2022 Volume 45 Issue 11 Pages 1706-1715
Wear-resistant polymers and ceramics-based media have been used to pulverize the bulk powder of poorly water-soluble drugs to nanoscale size in conventional milling; however, contamination of such media is still an issue in the context of drug formulation manufacturing. In the present study, we developed a novel method for pulverizing the particles of a poorly water-soluble drug, ketoprofen, to nanoscale size by mixing mannitol and polypropylene glycol as a safe pulverizing medium. The ketoprofen nanoparticles were prepared using a Hoover automatic muller, equipment that traditionally has been used for the mixing of paint and ink. This process represents a novel application of this machine for the on-demand preparation of nanoparticulate formulations for use in the clinical setting. The optimal composition of the drug formulation was determined by designing an experiment consisting of the central composite design and responsive surface method. We obtained a design space that yielded ketoprofen nanoparticles with targeted particle size, poly-dispersity index, and drug release properties. We validated the manufacturing conditions by preparing ketoprofen nanoparticles in four compositions. Thus, the present study provided useful information regarding not only simple and effective contamination-free milling but also the experimental conditions need to produce nanoparticles of a poorly water-soluble drug.
Improvement of the drug dissolution and bioavailability properties of poorly water-soluble drugs (e.g., compounds belonging to classes 2 and 4 of the biopharmaceutics classification system) has been attempted for many decades by the pharmaceutical industry and researchers; indeed, approximately 40% of marketed drugs and 70–90% of future drug candidates exhibit poor water solubility.1) The typical technique for improving these properties is the downsizing of drug particles.2) The increased surface area of drug particles that results from decreasing particle size can increase the rate of drug dissolution, as explained by the Noyes–Whitney equation.3) Build-down methods (e.g., dry & wet milling4) and high-pressure homogenization5) and bottom-up methods (e.g., nanoprecipitation6)) have been employed and serve as familiar strategies for the preparation of drug nanoparticles.7,8) Various novel and efficient methods to prepare drug nanoparticles have been explored as part of the search for techniques capable of altering the innate physical properties of drugs.
Conventional downsizing methods for the preparation of poorly water-soluble drugs and drug candidates employ media wet milling due to the scalability and universality of these techniques. For example, Elan’s nanocrystal technology has been used to prepare drug nanoparticles with improved bioavailability as oral drug formulations.9,10) However, contamination due to the effects of wear on beads of medium is a potential problem in the process of wet milling.11) Given that bead-bead collisions are a dominant reason for bead wear, wear-resistant polymers and ceramics (e.g., polystyrene, zirconia) have been applied, but for such high-energy interactions, wear still occurs even with these wear-resistant materials. In contrast, novel milling methods using safe pulverizing media have been reported to prevent contamination with pulverizing medium. For example, Sugimoto et al. conducted cryo-milling of a poorly water-soluble drug using dry ice beads as the pulverization medium; the distinct advantage of such beads is that the dry ice readily can be removed or sublimated in the final steps of the process.12) A patent granted to Activus Pharma mentions the use of sugar alcohols as the pulverizing medium13); sugar alcohols are medical excipients and can be removed by “washing out” of the formulation as necessary. These methods are ideal solutions to avoid the contamination intrinsic to strategies that use conventional milling methods. In extension of this approach, we tested use of the sugar alcohol mannitol (Man) as a contamination-free grinding medium for the pulverization of ketoprofen (Ket), a poorly water-soluble drug.
In this study, a Hoover muller (HM) was used for the milling of drug particles (Fig. 1). The HM is a pigment-dispersing machine that conventionally has been used for the mixing of ingredients in the preparation of paints and inks.14) In these processes, the pigment and its vehicle are placed between two disks, and rotation of the disks provides mixing, kneading, and dispersing (i.e., mulling) of the components. Analysis of the mechanics of materials have shown that the properties of fluids (e.g., pastes) in the HM, including the shear stress (σ), shear rate (D), and torque (M), are described by Eqs. 1, 2, and 3, respectively.15) The relevant parameters, as shown in Fig. 1A, include the gap between disks (h), the angular velocity (ω), the radius of the fluid (R), the arbitrary radius of the fluid (r), and the apparent viscosity of fluid (η). The resulting properties of the fluid then can be used to calculate the shear force (f), as shown in Eq. 4.
