AAPS PharmSciTech ( # 2016) DOI: 10.1208/s12249-016-0580-5

Research Article Preparation and Optimization of Immediate Release/Sustained Release Bilayered Tablets of Loxoprofen Using Box–Behnken Design Jin Wook Tak,1 Biki Gupta,1 Raj Kumar Thapa,1 Kyu Bong Woo,1 Sung Yub Kim,1 Toe Gyeong Go,1 Yongjoo Choi,1 Ju Yeon Choi,1 Jee-Heon Jeong,1 Han-Gon Choi,2 Chul Soon Yong,1,3 and Jong Oh Kim1,3

Received 8 April 2016; accepted 20 June 2016 Abstract.

The aim of our current study was to characterize and optimize loxoprofen immediate release (IR)/sustained release (SR) tablet utilizing a three-factor, three-level Box– Behnken design (BBD) combined with a desirability function. The independent factors included ratio of drug in the IR layer to total drug (X1), ratio of HPMC to drug in the SR layer (X2), and ratio of Eudragit RL PO to drug in the SR layer (X3). The dependent variables assessed were % drug released in distilled water at 30 min (Y1), % drug released in pH 1.2 at 2 h (Y2), and % drug released in pH 6.8 at 12 h (Y3). The responses were fitted to suitable models and statistical validation was performed using analysis of variance. In addition, response surface graphs and contour plots were constructed to determine the effects of different factor level combinations on the responses. The optimized loxoprofen IR/SR tablets were successfully prepared with the determined amounts of ingredients that showed close agreement in the predicted and experimental values of tablet characterization and drug dissolution profile. Therefore, BBD can be utilized for successful optimization of loxoprofen IR/SR tablet, which can be regarded as a suitable substitute for the current marketed formulations. KEYWORDS: Box–Behnken design; loxoprofen; optimization; sustained release.

INTRODUCTION Loxoprofen, chemically designated as (RS)-2-{4-[(2oxocyclopentyl)-methyl]-phenyl}-propionic acid, is a nonsteroidal anti-inflammatory drug (NSAID) belonging to the propionic acid derivative group (1). It is a nonselective COX inhibitor possessing anti-inflammatory, analgesic, and antipyretic effects (2). As a prodrug, loxoprofen is metabolized to its active form, i.e., its trans-alcohol form, which exerts an inhibitory action on prostaglandin E2 production almost three times as strong as that of indomethacin (3). Hence, it was found to be effective against chronic rheumatoid arthritis, osteoarthritis, back pain, periarthritis of shoulder, and postoperative or posttraumatic pain after tooth extraction (4–6). The drug has less gastric irritation and less ability to Electronic supplementary material The online version of this article (doi:10.1208/s12249-016-0580-5) contains supplementary material, which is available to authorized users. 1

College of Pharmacy, Yeungnam University, 214-1, Dae-Dong, Gyeongsan, 712-749, South Korea. 2 College of Pharmacy, Institute of Pharmaceutical Science and Technology, Hanyang University, 55, Hanyangdaehak-ro, Sangnokgu, Ansan, 426-791, South Korea. 3 To whom correspondence should be addressed. (e-mail: [email protected]; [email protected])

produce gastric lesions and toxicity compared to other NSAIDs, making it further suitable for use (3). Regulatory authorities, including US FDA, have recently strongly underlined the implementation of a quality by design (QbD) approach to counter suboptimal quality management of drugs (7). QbD is designated as a systematic approach to development of pharmaceutical products beginning with predefined objectives and underlining product and process understanding and process control, based on sound science and quality risk management (8). Construction of design spaces around individual or multiple unit operations is an important component of the control strategy for QbD. Design of experiments (DoE) has proved an extremely useful tool for construction of design spaces. Experimental design can be employed for construction of a predictive model of all critical responses, facilitate the identification of all potential independent factors, and evaluation of those factors in a systematic and rapid manner (9, 10). Tablet characteristics are affected by a number of different parameters during their formulation. Thus, optimization of the important parameters, using a suitable experimental design approach, can aid in preparation of tablets with optimal characteristics. Conventional optimization technique requires a single parameter to be varied at a time, resulting in unreliable data and improper conclusions in addition to wastage of excipients resulting from a large set of runs. 1530-9932/16/0000-0001/0 # 2016 American Association of Pharmaceutical Scientists

