Use of fine particle ethylcellulose as the diluent in the production of pellets by extrusion-spheronization

Abstract

The effect of small ethylcellulose particle size on the manufacture and properties of pellets produced by extrusion-spheronization was investigated. A factorial design revealed the effects of microcrystalline cellulose (MCC), polyethylene oxide (PEO), water, and spheronization speed and time on pellet properties. Response surface modeling allowed optimization of the responses with expansion to a central composite design. Pellet yield, size, shape, friability and drug release profile were studied, along with surface and interior morphology. Pellets were spherical irrespective of the formulation and process variables and exhibited physical and mechanical characteristics appropriate for further processing. Yield in the 12/20 mesh cut was lower with FPEC than observed with coarse particle ethylcellulose (CPEC), but FPEC-containing pellets were more rugged and the PEO to obtain optimal pellets was lower for FPEC compared to CPEC. Immediate release products were obtained and ethylcellulose particle size was of no consequence to drug release. Observed responses for the optimized product agreed with predicted values, demonstrating the success of the optimization procedure. These results suggest that FPEC is a good diluent for extrusion-spheronization.

1. Introduction

Since an issue with the use of micronized versions of ethylcellulose (EC) is poor flow, it has been recommended that wet granulation be employed to increase the particle size intended for tableting (Pollock and Balwinski, 1997). Although formulations containing EC often used an organic solvent such as ethanol or isopropanol as the sole fluid or as part of the wet massing fluid (Khan and Meidan, 2007; Pollock and Balwinski, 1997), it has been shown that water is sufficient to prepare a wet granulation product when using a fine particle version of EC (Iqbal et al., 2002; Pollock and Balwinski, 1997). Dried granules thus produced have demonstrated a higher compactibility, even with a coarse version of EC. Recently, this was shown to be due, at least in part, to plasticization of ethylcellulose by bound water that makes the polymer less rigid (Agrawal et al., 2003a).

The hydrophobic nature of ethylcellulose has been utilized in pellets prepared by extrusion-spheronization to sustain drug release, whether by adding ethylcellulose in a powder form (Agrawal et al., 2003b) or as a pseudolatex dispersion (Gandhi et al., 2005), although some claim that wetted EC has low spheronizing capability (Goskonda et al., 1994). To achieve spherical product, then, microcrystalline cellulose (MCC) was included in the formulation to contribute its plasticity to the wetted mass during extrusion and to the extrudate during spheronization (Goskonda et al., 1994). However, several drugs have proved to be unstable in the presence of MCC (Brandl et al., 1995; Carstensen et al., 1969; Patel et al., 1988; Signoretti et al., 1986; Torres and Camacho, 1994), and its replacement, in whole or in part, is warranted.

MCC is typically a minimum of 20% w/w of the powder blend (Gandhi et al., 2005; Goskonda et al., 1994; Hileman et al., 1993; Jover et al., 1996;). It has been noted, however, that the minimal amount of MCC to form a continuous network in an extruded and spheronized mass is about 14% w/w (Kleinebudde et al., 1999). In the search for formulations devoid of MCC or with minimal MCC content, a coarse particle version of EC with poly(ethylene oxide) (PEO) has proved to be successful with only 8% w/w MCC necessary to improve sphericity (Mallipeddi et al., 2010).

The present study was conducted to investigate the effect of EC particle size on its use in extrusion-spheronization and on the pellet properties, including yield, size, and sphericity. In particular, the ability of smaller particle size EC to improve the smoothness of the pellet surface and the uniformity in the pellet interior was of interest. Inclusion of MCC in the formulation to improve the roundness was also investigated. For each batch of pellets, friability and the drug release profile were also studied to see if the smaller particle size of EC enhanced pellet ruggedness or slowed drug release.

2. Materials and methods

2.1. Materials

Fine particle ethylcellulose (FPEC) with an average particle size of 9.7 μm is available from Dow Chemical Company (Midland, MI) as Ethocel^® Standard FP Premium. A high molecular weight polyethylene oxide (PEO) of 1 × 10⁶ Daltons (PolyOx™ WSR N-12K) was also obtained from Dow Chemical Company. FMC Corporation (Philadelphia, PA) provided a 36,450 Dalton, 51.5 μm average particle microcrystalline cellulose (MCC) labeled Avicel PH-101. Caffeine, obtained from Sigma–Aldrich (St. Louis, MO), was used at 10% w/w as a model drug.

2.2. Methods

Desired masses of FPEC, PEO, MCC, and caffeine were blended for 5 min using a KitchenAid^® planetary mixer. The powder blend was wetted with Nanopure water and mixing was carried out for 10 min. The wetted mass was placed in an EXD60 twin screw extruder (Fuji Paudal Co., Osaka, Japan) and extruded at 38 rpm through a 1.2 mm axial screen. The extrudate was immediately placed in a Q230 spheronizer (Fuji Paudal Co.) that had a cross-hatched plate. Spheronization at the desired speed and for the desired time was completed and then the pellets were collected from the spheronizer and dried on trays in a 50 °C oven for 8 h.

A five factor, two level, half fractional factorial design with replicated centerpoints was used to identify the effects of the MCC content, PEO content, and the water amount used in the wet massing step, as well as the spheronizer speed and spheronization time, on pellet properties. Certain experiments were then conducted to expand the screening design to a central composite design to optimize the responses. Each factor range was determined in preliminary experiments which also revealed that varying the extrusion speed over the range studied was not an influential factor. Extruder speed was therefore set at 38 rpm. Actual factor levels, coded as −1.5, −1.0, 0.0, +1.0 and +1.5 corresponding to negative alpha, low, base, high, and positive alpha levels, respectively, are presented in Table 1. Responses, however, are not coded.

