET surface area measurements are especially sensitive to these fine particles with smaller particles having larger surface areas. Hence, the results may indicate that the three vendors contain similar numbers of coarse particles but different levels of fines, which contribute to the differences in the SSA data.

Inte l Plot of SSA rva

95% CI for theMe n a 0.6

0.5

SSA

0.4

0.3

0.2

77 12 42 10

73 81 42 10

67 37 16 17 21 46 99 00 00 00 0 1 1 1 44 00 00 00 00 10 32 32 32 32 1 1 1 1 ba re tch f

78 16 63

51 27 63

79 33 63

Lactose anhydrous vendor Batch reference 10421277 DMV 10428173 10444667 Borculo Domo 632751 633379 631678 Kerry Bioscience 1320010016 1320010017 1320010021 1320019937 SSA m/g 0.40 (0.00) 0.39 (0.01) 0.41 (0.01) 0.48 (0.00) 0.48 (0.01) 0.48 (0.01) 0.50 (0.00) 0.50 (0.01) 0.54 (0.00) 0.55 (0.01) Geometric particle size (m) Dv10 Dv50 Dv90 23.0 (2.5) 34.0 (8.5) 33.5 (7.1) 28.8 (4.5) 26.5 (1.6) 25.7 (1.5) 26.5 (2.9) 29.5 (3.9) 31.3 (2.7) 26.8 (0.8) 116.8 (14.8) 130.4 (16.8) 158.7 (2.0) 114.3 (5.5) 103.2 (16.4) 110.8 (10.4) 164.0 (25.1) 185.9 (11.0) 154.5 (18.6) 118.8 (17.6) 282.3 (14.0) 265.7 (15.6) 309.5 (23.7) 248.4 (52.8) 260.8 (86.2) 255.3 (34.6) 394.5 (59.3) 356.8 (47.6) 327.5 (3.7) 285.3 (14.4) Arithmetic particle size (m) Dn10 Dn50 Dn90 1.4 (0.1) 1.3 (0.0) 1.5 (0.1) 1.4 (0.0) 1.6 (0.1) 1.3 (0.1) 1.4 (0.2) 1.3 (0.1) 1.3 (0.0) 1.5 (0.0) 4.4 (0.4) 3.6 (0.1) 4.2 (0.6) 4.1 (0.3) 5.0 (0.1) 3.8 (1.0) 4.2 (1.1) 5.5 (0.3) 4.2 (1.1) 4.4 (0.6) 11.8 (0.6) 9.8 (1.0) 10.6 (1.0) 11.8 (1.3) 14.0 (0.5) 12.4 (1.1) 12.8 (1.1) 15.6 (0.3) 16.1 (1.3) 19.5 (0.7)

Surface energy Similarly to the SSA data, the dispersive surface energy (DSE) results indicate that the three vendors are significantly different from each others. The three DMV batches were observed to have noticeably higher DSEs than Borculo Domo while the lowest DSEs were for Kerry (with the exception of batch ref: 1320010021). Batch-to-batch variability of DMV and Borculo Domo was low but the Kerry batches were again found to be more variable with two lots (batch ref: 1320010021, 1320009937) were measured with higher DSEs and the other two (batch ref: 1320010016, 1320009937) being lower. The reason for the differences in DSEs of the materials is not fully understood, however, it may be explained by slight differences in mechanically induced amorphous content on the surface of the samples, which can be created during milling. Newell et al.14 reported that a small fraction (0.7%) of amorphous materials present on the surface of lactose can have a significant effect to the surface energies due to the very low hydrocarbon probe concentration (0.04p/po) used in the test. A slight variation in the polar surface energies between the three vendors was also reported. The variations in the heats of absorption for DMV and Borculo Domo batches was found to have a relationship with the measured SSA, probably due to varied availability of crystalline faces for probe interaction due to milling. The highest inter-batch variability was again for Kerry; however this material did not show the above relationship. The variations in surface energetics could be hypothesized to be due to the small changes in crystallinity or surface purity15.

Individua Va Plot of DSE l lue

95% CI for theMe n a 50.0 47.5 45.0 DSE 42.5 40.0 37.5 35.0
77 73 67 37 16 17 21 12 81 46 99 00 00 00 0 1 1 1 42 42 44 00 00 00 00 10 10 10 32 32 32 32 1 1 1 1 ba re tch f 78 16 63 51 27 63 79 33 63

Lactose anhydrous vendor Batch reference 10421277 10428173 10444667 632751 Borculo Domo 633379 631678 1320010016 Kerry Bioscience 1320010017 1320010021 1320019937
DSE (mJ/m2) 46.02 (0.22) 45.66 (1.35) 45.66 (0.56) 43.19 (0.54) 43.31 (0.54) 43.11 (0.68) 39.81 (0.44) 39.68 (1.78) 44.02 (0.81) 42.51 (0.83) Free energy of absorption (J/mol) Ethyl acetate Acetone Acetonitrile Ethanol 8970.0 8537.3 8624.9 8245.0 8340.3 8231.6 8616.1 9005.6 9173.1 9308.7 11924.1 11918.0 11875.0 11227.6 11275.0 11338.5 11426.3 11907.2 12537.3 12024.2 13266.2 13386.7 13262.7 12385.5 12492.4 12483.6 12945.4 13896.2 14375.5 14045.9 11638.5 11635.5 11683.4 10950.8 11123.4 11060.6 10973.1 11901.8 12441.3 11994.5