 | (1) |
 | (2) |
 | (3) |
 | (4) |
We anticipate that lab-scale and small-scale HM machines may be applicable in novel clinical settings in which hospital pharmacists prepare drug nanoparticle formulations, although pharmaceutical application of HM remains (to our knowledge) poorly investigated. For instance, one potential application would be the generation of an oral spray formulation containing a non-steroidal anti-inflammatory drug, for use in the clinical treatment of oral mucositis, a condition that frequently is reported as a side effect of chemotherapy in patients with cancer.16) Separately, a group described the preparation of a suspension of rebamipide colloid nanoparticles for use as a mouth wash for the treatment of stomatitis.17) In extension of these examples, we hypothesized that the on-demand preparation of Ket nanoparticles using a HM likely would be of use for the treatment of oral mucositis in patients with cancer, given that such a formulation would increase the analgesic effect of the dissolved drug in the patient oral cavity.
In the current study, milling of Ket was conducted with Man and a small amount of an auxiliary agent, propylene glycol (PG), using a HM; the optimal experimental conditions for nanosizing were investigated further using a quality-by-design (QbD) strategy. The QbD approach has been used not only for the reproducible manufacturing of medicines but also for accelerating the ability of formulations to pass quality inspection in pharmaceutical factories, permitting drugs to reach the market much more quickly. The QbD approach includes processes permitting clarification of quality target product profiles (QTPPs), decisions on critical quality attributes (CQAs) that are required for QTPPs, selection of sets of critical material attributes (CMAs) and of critical process parameters (CPPs) based on risk assessment, and the preparation of design space (DS) as the results of the design of experiment (DoE).18,19) The DS is the feasible region of optimized experimental conditions and ensures the quality of medicines and drug formulations. In the current study, this methodology was applied for the preparation of a drug nanoparticle-based formulation through HM. The pulverization of drug particles prepared by HM is expected to be affected by the drug composition and the drug’s innate physical properties; thus, optimization of the drug formulation will be necessary for each drug. The DS was constructed by DoE based on the responsive surface method (RSM), and the robustness of the DS was confirmed by assessing several aspects of the DS.
Ket, Man, PG (guaranteed reagent; mass fraction, 99%<), and sodium dodecyl sulfate were purchased from FUJIFILM Wako Pure Chemical Corporation (Osaka, Japan).
Preparation of Samples Treated with the HMThe sample was prepared using an automatic HM (Imoto Machinery Co., Ltd., Kyoto, Japan). In brief, appropriate ratios of Ket, Man, and PG were added to a mortar, and these components then were mixed manually in the mortar, in a light and gentle manner, to form a lump that was designated the physical mixture (PM). The samples were transferred to the center of the lower disk of the HM and the upper disk was set in place and fixed in position. The sample then was mulled (milled) and rotated at 100 rpm for 10 min without any load. HM samples were further dissolved with water and centrifuged to remove the Man and PG in the supernatant. This washing process was repeated several times. The resulting pellets were frozen in a deep freezer set to a temperature below −80 °C, and then freeze dried using an FD1000 lyophilizer (Eyela, Tokyo, Japan) to get the powder samples composed of aggregate of Ket nanoparticles.
Scanning Electron Microscope (SEM)Ket bulk powder and Ket nanoparticles were observed with an SEM (S-4300; Hitachi, Tokyo, Japan). The resulting powder samples were coated with Pt–Pd using an E-102 ion sputtering device (Hitachi).
Particle Size MeasurementIn a typical experiment, a small amount of HM sample (approximately 5 mg) prepared as described above was dissolved in 800 µL of 0.1% sodium dodecyl sulfate solution to allow the resulting Ket particles to disperse in a disposable cell. The particle size and poly dispersity index (PDI) of the samples were analyzed by the dynamic scattering method using a ZetaSizer (Malvern Instruments, Ltd., Malvern, U.K.).
Dissolution Test and Determination of Ket ConcentrationDissolution tests were conducted using a United States Pharmacopeia dissolution apparatus II (paddle method). In brief, 900 mL of pure water was added to the dissolution glass vessel, and the temperature was maintained at 37 ± 0.5 °C with stirring at 50 rpm. Then 10 mg of Ket, PM, or HM samples containing 10 mg Ket were added (separately) to the vessel. At the indicated times, 4-mL samples were collected. The Ket concentration was determined by measuring the absorbance at 254 nm using a UV-1800 spectrophotometer (Shimadzu, Kyoto, Japan).
