INTRODUCTION:

Multiparticulate dosage forms are gaining much favour over single-unit dosage forms because of their potential benefits like predictable gastric emptying, no risk of dose dumping, flexible drug release patterns, and increased bioavailability with less inter and intra-subject variability¹,².

Pellets, as a drug delivery system, can deliver a lot of benefits such as less irritation in GIT, lowered risk of side effects, less-friable dosage form and easy coating. Various methods can be utilized to prepare pellets. Among various methods, extrusion–spheronization is the most frequently and efficiently used methods³.

To date, microcrystalline cellulose (MCC) remains the most frequently used excipient for the production of pellets by wet extrusion and spheronization. The manufactured pellets are particularly characterized by a narrow particle size distribution, a high sphericity, and suitable mechanical properties. However, with regard to low soluble drugs, MCC-based pellets show a tendency to have a prolonged drug release profile because of a lack of disintegration⁴.

The important parameter in pellet is size and shape of the pellet which is influenced by number of parameters like type of binder, amount of binder, screen hole diameter, extruder speed, spheronization time, spheronization speed, spheronization load³. To achieve desired the pellet shape, size and yield, one of the step has been added in the classical process after extrusion step.

Globally, India houses the second most number of individuals suffering from dementia with an estimated 4.1 million people suffering from it as per the ‘Dementia India’ report published by the Alzheimer’s and Related Disorders Society of India. This is expected to double by 2035. Maharashtra and Uttar Pradesh alone are expected to house more than 5, 00, 000 patients by 2026⁵.

Acetylcholinesterase inhibitors, gingko biloba, methylphenidate, and a variety of nonpharmacological interventions were found to be successful in reducing apathy in patients with Alzheimer disease. Currently, among several drugs available for Alzheimer’s disease treatment, galantamine hydrobromide is the latest one recommended to improve the cognitive functions, and subsequently to treat Alzheimer’s patients (U.S. National Library of Medicine, 2007). Galantamine hydrobromide has recently been approved by the FDA for symptomatic treatment for Alzheimer’s disease and vascular dementia by oral or injectable administration. However, its pharmacological activities administered by oral method or injection would be likely to cause some severe adverse effects in the gastrointestinal tract, and some organs or systems outside the central nervous system. Therefore, a more efficient 2administration route and pharmaceutical preparation to enhance delivery ability are urgently needed⁶,⁷.

So, the objectives of the present study was to develop and characterize functional drug delivery system of Antialzheimer drug considering the concepts DOE and QbD using modified processing technique incorporating release retardants.

MATERIALS AND METHODS:

Materials:

Galantamine HBr was received as a gift sample from Zydus Research Centre, Ahmedabad, India. Microcrystalline cellulose (Avicel PH 101) was obtained from FMC Biopolymer (Philadelphia, PA, USA). Compritol 888 ATO was obtained from Gattefossé India Pvt. Ltd., Mumbai, India. Ethocel (Ethyl Cellulose) was kindly provided by Dow Chemical International Pvt. Ltd., Mumbai, India. Etahnol (99%) was procured from Ambika Enterprise, Vadodara, India. In-house ultrapure water was used in all experiments. All other chemicals used were of analytical reagent grade.

Methods:

Quality Target Product Profile (QTPP):

Quality target product profile (QTPP) describes the design criteria for the product and should therefore form the basis for determining the critical quality attributes, critical process parameters and control strategy. The first step in QbD is to identify the QTPP, i.e., decide what you want the product to do, the type of dosage form and manufacturing method⁸. The QTPP for controlled release pellets are shown in Table 1.

Risk assessment:

The Failure Mode Effect Analysis (FMEA) method was used to perform the qualitative risk assessment, which could identify the CQAs that have the greatest chance of causing product failure, i.e., not meeting the QTPP. Initially, all the possible factors affecting the product quality were identified from the literature search and past experience. Fish bone diagram was constructed to identify potential risk. All the factors were divided into six criteria measurement, process, machine, men and material⁹.

Each variable (potential failure mode) was scored in terms of failure mode severity (S), occurrence probability (O) and likelihood of detection (D). For each risk, O, S, D scores were multiplied together to produce a “Risk Priority Number” (RPN).

Where,

O is the occurrence probability; S, the severity, which is a measure of how severe of an effect a given failure mode would cause; D is the detectability or the ease that a failure mode can be detected, because the more detectible a failure mode is, the less risk it presents to product quality. We ranked these all parameters as described by Fahmy 2012 et al ¹⁰.

The RPN threshold was set at 60, and any formulation variable or process parameter with an RPN 60 or above was regarded as a potential critical factor, that is, potential risks are evaluated by subsequent process characterization studies since it possibly has a potential impact on CQAs and in consequence on product safety and efficacy, while factors with a lower RPN can be eliminated from further study ³. The concentration of Compritol and the concentration of ethyl cellulose were identified as the main parameters affect the quality of sustained release pellets as these parameters got highest RPN score in the FMEA model.

Central Composite Design (CCD):

The optimization of modified release pellets of GH was performed by CCD. From the risk assessment analysis, concentration of Compritol and concentration of EC were selected as independent variable, studied at three levels each. The central point (0) was studied in quintuplicate. All other formulation and processing variables were kept invariant throughout the study. Percentage drug release at 2h (Y₁), percentage drug release at 6h (Y₂) and percentage drug release at 10h (Y₃) were used separately as the responses in the CCD as shown in Table 2.

The quadratic model was generated for each response parameter using multiple linear regression analysis (MLRA). Each targeted response parameter was statistically analyzed by one way ANOVA at 0.05 probability level. Terms with higher P value (P > 0.05) than the critical significance, were removed in the backward elimination step. Each term in final regression equation was only included if the P value of student t-test was less than 0.05. The mathematical model was expressed as follows:

Second-order polynomial model:

y= b₀ + b₁X₁ + b₂X₂ + b₃X₁X₂+ b₄X₁X₃ + b₅X₂X_3
+b₆ X₁² + b₇ X₂² + b₈ X₃²

Where, y is the measured response, b₀ is an intercept and b₁–b₈ are the regression coefficients, X₁, X₂ represents the main effect, X₁², X₂² the quadratic effect, and X₁X_2,X₂X₃andX₁X₃ are the interaction effect.