Tak et al. Statistical optimization, using a suitable experimental design, could be an appropriate remedy for this problem (11). The experimental designs most commonly applied in pharmaceutical experiments are central composite design (CCD) and Box–Behnken design (BBD) (12). Compared to CCD, BBD employs a reduced number of experimental runs, making its application more convenient. Marketed loxoprofen tablets are available in 60-mg dose that must be taken three times a day (13). For treatment of chronic musculoskeletal disorders, loxoprofen tablets should be administered for a long period of time; hence, the available dosage regimen could potentially pose problems of patient compliance, owing to their dosing frequency. Sustained release dosage forms can lead to reduction in dosing frequency and thereby ensure better patient compliance (14). Thus, a sustained release tablet of loxoprofen that can reduce dosing frequency has become a necessity. Consequently, the main objective of this study was to prepare sustained release tablets of loxoprofen optimized with the utilization of BBD. Moreover, incorporation of initial immediate release characteristics into the formulation to be succeeded by the required sustained release characteristics would further enhance the overall formulation characteristics. Hence, we developed immediate release (IR)/sustained release (SR) combined bilayer matrix tablets of loxoprofen. Our ultimate goal was to optimize loxoprofen (90 mg) release from the IR/SR tablets for a period of 12 h to reduce the dosing frequency and aid in enhancement of patient compliance. MATERIALS AND METHODS Materials Loxoprofen sodium was obtained from Kyongbo Pharmaceutical Co., Ltd. (Asan, South Korea). Hydroxypropyl

methylcellulose (HPMC, Metolose® 90SH-100,000) was purchased from Shin-Etsu Chemical Co., Ltd. (Tokyo, Japan). Magnesium stearate and glyceryl behenate (Compritol 888ATO®, Gattefossé) were kindly gifted by Dong-A Pharmaceutical Company (Yongin-si, South Korea). Spray-dried lactose was supplied from Daejung Chemicals & Metals Co., Ltd. (Siheung, South Korea) and Eudragit RL PO® from Evonik Degussa (Hanau, Germany). Poly-vinyl-poly-pyrrolidone, also known as crospovidone (Kollidon CL®), and poly-vinylpyrrolidone K30 (PVP-K30) were purchased from BASF Co. Ltd. (Ludwigshafen, Germany) and ISP Technologies, Inc. (Texas, USA), respectively. All organic solvents were of HPLC grade and all chemicals were of analytical grade. Methods Box–Behnken Design A three-factor, three-level BBD was used to examine and optimize the main effects, interaction effects, and quadratic effects of the formulation ingredients on the drug release profiles. A three-factor, three-level BBD requires 15 experimental runs with three central points to determine the experimental error and the precision of the design (15). The 15 experimental runs were operated using Design-Expert software (V. 8.0.4; Stat-Ease Inc., Minneapolis, Minnesota). Based on preliminary experiments, three independent variables were selected and their ranges established, as shown in Table I, along with the dependent variables and the goals set for the dependent variables. The results obtained for each response were fitted to a linear model and a quadratic model, described by the following polynomial equations.

Linear equation : y ¼ a þ b1 X 1 þ b2 X 2 þ b3 X 3 Quadratic equation : y ¼ k0 þ k1 X 1 þ k2 X 2 þ k3 X 3 þ k4 X 1 X 2 þ k5 X 2 X 3 þ k6 X 1 X 3 þ k7 X 21 þ K 8 X 22 þ k9 X 23

where y is the measured response, b1, b2, b3, and k0–k9 are the regression coefficients, and X1, X2, and X3 are the independent factors. The models were validated by analysis of variance (ANOVA), lack of fit, and multiple correlation coefficient (R2) tests (16).