Table 1.

Levels of variables of the experimental design for FPEC pellets.

Levels	Factors
	A PEO (%w/w)	B MCC (%w/w)	C Water (ml)	D Spheronizer speed (rpm)	E Spheronization time (min)
−Alpha (−1.5)	2.5	7.0	187.5	450.0	7.0
Low (−1.0)	3.0	8.0	190.0	510.0	8.0
Base (0.0)	4.0	10.0	195.0	630.0	10.0
High (+1.0)	5.0	12.0	200.0	750.0	12.0
+Alpha (+1.5)	5.5	13.0	202.5	810.0	13.0

2.3. Pellet characterization

2.3.1. Size, yield, and shape

Using a nest of U.S. Standard Sieves with a 2^1/2 progression of the aperture that could capture the largest and smallest pellets on a sieve and a Retsch Vibrotronic VE1 sieve shaker (Brinkmann Instrument Co., Westbury, NY), sieve analysis was completed by screening subsets of about 35 g from each batch of pellets for 5 min until the entire mass from that batch was measured. Knowing the mass of pellets retained on each sieve, an average pellet size, d_avg, was calculated:

$$d_{\text{avg}}=\frac{(\sum \%\text{retained})(\text{average sieve aperture})}{100\%}$$ (1)

The cumulative mass of pellets in the 0.84–1.68 mm range (the 12/20 mesh cut), when expressed as a percentage of the total mass of pellets from a batch, was reported as yield. Any further evaluation of pellet characteristics was conducted on pellets from the 12/20 mesh cut.

A QICPIC Dynamic Image Analysis System (Sympatec Inc., Clausthal-Zellerfeld, Germany) was equipped with the RODOS/L dry dispersing unit to evaluate the shape of pellets. Once fed into the dry disperser, pellets were accelerated in a Venturi tube up to 100 m/s. The velocity gradients cause pellets to become dispersed and aerosolized. Using a high-speed digital camera with an exposure time of about 1 ns and a pulsed light source, motion blur was minimized during imaging. Windox 5.0 software was used to measure the sphericity of individual pellet images. Projection sphericity is defined as the ratio of the perimeter of a circle with an area equivalent to that of the pellet image (P_EQPC) to its actual perimeter (P_real):

$$\text{Projection Sphericity}=\frac{P_{\text{EQPC}}}{P_{\text{real}}}=\frac{2\sqrt{\pi A}}{P_{\text{real}}}$$ (2)

where A is the surface area of the pellet image. The ratio of the minimum to maximum Feret diameter of each pellet was defined as its aspect ratio (AR). Based on these calculations, sphericity and AR will be found in the range of 0–1. A higher value indicates that the shape of the pellet is more spherical.

2.3.2. Friability

A pellet sample of 3 g was placed in a Model 1805 Roche friabilator (Vankel Industries, Inc., Edison, NJ) along with twenty-five 3-mm glass beads. The friabilator was operated at 25 rpm for 100 revolutions. The glass beads were then separated using a 12 mesh sieve and the pellets and smaller particles were allowed to pass through. Below the 12 mesh sieve, a 20 mesh sieve captured the pellets while letting smaller particles pass through. The mass remaining on the 20 mesh sieve was weighed. The percentage loss of pellet mass is reported as the friability. The procedure was repeated to allow duplicate results for each batch of pellets.

2.3.3. Scanning electron microscopy

Samples of pellets from each batch were mounted on metal discs using a silicon adhesive. After sputtercoating with gold (Denton Desk II Vacuum, Moorestown, NJ), the samples were observed using an S-530 Scanning Electron Microscope (Hitachi High Technologies America, Inc., Pleasanton, CA) operated at a 15 kV accelerating voltage. Orion software allowed the capture of digital images.

2.3.4. Release studies

A 200 mg sample of pellets was placed in the vessel of a USP type I Model No. 2100 C dissolution apparatus (Distek Inc., North Brunswick, NJ) that contained 900 ml of 0.05 M, pH 6.8, phosphate buffer. Baskets were rotated at 100 rpm. Drug release was assessed by sampling at specific times and measuring the drug concentration at 273 nm using a model 1601 UV/Vis spectrophotometer (Shimadzu Scientific Instruments, Inc., Columbia, MD).

2.3.5. Screening design and response surface method

A five factor, two level, half fractional factorial (2^5-1) design (Resolution V) was employed as a screening design to identify the significant formulation and process variables that affect pellet size, shape, yield, and friability. Detection of curvature in the responses is possible by the addition of three center points that also allow an estimation of pure error. Analysis of the results from this screening design assumes a linear response to increases in the factor levels, as given in this model equation:

$$Y={\text{B}}_o+{\displaystyle \sum _{i=1}^5}{\text{B}}_i{\text{X}}_i+{\displaystyle \sum _{i=1}^4}{\displaystyle \sum _{j=\text{i}+1}^5}{\text{B}}_{\mathit{ij}}{\text{X}}_{\mathit{ij}}$$ (3)

where Y is the response, X_i are the factors at their coded levels, X_ij are the two factor interactions using the coded levels, B_o is the overall average for the response, B_i are the coefficients for linear effects, and B_ij are the coefficients for interaction effects.