DMV

Function-related properties Flowability A minimum degree of vibration was needed to aid the flow of all the ten batches of lactose. Despite the difference in the levels of fines and surface energies observed, the AOR results show similar flow properties (AOR= 38-41o) in the ten batches of lactose, which would indicate that the materials are on the borderline between free flowing and a cohesive powders according to the manufacturers guide. This may indicate that the differences in particle sizes are not significant enough to affect a change in powder flow properties. However, it is important to note that the AOR tester is generally considered to be an unreliable tool for the prediction of powder flowability due to the lack of sensitivity to distinguish between two slightly different flowing materials4. Therefore, the results obtained in the study may not be able to relate to the real life powder flow properties. Compressibility and compactability Inter-batch variability appears to be low for all the vendors in terms of both compressibility and compactability. However, statistically significant differences between the suppliers were observed. The compressibility (yield pressure) data indicates that DMV and Borculo Domo batches possessed similar particle deformability behaviour while the four Kerry batches required a distinctly higher stress to initiate deformation. This could be resulted from a number of factors, including the fines content. As reported earlier, SSA data suggested that the four Kerry batches contained the highest levels of fines, which could accommodate and fill up the air gaps between the coarse particles. This may lead to a reduction in particle rearrangement during compression and therefore a greater pressure would be required to cause particle deformation. Moreover, it is generally concluded that lactose deforms predominantly by brittle fracture with some plastic deformation at the contact points. Therefore, a reduction in particle size may also contribute to the increased yield pressure (Py) by reducing the amount of fragmentation16. In terms of compactability, Kerry batches produced harder tablets with greater tensile strengths than DMV and Borculo Domo. Generally, the tensile strength increases with increasing bonding capacity of the powder. It has been demonstrated that the presence of small particles can fill up and reduce the void spaces, leading to the greater interparticulate contact areas and

hence producing higher compacted strengths within the tablets17. The highest tensile strength of tablets produced of Kerry batches may be due to its increased fines content.

vendor

Batch reference 10421277 10428173 10444667 632751 633379 631678 1320010016 1320010017 1320010021 1320019937

Lactose anhydrous AOR Yield pressure (MPa) (o)

39 2 39 1 38 1 39 1 41 2 40 1 40 2 39 1 40 1 40 2

Tensile strength (MPa) 1.7 0.1 1.8 0.0 1.9 0.1 1.6 0.1 1.8 0.1 1.4 0.1 2.4 0.2 2.4 0.2 2.5 0.1 2.5 0.2

136.9 0.7 135.2 1.2 136.1 1.7 133.2 1.9 134.3 1.6 130.4 2.1 148.4 1.4 144.7 1.5 146.2 2.3 144.0 2.1

DMV

Borculo Domo Kerry Bioscience

Microcrystalline cellulose (MCC) Physical properties

Particle size PH-102 The geometric (volume based) and arithmetic (number based) particle size distributions for the 12 batches of MCC PH-102 were also found to be indistinguishable and the two vendors were noted to be statistically equivalent (P=0.659). However, the geometric distributions were observed to be wide with a high degree of skewing to the fines end of the particle size range. When the particle images were viewed, two distinct classes of particles, fine primary particles and coarser agglomerates, were identified. This phenomenon was further analysed on another two grades of Avicel, PH101 and PH200, and the degrees of agglomeration were also compared.

Comparison of grades of Avicel (PH101, PH102 and PH200) According to the manufacturers particle size data determined by using laser light diffraction, the three Avicel grades, PH101, PH102 and PH200, are reported as having mean geometric particle sizes of 50m, 100m and 180m respectively. Using the Morphologi image-based analyzer, the median geometric particle size (Dv50) results obtained were generally in agreement with the manufacturers laser light diffraction data, indicating that both methods were measuring comparable particle populations. The only exception is one of the PH200 batch (ref.: M0913C) which has a significantly lower Dv50 than the one (Dv50= 224m) in the certificate of analysis (COA). A Dv90 value of 403m is noted on COA whilst no particle greater than 400m was identified in the three replicate samples. It is believed, and the trends in the arithmetic size data appear to concur, that the very coarse (>400m) particles, which have significant weighting in the geometric data, are present in such low numbers that relatively large sample sizes would be required to ensure detection. As such the Morphologi is much less likely to observe one of these large particles than a typical laser light scattering method. Nevertheless, the geometric particle size data show clearly that the geometric distributions shift to the coarse end of the scale as the Avicel series ascends from PH101 to PH200. The arithmetic distributions however do not show the equivalent shift in particle