Three-dimensional (3D) response surface plots and two dimensional (2-D) contour plots were constructed based on the model polynomial functions using Design Expert software. These plots are very useful to see interaction effects on the factors on the responses¹¹.

One optimum batch and two check-point batches were selected by intensive grid search analysis, performed over the entire experimental domain, to validate the chosen experimental design and polynomial equations. The criterion for selection of optimum was primarily based on the desired values of the response parameters, i.e. the drug release at 2h, 6h and 10h. The formulations corresponding to this optimum were prepared and evaluated for various response properties. Resultant experimental data were quantitatively compared with predicted values and percentage prediction error was calculated ¹².

Table 2: Coded and actual level of independent variables used for Central Composite Design and Responses (Dependent variable) to be measured

Variables	Symbol Xi	Coded values
Variables	Symbol Xi	-1.414	-1	0	1	1.414
Concentration of Compritol 888 ATO (%)	X₁	4.36	10	15	20	25.63
Concentration of Ethocel (%)	X₂	4.36	10	15	20	25.63
Responses to be studied
Response Y	Acceptable range
Y₁=Drug release at 2 h (% CDR)	40-50%
Y₂= Drug release at 6 h (% CDR)	70-80%
Y₃= Drug release at 10 h (% CDR)	NLT 85%

Preparation of Drug-loaded Pellet:

Pellets were prepared by laboratory scale extrusion/ spheronization process by adopting the following steps: Galntamine HBr and Avicel pH 101 were mixed in blender mixer for 5 min. Compritol and EC were dissolved in a blend of isopropyl alcohol and dichloromethane (1:1).organic solvent was added into the above powder blend to get wet mass. The mass was immediately transferred to the feeder and extruded through a mono-screw extruder with a die plate of 1 mm diameter (Chronimach India Pvt. Ltd., Ahmedabad, India) at a constant screw speed of 30 rpm. The extrudates were treated with pregelatinized starch to strengthen extrudes and to get maximum yield of required size pellets. Then extrudates were spheronized using a spheronizer (Chronimach India Pvt. Ltd., Ahmedabad, India). Spheronization was carried out at 1500 rpm for 10 minutes at room temperatures controlled by the inlet air. A radial plate spheronizer with a plate diameter of 45.0 cm was used. The wet pellets were dried in a hot air oven at 40°C for 1 hour. Finally, the pellets were passed through a 16–22 mesh and then stored in sealed glass bottles. (Table-1)

Characterization of Pellets:

Pellet size, size distribution and % yield:

The size distribution in relation of average diameter of the pellets was determined by an optical microscopic method. A compound microscope fitted with a calibrated ocular diameter and stage micrometer slide was used to count at least 100 particles [Parmar KV]. Additionally, Size distribution were carried out by Kalweka sieve shaker with a standard sieves between 500 and 4000 μm. The yield was defined as the percentage of pellets in the size range between 0.8 and 1.2 mm. All the batches were studied with regards to the morphological features such as aspect ratio, particle size and shape using photomicrograph ¹³.

Shape analysis:

The shape of pellets was determined in terms of aspect ratio. Aspect ratio is defined as longest ferret diameter over the ferret diameter perpendicular to longest. The diameter was calculated using optical microscope. Thirty pellets were placed on glass slide and observed under black background. A perfectly round pellet would have an aspect ratio value of <1.2 irrespective of size and the value >1.2 for pellets that were progressively non-spherical ¹⁴,¹⁵. Aspect ratio was calculated as follows:

Aspect ratio (AR) = d_max/d_min

Where,

d_max and d_min were the longest and shortest Feret diameters measured, respectively.

Flow property:

The prepared pellets were evaluated for bulk density, tapped density, compressibility index (Carr's index), angle of repose and Hausners’ ratio ¹³.

Surface characterization:

The surface of pellet was characterized by placing the pellets on one side of double adhesive stub which was then coated with gold (Fine coat, ion sputter, JFC-1100). The gold coated pellets were then observed using scanning electron microscope (SEM) (ESEM TMP+EDAX, Philips, Netherland) at 30 kv. SEM images of pellets were captured and observed for size, shape and surface at suitable magnification ¹⁶.

Friability:

Friability test (n=3) was performed using friabilator (EF-2, Electrolab India Pvt. Ltd., Mumbai). A preweighed sample (3 g) was placed in the friabilator along with 25 steel balls, each 2 mm in diameter. After 100 revolutions at 25 rpm, the mass retained on 20 mesh sieve was weighed, and friability was calculated as the percentage loss of mass between the initial and final weights of each bead sample ¹⁷. The Friability (%) was calculated as:

(Initial weight – final weight)

% Friability= -------------------------------------- X 100

Initial weight

Each batch was evaluated in triplicate.

Table 1: Formulation of galantamine HBr pellets

Batch	Galantamine HBr (g)	Amount of Compritol 888 ATO (g)	Amount of Ethocel (g)	Pregelatinized Starch (g)	IPA: DCM	Avicel PH 101 (g)
G 1	1.6	2	2	0.4	1:1	14
G2	1.6	4	2	0.4	1:1	12
G 3	1.6	2	4	0.4	1:1	12
G 4	1.6	4	4	0.4	1:1	10
G 5	1.6	0.87	3	0.4	1:1	14.13
G 6	1.6	5.12	3	0.4	1:1	9.88
G 7	1.6	3	0.87	0.4	1:1	14.13
G 8	1.6	3	5.12	0.4	1:1	9.88
G 9	1.6	3	3	0.4	1:1	12
G 10	1.6	3	3	0.4	1:1	12
G 11	1.6	3	3	0.4	1:1	12
G 12	1.6	3	3	0.4	1:1	12
G 13	1.6	3	3	0.4	1:1	12

Batch size = 20g

Each capsule contained 200 mg pellet eq.to 16 mg galantamine Hbr

Drug Content study:

The content of GH was analyzed spectrophotometrically at 288 nm using UV spectrophotometer (UV 1800, Shimadzu analytical Pvt. Ltd. India). Twenty mg equivalent weight of GH in pellet was accurately weighed, powdered and transferred to a 100 ml volumetric flask containing distilled water. The solution in the volumetric flask was filtered, diluted suitably and was analyzed ¹⁸.