Software and DoEDoE was conducted using Design Expert software (Version 12; StatEase, Minneapolis, MN, U.S.A.). For the RSM, a face-centered central composite design (CCD) was applied by selecting a value of alpha = 1. Then two factors with three levels (Man at 400 mg, 1200 mg, or 2000 mg; PG at 150 µL, 250 µL, or 350 µL) were selected after preliminary experiments and a total of 13 experimental points was set, while three responses (particle size, in nm; PDI, no units; drug dissolution amount, in %) were assigned. Following the compositions, the samples were prepared using the HM, and data for the particle diameter, PDI, and drug release at 15 min were collected. Contour plots for a fitted model were obtained against particle size, PDI, and drug release. The DS then was constructed by overlaying the three contour plots.
Powder X-Ray Diffraction (PXRD)The PXRD peak pattern was obtained using a benchtop Rigaku MiniFlex 600 instrument with Cu Kα radiation (Rigaku, Tokyo, Japan). The tube voltage and tube current were set at 40 kV and 15 mA, respectively, and the PXRD data were recorded between 5 and 33°.
Differential Scanning Calorimetry (DSC)The DSC peaks of the samples were obtained using a DSC-60 instrument (Shimadzu). Using a micro-spatula, samples were transferred with care into a small aluminum pan (Shimadzu), which then was covered with an aluminum lid (Shimadzu). The container was closed tightly using an SSC-30 sealer & crimper (Shimadzu). The sample then was placed in a furnace, and the atmosphere was purged and maintained with N2 (20 mL/min). The sample was heated at 10 °C/min and analyzed over a range of temperatures from 30 to 200 °C.
In the current study, a contamination-free milling method was employed to prepare nanoparticles of a poorly water-soluble drug (Ket nanoparticles) using Man and PG, components that are considered safe pharmaceutical excipients. Milling using an HM is a useful method for preparation at low energy, given that this procedure does not produce heat while permitting the generation of particles smaller than 500 nm under some experimental conditions. A SEM image of Ket bulk powder before pulverization is provided in Fig. 2A, and a SEM image of Ket nanoparticles after pulverization and removal of Man and PG after washing with water is shown in Fig. 2B. On-demand preparation of Ket nanoparticle suspension (milling time, 10 min) may be a useful option for the treatment, in clinical settings, of patients with oral mucositis.
(A) Ket bulk powder. (B) Ket nanoparticles treated with HM. The milling ratio of Ket : Man : PG was 100 : 1699 : 221 (mg/mg/µL), which was the optimized ratio as explained below.
As shown in Fig. 1B, Ket particles were pulverized by milling using a HM in the presence of Man with a small amount of PG. While the illustration of milling process in the disk of HM is shown in Fig. 1A and described by Eqs. 1–4, it is difficult to estimate the shear force at the microscale because the actual event of milling is dynamic, given the change in particle size that occurs during milling. Presumably, PG was adsorbed by and/or coated onto the surface of Man particles and Ket particles, which then were attached to each other and ground by collisions resulting from the shear stress and by shearing action occurring between the walls (disk surfaces) in the HM.15) We consider that the solubility of Ket in PG is minimum. Filippa et al. exhibited the solubility of Ket in organic solvent (Log S; water, −3.203 < PG, −0.302 < methanol, 0.232 < ethanol, 0.267).20) The milling of the Ket particles presumably was ended by the disintegration of the Man particles themselves, resulting in the depletion of larger Man particles during pulverization of the Ket particles. Adequate amounts of PG coating on the surface of powder particles likely lead to the efficient milling of Ket particles. In contrast, the increased surface area of the Man particles and Ket particles by milling may reduce the amount of PG coating the particles, reducing the efficiency of Ket particle pulverization. While the milling method in the present study is very different from conventional milling methods (in which the size of beads of the medium is constant), the mannitol-based medium employed in the present study changes its size depending on the milling. Thus, the use of mannitol may contribute to increased milling efficiency of the drug particles, resulting in the production of drug nanoparticles at a lower energy. The net effect is that the milling process is affected by both the composition of the drug formulation and the innate properties of the drug being formulated. However, the various aspects of this milling process remain (to our knowledge) poorly understood and investigated, given that this process has been referenced solely in a patent.13)
The present study focused on optimization of the experimental conditions used to prepare Ket nanoparticles; specifically, these conditions were investigated using DoE and QbD approaches that have been employed elsewhere for the robust control of manufacturing processes in industries. The use of DoE for the preparation of nanoparticles by milling has been reported in several instances in the literature.21–24) The DoE includes the output parameters of DoE such as particle size, particle size distribution, and the released amount at dissolving, and the input parameters derived from the milling equipment were set as the factors affecting the output. In this study, the QbD approach was employed based on our prior knowledge and preliminary experiments, as described below.