population, indicating that the shift in the geometric data is not due to an overall increase in size of all particles within the population. The distributions instead show very similar modes but varied degrees of skewing towards the coarse end of the distribution which would appear to indicate that the main population of particle remain unchanged although with the addition of a small population of coarser agglomerates. To further investigate this hypothesis, morphological filtering was used to differentiate between the fine primary particles and the larger agglomerated particles. The filtered geometric data was found to show two overlapping normal distributions (Fig). Both distributions were found to be very repeatable for all PH102 batches tested, independent of vendor. All grades of MCC show indistinguishable geometric particle size distributions for the primary particle population (Fig, table), which contains approximately 97-99% of the number of particles in the population. The grades of MCC were found to differ in the particle size distribution for the agglomerated particles. The arithmetic particle size distribution (Fig) indicates a higher degree of tailing to the coarse end of the distribution resulting in the geometric particle size data (Fig, table) having a wider distribution in the higher PH grades. Both findings can be related to the presence of a very small population of significantly larger agglomerates in higher PH grades. The percentage of agglomerated particles in the whole population was noted to be 0.6%-2.4% however an increase in numbers with the higher MCC grades was not found to cause the differences in the geometric size data, but rather the width of the geometric agglomerate distribution (table). Arithmetic span?! This hypothesis is further supported by the scanning electron micrographs (SEM) (Fig.), showing clearly the presence of fine primary particles in all three grades, but differences in the sizes of agglomerates which are found only in small numbers. Particle sizes in PH200 appear to be distinctly separated into fine primary particles and very large agglomerates, with the absence of middle-sized ones. Geometric data is known to be sensitive to the presence of large particles as a cubic function is applied to the particle diameters in order to obtain the equivalent spherical volumes and as such a small change in the fraction of agglomerates could have a significant impact on the particle size data obtained. This suggests that an increase in geometric mean particle size may not necessarily mean a shift in size in the whole particle population as may be assumed. In fact, the three Avicel grades investigated are generally very similar with the only differences seen being the width of the distribution of the small population of large, agglomerated particles within the population.

Microcrystalline cellulose Vendor MCC grade Avicel PH101 Batch reference 60811C 60727C Avicel PH102 70745C 70709C 70728C FMC XN07818924 P208819026 Avicel PH200 M0913C M0728C 5610291512 5610291009 5610290506 JRS Vivupur 102 5610290203 5610288045 5610291612 5610285531 Geometric particle size (m) Dv10 Dv50 Dv90 30.8 71.2 119.3 (3.0) (6.3) (6.0) 33.6 72.3 124.1 (3.7) (7.1) (15.7) 33.8 86.1 182.8 (4.6) (23.5) (24.2) 43.5 132.1 255.2 (5.1) (32.3) (39.7) 45.1 128.0 212.8 (3.3) (30.8) (22.6) 39.1 99.1 209.6 (3.2) (5.9) (26.5) 41.7 100.4 197.8 (2.7) (1.4) (12.9) 40.9 122.1 261.7 (14.0) (27.2) (25.4) 55.6 160.5 306.0 (14.5) (45.5) (88.0) 35.7 83.6 201.2 (1.7) (13.3) (18.2) 44.7 121.2 223.3 (2.9) (7.4) (19.8) 39.2 98.7 214.0 (1.5) (8.4) (17.6) 41.7 100.1 219.9 (1.1) (6.6) (22.2) 44.1 122.9 233.3 (2.2) (25.2) (20.6) 44.8 119.9 229.2 (2.8) (9.7) (7.7) 38.0 92.5 213.2 (3.2) (12.4) (22.4) Arithmetic particle size (m) Dn10 Dn50 Dn90 1.3 4.6 21.9 (0.2) (2.1) (7.4) 1.5 5.4 30.8 (0.2) (1.7) (2.9) 1.2 3.7 19.9 (0.1) (1.8) (11.3) 1.2 3.3 19.1 (0.1) (1.2) (8.1) 1.2 2.7 18.5 (0.0) (1.0) (15.9) 1.3 4.9 27.5 (0.1) (0.9) (1.5) 1.2 3.1 20.3 (0.0) (0.1) (1.6) 1.1 2.3 9.8 (0.0) (0.5) (4.6) 1.1 2.2 7.8 (0.0) (0.3) (2.0) 1.2 2.5 15.6 (0.0) (0.6) (11.5) 1.2 2.5 15.9 (0.0) (0.3) (6.2) 1.2 2.8 24.6 (0.0) (0.5) (8.5) 1.1 2.5 18.4 (0.0) (0.2) (7.9) 1.1 2.5 19.7 (0.0) (0.3) (3.7) 1.2 2.8 20.2 (0.1) (1.2) (18.1) 1.2 3.6 23.0 (0.1) (2.2) (7.2)

Grade PH101

Batch reference 60811C 60727C 70745C 70709C 70728C XN07818924 P208819026 5610285531 5610288045 5610291512 5610291009 5610290506 5610290203 5610291612 M0913C M0728C Batch reference

PH102

PH200

Dv50primary particles m 50.0 48.2 49.5 49.6 48.8 48.9 51.4 50.8 54.0 47.6 54.4 54.6 56.0 53.0 47.1 48.4

Dv50agglomerates m 103.3 106.3 114.6 135.8 145.7 142.5 121.1 155.8 146.1 143.1 145.2 146.1 155.9 155.0 176.5 219.6 Geometric span of the agglomerates 0.68 0.91 0.96 1.17 1.17 1.06 1.16 0.85 1.02 1.00 1.06 0.91 0.96 1.33 1.42 1.29

Grade PH101

PH102

PH200

Arithmetic span of the agglomerates 60811C 0.68 60727C 0.83 70745C 0.88 70709C 1.00 70728C 1.07 XN07818924 0.84 P208819026 0.92 5610285531 1.14 5610288045 1.14 5610291512 1.12 5610291009 1.05 5610290506 1.20 5610290203 1.06 5610291612 1.01 M0913C 1.02 M0728C 0.88

Volume transformation: CE Diameter (m) smoothed over 60 points

0.9 0.8 0.7 0.6 0.5 % 0.4 0.3 0.2 0.1 0.0 1 10 CE Diameter (m) 100 1000

Record 3: Avicel PH102 #70709C Record 68: Avicel PH102 #70709C Int<40 Conv>0.8

Record 67: Avicel PH102 #70709C Int>40

Geometric frequency plot of Avicel PH102 batch 70709C (----), with separated to show overlapping agglomerated (----) and primary particle (----) distributions
Volume transformation: CE Diameter (m) smoothed over 60 points