represent the minimum and maximum possible value. If Yi is equal to or less than Ymin, then, di = 0, if Yi is higher or equal to Ymax, di = 1. For the response to be minimized, the desirability function is defined as:

Optimization Using Desirability Functions

di ¼ ðY max −Y i Þ=ðY max −Y min Þ

All responses were optimized using the desirability function approach introduced by Derringer and Suich (17). Each response was associated with its partial desirability function (di), scaled from 0 to 1 (closest to the assigned target), and a utility function was computed to provide the overall or global desirability. For the response to be maximized, the desirability function can be defined as: di ¼ ðY i −Y min Þ=ðY max −Y min Þ where di is the individual desirability of the response to be maximized, Yi is the experimental result, and Ymin and Ymax

If Yi is higher than or greater than Ymax then di = 0, and if Yi is less than or below minimum then di = 1. Here, lower and upper limits for the responses were set from the highest and lower limits of the observed responses. After obtaining the individual desirability value for each response, the results were combined usually together to give overall desirable function (D) as the geometric mean, given by the following equation: D ¼ ðd1  d2  d3  d4  ::::::::::  dnÞ1=n where n specifies the number of responses.

Bilayered Tablets of Loxoprofen Using Box–Behnken Design Table I. Variables Used in the Box–Behnken Design Independent variables

X1: Ratio of drug in the IR layer to total drug X2: Ratio of HPMC to drug in the SR layer X3: Ratio of Eudragit RL PO to drug in the SR layer Dependent variables Y1: Drug release at distilled water at 30 min (%) Y2: Drug release at pH 1.2 at 2 h (%) Y3: Drug release at pH 6.8 at 12 h (%)

Preparation of Loxoprofen Sodium Granules and Tablets IR/SR bilayer tablets of loxoprofen were prepared using a wet granulation method. The weights of IR and SR layers were fixed in all formulations at 70 and 380 mg, respectively. For granulation of the IR layer, different amounts of loxoprofen were mixed with lactose, crospovidone, PVP-K30, and magnesium stearate. Similarly, for granulation of the SR layer, loxoprofen was mixed with lactose, Eudragit RL PO, HPMC 2208, PVPK30, Compritol 888-ATO, and magnesium stearate. Eudragit RL PO, HPMC 2208, and Compritol 888-ATO were used to sustain the drug release profiles. Weights of all excipients, except for those defined by the independent factors, were fixed. Ethanol was used as a binding solution. The kneaded mixture was sieved through an 18-mesh sieve. The detailed compositions of the prepared formulations are shown in Table II. Tablets were prepared using a single compression machine (ERWEKA AR401, Germany) with round punches, 1 cm in diameter. For preparation of the bilayer tablets, two compression steps were performed. First, the SR layer granules were compressed, followed by addition of the IR layer granules.

Levels, actual (coded) Low (−1)

Medium (0)

High (+1)

0.4 2.5 0 Constraints Range In the range In the range In the range

0.45 3 0.5

0.5 3.5 1

Goal 50% 45% Maximize

and the dissolution was continued up to 12 h. Dissolution profiles of all the prepared formulations were compared with a commercial product. HPLC Analysis Method of Loxoprofen Loxoprofen concentrations were determined using a validated HPLC method. The HPLC analysis was performed using a Hitachi HPLC system (Chromaster®, Hitachi, Japan), consisting of a pump, autosampler, column oven, and UV detector, and Open LAB® software was used for data analysis. Inertsil C8 (4.6 × 150 mm, 5 μm) column was used at a UV detection wavelength of 222 nm. Acetonitrile/water (50:50, v/v), pH adjusted to 3.5 with 10% H3PO4, was used as the mobile phase. Column temperature was maintained at 25°C. Flow rate and injection volume were 1.0 ml/min and 20 μl, respectively. Stability Test Stability test was performed by storing the prepared formulations at 40°C, RH 75% for a defined period of time. Content uniformity and dissolution tests were performed at specified time points to assess the stability over a storage period of 8 weeks.

