Effects of Processing on a Sustained Release Formulation Prepared by Twin-Screw Dry Granulation

Abstract

Dry granulation is an indispensable process used to improve the flow property of moisture-sensitive materials. Considering the limitations of currently available dry granulation techniques, it is necessary to develop a novel technique. In this study, a twin-screw dry granulation (TSDG) technology was successfully applied to produce a sustained-release dry granule formulation, which was subsequently compressed into sustained-release tablets. Based on a preliminary study, theophylline was selected as model drug, Klucel™ EF, Ethocel™, and magnesium stearate were selected as excipients. A Resolution V Irregular Fraction Design was applied to determine the effect of different processing parameters (screw speed, feeding rate, barrel temperature, and screw configuration) on product properties (flow properties, particle size distribution, and dissolution time). A reliable model was achieved by combining the data obtained, and processing parameters were automatically optimized to attain the setting goal. In general, TSDG was demonstrated to be an alternative method for the preparation of dry granules. The continuous processing nature, simplicity of operation, and ease of optimization made TSDG competitive compared with other conventional dry granulation techniques.

1. Introduction

Based on Perry’s Chemical Engineer’s Handbook, granulation process is defined as “any process whereby small particles are gathered into larger, permanent masses in which the original particles can still be identified” ¹ Granulation is widely utilized in numerous industries, such as agriculture, mining, chemical, pharmaceutical, and food industries ². Application of the granulation process in the pharmaceutical industry was triggered by the invention of tablet press in 1843 ³. Nowadays, granules are an important intermediary before tablet compression and capsule filling ^4-6. The main purpose of granulation in the pharmaceutical industry is to convert small drug and excipient particles into larger agglomerates, which have some beneficial physicochemical properties, such as better flow properties, more uniform drug distribution, reduced dust, prevention of segregation, increased compactability, and enhanced appearance of the product ^3,4,6-9.

Wet granulation, which is used to produce 70% of the worldwide industry’s granules ¹⁰, is the most popular granulation technique used in the pharmaceutical industry due to the available knowledge and equipment. In general, wet granulation comprises several steps: liquid binder addition, agglomeration, drying, and screening ^8,11-13. The aggregation of primary particulates is achieved with the liquid binder ⁸. Melt granulation is a particular form of wet granulation, and the mechanisms underlying these two methods are similar ^4,14. However, there are limitations of wet/melt granulation, such as multiple unit features, time inefficiency, high cost, large space requirement, unsuitability for heat and moisture-sensitive Active Pharmaceutical Ingredients (APIs), and stability issues ^8,11,12,15. Thus, an alternative technology of dry granulation is necessary ¹⁶.

Unlike wet granulation, dry granulation employs dry binders and high mechanical force to enable the agglomeration process ^4,8,17. Slugging and roller compaction are two dry granulation techniques used widely in the pharmaceutical industry ^8,10,16-19. Slugging was the favorite dry granulation technique in the 1950–1970s in the pharmaceutical industry ¹⁷. In this process, the primary powders are compressed into large tablets or slugs, followed by milling and sieving. Granule quality is highly dependent on the flow ability of raw materials, which may cause inhomogeneity ¹⁰. Other disadvantages of slugging include: single batch processing, low manufacturing throughput, and the need for frequent maintenance. Compared to slugging, an auger-feed system has been utilized in roller compaction to constantly deliver raw material, which ensures the good content uniformity of granule. In addition, roller compaction is a highly efficient and continuous process, which has a much higher output than slugging. Thus, roller compaction is considered to be a more competitive dry granulation technique in pharmaceutical plants. Although roller compaction has many advantages over slugging, it is not perfect. Disadvantages have also been reported with this method, including the relative high levels of fines and dust ^7,9,10,20, inferior tensile strength of the final tablet ^9,19,21,22, and the loud operation environment ¹⁰. Therefore, there is an increasing need for an alternative dry granulation technique that can solve these problems.

Twin-screw hot melt extrusion is a process whereby raw materials are propelled with spinning screws and products are collected through a die block ^23,24. This technique has attracted considerable attention in the pharmaceutical industry due to its advantages, such as being a solvent-free, continuous process with no time-consuming steps, compared to other conventional techniques. The benefits of twin-screw extruder have led to the widespread application of this equipment in the pharmaceutical industry, including wet granulation ^25,26 and melt granulation ^27,28. The only difference between twin-screw granulation and twin-screw extrusion is that the die is removed during the process to prevent the extreme densification of material inside the barrel. A twin-screw extruder has been successfully applied to wet/melt granulation. However, the application of twin-screw extruders for dry granulation processing is rarely reported ²⁹. To carry out the twin-screw dry granulation (TSDG) process, the temperature of the entire barrel was set below the melting point or glass transition temperature of all formulation ingredients. The granulation process was achieved by applying dry binders and high mechanical force, which is ultimately the same as conventional dry granulation processes.