1.0 0.9 0.8 0.7 0.6 % 0.5 0.4 0.3 0.2 0.1 0.0 1 10 CE Diameter (m) 100 1000

Record 63: Avicel PH101 #60811C Int<40 Conv>0.8 Record 68: Avicel PH102 #70709C Int<40 Conv>0.8 Record 80: Avicel ph101 #60806c Int<40 Conv>0.8

Record 65: Avicel PH102 #P208819026 Int<40 Conv>0.8 Record 72: Avicel PH200 #M0728C Int<40 Conv>0.8 Record 96: Avicel PH200 #M0913C Int<40 Conv>0.8

Geometric frequency plot of primary particles within MCC PH101, PH102 and PH200 samples

CE Diameter (m) smoothed over 60 points

1.2

1.0

0.8

0.6

0.4

0.2

0.0 10

100 CE Diameter (m)

1000

Record 69: Avicel PH102 #70728C Int>40 Record 115: Avicel PH101 #60727C Int>40

Record 71: Avicel PH200 #M0728C Int>40

Arithmetic frequency plot of agglomerated particles within MCC PH101 (----60727C), PH102 (----70728C) and PH200 (----M0728C) samples
Volume transformation: CE Diameter (m) smoothed over 90 points

1.0 0.9 0.8 0.7 0.6 % 0.5 0.4 0.3 0.2 0.1 0.0 10

100 CE Diameter (m)

1000

Record 69: Avicel PH102 #70728C Int>40 Record 115: Avicel PH101 #60727C Int>40

Record 71: Avicel PH200 #M0728C Int>40

Geometric frequency plot of agglomerated particles within MCC PH101 (----60727C), PH102 (----70728C) and PH200 (----M0728C) samples

PH101- 60811C

PH102- 70728C

PH200- M0728C

PH200- M0728C

Specific surface area Similarly to the particle size data, the SSA results for PH-102 indicate no noticeably difference between the two vendors, although JRS batches were shown to have slightly higher SSAs. Intra-vendor batch-to-batch variability was observed to be minimal. No significant difference in the SSAs of PH-101, PH-102 and PH-200 was recorded.

Surface energy The DSE results for the 12 batches of MCC appear to be variable with batch-to-batch variations observed for both vendors. The reason for the surface heterogeneity within and/or between suppliers is not fully understood, however the higher dispersive energy may be due to

differences in the availability of amorphous regions that are remnants from acid hydrolysis of cellulose. The thermodynamically less stable amorphous materials have been suggested to dominate the measured surface energy despite the fact that MCC is largely crystalline18.

Microcrystalline cellulose vendor Grade Avicel PH-101 Batch reference 60811C 60727C SSA (m/g) 1.07 (0.04) 1.16 (0.01) 1.09 (0.02) 1.16 (0.02) 1.12 (0.01) 1.05 (0.02) 1.05 (0.02) 1.15 (0.02) 1.20 (0.02) 1.22 (0.00) 1.26 (0.08) 1.17 (0.02) 1.12 (0.04) 1.26 (0.05) 1.25 (0.01) 1.25 (0.01) DSE (mJ/m2)

N/A 61.99 (0.39) 66.55 (0.58) 60.91 (2.10) 58.78 (0.56) 67.83 (0.38) N/A 60.63 (0.20) 59.36 (0.31) 59.71 (0.41) 55.84 (0.25) 64.56 (0.75) 59.87 (1.00) 60.48 (0.40)

70745C 70709C
FMC Avicel PH-102

70728C XN07818924 P208819026

Avicel PH-200

M0913C M0728C

5610291512 5610291009 5610290506

JRS

5610290203 Vivapur 102 5610288045 5610291612 5610285531

Function-related properties

Flowability All three grades of Avicel required a minimum degree of vibration to flow down the metal hopper, however two batches of JRS 102 (batch ref: 5610285531, 5610288045) were unable to form precise peaks as the powders accumulated due to static-like behaviour at the lower chute, blocking the passage to flow. This phenomenon was only observed on a day of relatively low humidity (RH=31%) but not observed when repeated on a day with a higher relative humidity (RH=59%), and was much improved using a static eliminator. The cause was believed to be electrostatic charge accumulation, which is dependent on various factors including a powders intrinsic properties and environmental conditions. Nevertheless, further investigation of this phenomonon was outside the scope of this report and so was not further investigated. The AOR results for PH102 show that FMC batches generally flowed slightly better than JRS batches but no significant inter-batch variability was reported in both vendors. Despite the presence in only small fractions within the samples, the agglomerates have shown to play a significant role in terms of flow. It is generally accepted that larger particles flow better than smaller ones and this is illustrated in the AOR study of Avicel, showing a rank order of flow properties: PH101<PH102< PH200, which is in agreement with the literature results. Taylor et al. observed a similar rank order but suggested that the relative difference was not as discriminating as predicted based on the large difference in the reported mean geometric particle sizes (50m vs 100m vs 180m). In fact, multiple studies have shown an improvement in flowability in line with increasing geometric particle size. Instead of shifting the size of the whole particle population, the difference in relative flowability may also be described by the hypothesized shift in only a small sub-population of large agglomerates discussed earlier. Rather than behaving as a glidant like silica dioxide, Jones19 suggested an improvement in flow properties in the presence of larger glidant particles by the mechanism of physically separating the finer particles and hence reducing the cohesive bridging. Lahdenp et al.20 observed that an addition of only small amount of PH200 to PH101 can be sufficient to improve flow of the latter.