Tablet Evaluation RESULTS AND DISCUSSION Weight Variation and Hardness Testing The weight of the tablets (n = 20) was measured using an electronic balance (RADWAG PS 360/C/2, Poland), and their hardness (n = 10) was determined using a tablet hardness tester (ERWEKA TBH 100, Germany). Dissolution Profiles of Formulations Dissolution test for the prepared formulations was performed in specified dissolution media using a paddle method (100 rpm) at 37 ± 0.5°C (TW-SM, Universal Scientific Co. Ltd., South Korea). Dissolution media was comprised of pH 1.2 and pH 6.8 solutions. Six tablets of each formulation were added separately in 900 ml of dissolution medium. Initially, the tablets were added to 750 ml of 0.1 mol/l HCl solution maintained at pH 1.2 for 2 h. After 2 h, 250 ml of 0.2 mol/l Na3PO4 solution was added to the pH 1.2 solution, to change the pH to 6.8,

Application of BBD and Preparation of IR/SR Bilayer Tablets A schematic representation of the prepared IR/SR loxoprofen tablet is presented in Fig. 1. By application of BBD, 15 different formulations were prepared based on the composition shown in Table II, which was based on the corresponding factor distribution in the BBD matrix shown in Table III. A simple design of the ranges of the three independent factors and the application of BBD in loxoprofen IR/SR bilayer tablets are shown in supplementary Figs. S1 and S2. The purpose of employing BBD was to optimize the drug fraction in the IR and SR layers, and the amounts of HPMC and Eudragit RL PO in the SR layer with constraints on percentage drug release in distilled water at 30 min, percentage drug release in pH 1.2 at 2 h, and percentage drug release in pH 6.8 at 12 h, and to prepare loxoprofen IR/SR bilayer tablets

450 450 450 450 450 450 450 450 450 450 450 450 450 450 450 380 380 380 380 380 380 380 380 380 380 380 380 380 380 380 5.00 5.00 5.00 5.00 5.00 5.00 5.00 5.00 5.00 5.00 5.00 5.00 5.00 5.00 5.00 47.50 47.50 47.50 47.50 47.50 47.50 47.50 47.50 47.50 47.50 47.50 47.50 47.50 47.50 47.50 7.20 7.20 7.20 7.20 7.20 7.20 7.20 7.20 7.20 7.20 7.20 7.20 7.20 7.20 7.20 196.64 140.46 214.52 127.69 183.87 168.55 153.23 168.55 153.23 140.46 178.76 153.23 183.87 168.55 196.64

Total weight (mg) Mg. stearate (mg) Compritol (mg) PVPK30 (mg) HPMC2208 - 100000 (mg)

Fig. 1. Schematic representation of the developed loxoprofen formulation

possessing similar characteristics to the commercial product of loxoprofen. Previously, BBD has been utilized to successfully optimize oral sustained release dosage forms (18, 19).

0.00 0.00 30.65 25.54 61.29 28.09 0.00 28.09 51.08 56.18 25.54 30.65 0.00 28.09 56.18 67.48 123.66 13.85 116.00 13.85 67.48 116.00 67.48 64.93 67.48 64.93 75.14 75.14 67.48 11.30 56.18 56.18 61.29 51.08 61.29 56.18 51.08 56.18 51.08 56.18 51.08 61.29 61.29 56.18 56.18 70 70 70 70 70 70 70 70 70 70 70 70 70 70 70 3.00 3.00 3.00 3.00 3.00 3.00 3.00 3.00 3.00 3.00 3.00 3.00 3.00 3.00 3.00 3.00 3.00 3.00 3.00 3.00 3.00 3.00 3.00 3.00 3.00 3.00 3.00 3.00 3.00 3.00

Lactose (mg) Loxoprofen sodium hydrate (mg) Total weight (mg) M g . stearate (mg) P V P K30 (mg)