The objective of this study was to apply TSDG technology to produce a sustained-release dry granule formulation, and subsequently compress the sustained release tablets. Theophylline was used as a model drug in this study. Design of experiment (DoE) was applied to determine the effect of different processing parameters on the dry granule properties, and to optimize these factors.

2. Materials and Methods

2.1. Materials

Theophylline was purchased from Ria International LLC (Saint Louis, MO, USA). Caffeine citrate was supplied by Fisher Scientific (Pittsburgh, PA, USA). Acetaminophen was kindly contributed by Mallinckrodt Inc. (Raleigh, NC, USA). Chlorpheniramine maleate was obtained from Medchemexpress LLC (Princeton, NJ, USA). Hydroxypropyl cellulose (Klucel™ EF) was a generous gift from Ashland Specialty Ingredients (Wilmington, DE, USA). Ethyl cellulose (Ethocel™ Std. 10) was kindly donated by Dow chemical company. Magnesium stearate was purchased from Spectrum Laboratory Products Inc. (Gardena, CA, USA). All other reagents used in this study were of analytical grade.

2.2. Methods

2.2.1. Preliminary Study

The feasibility of four different APIs (theophylline, caffeine citrate, acetaminophen, and chlorpheniramine maleate) for TSDG application was tested at three levels of drug loading (30, 40, and 50%, respectively). Different formulation parameters, including polymer ratio and lubricant concentration were pretested for their influence on granule quality.

2.2.2. Design of Experiment (DoE)

A Resolution V Irregular Fraction Design was created using Design-Expert^® 8.0.6 (Stat-Ease, Inc., Minneapolis, MN, USA), with three numeric factors (screw speed, feeding rate, and barrel temperature) and one categorical factor (screw configuration), which all varied over only two levels. Flow properties (angle of repose and flowability index), particle size distribution, and dissolution time were selected as response variables to optimize the process.

2.2.3. Twin-Screw Dry Granulation

Theophylline was selected as a model drug for this study based on the results of a preliminary study. All the ingredients were passed through a US # 35 mesh screen in order to remove any formed aggregates. Theophylline (50% w/w) was premixed with Klucel™ EF (25% w/w) and Ethocel™ (25% w/w) using a Maxiblend™ V-shell blender (GlobePharma, New Brunswick, NJ) at 25 rpm for 20 min. Magnesium stearate (0.2% w/w) was then added and mixed for 4 min. Dry granulation was prepared using an 11 mm co-rotating twin-screw extruder (Thermo Fisher Scientific, Waltham, MA, USA) without a die. The extruder consisted of eight zones, the individual temperature of which, excluding zone 1, could be precisely controlled. The barrel was maintained within a temperature range of 70–110°C (below the melting point or glass transition temperature of all ingredients). The physical mixture was fed into the extruder using an 11 mm single-screw feeder (Thermo Electron, Karlsruhe, Germany) at a selected feeding rate of 3 or 10% (approximately 1.42 or 5.09 g/min). The screw speed was set at either 50 or 100 rpm. The materials were collected at the end of the extruder after the steady state was achieved. The collected samples were stored in polyethylene bags for further processing and analysis.

2.2.4. Drug Content

Granules equivalent to 10 mg of theophylline were accurately weighed and dissolved in 100 mL of 0.1N HCl. The sample was analyzed by UV–vis spectrophotometry at λmax of 270 nm. All samples showed uniform drug content.

2.2.5. Particle Size Distribution

The particle size distribution was determined using the sieve analysis method. The granules were divided into three portions by two USA standard test sieves, # 35 (500 μm) and # 12 (1.68 mm). The large size granules (over 1.68 mm) were retained in the upper sieve (# 12), the medium size granules (between 500 μm and 1.68 mm) were retained in the lower sieve (# 35), and the fine granules (under 500 μm), which were not retained by either sieve, were collected by the wax paper under the sieves. Each portion of granules was weighed, and the percentile weight distribution was calculated accordingly.