Compressibility and compactability

The compressibility data indicates a lower batch-to-batch variability in Vivupur 102, which is also noted to have a statistically higher yield pressure than Avicel PH102. No correlation with the physical properties measured was observed. One hypothesis for this variation in compressibility is the different degree of polymerizations which can be resulted from the different wood types used for the production of MCC in the two suppliers21. MCC is well-known to deform primarily by plastic deformation and its compressibility is noted to be independent of particle size22. It is illustrated in the yield pressure results that no significant difference was observed between the three Avicel grades.

Looking at the compactability of the tablets produced of PH102, no significant difference was identified between FMC and JRS. However, inter-batch variability was noted for both vendors. In addition to plastic deformation, mechanical interlocking also plays a major role in the compactability of MCC. It is expected that an increase in particle size can reduce the interparticulate bonding area and so produce weaker tablets23. However, this trend was not observed in the three Avicel grades where the differences were generally insignificant such that they could be considered of similar compactability. Comparable results were also reported by Doelker et al., who observed identical compactability in unlubricated tablets produced of the three Avicel grades, particularly when taken into account of inter-batch variability 24. This also correlates with the surface area findings and possibly supports the hypothesis of similar sizes in the majority particle populations within the three grades. Another possible explanation is that the large agglomerates, consisting of small MCC primary particles, are weak enough to undergo fragmentation during compression, leading to the increased binding points and therefore no difference in compatibility would be detected. However, similar intrinsic compaction properties do not necessary mean similar compactability in lubricated MCC. A greater lubricant sensitivity has been noted in higher particle size grades, leading to the formation of tablets with lower tensile strengths24,25

Microcrystalline cellulose Vendor Grade Avicel PH102 Batch reference 70745C 70709C AORo 33 2 34 0 Yield pressure (MPa) 62.1 (0.69) 59.5 (1.15) Tensile strength (MPa) 8.11 (0.15) 7.74 (0.38)

70728C FMC XN07818924 P208819026 5610291512 Vivupur 102 5610291009 5610290506 JRS 5610290203 5610288045 5610291612 5610285531 Avicel PH101 FMC Avicel PH200 60811C 60727C M0913C M0728C

34 1 35 1 35 0 36 0 36 1 36 1 36 1 36 0 34 1 37 1 37 3 37 1 30 2 32 0

59.2 (2.14) 59.0 (1.08) 61.2 (0.61) 64.8 (1.00) 66.0 (1.31) 65.7 (0.66) 65.9 (0.42) 66.2 (0.65) 66.1 (0.85) 65.8 (0.76) 63.5 (0.62) 65.3 (0.51) 67.2 (0.81) 63.2 (0.62)

7.71 (0.28) 7.30 (0.12) 7.97 (0.12) 8.27 (0.20) 7.45 (0.13) 7.95 (0.23) 7.35 (0.18) 7.95 (0.18) 8.37 (0.36) 8.12 (0.15) 8.44 (0.19) 8.38 (0.30) 8.61 (0.21) 8.03 (0.24)

Croscarmellose sodium Physical properties

Specific surface area A notable difference in SSA results between the three vendors was reported with distinctly higher SSAs for the three Ac-Di-Sol (FMC) batches and the lowest for Primellose (DMV). Some degrees of batch-to-batch variations were observed in all three vendors but DMV batches were generally more consistent.

Particle size A statistically significant difference (P = 0.00?) in particle size distribution between the three vendors was observed. An ascending order in the geometric and arithmetic particle sizes was noted to be FMC<JRS<DMV, which correlates inversely with the SSA results. This relationship is expected as the smaller the particles, the greater the expected surface area. Intravendor batch-to-batch variability was however acceptable.

Croscarmellose sodium vendor Batch reference FN00023 DMV FN00005 10417292 3201083113 3201083167 3201082080 JRS 3201082082 3201083104 3201084219 3201091027 T0937C FMC T0731C T0850C SSA m/g 0.33 (0.01) 0.33 (0.02) 0.28 (0.01) 0.42 (0.01) 0.44 (0.01) 0.50 (0.01) 0.49 (0.01) 0.38 (0.01) 0.36 (0.01) 0.52 (0.02) 0.66 (0.04) 0.67 (0.01) 0.58 (0.01) Geometric particle size (m) Dv10 Dv50 Dv90 29.8 67.7 106.9 (1.5) (3.7) (6.0) 34.8 62.2 98.5 (2.4) (1.2) (1.8) 39.9 67.7 106.9 (1.9) (3.7) (6.0) 27.0 48.4 67.4 (2.5) (2.1) (3.3) 29.1 50.9 71.7 (0.8) (1.2) (0.7) 25.7 50.0 71.4 (3.1) (2.2) (1.1) 29.2 50.5 71.3 (1.8) (1.4) (1.6) 29.5 50.9 70.5 (1.3) (1.1) (1.4) 29.1 51.2 71.1 (0.3) (0.1) (0.9) 30.1 52.3 72.8 (1.1) (0.6) (1.1) 26.4 45.1 67.7 (2.0) (0.7) (0.5) 25.2 43.4 66.0 (0.2) (0.9) (1.4) 28.8 47.5 72.4 (1.9) (0.4) (1.4) Arithmetic particle size (m) Dn10 Dn50 Dn90 5.6 14.4 44.1 (0.5) (0.4) (1.6) 3.3 15.2 50.9 (1.1) (0.7) (2.9) 3.4 17.6 58.2 (0.6) (2.8) (0.9) 3.9 11.1 39.3 (0.2) (0.3) (3.1) 2.7 9.5 39.8 (0.5) (1.0) (2.0) 3.7 10.4 36.1 (0.2) (0.7) (4.8) 3.0 10.1 41.1 (0.4) (0.5) (1.7) 3.6 10.9 42.0 (0.1) (0.5) (1.8) 3.9 11.3 41.8 (0.3) (1.1) (1.4) 2.9 10.2 42.0 (0.3) (0.6) (1.0) 2.3 8.1 34.6 (0.1) (0.4) (3.0) 2.2 7.4 32.3 (0.4) (1.3) (2.4) 2.1 8.0 38.1 (0.1) (0.0) (2.7)