QbD Approach and the Construction of DS to Optimize Ket Nanoparticle Formulation by HMQTPPs and CQAs are shown in Table 1. QTPPs are intended to optimize the final quality of the drug formulation, and CQAs are defined to achieve QTPPs. In the present work, the QTPPs included not only the preparation of drug particles with nanoscale size and homogeneity but also the improvement of drug dissolution. To these ends, particle size, PDI, and the amount of drug dissolution were identified as the parameters to be employed in our CQAs. The Ket particle size was an important parameter because the dissolution of this drug, which is otherwise poorly water-soluble, is expected to be improved by downsizing the particle to the submicron- and nano-scale, per the Noyes–Whitney equation.25) PDI is an essential parameter for evaluating the size deviation of nanoparticles. Nanoparticle suspensions with large size distributions tend to exhibit aggregation due to Ostwald ripening, creating a drug concentration gradient between the small and large particles.11,26) The aggregation of nanoparticles affects, in turn, drug dissolution. The third parameter of the CQAs was the amount of drug dissolution. Drug dissolution is a critical factor, affecting the absorption, bioavailability, therapeutic effect, and side effects of the drug. In the present study, we assumed that the Ket nanoparticles would be employed in the future for use in an oral spray for the clinical treatment of oral mucositis. The dissolution of drug molecules from drug nanoparticles is important for estimating the local analgesic effect. After the preliminary experiment, we collected the samples at 15 min after the dissolution test to identify improvements in drug dissolution.
 |
In the present study, we considered CMAs and CPPs, which are known to have an important influence on CQAs. As shown in Table 2, regarding the composition of the drug formulation, the amount of Man (Factor 1) and PG (Factor 2) were incorporated as the factors of the CMAs. While three levels of both Man (400, 1200, 2000 mg) and PG (150, 250, 350 µg) were set after the preliminary experiments, the amount of Ket was fixed at 100 mg to understand the influence of the ratios of Ket: Man: PG. Regarding the CPPs, the rotation speed was fixed at 100 rpm due to the nature of the HM, and the total time was set 10 min (i.e., 1000 rotations total) based on the results of our preliminary experiment.
 |
For example, the composition (ratio) of Run 1 was Ket: Man (Factor 1): PG (Factor 2) = 100 mg: 2000 mg: 150 µL. A total of 13 experiments were conducted by the CCD as DoE. The size (Response 1) and PDI (Response 2) of Ket nanoparticles and amount of released drug at 15 min (Response 3) were determined.
Face-centered CCD was adopted for the DoE in the current study due to the simple experimental design, permitting the selection of three levels for both factors. Experiments were conducted using the assigned conditions listed in Table 2, and the responses (particle size, PDI, and the amount of drug dissolution at 15 min) were obtained. As shown in Fig. 3, contour plots were obtained as responses to the factors, based on RSM, which was suitable for the visually understanding of the relationship between the factors and responses. Incorporation of larger amount of Man enhanced the pulverization of Ket particles at the nanoscale (<300 nm) (Fig. 3A). According to Eq. 4, viscosity (η) of lump sample increases with an increase of Man, and thereby, the increased viscosity of lump sample produces the shear stress (f), resulting in the smaller Ket nanoparticle. The tread was observed in Fig. 3A. In contrast, we guess that excess amount of Man may decrease the actual radius of the lump (R) which are mulled, resulting in the decrease of shear stress. This means that the amount to be loaded between the disks of HM is limited. In contrast, changes in the amount of PG had minimal effects, and the use of excess PG appeared to reduce the degree of pulverization in the presence of larger amounts of Man. Regarding the PDI, the optimal ranges of the amounts of Man and PG could be determined (Fig. 3B). The trend of PDI was partially similar to that of particle size, the optimal area seems to be shifted to the left (smaller amount of Man). We guess that nano-pulverization with higher shear stress (upper right area in Fig. 3B) may produce nanoparticle with larger particle size deviation in this experimental condition. Drug release was highly affected by the amount of Man, with the higher incorporation of Man increasing drug release; in contrast, the influence of PG on drug release was minimal (Fig. 3C). Overall, elevation of the amount of Man in the milling process with the HM tended to reduce the particle size and increase drug release.