2.2.6. Flow Properties

2.2.6.1. Angle of Repose

The fixed funnel method was applied to measure the angle of repose. The material was poured through a funnel to form a cone. To minimize the impact of falling particles, funnel height was gradually elevated to ensure the tip was held close to the rising cone without touching it. The material was poured until the predetermined width of the pile was reached. The height and radius of the base of the cone was measured. The following equation was used to calculate the angle of repose (θ):

$$Tan(\theta )=\frac{h}{r}$$

Where, h is the height of the formed cone and r is the radius of the base of the cone, respectively.

2.2.6.2. Flowability Index

The flowability index was measured using a Flodex powder flowability index test instrument (Hanson Research, Chatsworth, CA, USA). Funnel height was adjusted such that the bottom of the funnel was close to but not touching the surface of the material. Ten grams of material were poured into the middle of the cylinder through the funnel. After loading was completed, 30 seconds was permitted for the possible flocculation of individual material or the whole load. A 16 mm flow disk was used for previously untested material. The release lever was turned on slowly to open the hole in the disk without vibration. The test was considered positive when the open hole on the disk could be observed from the top of cylinder. Positive tests were repeated with disks possessing smaller holes until a negative result was obtained. Negative tests were repeated with disks possessing larger holes until a positive result was achieved. The smallest hole thought which the sample could fall freely for three successive tests was considered to represent the flowability index ³⁰.

2.2.7. Differential Scanning Calorimetry (DSC)

Approximately 2–5 mg of each sample was weighed and hermetically sealed in an aluminum pan. A Diamond DSC (PerkinElmer, Shelton, CT, USA) was used to measure the degree of crystallinity of the samples. The instrument’s Pyris manager software (Shelton, CT, USA) was utilized to analyze the data. The heating rate was set at 20°C/min from 20 to 300°C under an inert atmosphere of nitrogen at a flow rate of 20 mL/min.

2.2.8. Tablet Compression

Obtained granules (200 mg) were mixed with 0.2% magnesium stearate immediately prior to manual direct compression on a MCTMI single punch tablet press (GlobePharma Inc., New Brunswick, NJ, USA). A 0.375-inch flat round punch was used, and a compression force of 150 kg/cm² was applied.

2.2.9. Scanning Electron Microscopy (SEM)

The morphology of all samples (granules, tablets, and raw materials) was determined by utilizing SEM. Samples were mounted onto an aluminum stage with an adhesive carbon tape, and then sputter-coated with gold under an argon atmosphere using a Hummer 6.2 Sputter Coater (Ladd Research Industries, Williston, VT, US). The coater was kept in a high-vacuum evaporator equipped to guarantee a uniform coating. Finally, images were captured using a JSM-5600 SEM (JEOL USA, Inc., Waterford, VA, USA) at an accelerating voltage of 5 kV.

2.2.10. In Vitro Drug Release Study

The in vitro drug release of tablets and granules equivalent to 100 mg of theophylline was confirmed using dissolution apparatus II and 900 mL of 0.1 N HCl (pH 1.2) simulated gastric medium. SR8-plus™ dissolution apparatus (Hanson, Chatsworth, CA, US) was maintained at 37 ± 0.5°C and the paddle speed was set at 50 rpm. Granules were filled into gelatin capsules (size 0) for dissolution. The UV spectra were collected by an in-site Rainbow UV-Vis probes (Pion Inc., Billerica, MA, USA) at λmax 270 nm every 5 min for the first 1 h, then every 15 min up to 2 h. Tablets were dropped directly into the dissolution vessel for the test. The UV spectra were collected at λmax 270 nm using the same instrument every 15 min for the first 1 h and then every 30 min up to 24 h.

3. Results and Discussion

3.1. Preliminary Study

3.1.1. Lubricant

While conducting the preliminary study, a grinding noise, which did not occur during the hot-melt extrusion (HME) process, was heard. This suggested the existence of friction in the twin-screw extruder. In HME, the flexible nature of the melt will not cause any rubbing at powder-instrument interfaces (wall friction) ³¹. The lubrication effect of the melt itself will also prevent abrasion between the screw and barrel. However, during the TSDG process, theophylline was kept in a solid state, and friction could occur universally where the equipment and material were in contact. Thus, the addition of dry lubricant was necessary to protect the instrument and prevent product contamination. Magnesium stearate, which is a widely used dry lubricant in the pharmaceutical industry, was applied to the formulation. The addition of magnesium stearate eliminated the noise mentioned above. Nevertheless, too much magnesium stearate (0.5%) may significantly decrease granule yield and result in samples containing only fines. This phenomenon results from the nature of the lubrication effect. The lubricant molecules will form some oriented monolayer or multilayer boundary, which will prevent additional contact between instruments and powder particles. However, too much lubricant or long-term mixing with API and excipients results in the formation of a boundary between particles, prevention the agglomeration of powder particles. In this case, a moderate quantity of 0.2% magnesium stearate was selected and added to the formulation at the final stage of mixing.