A more detailed description of curvature in the responses and optimization of the responses are facilitated by augmenting the factorial design to a central composite design (CCD). To accomplish this, additional experiments were conducted as an axial block with an alpha of 1.5. The following second order polynomial model equation was fitted to the CCD results:

$$Y={\text{B}}_o+{\displaystyle \sum _{i=1}^5}{\text{B}}_i{\text{X}}_i+{\displaystyle \sum _{i=1}^4}{\displaystyle \sum _{j=i+1}^5}{\text{B}}_{\mathit{ij}}{\text{X}}_{\mathit{ij}}+{\displaystyle \sum _{i=1}^5}{\text{B}}_{\mathit{ii}}{\text{X}}_i^2$$ (4)

where B_ii are coefficients for quadratic effects, and the other parameters are described above for Eq. (3). Reverse hierarchical regression analysis determines which terms describe statistically significant effects and thus appear in the model equations. Significance was determined by analysis of variance (ANOVA) with an error probability of 0.05.

A simultaneous optimization technique based on desirability functions (Derringer and Suich, 1980) was chosen. After each response is assigned an individual desirability function, d_i, found in the 0–1 range, coded factor levels are revealed that maximize the overall desirability, D:

$$D={(d_1{d}_2{d}_3\dots d_m)}^{1/m}$$ (5)

where m is the number of responses (Montgomery, 2001). The screening design, the response surface approach, and data analysis were accomplished using Design-Expert 7.0.3 software (Stat-Ease Inc., Minneapolis, MN).

3. Results

Preliminary studies revealed that FPEC and the model drug alone or in combination resulted in the formation of a hard slug in the extruder, irrespective of the water level. Addition of PEO provided the lubrication and plasticity necessary to extrude the wetted mass. However, in the spheronizer, the extrudate tended to form aggregates with minor increases in water levels. Hence, MCC was included as it was in formulations based on a coarse particle ethylcellulose (CPEC) to expand the working range of the water level (Mallipeddi et al., 2010). With MCC as low as 8% w/w in the FPEC formulation, a good yield of spherical pellets with low friability was achieved. In the absence of PEO, addition of MCC at levels even as high as 20% w/w did not yield any extrudate. This confirms that PEO is an extrusion aid.

Results from the pellet characterization studies are presented in Tables 2 and 3. Because each of the factors exhibited a significant effect on the responses and data analysis revealed significant curvature (Table 4), a central composite design was pursued with each of the factors retained (Table 3). In the data analysis of the central composite design data, the selected mathematical model was considered appropriate when a high coefficient of determination was reported, there was an absence of a lack-of-fit of the model equation to the data, and the residuals were random.

Table 2.

Half fractional factorial (2^5-1) block (screening design) for FPEC pellets.

Standard Run	Factors					Mean Responsesa
	A PEO	B MCC	C Water	D Spheronizer speed	E Spheronization time	Yield (%)	Pellet size (mm)	Pellet shape		Friability (%)
								PSb	ARc
1	−1	−1	−1	−1	+1	73.34	1.28	0.95	0.94	1.13
2	+1	−1	−1	−1	−1	52.97	1.43	0.94	0.93	0.41
3	−1	+1	−1	−1	−1	86.80	1.23	0.95	0.93	1.20
4	+1	+1	−1	−1	+1	44.31	1.63	0.94	0.92	0.58
5	−1	−1	+1	−1	−1	42.09	1.60	0.94	0.94	0.96
6	+1	−1	+1	−1	+1	1.05	2.35	0.94	0.89	0.04
7	−1	+1	+1	−1	+1	71.83	1.35	0.94	0.95	1.14
8	+1	+1	+1	−1	−1	9.96	2.10	0.94	0.90	0.16
9	−1	−1	−1	+1	−1	40.37	1.62	0.95	0.96	0.05
10	+1	−1	−1	+1	+1	40.24	1.60	0.94	0.92	0.09
11	−1	+1	−1	+1	+1	60.05	1.42	0.95	0.97	1.09
12	+1	+1	−1	+1	−1	37.44	1.62	0.94	0.91	0.30
13	−1	−1	+1	+1	+1	41.90	1.58	0.94	0.95	0.96
14	+1	−1	+1	+1	−1	1.50	2.43	0.93	0.85	0.05
15	−1	+1	+1	+1	−1	58.48	1.47	0.94	0.94	1.00
16	+1	+1	+1	+1	+1	0.95	2.45	0.93	0.85	0.37
17	0	0	0	0	0	67.33	1.37	0.95	0.93	0.34
18	0	0	0	0	0	57.48	1.45	0.94	0.94	0.31
19	0	0	0	0	0	64.15	1.59	0.94	0.94	0.26

^aAveraged from two determinations.

^bPS = projection sphericity.

^cAR = aspect ratio.

Table 3.

Runs to augment the screening design for FPEC pellets to a central composite design for five variables: axial block with alpha = 1.5.