In-vitro drug release and drug release kinetic study:

Pellets equivalent to 16 mg GH were studied for release test in USP type II (paddle) dissolution apparatus (EDT-08Lx, Electrolab India Pvt. Ltd., Mumbai) at 100 rpm. The release medium was 900 mL of distilled water maintained at 37±0.5 °C. At pre-determined time points, samples were withdrawn and replaced with an equivalent-volume of fresh medium. The samples were filtered through 0.45 μm filter and then analyzed for GH as the amount of drug was measured at absorption maxima of 288 nm using UV-visible spectrophotometer (UV 1800, Shimadzu analytical Pvt. Ltd. India)¹⁹,²⁰. The similarity factor (f2) method was used to assess the similarity of two release profiles ²¹. In vitro drug release of optimized GH sustained release Pellets were fitted using different mathematical models like zero order kinetics, first order kinetics, Korsmeyer–Peppas model ²² and Higuchi model ²³. The parameters in each equation were determined by using DDSolver software ²⁴. The correlation coefficient (R) was used to evaluate the applicability of these release models. The model with a maximum R value was the best fit one.

Prediction of in vivo data from in vitro data:

In vivo pharmacokinetic parameters were predicted from in vitro drug release data of optimized batch by the use of back calculation of Wagner Nelson Method using the reported pharmacokinetic parameters like volume of distribution (V_d), oral bioavailability (F) and elimination rate constant (K_e) of drug ²⁵. The derived equation of Back calculation Wagner Nelson is as follows:

Where, C_{t+1 =}Predicted Plasma Concentration at time t+1, D = Dose of drug administered, V_d= Apparent volume of distribution, C_t= Plasma Concentration at time t, ΔF = Fraction of dose absorbed, Δt = Time interval between t and t+1, K_e = Elimination rate constant of the drug ²⁶.

The predicted in vivo data of the optimized batch was then compared to the reported in vivo data and % prediction error (%PE) was calculated. % PE for each formulation should not exceed 15% ²⁷.

RESULT AND DISCUSSION:

Risk Assessment Analysis:

A fishbone diagram was constructed in accordance with ICH Q8 R2 guideline, to identify an initial list of potential high risk factors that might influence the quality of the GH loaded pellets as shown in Fig. 1. Six main causes (material, man, measurement, environment, machine and process factors) were identified. All the factors except material and process had non-significant effect as all research work done under standard conditions which were fixed by preliminary trials or prior knowledge. Risk Priority Number (RPN) was calculated based on occurance, severity and detectability as shown in Table 3 ³,²⁸. The factors which has high impact on in vivo performance or in vitro drug release, given high RPN score (>60) compared to others. The concentration of Compritol and concentration of ethyl cellulose were identified as potentially important factors. Potential risks are evaluated by subsequent process characterization studies since it possibly has a potential impact on CQAs and in consequence on product safety and efficacy, while factors with a lower RPN can be eliminated from further study.

Fig. 1: Fishbone diagram for pellet formulation and development

Table 4: FMEA analysis: RPN Score of individual parameters

	Failure mode	Impact of change	Occurrence	Route of failure	Severity	Control Measure	Detection	RPN	Rank
Material	Conc. of Ethyl Cellulose	In vitro drug release, Friability	5	Different concentration, Material variation	5	Dissolution apparatus, Friabilator qualification	5	125	1
	Avicel ph 101 amount	Sphericity of pellet	4	Material variation	1	Optical Microscope, Sieve analysis	3	12	8
	Conc. of Compritol 800 ATO	Dissolution, Friability	5	Different Concentration, Material Variation	5	Dissolution, Friability tester	5	125	1
	Conc. of Pregelatinized starch	Pellet shape and size, % yield	3	Different Concentration, Material Variation	4	Proper blending	3	36	4
Environment	Humidity	Pellet quality	1	Change in atmospheric condition	1	Hygrometer	1	1	11
	Pressure	No effect	1	Change in atmospheric condition	1	Manometer	1	1	11
	Temperature	Pellet quality	1	Change in atmospheric condition	1	Thermometer	1	1	11
Man	Analyst	Product Quality Testing	1	Manual error	1	Audit, SOP	1	1	11
	Stockiest	Product Quality	1	Manual error	1	Audit, SOP	1	1	11
	Retailer	Product Quality	1	Manual error	1	Audit, SOP	1	1	11
	QC Person	Product Quality Testing	1	Manual error	1	Audit, SOP	1	1	11
	Formulator	Product Quality	1	Manual error	1	Audit, SOP	1	1	11
Measurement	Particle size and size distribution	Pellet quality	3	Machine Failure, Poor development	3	Optical Microscope, Sieve analysis	1	9	9
	In vitro drug release	Dissolution	4	Machine Failure, Poor development	5	Dissolution tester	5	100	2
	Alcohol induced dose dumping	Dissolution	4	Machine Failure, Poor development	3	Dissolution tester	3	36	4
	IVIVC	Poor Prediction	4	Machine Failure, Poor development	3	Dissolution	3	36	4
	Drug release kinetics	Variable drug release	4	Machine Failure, Poor development	3	Dissolution	3	36	4
	Micromeritic Property	Poor Pellet quality	3	Machine Failure, Poor development	1	Carr's Index, Angle of Repose	3	9	9
	Physicochemical Parameter	Friability, Weight variation	3	Machine Failure, Poor development	1	Roche friabilator, Weight balance	4	12	8
Machine	Dissolution apparatus	Dissolution, IVIVC	4	Operator's Error, Equipment failure	3	Validation and qualification of instrument	1	12	8
	UV Spectrophotometer	Drug Estimation	3	Operator's Error, Equipment failure	3	Validation and qualification of instrument	1	9	9
	Extruder Spheronizer	Sphericity of Pellet	3	Operator's Error, Equipment failure	1	Validation and qualification of instrument	1	3	10
	Dryer	Physical Property	3	Operator's Error, Equipment failure	1	Validation and qualification of instrument	1	3	10
Process	Sieve Number	Particle size	3	Operator's Error, Equipment failure	1	Validation	1	3	10
	Speed of Extruder	Sphericity of Pellet	4	Machine failure, Poor development, Operator's error	4	Automation of Machine	3	48	3
	Drying Method	Non-uniform drying	3	Machine failure, Poor development, Operator's error	3	Automation of Machine	1	9	9
	Mixing method	Non-uniform mixing	3	Machine failure, Poor development, Operator's error	3	Automation of Machine	1	9	9
	Mixing Time and speed	Non-uniform mixing	3	Machine failure, Poor development, Operator's error	3	Automation of Machine	3	27	6
	Drying Temp. and Time	Non-uniform drying	3	Machine failure, Poor development, Operator's error	3	Automation of Machine	3	27	6
	Speed of Spheronizer	Non-uniform particle size	4	Machine Failure, Poor development, Operator's error	3	Automation of Machine	2	24	7
	Load in Extruder	Sphericity of pellet	4	Machine Failure, Poor development, Operator's error	4	Automation of Machine	2	32	5
	Spheronization time	Sphericity of pellet	4	Machine Failure, Poor development, Operator's error	4	Automation of Machine	2	32	5