3.1.2. Polymer and Excipient Ratio

Hydroxypropyl cellulose (HPC), which is a thermoplastic polymer, has been previously studied as a dry binder for both tablet compression ^32-34 and roller compaction ³⁵. Ethyl cellulose (EC), which is a water-insoluble polymer, has been applied to prepare multiple sustained-release formulations ^36-38. Thus, a combination of HPC (dry binder) and EC (release controller) was selected for the excipients to prepare sustained release dry granulation. Two ratios of HPC and EC (1:1 and 1:3) were studied to optimize the polymer blend. The 1:1 combination exhibited the best probability of producing medium size granules. Therefore, a 1:1 ratio of EC and HPC was used as excipients for this study.

3.1.3. API and Drug Loading

Four different APIs (chlorpheniramine maleate, acetaminophen, caffeine citrate, and theophylline) with melting points ranging from 140 to 270°C were selected as candidates for the TSDG process. Physical mixtures of each API at three drug loadings (30, 40, and 50%, respectively) were prepared to test their feasibility of application in TSDG (supplementary data Table 1). All the tests were conducted at the same feeding rate, screw speed, screw configuration, and barrel temperature.

Each product after TSDG was tested with DSC to determine the crystallinity of API (supplementary data Figure 1). The DSC thermograms showed a single sharp endothermic peak, which demonstrated the preservation of crystallinity after TSDG. However, the lower drug load formulations demonstrated a melting point depression. This may be attributed to the higher polymer content (polymer Tg around high processing temperatures) within the lower drug load formulations. The high drug load formulations retained crystallinity after TSDG and were thus selected for further investigation of process parameters by DoE.

The morphology of products obtained is summarized in supplementary data Table 2. The low drug loading PM tended to achieve more robust and larger granules after TSDG. However, none of these four APIs could form granules at 50% drug loading. This was because more binder was present in the low drug loading formulation. The higher binder ratio could more effectively assist the agglomeration of raw materials, and thus, granulation. Due to the low binder ratio in the 50% formulation, the raw material could not accumulate to form granules, and hence presented as fine powders. From the API aspect, acetaminophen and caffeine citrate showed the best granulation ability. More sturdy and bigger size granules were attained at 30% drug loading. In contrast, 30% chlorpheniramine maleate and theophylline achieved smaller and long striped shaped granules. The theophylline granules were weaker than the chlorpheniramine maleate granules. With 40% drug loading, acetaminophen and caffeine citrate could still form some long striped shaped granules. However, few granules could be seen from the chlorpheniramine maleate and theophylline formulation. In summary, higher drug loading hindered granulation, and granulation was most difficult for theophylline out of the APIs tested. Successful dry granulation with difficult formulation combinations indicates that this process would be easier to achieve with other formulations. Therefore 50% drug loading of theophylline was used for the DoE study to optimize processing parameters.

3.2. DoE Set-up

In addition to formulation parameters, processing parameters are also critical for the quality of the final product. To determine the appropriate experimental design, a series of processing parameters were tested. From these initial results, four processing parameters (screw configuration, screw speed, barrel temperature, and feeding rate) were selected and the operation ranges are shown in supplementary data Table 3. Two screw configurations utilized in the experimental design are shown in Figure 1. For screw configuration 1, only one mixing zone, # comprised of seven mixing elements with 60° offset angles, was presented at zone 5. For screw configuration 2, another small mixing zone, comprised of five mixing elements with 30° offset angles, was added at zone 7. The barrel temperature profile is shown in supplementary data Table 4. The highest temperature was set at zone 5 to assist the granulation process.

Figure 1.

Screw configuration utilized in TSDG.

Twelve runs of experimental design were proposed using a Resolution V Irregular Fraction Design (supplementary data Table 5). This experimental design was selected because it allows the estimation of all main effects. The two-factor interactions will be aliased only by three-factor or higher interactions ³⁹. This design was considered to be excellent, and a worthwhile alternative to the full factorial design, in order to reduce the number of runs and obtain clean results.