Surface energy Batch-to-batch consistency was reported in both DMV and JRS batches. The three DMV batches were observed to have significantly higher DSEs than JRS while the three FMC batches were too variable that no pattern could be seen. Despite the variability, the results show no correlation with the SSA or the particle size data.

Function-related properties

Flowability Although the three vendors were found to have statistically significant different particle size and surface area, no notable difference between the batches and/ or vendors was noted in the AOR data. Again, this might be due to the insensitivity of AOR tester as discussed previously; or it might simply because the differences in the particle size were not significant enough to cause a change in flowability. Compressibility and compactability Although croscarmellose sodium is usually used in low levels, typically less than 2%, as a disintegrant in tablet production, its compressibility and compactability have been tested to check for inter and intra vendor variability. A software error was encountered doing the Heckel analysis which might be resulted from the high compression forces (3000daN) required to produce 0.85 solid fraction croscarmellose tablets. Therefore, yield pressure (Py) values could not be measured accurately. The compactability results show a low degree of vendor batch-to-batch variability in FMC and JRS batches. Surprisingly, the data obtained for DMV batches were variable despite the reported consistency in the physical properties. Significant inter-vendor variability was also noted. FMC batches produced much harder tablets than the other two suppliers, which might be explained by the smallest particle size distributions measured. The finer the particles the greater the inter-particulate areas and the stronger the tablets produced. Nevertheless, croscarmellose sodium is normally incorporated at low level (2-3%) within a tablet and so it would not be expected that these differences would be significant enough to impact the compactability of a formulated blend.

vendor
DMV

JRS

FMC

Croscarmellose sodium Batch AORo Tensile strength reference (MPa) FN00023 31 2 2.7 0.2 33 2 FN00005 3.9 0.2 10417292 33 1 3.3 0.2 34 0 3201083113 2.8 0.1 3201083167 32 0 2.4 0.2 32 1 3201082080 2.9 0.1 3201082082 34 1 2.6 0.1 34 2 3201083104 2.8 0.2 3201084219 33 1 2.4 0.2 33 1 3201091027 3.2 0.1 T0937C 33 0 6.9 0.3 32 1 T0731C 6.7 0.2 T0850C 35 1 6.9 0.2

Disintegration Croscarmellose sodium works as a superdisintegrant mainly due to its water uptake and swelling properties. Particle size is one of the most important parameters affecting the swelling properties of a disintegrant. Larger particles are noted to be able to swell more extensively compared to smaller particles, and hence a shorter disintegration time may be expected26. Disintegration testing was not performed for all 13 bathces due to time limitations, however one batch of Primellose from DMV (batch ref: 10417292) and Ac-Di-Sol from FMC (batch ref: T0937C), which were noted to have particle sizes at the extremes of the data set, were tested for disintegration. Tablets compressed from pure dicalcium phosphate, which were used as a control, did not disintegrate at all (>1500s). This is expected due to the fact that dicalcium phosphate is practically insoluble in water and so will not disintegrate without the addition of disintegrant27. In the presence of 5% of croscarmellose sodium, both vendors functioned equivalently and the tablets disintegrate completely in few seconds as reported previously27. No differentiation could be observed even when the disintegrants were used in low level (0.1%). This would suggest that the differences in particle size were not significant enough to cause a change in disintegration. Similar observations by Zhao & Augsburger26 reported that Primellose and Ac-Di-Sol were functionally equivalent despite a significant difference in particle sizes. They suggested that the ratio of basic to acidic substituents is also important to

disintegrant swelling apart from particle size. Despite having smaller particle sizes, Ac-Di-Sol was reported with a high degree of basic substitution, leading to its increased hydrophilicity and hence swelling ability.

Conclusion Although the commercially available excipients achieve compendial specifications, this does not necessarily mean equivalent performance during processing and in the drug formulation. This report has started the work on the investigation of excipients batch-to-batch and brand-tobrand variability in terms of physical and functional-related properties. Some significant batchwise (such as lactose from Kerry) and vendor-dependent (as seen in lactose and croscarmellose) variations have been demonstrated. Despite of the inherent variability, the significance of their effects during real life processing and on the final dosage from cannot be confirmed and therefore, an advanced investigation, including testing on excipient-placebo or excipient-API, is required. Moreover, a continuous study on other commonly used excipients, for instance silicon dioxide and magnesium stearate, is also valuable to further enhance the understanding of materials and hence the quality of the final products.