Optimization of Critical Process Parameters (CPP) of Pellet Formulation:

Central composite design was employed for the optimization of responses. Thirteen batches were prepared which 2² factorial points, four star points and five replicates at the centre point. Polynomial equation was constructed using design expert (version 8.0.7.1). Response surface plot and contour plots were drawn. The experiment runs with independent variables and the observed responses are shown in Table 5.

Polynomial equation including the main effects, interaction terms and polynomial terms were selected based on the estimation of several statistical parameters, such as the multiple correlation coefficient (R²), adjusted multiple correlation coefficient (adjusted R²) and the predicted residual sum of squares (PRESS), provided by all responses were best fitted quadratic model. Less significant (P >0.05) were eliminated from full model. Coefficients with negative sign signify the antagonistic effect, while positive sign signifies the synergistic effect to the responses.

Table 5: Design layout responses obtained in the Central Composite Design batches

Batch	X₁	X₂	Y₁(%)	Y₂(%)	Y₃(%)
G 1	-1.00	-1.00	56.57	78.36	94.2
G2	1.00	-1.00	50.98	73.52	90.41
G 3	-1.00	1.00	51.37	73.07	86.18
G 4	1.00	1.00	32.85	55.34	70.23
G 5	-1.41	0.00	56.37	78.88	93.53
G 6	1.41	0.00	30.7	56.85	74.38
G 7	0.00	-1.41	53.8	79.31	97.5
G 8	0.00	1.41	36.34	58.89	73.83
G 9	0.00	0.00	45.26	73.54	91.16
G 10	0.00	0.00	43.12	73.91	91.53
G 11	0.00	0.00	45.05	74.55	86.08
G 12	0.00	0.00	47.18	76.91	94.35
G 13	0.00	0.00	43.14	74.02	90.24

Multiple linear regression analysis (MLRA) and analysis of variance (ANOVA) were performed to identify the relationship between two independent factors (X₁ and X₂) and three dependent variables (Y₁, Y₂ and Y₃). The results of MLR (the value of correlation coefficient and the values of coefficients) and ANOVA (Fisher’s ratio and P values) are summarized in Table 6 for the three responses.

Table 6: ANOVA for Response Y_1,Y₂ and Y₃ Surface Quadratic Model

Factors	Y₁			Y₂			Y₃
	F value	P value	Coded Coefficient	F value	P value	Coded Coefficient	F value	P value	Coded Coefficient
Model	19.57	0.0006	44.75	46.29	< 0.0001	74.59	26.78	0.0002	90.67
X₁– A	56.21	0.0001	-7.55	100.23	< 0.0001	-6.72	41.35	0.0004	-5.85
X₁- B	35.51	0.0006	-6.00	95.15	< 0.0001	-6.54	71.75	< 0.0001	-7.71
AB	5.15	0.0575	-3.23	11.54	0.0115	-3.22	5.58	0.0502	-3.04
A2	0.078	0.7875	0.30	16.97	0.0045	-2.96	11.07	0.0126	-3.25
B2	0.98	0.3549	1.07	10.63	0.0139	-2.35	6.01	0.0440	-2.39
Lack of Fit	5.25	0.0715		3.29	0.1403		0.40	0.7640
R² Value	0.9333			0.9706			0.9503

Response Y₁ - drug release at 2hr:

Large model R² value (0.9333) and high F value (19.57) with very low p value (0. 0006) for drug release at 2h indicates that the model is significant for this response. The results which are summarized in Table 5 indicate that at least one of two factors had a significant effect on X₁. Here, both the main effect i.e. concentration of ethyl cellulose (P = 0.0006) and concentration of Compritol (P=0.0001) had significant effect on % drug release at 2 h as P value is less than 0.005. The interaction terms had non-significant effect on % drug release 2h as P> 0.005. The lack of fit was found non-significant, which was desired. Compritol 888 ATO has more pronounced hydrophobic property attributed to longer fatty acid chain length in behenic acid. Increasing Compritol 888 ATO concentration caused reduction in the drug release due to decreased permeation of dissolution medium into the pellets resulting from increased lipophilicity of the waxy substance. In addition to Compritol, EC also effectively retarded the drug release. This is obvious since ethyl cellulose is practically insoluble in aqueous medium and therefore the drug is released by erosion mechanism. It has been previously reported that high levels of EC reduce drug release rates on account of formation of a strong matrix with reduced porosity. This increases diffusional path length leading to reduced water penetration through the micropores resulting in slower drug release (S. Chandran). From the polynomial equation, it is also seen that the factor X₁ and X₂ exhibit negative effect on Y₁. Significant change in % drug released at 2 h can be obtained by making slight change in the concentration of both polymers. The relationship between the dependent and independent variables was also elucidated by constructing response surface plots as shown in Fig. 2. The contour lines are non-linear in nature in the contour plot since only the main terms (X₁ and X₂) and interaction effects were found significant for % CDR at 2h. The equation in terms of un-coded factors is:

Y₁=+44.75 - 7.55X₁ - 6.00X₂ - 3.23 X₁ X₂ + 0.30 X₁²+ 1.07 X₂²

Response Y₂ - Drug Release at 6h:

Large model R² values (0.9706) and high F value (46.29) with very low p value (0. 0001) for Drug Release at 6h indicated that the model was predictive and significant for this response. The main effect and interaction terms both have significant effect on response Y₂ as p value is <0.05. Therefore, conclusions cannot be drawn from the numerical values of the coefficients of the main effects and interaction effect. In such instances, it is common to draw the conclusions from the contour plot rather the values of coefficients. Figure 2 shows the contour plot for % CDR at 6 h. The contour lines are non-linear in nature. Factor X₁-Concentration of Compritol was the most significant factor (F=100.23, p=0.0001) affecting the drug release at 6h negatively; increase in the concentration of polymer reduced the drug release at 6h. Also Factor X₂- concentration of ethyl cellulose has the negative effect on drug release at 6h (F=95.15, p=0.0001). From the equation it is also observed that interaction terms have also the negative effect on drug release at 6 hrs. The lack of fit was found non-significant, which is desired. The reason cited above for release retardation is also applicable at this sampling time. The equation in terms of un-coded factors is:

Y₂ =+74.59 - 6.72 X₁ - 6.54 X₂ - 3.22 X₁ X₂ - 2.96 X₁²- 2.35 X₂²

Response Y₃ - Drug Release at 10h:

The main effect and interaction terms both have significant effect on response Y₂ as p value is <0.05. Therefore, conclusions cannot be drawn from the numerical values of the coefficients of the main effects and interaction effect. In such instances, it is common to draw the conclusions from the contour plot rather the values of coefficients. Figure 2 shows the contour plot for % CDR at 10h. The contour lines are non-linear in nature. Factor X₂- concentration of ethyl cellulose was the most significant factor (F=71.75, p=0.0001) affecting the drug release at 10h negatively; increase in the concentration of polymer reduced the drug release at 10h. Also X₁-Concentration of Compritol has the negative effect on drug release at 10h (F=41.35, p=0.0004). From the equation it is also observed that interaction terms have also the negative effect on drug release at 10h. The lack of fit was found non-significant, which is desired. The reason cited above for release retardation is also applicable at this sampling time. The equation in terms of un-coded factors is:

Y₃ = +93.90 - 2.70 X₁ - 8.08 X₂ + 6.40 X₁ X₂ - 4.16 X₁²- 3.52 X₂²

Design space and validation of response surface methodology:

Design space is identified using overlaid plot which gives acceptable region combining the all contour plots of each responses. Overlaid plot was drawn in design expert software which is shown in Fig. 2. An ideal product is one which satisfies the requirements of drug released at 2, 6 and 12 h. The yellow colored region shown in overlaid plot is optimized region. The scientist is free to choose any point within the design space. It is worthwhile to note that FDA requires that the design space be clearly defined in ANDA. Optimization was achieved by computing overall desirability. The software suggested that when the concentration of Compritol and ethyl cellulose was 0.84 and -0.70 (see the square within the overlaid plot), the three requirements, outlined in Table 7 are satisfied.

The validation of model was performed by grid search analysis method. For this, additional two random check point batches were selected from extreme points of the design space. The responses of both the batches were predicted using software. These additional two batches were formulated and responses were observed or recorded. The observed responses were compared with the predicted responses using percentage prediction error. The observed responses, predicted responses and percentage prediction error were shown in Table 7. The percentage prediction error was found to be in the range of -4.92 to 2.40 which is highly desirable. . The low magnitudes of error as well as the significant values of R² in the present study prove high predictive ability of RSM.

Table7: Validation of model

Check point Batches	M₁	M₂	M₃
X₁(%)	0.84	0.59	0.55
X₂(%)	-0.70	-0.17	-0.60
Y₁^p(%)	44.99	44.99	44.99
Y₁^a(%)	42.27	43.29	46.12
% Error^#	-6.43	-3.92	2.45
Y₂^p(%)	68.63	68.27	68.98
Y₂^a(%)	75.08	65.25	66.67
% Error^#	8.59	-4.62	-3.46
Y₃^p(%)	87.7	86.94	88.4
Y₃^a(%)	92.26	85.29	87.14
% Error^#	4.94	-1.93	-1.44

^a Actual Value, ^p Predicted Value, M₁Optimized Batch, M₂ and M₃ Check point batches having desirability near 0.99 from grid search

Risk mitigation and control strategy:

The risk mitigation and control strategy demonstrates product knowledge of the current process and assures that quality is built into the product and not just tested- the objective of QbD. As seen in the initial risk assessment, the attribute which secure less than 60 RPN score had least chances of affecting final product quality (CQA). Hence it was not further investigated and marked as low risk in risk control strategy. Concentration of compritol and ethyl cellulose secured more than 60 RPN score, so they were considered more significant. These both factors were studied further in optimization study by central composite design. Central Composite Design was used to investigate the multidimensional interaction of X₁, X₂ and X₃ on Y₁ and Y₂. A design space was established for these input variables which were ranked as high risk in the initial risk assessment.

Shape analysis:

Shape analysis or roundness of pellet was observed by aspect ratio parameter. A perfectly round pellet would have an aspect ratio value between 0.8 to 1.2 irrespective of size and the value >1.2 or <0.8 for pellets that were progressively non-spherical. The aspect ratio values of all the batches were found between 0.8 to 1.2 except Batch G4, G6 and G8. The values of aspect ratio of G4, G6 and G8 were found to be > 1.2 due to higher concentration of Compritol 888 ATO and ethyl cellulose. The aspect ratio of optimized batch was found to be 0.99 so found spherical in shape. Loading of extrudates, spheronization time and speed had also great influence over the pellet shape but they were kept constant during process. Table 7 shows the shape analysis and data of aspect ratio for factorial designed batches.