3.3. Characterization of Crystallinity

Theophylline is categorized as a Biopharmaceutics Classification System (BCS) I drug. Thus, the crystallinity of theophylline was expected to be maintained in order to prevent potential thermal stability problems ^40,41. The crystallinity of theophylline as a pure drug, and in all twelve runs was confirmed by DSC (Figure 2). A single sharp endotherm peak at around 275°C, and a small endotherm peak at about 288°C, were observed for pure theophylline. This demonstrated the API we used was theophylline anhydrous form I ⁴². After TSDG, the large endotherm peaks of all runs were preserved; however, they were compressed to approximately 270°C. In addition, the small endotherm peak shifted slightly to a lower temperature. These results confirmed that the crystallinity of theophylline was preserved after TSDG, Nevertheless, some interaction between theophylline and polymers is likely ⁴³.

Figure 2.

DSC thermograms of pure theophylline and twelve DoE formulations.

3.4. The Effects of the Processing Parameters on the Particle Size Distribution

Each run of samples was separated into three portions: the large size granules (over 1.68 mm), the medium size granules (between 500 μm and 1.68 mm), and fines (under 500 μm). The granule size distribution results for all runs are shown in Table 1. The medium size granules are the most important portion for the tableting process. In addition, fine particles are very important for compression and compaction features, since the vacant space between granules could be occupied by them. However, the amount of fines should not exceed 10% ⁴⁴. It has also been reported that a small amount of fines could improve the appearance of the tablet ⁴⁵. Thus, both the medium size granule and fines were selected as the dependent variable (response) in the Resolution V Irregular Fraction Design. The regression equations for both medium size granules and fines fraction were calculated from the experimental results. The coefficients and statistical significance are summarized in supplementary data Table 6. Only the parameters related to independent variables (factors) and models are given. The full regression equation is shown under each section.

Table 1.

DoE results.

Run	Fines (< 500μm)/%	Medium Size Granule (500 μm - 1.68 mm)/%	Large Size Granule (> 1.68 mm)/%	Angle of Repose/°	Flowability Index/mm	Dissolution Time for Maximum Drug Release/h
1	2.69	32.29	65.02	28.92	20	13
2	74.63	25.37	0	31.23	22	23
3	6.88	70.42	22.7	27.6	14	16
4	13.74	59.23	27.03	30.65	24	12
5	25.24	56.15	18.61	27.02	20	23
6	12.41	51.49	36.1	26.67	22	16
7	12.46	64.76	22.78	25.49	9	15
8	4.98	53.84	41.18	26.93	14	10
9	29.39	57.93	12.68	29.18	20	18
10	23.71	66.98	9.31	28.48	22	20
11	0.61	11.45	87.94	30.14	34	16
12	30.36	59.43	10.21	29.24	20	21

3.4.1. Medium Size Granules

The percentage of medium size granules in all 12 runs varied from 11.45 to 70.42%, whereby the lowest medium-sized grain yield was obtained on runs 1, 2, and 11 (less than 35%), and the highest on runs 3, 7, and 10 (over 60%). The regression equation for the percentage of medium size granules was as follows:

$$\begin{gathered} \%\mathrm{Medium}\mathrm{size}\mathrm{granule} \\ =+56.07-10.25\ast \mathbf{A}-0.69\ast \mathbf{B}+6.83\ast \mathbf{C}+3.15\ast \mathbf{D}+3.16\ast \mathbf{A}\ast \mathbf{C}+5.19 \\ \ast \mathbf{B}\ast \mathbf{C}-6.35\ast \mathbf{B}\ast \mathbf{D}+7.00\ast \mathbf{C}\ast \mathbf{D}+8.87\ast \mathbf{A}\ast \mathbf{B}\ast \mathbf{C} \end{gathered}$$

Based on ANOVA analysis, the p value of the model was 0.0347, which indicated that the model was reliable. The most significant factors for the percentage of medium size granule was screw configuration and screw speed. The relationship between response and these two factors could be better understood by the contour plot shown in Figure 3. Figure 3a and 3b were both obtained at a feeding rate of 10%. The only difference was the screw configuration: one mixing zone for Figure 3a and two mixing zones for Figure 3b.

Figure 3.

Contour plots showing the effect of screw speed on percentage of medium size granules: (a) one mixing zone configuration; (b) two mixing zone configuration.