Standard Run	Factors					Mean Responsesa
	A PEO	B MCC	C Water	D Spheronizer speed	E Spheronization time	Yield (%)	Pellet size (mm)	Pellet shape		Friability (%)
								PSb	ARc
20	−1.5	0	0	0	0	63.69	1.36	0.95	0.96	1.27
21	+1.5	0	0	0	0	6.86	2.18	0.94	0.89	0.07
22	0	−1.5	0	0	0	26.43	1.77	0.94	0.92	0.54
23	0	+1.5	0	0	0	42.85	1.60	0.94	0.92	0.80
24	0	0	−1.5	0	0	60.74	1.32	0.94	0.94	0.67
25	0	0	+1.5	0	0	9.88	2.13	0.95	0.90	0.30
26	0	0	0	−1.5	0	63.84	1.42	0.94	0.93	0.64
27	0	0	0	+1.5	0	45.78	1.60	0.94	0.92	0.57
28	0	0	0	0	−1.5	55.37	1.46	0.94	0.92	0.47
29	0	0	0	0	+1.5	46.35	1.57	0.94	0.94	0.44
30	0	0	0	0	0	58.63	1.53	0.94	0.93	0.32
31	0	0	0	0	0	59.23	1.57	0.94	0.93	0.33

^aAveraged from two determinations.

^bPS = projection sphericity.

^cAR = aspect ratio.

Table 4.

Analysis of variance: fractional factorial design for FPEC pellets.

	Source	Response
		Yield	Pellet size	Pellet shapea	Friability
p-value	Model	<0.0001	<0.0001	<0.0001	<0.0001
	Lack of fit	0.9295	0.5704	0.7365	0.8706
	Curvature	0.0002	0.0082	0.0009	<0.0001
R2 value	R2	0.9943	0.9306	0.9886	0.9980
	Adjusted for degrees of freedom	0.9807	0.9093	0.9785	0.9953
	Predicted	0.9618	0.8519	0.9513	0.9887

^aPellet shape was modeled in terms of AR.

The variation in the average diameter of the pellets (1.23–2.45 mm), as seen in Tables 2 and 3, reveals the influence of the levels of formulation and process variables on the pellet size. ANOVA of the screening design data suggested that the fit of the model to the data (Table 4) was significant (p < 0.0001) and this resulted in a correlation coefficient that was reasonable $\left(R_{\text{adjusted}}^2=0.9093\right)$. A lack-of-fit that was not significant (p = 0.5704) and randomly distributed residuals confirm that the model was appropriate. The linear model equation for pellet size using coded factors is:

$$\text{Pellet size}=1.70+0.25\text{A}+0.22\text{C}+0.076\text{D}+0.16\text{AC}$$ (6)

The presence of a significant curvature (p = 0.0082) in this response (Table 4) necessitated a higher order model for a better description of the influence of the factors on this response. After augmentation to a central composite design, the quadratic model was found to be appropriate (Table 5). ANOVA using backward hierarchical regression (Table 6) revealed that the quadratic model has a good fit to the data $\left(R_{\text{adjusted}}^2=0.9711\right)$, a lack-of-fit that is not significant (p = 0.9618), and randomly distributed residuals with no outliers. The second order polynomial regression equation for pellet size using coded factors is:

$$\text{Pellet size}=1.50+0.26\text{A}-0.044\text{B}+0.23\text{C}+0.072\text{D}+0.016\text{E}+0.038\text{AB}+0.16\text{AC}+0.047\text{AE}-0.035\text{BC}+0.044\text{BE}+0.10{\text{A}}^2+0.062{\text{B}}^2+0.081{\text{C}}^2$$ (7)

Table 5.

Central composite design: model fit summary for the FPEC pellet responses.

Source	Responses
	Yield		Pellet size		Pellet Shapea		Friability
	Sum of squares	p-Value	Sum of squares	p-Value	Sum of squares	p-Value	Sum of squares	p-Value
(a) Model analysis
Mean vs total	62497.08		84.1248		26.4966		10.2897
Block vs mean	0.1016		0.0089		<0.0001		0.0331
Linear vs block	11981.7	<0.0001	2.5971	<0.0001	0.0169	<0.0001	3.6805	<0.0001
2FI vs linear	1818.114	0.4910	0.5433	0.0552	0.0050	0.0004	0.1201	0.9616
Quadratic vs 2FI	2338.743	0.0002	0.2604	0.0010	0.0008	0.0039	0.4942	<0.0001
Cubic vs quadratic	98.5098	0.6879	0.015	0.7863	0.0001	0.4430	0.0139	0.6509
Residual	124.0061		0.0256		<0.0001		0.0159
Total	78858.25		87.575		26.5194		14.6474
(b) Lack of fit
Linear	4328.721	0.0029	0.8187	0.1170	0.0058	0.0300	0.6408	0.0095
2FI	2510.607	0.0033	0.2754	0.2024	0.0009	0.1657	0.5207	0.0051
Quadratic	171.8639	0.1789	0.0151	0.9049	0.0001	0.6464	0.0265	0.1416
Cubic	73.3541	0.0912	<0.0001	0.9268	<0.0001	0.8767	0.0125	0.0435
Pure error	50.6519		0.0255		0.0007		0.0033
(c) R2 analysis
	Adj R2	PRESS	Adj R2	PRESS	Adj R2	PRESS	Adj R2	PRESS
Linear	0.6766	7135.821	0.7036	1.511	0.6877	0.0113	0.8200	0.9285
2FI	0.6757	36594.10	0.8189	4.1889	0.9163	0.0188	0.7490	9.2306
Quadratic	0.9562	5448.375	0.9621	0.7983	0.9762	0.0052	0.9778	0.7761
Cubic	0.9451	1059261	0.9462	1.2707	0.9786	0.0093	0.9734	181.0333

^aPellet shape was modeled in terms of AR.