Table 8: Characterization of Pellets

Batch	Aspect Ratio	Shape	Pellet Size (mm)	% Yield (%)	Carr's Index (%)	Hausner's Ratio	Angle of repose (˚)	Drug Content (%)	Friability (%)
G 1	0.90±0.12	Spherical	1.20±0.22	98.00±0.40	6.38±0.19	1.06±0.002	18.33±0.47	98.68±1.41	0.79±0.04
G 2	0.91±0.10	Spherical	1.04±0.12	96.67±1.24	6.98±0.22	1.07±0.002	17.33±0.48	98.03±2.14	0.41±0.06
G 3	1.10±0.2	Spherical	1.07±0.12	98.50±1.08	6.82±0.22	1.073±0.002	17.00±0.82	99.34±1.93	0.38±0.04
G 4	1.33±0.02	Spherical + Oval	0.89±0.01	97.17±1.64	6.98±0.22	1.07±0.003	17.84±0.85	99.34±1.61	0.01±0.01
G 5	1.00±0.25	Spherical	1.18±0.02	98.67±0.84	6.68±0.36	1.07±0.004	18.66±0.24	99.12±1.24	0.74±0.06
G 6	1.23±0.5	Spherical + Oval	0.90±0.01	96.33±1.24	6.00±0.17	1.06±0.001	18.84±0.23	101.97±1.61	0.06±0.04
G 7	1.01±0.24	Spherical	1.13±0.05	96.67±0.47	6.38±0.19	1.06±0.002	19.17±0.24	101.75±2.17	0.68±0.05
G 8	1.21±0.30	Spherical + Oval	0.88±0.02	97.00±0.81	6.52±0.19	1.06±0.002	18.66±0.48	98.25±2.75	0.13±0.14
G 9	0.91±0.24	Spherical	0.96±0.04	97.00±1.47	6.520.19	1.06±0.002	19.90±0.54	100.88±2.03	0.23±0.07
G 10	0.97±0.30	Spherical	1.06±0.10	97.67±1.24	6.840.42	1.07±0.004	19.34±0.47	101.32±2.46	0.19±0.08
G 11	1.13±0.90	Spherical	0.96±0.02	97.33±1.64	6.520.19	1.06±0.002	19.00±0.82	100.88±2.64	0.23±0.09
G 12	1.10±0.78	Spherical	0.98±0.03	95.83±1.17	6.12±0.17	1.06±0.001	16.67±0.47	98.90±2.64	0.12±0.04
G 13	1.00±0.98	Spherical	1.03±0.05	97.00±1.47	6.26±0.32	1.06±0.003	19.50±1.08	99.78±2.48	0.13±0.05
Optimum Batch	0.99±0.35	Spherical	0.98±0.14	98.24±1.20	7.07±1.24	1.08±0.014	18.66±0.94	98.69±0.31	0.2±0.07

n=3

Pellet Size and % yield of pellets:

Pellet size was performed optical microscopy method which was depicted in Table 8. Number of other process parameters like orifice diameter, spheronization time, spheronization speed and load of extrudates also affects the pellet size but they all kept constant during formulation of all batches. The results of pellet size of different batches demonstrate that as concentration of polymer increase, the pellet size decreases and vice versa. Batch G1, G5 and G7 had high pellet size due to lower concentration of polymer while batch G4, G6 and G8 showed less pellet size due to high concentration of polymers. The Pellet size of optimized batch was found to be 0.98±0.14 mm. Yield is calculated only from the pellet size in the range of 0.85 to 1.2mm. Yield of all the batches were found very high in the range of 95.83 to 98.50 % mainly due to the effect of pregelatinized starch which is sprinkled after the stage of extrusion.

Micromeritic Property:

Flow properties are the significant concern in the formulation and industrial production of oral solid dosage form .All the results of micromeritic properties were shown in Table 8. The Carr’s index of different batches of pellets was found to be in the range of 6.00 to 7.07 which showed excellent flow property and compressibility. Angle of repose is characteristic to predict the flow rate of powder. Angle of repose of pellets was found to be in the range of 16.67 to 19.50º which indicated Good flow property. Hausner’s ratio was found to be in the range of 1.06-1.08 which indicated Good flow property.

Surface characterization by SEM:

SEM images were taken for the optimized batch pellets. The pellets were found to be spherical and surface was found to be smooth as shown in Fig. 3. The smooth surface as well as desired pellet size was obtained due to the effect of pregelatinized starch.

Fig. 3: SEM photograph of pellet

Friability of pellets (%):

In general, friability indicates the ability of pellets to withstand the shear forces during handling and various pharmaceutical procedures. All thirteen batches of pellets were found to have high mechanical strength, as indicated by their friability values (<1% w/w). Friability of all thirteen batches was found to be less than 1 % ranging between 0.01- 0.79 % and hence were found to be within the limits. As the concentration of polymer increases, the pellets were become less friable so finally friability decreases. G4 and G6 batches have very less friability as both batches contain high amount of polymers i.e. Compritol and Ethyl Cellulose. Vice versa, G1 and G7 batches contain low amount of polymers which yield slightly high friability value but it was less than 1%.

Drug content of pellets:

Drug content of Batch G1 to G13 was ranging between 98.03 to 101.97 %w/w as shown in Table 8. Drug content of Optimized batch was found to be 98.69%w/w.

In vitro Drug Release Study and drug release kinetic:

In vitro drug release profiles of all GH pellet Batches as well as marketed formulation of GH and optimized GH Pellet were depicted Fig. 4. The batches which contained lesser amount of Compritol and Ethyl Cellulose showed higher drug release. As shown in figure 2 as ratio of polymers increased from 5 % to 25 %, it resulted into greater retardation of drug release. This might be due to an increased polymer concentration resulting in a decrease in the total porosity of the matrices (initial porosity plus porosity due to dissolution of the drug), decreasing the penetration of the dissolution medium into the matrix system and thus reducing drug dissolution. In addition, increasing the polymer content led to an increase in the drug diffusion path length, which in turn retarded drug diffusion from the matrix ²⁹. Batch G 4, G 6 and G 8 which contains high amount of polymer showed less drug release compared to marketed product. Drug release was found incomplete in those all the batches. Batch G 5 and G 7 showed high drug release as concentrations of polymers were less in both batches. Both the batches released the drug more than 90% in 10 hrs.