Figure 3a shows more medium size granules than Figure 3b; therefore, one mixing zone configuration is beneficial for the generation of mid-sized particles. This result was consistent with the negative coefficient (−10.25) in front of A. This phenomenon could be explained by the effect of the second mixing zone. Dhenge et al. reported that the second mixing zone could either decrease or increase the size of agglomerated granules by the first mixing zone. This change in size was due to several functions of the second mixing zone: shearing, deformation, breaking, mixing, re-granulation, and consolidation ⁴⁶. Under our conditions, if the granule generated by the first mixing zone was not strong enough, it could be broken down into smaller granule or fines. In contrast, if the previously obtained granule had adequate strength, it would be further combined to generate larger granules. In either situation, the amount of medium size granules would decrease. Hence the second mixing zone was considered to hinder the generation of medium size granules.

Figures 3a and 3b show the same trend; increased screw speed leads to a higher percentage of medium size granules. At first glance, this finding appeared to conflict with those of Tu et al., who claimed that increasing screw speed would decrease the granule size ⁴⁷. However, this was consistent when the large size granules were considered. Based on the regression equation of the large size granule (data not shown), a negative correlation was found between the amount of large-size granules and screw speed. Hence, this is very similar to the condition whereby the increased screw speed would decrease the number of large size granules and increase the number of medium size granules, i.e., decrease the whole granule size. A possible mechanism is that the mean residence time decreases with increasing screw speed, reducing the kneading effect on powder particles, thus generating smaller granules.

3.4.2. Fines

The number of fines also varied largely among the 12 experimental designs (0.61–74.63%). Similar to the quantity of medium size granules, a regression equation was also given for percentage of fines. However, based on the Box-Cox Plot the software recommended a square root transformation for the response, which was presented as $y^{’}=\sqrt{y+k},\mathrm{k}=0$ in this case. Hence the new regression equation was as follows:

$$\sqrt{\%\mathbf{Fines}}=+3.96+0.19\ast \mathbf{A}-1.11\ast \mathbf{B}+0.75\ast \mathbf{C}-0.83\ast \mathbf{D}-0.43\ast \mathbf{A}\ast \mathbf{B}+0.62\ast \mathbf{A}\ast \mathbf{C}$$

A p-value of 0.0018 indicated that the model was dependable. Barrel temperature, screw speed, and feeding rate were found to have a significant effect on the amount of fines (supplementary data Table 6; p < 0.05). The regression results can be observed on the contour plot in Figure 4.

Figure 4.

Contour plots showing the effect of barrel temperature, screw speed and feeding rate on percentage of fines: (a) & (c) one mixing zone configuration; (b) & (d) two mixing zone configuration.

Figure 4a and 4b was generated under a screw speed of 100 rpm with two different screw configurations. A negative correlation was found for the percentage of fines between barrel temperature and feeding rate. The decreased quantity of fines may be considered as an increase in granulated particles. Therefore, this phenomenon was equivalent to the situation that higher feeding rate and barrel temperature would lead to more granules. Osborne et al. ⁴⁸ and Palzer et al. ⁴⁹ stated the contact points between powder particles might sinter together to form solid bridges when the temperature was close to the glass transition temperature. The solid bridge would bond the particles to form strong agglomerates, or granules. In this case, the increased temperature would support the formation of granules. This relationship between granules and feeding rate has been reported previously ^50-52. A high feeding rate would increase the amount of powder in the barrel, thus increasing the likelihood of them condensing and compacting into granules.

Unlike the factors discussed above, screw speed positively affected the percentage of fines (Figure 4c and 4d). This effect could be explained by the mechanism discussed for the medium size granules. The increased screw speed would decrease the particle size, thus generating more fine particles.

3.5. The Effects of Processing Parameters on the Flow Property

One of the most important objectives of granulation is to improve the flow property of raw materials. This increased flow property can bring numerous advantages, such as consistent tablet properties, more uniform formulation, and high production speed. Therefore, assessment of the flow property of granules is important for examining the effects of granulation. Angle of repose was the most widely used method to determine flow property. However, it has been noted that data obtained from angle of repose is not dependable because this method does not mimic the behavior of the powder in the instrument ³⁰. To overcome this, another technique, the flowability index, was developed to characterize the flow property. In this study, both methods were utilized to test the flow property of samples. The data obtained were further applied to the experimental design to generate a regression equation. The coefficients and the statistical significances are summarized in supplementary data Table 7.