The observed hardness and weight variation of the 15 tablet formulations are shown in supplementary Table S1. The limits for hardness and tablet weight variation were ≥80 N and 450 mg ± 5% (427.5–472.5 mg), respectively. All formulations possessed hardness greater than 90 N with minimal variation in tablet weights (less than 0.5%), suggesting that all were in compliance with the set standards. Dissolution Profiles of Loxoprofen from Different Formulations Dissolution tests were performed for each formulation, and the dissolution profiles were compared with the dissolution profile of the commercial product. From direct comparison of the dissolution profiles, it was difficult to draw any

Table III. Experimental Matrix and Observed Responses from Randomized Runs in the Box–Behnken Design

45.97 45.97 40.86 51.08 40.86 45.97 51.08 45.97 51.08 45.97 51.08 40.86 40.86 45.97 45.97

15.93 15.93 21.04 10.83 21.04 15.93 10.83 15.93 10.83 15.93 10.83 21.04 21.04 15.93 15.93

2.10 2.10 2.10 2.10 2.10 2.10 2.10 2.10 2.10 2.10 2.10 2.10 2.10 2.10 2.10

Run

F1 F2 F3 F4 F5 F6 F7 F8 F9 F10 F11 F12 F13 F14 F15

Loxoprofen s o d i u m hydrate (mg)

IR

Lactose (mg)

Crospovidone (mg)

SR

Eudragit RL PO (mg)

Evaluation of the Tablet Characteristics

Formulation code

Table II. Composition of Loxoprofen IR/SR Bilayer Tablets

Total I R / S R weight (mg)

Tak et al.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

Independent variables

Dependent variables

X1

X2

X3

Y1 (%)

Y2 (%)

Y3 (%)

0.45 0.45 0.4 0.5 0.4 0.45 0.5 0.45 0.5 0.45 0.5 0.4 0.4 0.45 0.45

3.5 2.5 3.5 2.5 3 3 3 3 3 2.5 3.5 2.5 3 3 3.5

0 0 0.5 0.5 1 0.5 0 0.5 1 1 0.5 0.5 0 0.5 1

55.22 55.64 46.42 58.28 45.12 52.68 59.48 52.3 56.49 51.33 57.45 47.88 49.76 52.16 50.59

47.47 48.16 35.2 51.16 32.78 44.18 53.33 43.94 48.95 43.67 49.56 35.37 38.79 43.45 38.79

86.11 91.4 75.74 95.84 77.4 90.42 92.45 90.55 91.42 90.75 90.55 85.47 79.08 89.57 82.21

Bilayered Tablets of Loxoprofen Using Box–Behnken Design inference; thus we focused on the application of BBD to the dissolution data. Figure 2 shows the dissolution profiles of the drug at pH 1.2 and 6.8 (Fig. 2a), and in distilled water (Fig. 2b). Percentages of drug release at pH 1.2 at 2 h, with a target value of 45%, and percentage drug release at pH 6.8 at 12 h, with a target value of ∼100%, were chosen as two of the dependent variables. Gastric emptying time is generally 2 h, and thus dissolution at gastric pH (pH 1.2) was performed for 2 h (20). During this period, 40∼50% of drug release are desirable for the IR/SR bilayer tablet; therefore, a value of 45% was set as desirable. From the results, it was observed that tablets of F6 showed a drug release percentage of 44.18% at pH 1.2 at 2 h, which was pretty close to the desired release percentage. In addition, to mimic the intestinal pH condition, dissolution at pH 6.8 for 12 h was performed wherein maximum drug release is optimal (21). Tablets of F4 showed the highest dissolution percentage of almost 96% during this latter period. Likewise, the drug dissolution percentage in distilled water at 30 min was included as the third dependent variable for comparison with the commercial product. The commercial product showed almost 50% release of loxoprofen at the test time point of 30 min (Fig. 2b). To

prepare a formulation possessing similar characteristics, we aimed at a drug release of 50% in distilled water at 30 min for our formulation. Tablets of F13 and F15 showed desirable drug release in distilled water at 30 min (∼50%). Statistical Analysis of the Designed Experiment BBD, with ratio of drug in the IR layer to total drug (X1), ratio of HPMC to drug in the SR layer (X2), and ratio of Eudragit RL PO to drug in the SR layer (X3) as independent variables, was applied to optimize the formulation with constraints on drug release in distilled water (Y1) and drug release at pH 1.2 and pH 6.8 (Y2 and Y3) at specified time points. A total of 15 experimental runs were carried out based on the experimental matrix of BBD, which is shown in Table III along with the observed responses. Data of the responses Y1 and Y2 were fitted to the linear model, whereas those of Y3 were fitted to the second-order quadratic model. Validations of the models were performed by analysis of variance (ANOVA) test, lack of fit tests, and correlation coefficients (R2). The