Acknowledgements I would like to thank John Gamble for his constant help and support throughout the project. Mike Tobyn is thanked for his advice on the study and especially on statistical analysis. Mridul Majumder, from Pharmaceutical Ltd, UK, is also thanked for all the hard work on IGC. I am very grateful to Helen Toale, Vivienne Gray, and Michael Leanne for their help and contributions to this study.

Appendix Morphologi G3 system28 The morphologi G3 particle characterization system is used to measure the morphological characteristics, including size and shape of the particles. Using the image analysis technique, the sample is scanned and digital images are produced. Each particle within the sample is individually measured and recorded, allowing particle count and foreign particle detection. In addition to the volume based distribution as reported by the ensemble particle sizing methods, the morphologic instrument is particularly useful for detecting the relatively small particles (fines) by the application of number-based resolution. Volume basis means that the contribution each particle makes is proportional to its volume, i.e. large particles dominate the distribution and sensitivity to small particles to small particles is reduced as their volume is so much smaller than the larger ones. Number basis means that the contribution each particle makes to the distribution is the same, i.e. a very fine particle has exactly the same weighting as a very large one. Standard percentile readings, D0.1, D0.5 and D 0.9 are reported for both distributions. D0.1 is the size of particle below which 10% of the sample lies. D0.5 is the size in microns at which 50% of the sample is smaller and 50% is larger. D0.9 is the size of particle below which 90% of the sample lies. As the particle is not a perfect sphere, there are many ways to describe particle size as a single number. In the Morphologi system, particle size is described as Circle Equivalent (CE) diameter. The 3D particle image is captured as a 2D image and converted to a circle of equivalent area to the 2D image. CE diameter is reported as the diameter of this circle. Morphological filtering Morphological filtering was performed based on the following parameters: Convexity - the perimeter of the convex hull of the object divided by its perimeter An example is illustrated in figure and it is calculated as: Convexity = (Perimeter of A+B)/ Perimeter of A. It is a measure of how spiky a particle is and is used to remove any partially imaged or overlapping particles.

Circularity- the ratio of the circumference of a circle equal to the objects projected area to the perimeter of the object. It is calculated as: Circularity = (2 X ( X Area))/ Perimeter A perfect circle has circularity of 1.0 while a very narrow elongated object has circularity close to 0. It is also employed in the study to exclude any partially imaged and overlapping particles in MCC. Intensity Mean- the average of the pixel greyscale levels in the object and is calculated as: Intensity mean = ( Ii)/ N Where Ii Is the intensity value of pixel (i) and N is the total number of pixels in the particle An intensity mean filter is employed for the characterisation of agglomeration in MCC samples. The more agglomerated the particle the higher the intensity mean.

Stylcam compression replicator Compaction analysis can be performed using The Stylcam compression replicator (Medelpharm, France). Replicating the production tablet presses, tablet compression data can be obtained and hence the performance of the formulations at full scale production can be predicted. Having this information at early can allow formulator to improve the formulation, enhancing its performance and avoiding possible issues during production. Heckel Analysis Compression data (reduction in tablet porosity with applied pressure) is interpreted using Heckel analysis. The Heckel equation assumes that compression follows a first-order chemical reaction where the inter-particulate pores are the reactant whilst the densification is the product. The rate of change in density with respect to the applied pressure is therefore assumed to be directly proportional to the tablet porosity29. The equation is described as the followings: In (1/ (1-E)) = kP + A Where E is the tablet porosity and P is the applied pressure. A is a constant which is related to the particle rearrangement and fragmentation and k is the slope of the linear part of the relationship which represents the ability to densify by deformation. The reciprocal of k is calculated to represent the yield pressure (PY) for the particles, which allows an interpretation of the mechanism of deformation. The PY is defined as the stress at which particle plastic deformation is initiated (Aulton, 2002). This is the minimum pressure required to cause deformation of the material undergoing compression. A typical Heckel profile30 is illustrated in Figure and three phases can be separated: Phase I initial curvature representing particle fragmentation and inter-particle motion Phase II linear portion over a substantial range of applied pressure. The PY is calculated from the gradient of this portion Phase III decompression stage where tablet height expands, representating the increased tablet porosity

The Stylcam uses in-die method for Heckel analysis. The in-die method collects data during compaction of the powder and Heckel plots are automatically generated on the Stylcam for every compact, given that true density values and all relevant thickness, weight, diameter and hardness data has been entered. The Py is determined automatically by selecting a line of best fit at the Phase II linear section. Heckel equation is one of the most frequently used equations in pharmaceutical research for the prediction of tablet compression profile, however it is important to note that there is a certain amount of associated errors and limitations. It has been reported that the Heckel plot and the derived parameters are extremely sensitive to small errors in experimental conditions in variations such as true density values and so the results and conclusions should be carefully considered29.