3.5.1. Angle of Repose

The angle of repose for samples of 12 runs was dispersed across a very narrow range. Except for runs 2, 4, and 11, all runs presented excellent flow properties (25° < angle of repose < 30°). The resultant regression equation was created as follows:

$$\begin{gathered} \mathrm{Angle}\mathrm{of}\mathrm{repose} \\ =+28.01+0.42\ast \mathbf{A}+0.50\ast \mathbf{B}+0.75\ast \mathbf{C}-0.54\ast \mathbf{D}+0.52\ast \mathbf{A}\ast \mathbf{D}-0.67\ast \mathbf{B} \\ \ast \mathbf{C}-0.60\ast \mathbf{C}\ast \mathbf{D}-0.77\ast \mathbf{A}\ast \mathbf{B}\ast \mathbf{C} \end{gathered}$$

The p-value was 0.0332, which demonstrated that the model was reliable. Screw speed had the most significant effect on angle repose (p < 0.05). However, the effects of barrel temperature and feeding rate were not negligible (p < 0.10).

The positive coefficient for screw speed implied that increasing the screw speed would increase the angle of repose, which represented a low flow property. Feeding rate exhibited a negative coefficient; thus, a higher feeding rate would decrease the angle of repose and improve the flow property. Both parameters corresponded well with the amount of fines. Increasing screw speed or decreasing feeding rate would result in more fines, leading to a poorer flow property and a higher angle of repose. Nevertheless, the positive constant in front of barrel temperature, representing a higher temperature, would increase the angle of repose and decrease the flow property of granules, which seemed to conflict with data achieved from the percentage of fines. This could be explained by the granule shape. In addition to particle size, particle shape was also a critical parameter affecting the flow property. Granules produced under higher temperatures tended to form a long stripe shape, which was associated with a poor flow property. The effect of the parameters on angle of repose was plotted on a contour, as shown in supplementary data Figure 2.

3.5.2. Flowability Index

Flowability index was defined as the smallest hole through which the core cylinder of powder could overcome the side internal friction and pass. Based on this definition, flowability index was similar to the angle of repose: a higher number was associated with a lower flow property. A flowability index of 10–24 mm has been reported as good ³⁰. For this parameter, the software used suggested an inverse square root transformation, resulting in the following regression equation:

$$\begin{gathered} \frac{1}{\sqrt{\mathrm{Flowability}\mathrm{index}}} \\ =+0.23-0.027\ast \mathit{A}-0.016\ast \mathit{B}+0.016\ast \mathit{D}+5.482\mathit{e}-003\ast \mathit{A}\ast \mathit{B}-0.027 \\ \ast \mathit{A}\ast \mathit{D}-2.602\mathit{e}-003\ast \mathit{B}\ast \mathit{C}-0.014\ast \mathit{B}\ast \mathit{D}+2.296\mathit{e}-003\ast \mathit{C}\ast \mathit{D} \end{gathered}$$

This model was accompanied with a p-value of 0.007, implying a high level of confidence. Screw configuration, barrel temperature, and feeding rate w had significant effects on flowability index. However, the complicated transformation made the relationship between different factors unclear. Hence, the contour plot should be examined to better understand the effect of different factors (Figure 5).

Figure 5.

Contour plots showing the effect of barrel temperature and feeding rate on percentage of medium size granules: (a) one mixing zone configuration; (b) two mixing zone configuration.

Figure 5a (one mixing zone) shows that a higher temperature and lower feeding rate would result in a higher flowability index, and thus, a poorer flow property. This result was consistent with the angle of repose. In Figure 5b, the effect of barrel temperature is consistent. Interestingly, the effect of feeding rate differed at different conditions. When barrel temperature was below 95°C, feeding rate had a limited effect on flowability index. This is because the ability of the higher feeding rate to decrease fines was offset by the second mixing zone. However, when the temperature exceeded 95°C, a higher feeding rate tended to increase the portion of large size granules, thus leading to a higher flowability index. Comparing Figures 5a and 5b shows that the granules produced with the two mixing zone configuration would have a lower flow property; this is consistent with the second mixing zone creating more fines or large size granules.

3.6. In Vitro Drug Release Study

A simulated gastric solution (0.1 N HCl, pH = 1.2) was used as a dissolution medium in the dissolution study. Ten sample runs released more than 95% of drug in 23 hours (Run 2 and Run 5: > 94%). The maximum drug release time for tablets varied from 10 to 23 hours (supplementary data Figure 3 and Table 1). The regression equation suggested by Design Expert software was shown as follows:

$$\begin{gathered} \mathrm{Dissolution}\mathrm{Time} \\ =+16.92+2.13\ast \mathbf{A}-2.42\ast \mathbf{B}+1.62\ast \mathbf{C}-0.75\ast \mathbf{D}+1.37\ast \mathbf{A}\ast \mathbf{B}+1.63\ast \mathbf{B} \\ \ast \mathbf{C}+2.50\ast \mathbf{B}\ast \mathbf{D} \end{gathered}$$