Fig. 2. Dissolution profiles of loxoprofen tablets at a pH 1.2 (2 h) and pH 6.8 (10 h), and b distilled water. Values are mean ± standard deviation (SD, n = 6)

Tak et al. results of ANOVA and lack of fit tests for evaluation of the models for each of the three responses are shown in Table IV. ANOVA was used to test the statistical significance of the ratio of mean square variation due to regression and mean square residual error. As shown in Table IV, the models representing the responses Y1 and Y2 showed good fit to the linear model at 5% significance level, as confirmed by their model p values (both <0.0001) However, the model representing the response Y3 showed good fit to the quadratic model at 5% significance level, as confirmed by its model p value (p < 0.0001). The corresponding large values of F (346.68, 134.43, and 159.18 for Y1, Y2, and Y3, respectively) indicate that most of the variation in the responses can be explained by the regression equation. At 5% significance level, the model was considered significant if p < 0.05 for the model and p > 0.05 for lack of fit. The lack of fit measures the failure of the model to represent the data in the experimental domain at points not included in the regression. Nonsignificant lack of fit is a desirable statistical parameter to prove the fitting of the model to the responses. As shown in Table IV, all models showed nonsignificant lack of fit. In addition, Table IV shows that all three factors (X1, X2, and X3) appeared to have significant effects on Y1 (p < 0.05 for all three), with X1 having the most significant effect (evident from its highest F value), followed by X3 and then X2. The effects of the three independent factors on response Y2 followed a trend similar to that of Y1. On response Y3, all three factors

had significant direct effects, X1–X2 and X2–X3 had significant interaction effects, and X 1 and X 3 had significant quadratic effects. As with the other two responses, X1 had the most significant direct effect on Y3 as well. However, it was followed by X2 and then X3, unlike the previous cases. These results were comparable with those of previous reports (18, 19). The correlation coefficients (R2) for the responses Y1, Y2, and Y3 are shown in supplementary Table S2. The correlation coefficients represent the confidence that the regression equations would predict the observed value better than the mean. The correlation coefficients (R2) for the responses Y1, Y2, and Y3 were appreciably high with 0.9895, 0.9734, and 0.9965, respectively. The adjusted R2 and the predicted R2 values were also in good agreement (Table S2). Factor coefficients for the three responses are included in supplementary Table S3. The factor coefficients indicate the quantitative effects of the independent variables (X1, X2, and X3) and their interactions on the responses. Coefficients (factor intercepts) with more than one term (X1·X2, X1·X3, and X2·X3) and those with the higher order terms (X12, X22, and X32) indicate the interactions and quadratic effects, respectively. Positive signs represent the synergistic effects of the factors, whereas negative signs represent the antagonist effects of the factors on the responses. From Table S3, we can infer that X1 had the most dominant effect on Y1, as suggested by its highest coefficient value, followed by X3 and then by X2. The effect of X1 on Y1 was synergistic, while those of X2 and X3 on Y1 were antagonistic. X1, X2, and X3

Table IV. Analysis of Variance and Lack-of-Fit Tests for the Response Surface Models Source Y1: Drug release at distilled water at 30 min (%)

Y2: Drug release at pH 1.2 at 2 h (%)

Y3: Drug release at pH 6.8 at 12 h (%)

Model X1 X2 X3 Residual Lack of fit Pure error Cumulative total Model X1 X2 X3 Residual Lack of fit Pure error Cumulative total Model X1 X2 X3 X1X2 X1X3 X2X3 X12 X22 X32 Residual Lack of fit Pure error Cumulative total