Table 6.

ANOVA for response surface reduced quadratic model for FPEC pellets: yield.

Source	p-Value	Source	p-Value
Model	<0.0001	CD	0.0121
A-PEO	<0.0001	A2	<0.0001
B-MCC	0.0004	B2	<0.0001
C-Water content	<0.0001	C2	0.0001
D-Spheronizer speed	<0.0001	Lack of fit	0.3746
AB	0.0008	R2	0.9711
AC	<0.0001	Adj R2	0.9534
AD	0.0413	Pred R2	0.9067

Ranging from 1.06% to 86.80%, yield in the 12/20-mesh fraction reflects a profound influence of the variations in the levels of the formulation and process variables. ANOVA of the response data from the screening design experiments (Table 4) indicates that the model equation was significant (p < 0.0001) and provides a good prediction of the data $\left(R_{\text{adjusted}}^2=0.9807\right)$. Lack-of-fit was not significant (p = 0.9295), meaning that the model equation can adequately predict the screening design data. For yield, the linear regression equation using coded factors is:

$$\text{Yield}=41.45-17.90\text{A}+4.77\text{B}-12.98\text{C}-6.34\text{D}+0.25\text{E}-5.16\text{AB}-7.20\text{AC}+2.82\text{AD}-2.17\text{AE}+2.06\text{BC}-2.20\text{BE}+3.58\text{CD}$$ (8)

Significant curvature (p = 0.0002) was detected.

When the design was augmented to a central composite design, the results (Table 3) enabled description of the response surface using a higher order model that accounts for the curvature. The quadratic model was suggested by the high correlation and low predicted residual sum of squares (PRESS) (Table 5). ANOVA for the quadratic model was performed by backward hierarchical regression and the results are presented in Table 7. The significance of the model fit to the data (p < 0.0001) was confirmed by the high coefficient of determination $\left(R_{\text{adjusted}}^2=0.9534\right)$ and a lack-of-fit that was not significant (p = 0.3746). Residuals were normally distributed and random. The second order model equation for yield using coded factors is:

$$\text{Yield}=59.20-18.13\text{A}+4.93\text{B}-13.86\text{C}-6.27\text{D}-5.16\text{AB}-7.20\text{AC}+2.82\text{AD}+3.58\text{CD}-7.74{\text{A}}^2-8.02{\text{B}}^2-7.72{\text{C}}^2$$ (9)

Table 7.

ANOVA for the response surface reduced quadratic model for FPEC pellets: pellet size.

Source	p-Value	Source	p-Value
Model	<0.0001	BC	0.0293
A-PEO	<0.0001	BE	0.0088
B-MCC	0.0036	A2	<0.0001
C-Water content	<0.0001	B2	0.0030
D-Spheronizer speed	<0.0001	C2	0.0003
E-Spheronization time	<0.0001	Lack of fit	0.9618
AB	0.0191	R2	0.9841
AC	<0.0001	Adj R2	0.9711
AE	0.0055	Pred R2	0.9431

Pellet shape was evaluated by assessing the projection sphericity (PS) that measures the circularity of the two dimensional images and the aspect ratio (AR) that simply reports the ratio of the minimum to maximum Feret diameters. PS and AR values were high (p ⩽ 0.85) in the screening design results (Table 2). AR proved to be a more discriminating result than PS for revealing the difference in pellet roundness from batch to batch. ANOVA suggested that the model equation for AR was significant (p < 0.0001) with a good fit $\left(R_{\text{adjusted}}^2=0.9785\right)$ to the data (Table 5). Lack-of-fit was not significant (p = 0.7365), but curvature was significant (p = 0.0009). The linear model equation for aspect ratio using coded factors is:

$$\text{Aspect ratio}=0.92-0.026\text{A}-0.013\text{C}-0.0031\text{D}+0.0019\text{E}-0.011\text{AC}-0.011\text{AD}-0.0031\text{AE}-0.0081\text{CD}$$ (10)

Following completion of the central composite design experiments, the sequential sum of squares, correlation values, and PRESS suggested that the quadratic model (Table 5) would be adequate to describe the aspect ratio. ANOVA of the quadratic model (Table 8) indicated a good fit to the data $\left(R_{\text{adjusted}}^2=0.9841\right)$ and a lack-of-fit that was not significant (p = 0.5211). There were no outliers and the residuals were randomly distributed. The second order polynomial regression equation for aspect ratio using coded factors is:

$$\text{Aspect ratio}=0.93-0.025\text{A}-0.0005\text{B}-0.013\text{C}-0.0032\text{D}+0.0029\text{E}-0.011\text{AC}-0.011\text{AD}-0.0031\text{AE}-0.0081\text{CD}-0.0051{\text{B}}^2-0.0051{\text{C}}^2$$ (11)

Table 8.

ANOVA for the response surface reduced quadratic model for FPEC pellets: pellet shape.