The p-value provided for this model was 0.0056 (supplementary data Table 8). Three factors (screw configuration, barrel temperature, and screw speed) were found to have a significant effect on dissolution time. The positive coefficient of screw configuration and screw speed implied that the application of two mixing zone configurations and high screw speed, would increase the dissolution time of tablets, resulting in slower release. However, when a higher temperature was utilized, release would be faster. This finding was in contrast to what was expected, based on the granule properties mentioned above. Therefore, the tableting process is likely to have affected the granule property. To confirm this, the dissolution of primary granules was evaluated. The time taken for 80% drug release of both tablets and granules was compared. Eighty percent drug release was selected because the release time for different runs of samples at this level were sufficiently distinct. The results are summarized in Table 2. Notably, dissolution time could be divided into two groups. In addition, the release time for tablets and granules from the same run of the experiment contrasted. Thus, the counterpart (granules) of the fast release tablets will release slow, vice versa.

Table 2.

80% drug release time for tablet and capsule.

		Tablet (h)		Capsule (min)
F1	3.2		25
F3	3.1		15
F7	2.5	Release time < 4 h	25	Release time > 15 min
F8	2.9		17.5
F11	3.8		29
F2	6.9		7
F4	4.5		14
F5	6		7
F6	4.7	Release time > 4 h	13	Release time < 15 min
F9	4.5		9
F10	6.3		10
F12	7		14

This negative correlation between tablet and granule release time was notable, which could be explained by the effect of the binder. When a large or strong granule was formed (based on our observations, these two parameters are usually positively correlated), which have a relatively slow release, more binder from the formulation would be used by the granule. Small amounts of binder would be present outside the granule. In this case, when the tablet was compressed, the binder used to bind granules together would not be sufficient, and the adhesion of granules would be weak. The tablet would be more easily disintegrated into granules, and release faster because of the enhanced surface area. This hypothesis was supported by the SEM photograph taken for granules and tablets (Figure 6). All the images were taken at 25-times magnification. As shown in the figure, from run 11 to 2, the granule size decreased gradually, and reduced cleavage on the counterpart tablet was apparent. Cleavage was a clear sign of the weak adhesion among granules. Thus, this hypothesis is reasonable.

Figure 6.

SEM photograph of granules and tablets: (a) & (b) Run 11; (c) & (d) Run 3; (e) & (f) Run 2.

3.7. Optimization of Processing Parameters

From a series of experiments, the relationship between different factors and responses was quite clear. However, these were too complicated to analyze together (supplementary data Table 9). In addition, some of the parameters required by different responses were in conflict. For example, lower quantities of fines and higher quantities of medium size granules are preferred, but screw speed has the opposite effect on these two responses. Therefore, a sophisticated model was necessary to optimize the factors. All the models we obtained from the software used had a high R² and pred R², making it possible to utilize these models for optimization. Based on previous descriptions and experimental data, the goal was set as, fines (5–10%), medium size granules (60–80%), angle of repose (25–30°), flowability index (10–24 mm), and dissolution time (12–16 h). The desirability plot is shown in supplementary data Figure 4. In conclusion, high feeding rate and screw speed should be utilized. Relatively low barrel temperature was beneficial for this goal. One mixing zone configuration should be applied.

4. Conclusions

In this study, application of a twin-screw extruder for the dry granulation process was successful. The effects of various factors on granule properties were evaluated by performing a Resolution V Irregular Fraction Design study. Automatic optimization of processing parameters to achieve the setting goal was achieved by utilizing the obtained model. In general, TSDG was found to be an alternative method for preparing dry granules. The continuous processing nature, simplicity of operation, milling process not needed, and ease of optimization make TSDG competitive compared to other conventional dry granulation techniques.

Supplementary Material

1

Supplementary Data Figure 1. DSC thermograms of pure APIs and products (30-50% drug loading) after TSDG: (a) acetaminophen; (b) chlorpheniramine maleate; (c) caffeine citrate; (d) theophylline.

2

Supplementary Data Figure 2. Contour plots showing the effect of barrel temperature, screw speed and feeding rate on angle of repose: (a) & (c) one mixing zone configuration; (b) & (d) two mixing zone configuration.

3

Supplementary Data Figure 3. In vitro release profiles of a tablets from 12 runs of samples.

4

Supplementary Data Figure 4. Desirability plot of set goals.