References

Reier, G.E., Excipient Standardization: Users Viewpoint, Drug development and Industrial Pharmacy, 13(13), 2389-2407 (1987) 2 Rios, M., Debating Excipient Functionality, Pharmaceutical Technology, 2006 3 Moreton, R.C., Excipient Functionality, Pharmaceutical Technology, 98-119 May 2004 4 Amidon, G.E., Physical and Mechanical Property Characterization of Powders, in: Physical Characterization of Pharmaceutical Solids (Drugs and the Pharmaceutical Sciences, Volume 70) Eds: Brittain, H.G. 1995 5 Handbook of Pharmaceutical Excipients, 5th Edition, Pharmaceutical Press, 2006 6 Guo, J., Lactose in Pharmaceutical Applications, Drug Del. Technol. 2004 7 Landin, M., Martinez-Pacheco, R., Gomez-Amoza, J.L., Souto, C., Concheiro A., Towe, R.C., Effect of coutry of origin on the properties of microcrystalline cellulose, International Journal of Pharmaceutics, 91 (1993) 123-131 8 Rowe, R.C., McKillop, A.G., Isaka, H., The effect of batch and source variation on the crystallinity of microcrystalline cellulose, Int J. Pharm. 101 (1-2): 169-172 (1994) 9 Albers, J., Knop, K., Kleninebuddle, P., Brand-to-brand and batch-to-batch uniformity of microcrystalline cellulose in direct tableting with a pneumohydraulic tablet press, Pharm. Ind. 68, Nr. 12, 1420-1428 (2006) 10 Steele, D.F., Moreton, R.C., Staniforth, J.N., Young, P.M., Tobyn, M.J., Edge, S., Surface energy of microcrystalline cellulose determined by capillary intrustion and inverse gas chromatography, The AAPS Journal, Vol. 10, No. 3 (2008) 11 Williams, R.O., Sriwongjanya, M., Barron, M.K., Compaction properties of microystalline cellulose using tabletting indices, Drug development and industrial pharmacy, Vol 23, No. 7, 695-704 (1997) 12 Landin, M., Martinez-Pacheco, R., Gomez-Amoza, J.L., Souto, C., Concheiro, A., Rowe, R.C., Influence of microcrystalline cellulose source and batch variation on the tabletting behaviour and stability of prednisone formulations, International Journal of Pharmaceutics, 91 (1993) 143-149 13 Steele, D.F., Edge, S., Tobyn, M. J., Moreton, R.C., Staniforth, J.N., Adsorption of an amine drug onto microcrystalline cellulose and silicified microcrystalline cellulose samples, Drug development and industrial pharmacy, Vol. 29, No. 4 (2003) 475-487 14 Newell, H.E., Buckton, G., Butler, D.A., Thielmann, F., Williams, D.R., The use of inverse Phase Has Chromatography to Measure the Surface Energy of Crystalline, Amorphous, and Recently Milled Lactose, Pharmaceutical Research, Vol. 18, No. 5, 2001 15 Grimsey, I.M., Feeley, J.C., York, P., Analysis of the Surface Energy of Pharmaceutical Powders by Inverse Gas Chromatography, Journal of Pharmaceutical Sciences, Vol. 91, NO. 2 (2002) 16 Roberts, R.J., Rowe, R. C., The effect of the relationship between punch velocity and particle size on the compaction behaviour of materials with varying deformation mechanisms, J, Pharm. Pharmacol. 38(8): 567-571 (1986) 17 Rahmouni, M., Lenaerts, V., Massuelle, D., Doelker, E., Leroux, J., Influence of Physical Parameters and Lubricants on the Compaction Properties of Granulated and Non-granulated Cross-linked High Amylose Start, Chem. Pharm. Bull. 50(9) 1155-1162 (2002) 18 Swaminathan, V., Cobb, J., Saracovan, I., Measurement of the surface energy of lubricated pharmaceutical powders by inverse gas chromatography, International Journal of Pharmaceutics 312: 158-165 (2006) 19 Jones, T. M. Mechanism of flow improvement by addition of fine particles to bulk solids J Pharm. Sci. 1968; 57:20152016 20 Lahdenp, E., Niskanen, M., Yliruusi, J. Crushing strength, disintegration time and weight variation of tablets compressed from three Avicel PH grades and their mixtures, European Journal of Pharmaceutics and biopharmaceutics 43 (1997) 315-322 21 Shlieout, G., Arnold, K. Muller, G. , Powder and Mechanical Properties of Microcrystalline Cellulose With Different Degrees of Polymerization, AAPS PharmaSciTech 2002; 3 (2) 22 Celik, M. Overview of compaction data analysis techniques, Drug Dev. Ind. Pharm. 18 (1992) 767-810 23 Nystrom, C. , Alderborn, G., Duberg, M., Karehill, P.G., Bonding surface area and bonding mechanism-two important factors for the understanding of powder compactibility, Drug Dev. Ind. Pharm. 19 (1993) 2143-2196 24 Doelker,E., Massuelle, D., Veuillez, F., Humbert-Droz, P. Morphological, packing, flow and tableting properties of new avicel types, Drug Dev. Ind. Pharm. 21(6), 643-661 (1995)
25

Jones, T.M. Mechanism of flow improvement by the addition of fine particles to bulk solids J. Pharm. Sci. 1968; 57: 2015-2016 26 Zhao, N., Augsburger, L.L., The Influence of Product Brand-to-Brand Variability on Superdisintegrant performance A Case Study with Crocarmellose Sodium, Pharmaceutical Development and Techonology, 11: 179-185, 2006 27 Bolhuis, G.K., Eissens, A.C., Zoestbergen, E., DC Calcium lactate, a new filler-binder for direct compaction of tablets, International Journal of Pharmaceutics, 221: 77-86 (2001) 28 Morphologi G3 User manual , Malvern Instruments Limited, 2008 29 Sonnergaard, J.M., A critical evaluation of the Heckel equation, International Journal of Pharmaceutics 193, 63-67 (1993) 30 Comoglu, T, An overview of compaction equations, J. Fac. Pharm, Ankara, 36(2) 123-133, 2007