Source	p-Value	Source	p-Value
Model	<0.0001	AE	0.0223
A-PEO	<0.0001	CD	<0.0001
B-MCC	0.6641	B2	0.0036
C-Water content	<0.0001	C2	0.0036
D-Spheronizer speed	0.0102	Lack of fit	0.5211
E-Spheronization time	0.0163	R2	0.9802
AC	<0.0001	Adj R2	0.9841
AD	<0.0001	Pred R2	0.9861

The friability values for the pellets in the present study were below 2%, suggesting rugged pellets. In the screening design results, friability ranged from 0.04% to 1.20% (Table 2), indicating that the levels of the variables influence the ruggedness of the pellets. Indeed, each of the factors had a significant influence on pellet friability. Several two factor interactions also had significant effects on friability. A significant linear model was found (p < 0.0001) that provided an excellent prediction of the data $\left(R_{\text{adjusted}}^2=0.9953\right)$. Although Lack-of-fit was not significant (p = 0.8706), curvature in the response was significant (p < 0.0001), suggesting that a higher order model should be sought to describe this data. The linear model equation using coded factors is:

$$\text{Friability}=0.66-0.41\text{A}+0.072\text{B}-0.073\text{C}-0.044\text{D}+0.017\text{E}+0.031\text{AB}-0.022\text{AC}+0.048\text{BE}+0.054\text{CD}+0.026\text{CE}$$ (12)

After augmenting the experiments to the central composite design, the quadratic model was found to be the most appropriate (Table 5). Each factor except spheronizer time had a significant influence (p < 0.0143) on friability. Spheronizer time is retained as a term in the model equation because it is involved in a significant two factor interaction with MCC. Statistical analysis of the CCD data is presented in Table 9. The quadratic model is significant (p < 0.0001) with a good fit $\left(R_{\text{adjusted}}^2=0.9701\right)$ to the data. Lack-of-fit was not significant (p = 0.1170). For friability, the second order equation using coded factors is:

$$\text{Friability}=0.37-0.41\text{A}+0.075\text{B}-0.084\text{C}-0.040\text{D}+0.011\text{E}+0.048\text{BE}+0.054\text{CD}+0.12{\text{A}}^2+0.12{\text{B}}^2+0.092{\text{D}}^2$$ (13)

Table 9.

ANOVA for the response surface reduced quadratic model for FPEC pellets: friability.

Source	p-Value	Source	p-Value
Model	<0.0001	A2	<0.0001
A-PEO	<0.0001	B2	<0.0001
B-MCC	<0.0001	D2	0.0002
C-Water content	<0.0001	Lack of fit	0.1170
D-Spheronizer speed	0.0143	R2	0.9804
E-Spheronization time	0.4659	Adj R2	0.9701
BE	0.0095	Pred R2	0.9539
CD	0.0042

Based on the second order equations, the formulation and process variables were optimized for pellet friability, shape, size, and yield. The optimized factors were those that could minimize MCC content, maximize pellet ruggedness, maximize the pellet yield and roundness, and yet also target a 1.2 mm pellet diameter. Factor levels were limited to the central composite design space to avoid over-predicting of responses by venturing into an unknown factor space. The optimum factor levels were found to be 3.4% PEO and 9% MCC in the powder blend and 192 ml water for the wet massing step; 510 rpm for the spheronizer speed and 12 min spheronization time were the processing conditions.

Pellet shape was retained throughout the release study, but pellets removed from the release medium at the end of the study would disintegrate readily. A higher MCC level made it more likely for pellets to retain their shape during the release. In the release medium, pellets with a higher PEO level would become slightly swollen.

4. Discussion

In preliminary studies, in order to assess the effect of particle size of ethylcellulose on the properties of pellets produced by extrusion-spheronization, coarse particle ethylcellulose (CPEC) was replaced with fine particle ethylcellulose (FPEC) in the formulation. The fact that FPEC and the model drug alone or in combination resulted in the formation of a hard slug in the extruder, irrespective of the water levels, is similar to what was reported with CPEC (Mallipeddi et al., 2010). MCC had to be included in the formulation to deal with the formation of aggregates in the spheronizer after the addition of PEO made the wetted mass sensitive to water levels. Ultimately, the formulation and process parameters, with the exception of the water level, were in comparable ranges to those studied in the manufacture of CPEC pellets and, hence, the same ranges were pursued in the screening design.

When FPEC replaced CPEC in the formulation, the water level had to be increased before a proper extrudate could be obtained without forced flow in the extruder. When introduced into the spheronizer, the extrudate at this higher water level exhibited a toroidal movement initially. As time progressed, particles began to stick to the wall. Adhering material picked up more mass which obstructed the toroidal movement and a low yield was obtained. Low water levels did not result in appreciable fines, but rather a loss of yield due to pellets in a size range below the yield cut-off. When water levels were increased, the amount of material adhering to the wall reduced and eventually, at appropriate water levels, no material was lost, enabling the smooth rope motion in the spheronizer.

During spheronization, collisions cause some of the water to be squeezed out of the extrudate to form a layer of water on the surface. That water provides lubrication upon collisions against the spheronizer wall, thus reducing the amount of mass adhering to the spheronizer wall. At lower water content, insufficient lubrication is evident because at least some of the mass sticks to the spheronizer wall. If the water content is further lowered and extrudate is yet obtained, the extrudate would then produce finer particles in the spheronizer as the material would not have enough water to impart the cohesive and plastic properties necessary to form pellets of a desired size.

4.1. Release studies

An immediate release product was obtained in each case (data not shown), indicating that neither FPEC nor the formulation and process factors in the studied ranges can slow drug release to a profound extent. FPEC pellets also did not show marked differences in drug release rates with changes in the factors studied such that the release profiles were essentially superimposable.