It shows that the best-fit release kinetic data with the highest values of regression coefficient (r²) were shown by Higuchi models and zero order. The values of n were in the range of 0.5 to 0.6088 (i.e., between 0.50 – 0.89) exhibits non-fickian (anomalous) transport in which the drug was delivered by combined effect of diffusion and polymer relaxation (21). Data of r² indicate that Peppas models also suitably described the release of Galantamine from the pellets. None of the formulation exhibited first order kinetics.

All the batches were compared with marketed formulation and similarity factor f₂ was calculated for each batch. Batch G4 (20 % Compritol and 20% EC), G6 (25% Compritol and 15% EC) and G8 (15% Compritol and 25 % EC) which contains high amount of polymers showed f₂ value < 50 means they were not similar to the marketed formulation. Remaining all the batches had the f₂value between 50 to 100 means showed similar drug release to that of marketed formulation. The f₂value of optimized batch was found to be 77.03.

Fig. 4: In vitro drug release of GH pellet

Prediction of in vivo plasma concentration–time profile from in vitro release data:

The bioavailability parameters C_max, T_max and AUC, predicted from back calculation of Wagner Nelson approach method were found to be 44.59ng/mL, 2h and 299.15 respectively for optimized formulation. These parameters were found to be similar with reported in vivo PK data (42.3 ng/ml, 1.9 hrs and 277 respectively). The percentage predication error (% PE) values for C_max, T_max and AUC were found to be –5.42%, -5.26% and -8.00% for GH 16 mg CR Pellets. It is clear that %PE for each parameter is not exceeding 15%. These results indicate the ability and validation of numerical convolution techniques to predict the plasma drug profile from in vitro release data. The predicted plasma concentration time is shown in Fig. 5.

Fig. 5: Predicted in vivo plasma concentration time profile from in vitro drug release data

CONCLUSION:

The current pharmaceutical quality system incorporates the concepts of quality metrics, pharmaceutical development (ICH Q8), quality risk management (QRM, ICH Q9), and pharmaceutical quality system (ICH Q10). Based on the prior knowledge and available literature, detailed failure mode and effect analysis, a popular QRM technique, was carried out to identify potentially critical factors for identifying design space. One of the catchy finding of the present study was to use pregelatinized starch in the post extrusion step to obtain superior product and high yield. The drug release was significantly influenced by the amount of Compritol and ethyl cellulose. The findings of the current study can be useful to draft the control strategy for the life cycle management of the product.

ACKNOWLEDGEMENT:

The authors are grateful to the authorities of Anand Pharmacy College, Anand for the facilities.

CONFLICT OF INTEREST:

The authors do not have any conflict of interest.

REFERENCES:

1. Muley S, Nandgude T, Poddar S. Extrusion – spheronization a promising pelletization technique : In-depth review. Asian J Pharm Sci [Internet]. 2016;11(6):684–99. Available from: http://dx.doi.org/10.1016/j.ajps.2016.08.001

2. Roy P, Shahiwala A. Multiparticulate formulation approach to pulsatile drug delivery : Current perspectives. J Control Release [Internet]. 2009;134(2):74–80. Available from: http://dx.doi.org/10.1016/j.jconrel.2008.11.011

3. Wang J, Kan S, Chen T, Liu J. Application of quality by design (QbD) to formulation and processing of naproxen pellets by extrusion – spheronization. Pharm Dev Technol. 2014;1–11.

4. Bölcskei É, Regdon G, Sovány T, Ghanam D, Knop K, Kleinebudde P, et al. Preparing of pellets by extrusion/spheronization using different types of equipment and process conditions. Drug Dev Ind Pharm. 2014;40(6):762–4.

5. Poulomi Banarjee. Hindustan Times. India has second most number of individuals suffering from dementia, Hindustan Times [Internet]. 2016 Sep 20; Available from: http://www.hindustantimes.com/india/india-s-cities-are-faced-with-a-severe-waste-management-crisis/story-vk1Qs9PJT8l1bPLCJKsOTP.html

6. Kuna Y, Kumari N. Chronic effects of anti-Alzheimer ’ s drug, Galantamine hydrobromide on cholinergic system of mice brain. JOPR J Pharm Res [Internet]. 2013;6(7):714–9. Available from: http://dx.doi.org/10.1016/j.jopr.2013.06.010

7. Yang S. Preparation and evaluation of novel galantamine hydrobromide sustained-release capsule. Bangladesh J Pharmacol. 2017;11(2016):1–10.

8. FDA. Quality by Design for ANDAs: An Example for Modified Release Dosage Forms [Internet]. 2011. Available from: https://www.google.com/search?q=Qbd+mr+dosage+form&ie=utf-8&oe=utf-8

9. Stamatis DH. Failure Mode And Effect Analysis FMEA from Theory to Execution. 2nd ed. Milwaukee: ASQ Quality Press; 2003. 1-459 p.

10. Fahmy R, Kona R, Dandu R, Xie W, Claycamp G, Hoag SW. Quality by Design I : Application of Failure Mode Effect Analysis ( FMEA ) and Plackett – Burman Design of Experiments in the Identification of “ Main Factors ” in the Formulation and Process Design Space for Roller-Compacted Ciprofloxacin Hydrochloride I. AAPS PharmSciTech. 2012;13(4):1243–54.