(A) Particle size. (B) PDI. (C) Drug release. Experimental parameters are described in Table 2.
To understand the statistical reliability of the current results and the derived contour plots, a fitting summary of the model was generated (Table 3A). All responses of size, PDI, and drug release with various amounts of Man and PG were fitted with a quadratic model that reflected the curve derived from the contour plots. The p values of the quadratic model were <0.05 for each response, confirming the significance of the relationships suggested by the model. The responses of size and drug release were well-fitted, as shown by the similarity of the respective adjusted and predicted R2 values. In contrast, the response of PDI showed a high adjusted R2 but a low predicted R2, indicating that the model is not well-suited to the prediction of PDI.
 |
(A) Suggested-fit model and R2 values. (B) Regression coefficient of quadratic model and the p-values against the null hypothesis (coefficient = 0).
Analysis of the coefficients of intercept for Man and PG for the coded equations from RSM is shown in Table 3B. In this case, the coded equation was efficient in identifying the relative impact of the factors as assessed by comparing the factor coefficients, given that the standard partial regression coefficient of multiple regression analysis indicates the impact of the factors. Equations 5–7, which can be used to predict the output of samples, are provided below. (Note that the last digit after the decimal points represents rounding of the values).
 | (5) |
 | (6) |
 | (7) |
The coefficients of A in particle size; A2 in PDI; and A, B, and AB in drug release, were significant (p < 0.05), as shown in Table 3. These results mean that the Man amount affects the three parameters, such that high amounts of Man exhibit a positive effect in the preparation of the current drug formulation. In the present study, PG did not affect the response significantly under the tested experimental conditions. Nonetheless, we note that the PG amount is important for preparation of the current drug formulation, given that the use of lower amounts of PG (i.e., excess amounts of Man) prevents optimal grinding using the HM: PG is needed to help bind the Man and Ket powder sample, facilitating the formation of a lump. This point may be demonstrated by the fact that the p value of the coefficient of AB was less than 0.05 in drug release, a result that is reflected in the contour plots. In other words, while the PG amount did not affect the drug release in the presence of low amounts of Man, the PG amount did affect the drug release in the presence of high amounts of Man (Fig. 3C).
Next, the contour plots in Fig. 3 were overlaid, and the DS was obtained as shown in Fig. 4. The DS represents the feasible region of factors meeting the specified criteria of the CQA. In this study, the DS region was defined as follows: particle size <300 nm, PDI <0.35, and amount of drug dissolution >60% at 15 min. Four data points (Formulations A–D, as described in Table 4A) were used for the validation study below.
Points A–D correspond to Formulations A–D in Table 4A.
 |
Comparisons of predicted values and actual values of (B) size, (C) PDI, and (D) amount of released drug at 15 min.
After obtaining the DS (Fig. 4), the validity of the DS was investigated by choosing four random points of composition in the DS, including boundary points and inner points of the DS. Table 4 provides a summary of the four experimental points (Formulations A–D) in the DS based on the mean particle size, PDI, and drug release amount of the Ket nanoparticles obtained in these HM samples. The respective predicted and actual values of these parameters for each formulation of Ket nanoparticles were similar, such that the actual value of size, PDI, and drug release were within 95% prediction interval. The relative error (actual value/predicted value, expressed as %) of size in DS was smaller, while that of drug dissolution was larger, which is due to the limited data points. Then, the range of the 95% prediction interval of size was broader, while those of the PDI and drug release were relatively narrow and similar. Given that the production of nanoparticles less than 100 nm was difficult using the current method, we consider that the actual range to set is limited and that more data point can enhance the prediction.