FPEC has been used in wet granulation tablets as a matrix forming material to provide sustained release (Iqbal et al., 2002; Pollock and Balwinski, 1997). Its hydrophobic nature coupled with the increased tortuosity due to better packing of the smaller particles in comparison to CPEC retarded entry of the release medium into and diffusion of the dissolved drug out of the matrix. It was expected that FPEC would slow drug release in the present study. However, the far smaller diameter reduced the diffusion pathlength for the drug and higher overall surface area of pellets in comparison to a tablet, coupled with the presence of hydrophilic components such as PEO and MCC in the formulation, facilitated fluid penetration and resulted in an immediate release product. These results are in agreement with the report by Agrawal et al. (2004), where the use of chitosan with FPEC as a diluent resulted in immediate release from pellets. In an earlier report on wet granulation caffeine tablets containing FPEC, sustained release was achieved only when FPEC was at least 69.5% w/w of the powder blend (Agrawal et al., 2003a). In the current study, even though FPEC content was 73–79% w/w in the factorial design, the high surface area and shorter diffusion pathlength found with pellets as compared to tablets easily overcome the effect of the enhanced FPEC content.

4.2. Yield

An important goal in extrusion-spheronization is a high yield of pellets within a selected size range. From the model equations with coded factor levels (Eqs. (8) and (9)), it can be seen that PEO and water content have the strongest influences on the yield and they are involved in a significant two-factor interaction (Fig. 1a) as well as interactions with other variables (Fig. 1b–d). These figures indicate that it is difficult to discuss the significance of the influence of a single factor on a particular response when so many are involved in significant two-factor interactions.

Figure 1.

Response surface plots for yield as a function of PEO content and water level (a), PEO content and MCC content (b), PEO content and spheronizer speed (c), and water level and spheronizer speed (d). Variable values are in coded form (−1 to +1). The other variables were at the base level (coded value: 0).

PEO and water levels are also influential in quadratic effects. Moving toward the extreme values for these factors leads to a reduction in the yield. When PEO and water content are lower alone or in combination, the plasticity and lubricity of the wetted mass are lower, causing difficulties in extrusion and spheronization. Lower PEO coupled with lower water levels causes a reduction in the yield due to the loss of product as fines and to material adhering to the spheronizer wall. At higher PEO levels, where agglomeration tendencies are greater, increasing the water content dramatically reduces the yield (Fig. 1a) by the ready formation of agglomerates of a size greater than the yield cut-off.

MCC has linear and quadratic influences on yield that compete. This indicates that an increase in MCC initially improves the yield and this is evident across the entire range of PEO levels in Fig. 2b. The ability of MCC to take up water results in less hydrated PEO and less binding of discrete particles into aggregates in the spheronizer that results in oversized pellets. As the MCC content increases further, the binding capability of PEO has been reduced to the point that production of fines or undersized pellets is responsible for the reduction in the yield.

Figure 2.

Response surface plots for pellet size as a function of PEO content and water level (a), MCC content and water level (b), PEO content and MCC content (c), PEO content and spheronization time (d), and MCC content and spheronization time (e). The other variables were at the base level.

Spheronization time has no effect on yield across its range in the central composite design results, as evidenced by its absence in the main effects as well as in two factor interaction terms. Higher spheronizer speeds cause a reduction in yield both at high and low PEO and water levels (Fig. 1c and d) because the increased intensity of particle-to-wall interactions leads to the loss of product either by sticking to the wall or by generation of fines, depending on the water content. Increased particle-to-particle interactions at higher spheronizer speeds promote agglomeration at high water and/or high PEO content, resulting in oversized pellets. The factor effects on yield for FPEC pellets are in reasonable agreement with those observed with CPEC pellets (Mallipeddi et al., 2010).

4.3. Pellet size

Statistical analysis of the CCD data (Table 6) suggested that each factor had a significant influence on pellet size and multiple two factor interactions are significant. As expected, Fig. 2a shows that pellet size increases with an increase in PEO since wetted PEO is acknowledged as a binder (Howard et al., 2006; Maggi et al., 2000), although the MCC content has little effect (Fig. 2b and c). Confirmation that it is wetted PEO that is responsible for binding is evident in Fig. 2a, where an increase in PEO has a more profound effect on pellet size when the water level is higher. Indeed, the influence of an increase in the water level is profound in comparison to the effect of the MCC level on pellet size (Fig. 2b). Spheronization time is involved in significant two-factor interactions with PEO and MCC contents (Fig. 2d and e, respectively), with greater influence when PEO and/or MCC are high. Since spheronizer speed is absent from any binary interactions, an increase in this factor increases pellet size under all circumstances. At high spheronization speeds, the increased particle-to-wall interactions reduce loss of the smaller sized pellets as mass adhering to the wall at low water levels, whereas at high water and PEO levels, an increase in particle-to-particle interactions or agglomeration of mass from the wall is responsible for pellet growth in the spheronizer (Fig. 2a) and a reduction in yield (Fig. 1a and b).

Each formulation variable is involved in binary interactions and/or quadratic effects, suggesting that, at optimum values, pellets of a desired size can be produced. Pellet size reaches a maximum at a combination of high PEO and high water levels (Fig. 2a). The effect of an increase in the MCC content on a reduction in pellet size is more apparent at higher water content (Fig. 2b). This is likely because an increase in MCC content allows a greater uptake of water by this excipient and this allows less agglomeration to take place because PEO is less hydrated.