11. Bose A, Wong TW, Singh N. Formulation development and optimization of sustained release matrix tablet of Itopride HCl by response surface methodology and its evaluation of release kinetics. Saudi Pharm J [Internet]. 2013;21(2):201–13. Available from: http://linkinghub.elsevier.com/retrieve/pii/S1319016412000321

12. Singh G, Pai RS, Devi VK. Optimization of Pellets Containing Solid Dispersion Prepared by Extrusion / Spheronization Using Central Composite Design and Desirability Function. J Young Pharm [Internet]. 4(3):146–56. Available from: http://dx.doi.org/10.4103/0975-1483.100020

13. Gangurde AB, Sav AK, Javeer SD, Moravkar KK, Pawar JN, Amin PD. Modified extrusion-spheronization as a technique of microencapsulation for stabilization of choline bitartrate using hydrogenated soya bean oil. Int J Pharm Investig. 2015;5(4):275–83.

14. Akhgari A, Abbaspour MR, Pirmoradi S. Preparation and evaluation of pellets using acacia and tragacanth by extrusion-spheronization. DARU J Pharm Sci [Internet]. 2011;19(6):417–23. Available from: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3436078&tool=pmcentrez&rendertype=abstract

15. Kleinebudde P. Application of low substituted hydroxypropylcellulose ( L-HPC ) in the production of pellets using extrusion / spheronization. Int J Pharm. 1993;96:119–28.

16. Raval M, Sheth N, Ramani R. Formulation and evaluation of sustained release enteric-coated pellets of budesonide for intestinal delivery. Int J Pharm Investig [Internet]. 2013;3(4):203. Available from: http://www.jpionline.org/text.asp?2013/3/4/203/121294

17. Amin P, Deshmukh K. Meltlets ® of soy isoflavones: Process optimization and the effect of extrusion spheronization process parameters on antioxidant activity. Indian J Pharm Sci [Internet]. 2013;75(4):450. Available from: http://www.ijpsonline.com/text.asp?2013/75/4/450/119831

18. Desai N, Jain S, Pirthipal Singh PS, Amin P. Novel orodispersible compositions of nutraceuticals prepared by the technology of extrusion-spheronization. J Appl Pharm Sci. 2017;7(4):31–7.

19. Galantamine Extended-Release Capsules, USP Revision Bulletin Official. 2016.

20. Yan H, Zhang S, He J, Liu J. Application of ethyl cellulose, microcrystalline cellulose and octadecanol for wax based floating solid dispersion pellets. Carbohydr Polym [Internet]. 2016; Available from: http://dx.doi.org/10.1016/j.carbpol.2016.04.050

21. Yan X, He H, Meng J, Zhang C, Hong M, Tang X. Preparation of lipid aspirin sustained-release pellets by solvent-free extrusion / spheronization and an investigation of their stability. 2012;38(103):1221–9.

22. Korsmayer RW, Gumy R, Doelker E, Buri P, Peppas NA. Mechanisms of solute release from porous hydrophilic polymers. interantional jouranl Pharm. 1983;15:25–35.

23. Higuchi T. Mechanism of Sustained- Action Medication. J Pharm Sci. 1963;1145–9.

24. Zhang Y, Huo M, Zhou J, Zou A, Li W, Yao C, et al. DDSolver: An Add-In Program for Modeling and Comparison of Drug Dissolution Profiles. AAPS J [Internet]. 2010;12(3):263–71. Available from: http://www.springerlink.com/index/10.1208/s12248-010-9185-1

25. Singhvi G, Shah A, Yadav N, Saha RN. Prediction of in vivo plasma concentration – time profile from in vitro release data of designed formulations of milnacipran using numerical convolution method. Drug Dev Ind Pharm. 2013;1–4.

26. Mukesh Gohel , R. R. Delvadia, D. C. Parikh, M. M. Zinzuwadia, C. D. Soni, K. G. Sarvaiya, Neelima R. Mehta BRJ and ASD. Simplified Mathematical Approach for Back Calculation in Wagner-Nelson Method. Pharm Rev. 2005;3(2).

27. Cheng C, Wu P, Lee H, Hsu K. Development and validation of an in vitro e in vivo correlation ( IVIVC ) model for propranolol hydrochloride extended-release matrix formulations. J Food Drug Anal [Internet]. 2013;22(2):257–63. Available from: http://dx.doi.org/10.1016/j.jfda.2013.09.016

28. Kan S, Lu J, Liu J, Wang J, Zhao Y. ScienceDirect A quality by design ( QbD ) case study on enteric- coated pellets : Screening of critical variables and establishment of design space at laboratory scale. Asian J Pharm Sci [Internet]. 2014;9(5):268–78. Available from: http://dx.doi.org/10.1016/j.ajps.2014.07.005

29. Dhameliya P. Preparation and evaluation of sustained release tablet of eperisone hydrochloride by Compritol ATO 888 as a matrix forming agent. J drug deklvery Ther. 2014;4(3):132–7.

Received on 28.05.2018 Modified on 12.06.2018

Research J. Pharm. and Tech 2018; 11(11): 4899-4910.

DOI: 10.5958/0974-360X.2018.00892.2

Quality Attributes of the Drug Product	Target	Is it a CQA?	Justification
Physical Attributes (Appearance/ odour/ Size/ Micromeritic property/ Friability)	Physical attributes should be acceptable to the patient. No visual defects observed.	No	Physical attributes are not directly linked to safety and efficacy. Therefore, they are not critical. The target is set to ensure patient acceptability.
Route of Administration	Oral	No	Dosage form designed to be administered orally
Dosage form & Dosage strength	Pellets and 200mg	No	Uniform distribution in GIT and Therapeutic dose
Identification	Positive for drug substance	Yes	Though identification is critical for safety and efficacy, this CQA can be effectively controlled by the quality management system and will be monitored at drug product release. Formulation and process variables do not impact identity.
Assay (whole capsule)	100.0% of label claim	Yes	Variability in assay will affect safety and efficacy; therefore, assay is critical.
Content Uniformity	Conforms to USP <905> Uniformity of Dosage Units	Yes	Variability in content uniformity will affect safety and efficacy. Content uniformity of pellet is critical.
Drug Release	Similar drug release to marketed formulation: f2 > 50	Yes	For capsule containing a multi-particulate system, a non-uniform distribution of beads may cause different drug release profiles between whole and split tablets. Therefore, it is critical and the target is set in accordance with regulatory guidance.