Dissolution profiles of Ket from representative formulation samples are shown in Fig. 5A (for Formulation B) and Supplementary Fig. 1 (for Formulations A, C, and D). All four HM samples released more than 60% of Ket in the validated formulation at 15 min, meeting the established criteria. In contrast, Ket alone (bulk powder) exhibited less than 5% drug release after 180 min, consistent with the known poorly water-soluble nature of this drug. Although PM samples (i.e., samples obtained before HM milling) exhibited lower drug dissolution than HM samples, the PM sample exhibited remarkably higher drug dissolution than Ket alone (bulk powder). This difference reflected the wettability of PG, such that the air voids on the surface of the Ket particles were infiltrated by PG.
The data for drug dissolution are presented as the mean ± standard deviation (n = 3). (B) XRD peaks and (C) DSC peaks of Ket, Man, PM, HM, and Ket nanoparticles after HM treatment and washing. The composition of PM, HM, and Ket nanoparticle samples correspond to Formulation B in Table 4A.
The crystallinity of the Ket nanoparticles was investigated using XRD, as shown in Fig. 5B (for Formulation B) and Supplementary Fig. 2 (for Formulations A, C, and D). Ket bulk powder and Man powder exhibited high crystallinity. However, Ket-derived peaks were absent in the HM samples. As the Ket-derived peaks also were absent in the PM samples, we infer that the disappearance of the Ket-derived peak was due to the low-level incorporation of Ket in the drug formulation. To understand the crystallinity of Ket after HM processing, the HM samples were further dissolved with water to remove the Man and PG, permitting purification of the Ket nanoparticles away from these excipients, although we could not exclude the possibility that washing process may affect the crystallinity partially. The peaks of the purified Ket nanoparticles were similar to those of the Ket bulk powder. The presence of similar peaks was confirmed in Ket nanoparticles purified from Formulations A–D. These results suggested that the process of milling with the HM did not affect the crystallinity of the Ket component of the samples. DSC peak analysis was conducted to confirm the XRD results (Fig. 5C). The peaks of Ket and Man bulk powders occurred at 92 and 169 °C, respectively. The peaks in the PM and HM samples were similar to that of Man, except that the peaks in the PM and HM samples had broader peaks. We infer that the evaporation of PG (as expected for a polyol compound) may have partially affected the measurement, given that the boiling point of PG is 188 °C.27) In contrast, the peak of purified Ket nanoparticles exhibited a peak similar to that of Ket bulk powder. These XRD and DSC results suggested that the influence (on drug crystallinity) of pulverization of Ket particles to nanoscale size using the HM is minimal. Other reports of processes that used wet willing and cryomilling instruments have mentioned changes in the polymorphs and amorphous of the drugs being formulated, including albendazole,28) ranitidine,29) and indomethacin.30) Our method may have an advantage regarding the point of control of crystallinity of nanoparticles, given the relatively weak milling force employed in our process.
In conclusion, we demonstrated a unique milling method for preparing drug nanoparticles using a contamination-free pulverizing medium in HM equipment. The amount of Man and PG affected the resulting particle size and PDI of the Ket particles and the drug release as outputs. The production of Ket nanoparticles was controllable within the DS, and the range of suitable compositions of drug formulations to produce Ket nanoparticle was determined as the DS using DoE. The method described here is not expected to change drug crystallinity. This kind of DoE strategy likely will be of use for producing nanoparticles of poorly water-soluble drugs, although further experiments will be necessary to investigate the universality of this method and the application for other poorly water-soluble drugs. The results of current study provide useful information about the milling method, and the process may hold promise for the preparation of drug nanosuspensions in clinical settings.
Marin Ikuse: Investigation, Data curation, Visualization, Writing-original draft, Methodology, Formal analysis, Project administration. Tatsuaki Tagami: Conceptualization, Methodology, Formal analysis, Project administration, Writing-original draft, Writing-review & editing, Supervision. Koki Ogawa: Writing-review & editing, Supervision. Tetsuya Ozeki: Resources, Supervision.
The authors declare no conflict of interest.
This article contains supplementary materials.