4.4. Pellet shape

The uniformity in the shape of the pellets and their sphericity is important for further processing steps that rely on flow and packing properties, such as capsule filling. Aspect ratio (AR) and projection sphericity (PS) observations are almost identical to those for CPEC pellets (Mallipeddi et al., 2010). AR is more discriminating, revealing a wider variation and indicating that, even though visually spherical, factor levels influence this response.

In the screening design results, factors other than MCC affect AR and each was involved in significant binary interactions. Despite the fact that its main factor effect was not significant, MCC was retained in the quadratic model equation for hierarchical reasons. PEO and water content are the major factors influencing pellet shape, along with their interaction and those with process variables presented in the response surface plots (Fig. 3a–d). At lower PEO or water content, pellet shape improves with increasing spheronizer speed and spheronization time (Fig. 3b–d); AR decreases at higher PEO and/or higher water content (Fig. 3a–d). With higher PEO and water levels, the wetted mass is overplasticized due to the formation of a soft PEO hydrogel, and the extrudate flattens in the spheronizer. Flatter pellets are produced and evaporation of water from the hydrogel results in rougher surfaces. At higher spheronizer speeds, these influences are greater, causing less regularity in the pellet shape. In Fig. 3b and d, the effect of spheronizer speed is more evident at higher PEO and water levels, respectively.

Figure 3.

Response surface plots for pellet shape as a function of PEO content and water level (a), PEO content and spheronizer speed (b), PEO content and spheronization time (c), and water level and spheronizer speed (d). The other variables were at the base level.

4.5. Friability

A low friability reflects pellet ruggedness and suggests how well the pellets can withstand subsequent handling and processing. Friability is most influenced by PEO content followed by water content, the negative coefficients of which indicate that increasing PEO and/or water content decreases the friability of the pellets. This agrees well with the results for CPEC pellets (Mallipeddi et al., 2010) and is in agreement with the reports that wetted PEO acts as a binder (Howard et al., 2006; Maggi et al., 2000). However, with increasing PEO content, friability passes through a minimum, exhibiting a small increase at higher PEO content. Increased friability is attributed to irregular pellet shapes due to flattening and rougher surfaces observed at higher PEO content when the gel formed by PEO dries and shrinks. A two factor interaction between MCC and spheronization time was revealed (Fig. 4a) and the quadratic effect of MCC on friability is clearly more influential than is the spheronization time. An increase in MCC initially decreases friability, due to enhanced plasticity of the wetted mass allowing densification and a greater binding interfacial area within the pellet. Eventually, with further increases in MCC, less water is available to the remainder of the wetted mass that leads to reduced plasticity of the wetted mass in general. This reduction in plasticity of the wetted mass in general at higher MCC allows less molding of particles that is necessary to produce the bonding interfacial area during extrusion and spheronization that results in low friability. Increasing the water content can reduce the friability of the pellets at any spheronizer speed (Fig. 4b). At lower spheronizer speed, the influence of the water level on friability is more profound. Compaction forces are lower at lower spheronizer speeds and the material requires the enhanced plasticity offered by a higher water level to achieve the lower friability.

Figure 4.

Response surface plots for friability as a function of MCC content and spheronization time (a) and water level and spheronizer speed (b). The other variables were at the base level.

CPEC pellets exhibit slightly higher friability compared to FPEC pellets (Mallipeddi et al., 2010). Improved ruggedness of FPEC pellets should be observed because the lower particle size provides better packing properties, resulting in stronger binding properties, and relatively smoother surfaces.

4.6. Optimization

To assess the predictability of the response surface models derived from the central composite design results, a batch of pellets was prepared using the optimized factor levels and each of the responses was measured and compared to the predicted responses. The experimental results were in agreement with the predicted values (Table 10), which confirmed the validity and predictability of the model. SEM analysis of pellets from the optimized batch (Fig. 5) demonstrates the smooth texture and spherical shape of the pellets. The surface of the pellet was smoother than that of the optimized pellet in the coarse particle ethylcellulose study and the interior porosity of the optimized FPEC pellet is more evenly distributed (Mallipeddi et al., 2010). These two properties should lead to easier coverage with a film coat and a more uniform drug release rate, respectively.

Table 10.

Comparison between experimental and predicted values for the optimal formulation of FPEC.

	Responses
	Yield (%)	Pellet size (mm)	Pellet shape (AR)	Friability
Predicteda	74.32 ± 2.33	1.26 ± 0.03	0.94 ± 0.003	0.85 ± 0.033
Experimental	76.35	1.25	0.94	0.82

^aMean ± standard error.

Figure 5.

SEM micrographs of an FPEC pellet from the optimum batch of pellets (a and c) and the cross-section of another pellet from the optimum batch (b and d) at different magnifications (30X (a and b) and 80X (c and d)).

5. Conclusions

It is feasible to produce extruded and spheronized pellets with minimal amounts of MCC by using a fine particle version of ethylcellulose and a high molecular weight polyethylene oxide as the diluent and extrusion aid, respectively. The pellets can be highly spherical and exhibit the desired mechanical and physical properties that allow further processing, irrespective of the level of formulation and process variables within the ranges studied. Each batch of pellets provided immediate drug release and the effect of particle size of ethylcellulose on the drug release characteristics was of no apparent consequence. However, pellets prepared with this fine particle version of ethylcellulose were smoother and somewhat more rugged than those produced with a coarse particle version of ethylcellulose.