DF

Sum of squares

Mean square

F value

p value

3 1 1 1 11 9 2 14 3 1 1 1 11 9 2 14 9 1 1 1 1 1 1 1 1 1 5 3 2 14

261.80 225.99 1.49 34.32 2.77 2.62 0.14 264.57 539.11 462.99 6.73 69.38 14.70 14.43 0.28 553.82 509.24 345.45 104.04 6.59 4.93 0.11 2.64 31.16 0.52 17.67 1.78 1.21 0.57 511.02

87.27 225.99 1.49 34.32 0.25 0.29 0.072

346.68 897.78 5.91 136.34

<0.0001 <0.0001 0.0333 <0.0001

4.03

0.2147

179.70 462.99 6.73 69.38 1.34 1.60 0.14

134.43 346.36 5.04 51.91

<0.0001 <0.0001 0.0463 <0.0001

11.58

0.0820

56.58 345.45 104.04 6.59 4.93 0.11 2.64 31.16 0.52 17.67 0.36 0.40 0.28

159.18 971.83 292.69 18.53 13.86 0.30 7.43 87.66 1.46 49.70

<0.0001 <0.0001 <0.0001 0.0077 0.0137 0.6091 0.0415 0.0002 0.2809 0.0009

1.42

0.4378

Bilayered Tablets of Loxoprofen Using Box–Behnken Design had similar effects on the response Y2. X1 had the most dominant direct effect on Y3 as well, but in this case, it was followed by X2 and then X3. The synergistic/antagonistic effects of the responses in the case of Y3 as well were similar to those of Y1 and Y2. In addition, X1 and X2 had the most dominant interaction effects on Y3 (synergistic), followed by the interaction effects of X2 and X3 (antagonistic), and then by those of X1 and X3 (synergistic). Factors X1 and X3 exhibited strong quadratic effects as well on the response Y3 (both antagonistic).

Table V. Amount of Different Ingredients for the Optimized Tablet (F16) Ingredients (mg)

IR layer

SR layer

Loxoprofen Lactose Crospovidone PVP-K30 Eudragit RL PO HPMC 2208 Compritol Mg stearate Total

46.989 14.911 2.10 3.0

55.161 89.176 – 7.2 38.061 137.903 47.50 5.0 380

3.0 70

Response Surface Analysis The three-dimensional response surface plots and two-dimensional contour plots are graphical representations of the regression equations and express two independent variables at once against each response (Fig. 3). Thus, the statistically significant relationship between the dependent and independent variables was further interpreted using response surface analysis. In all the response surface and contour plots, the factors showing the least significant values were fixed at their middle levels. Figure 3a shows the response surface and contour plots for the effects of X1 and X3 on Y1 at middle level of X2. As discussed in the previous section, X1 showed the most significant effect on Y1, followed by X3, thus these two factors were plotted against Y1 in the response surface and contour plots. From Fig. 3a, we can observe

that drug release in distilled water at 30 min (Y1) increased as the ratio of drug in the IR layer to the total drug (X1) increased, while it decreased as the ratio of Eudragit RL PO to drug in the SR layer (X3) increased. Figure 3b shows the response surface and contour plots for the effects of X1 and X3 on Y2 at middle level of X2. Y2 represents the initial effect of drug upon activation and release in gastric condition. As shown in Fig. 3b, similar results were observed for drug release at pH 1.2 at 2 h (Y2) as with response Y1. The Design-Expert software suggesting linear model fitting for these two responses indicated that the interaction and quadratic effects were nonsignificant. Figure 3c shows the response surface and contour plots for the effects of X1 and X2 on Y3 at middle level of X3. These plots infer that drug release at pH

Fig. 3. Response surface and contour plots showing a the effects of ratio of drug in the IR layer to total drug (X1; d,IR/d,T) and ratio of Eudragit RL PO to drug in the SR layer (X3; Eud/d,SR) on drug release in distilled water at 30 min (Y1; Rel@30min,dw), b the effects of ratio of drug in the IR layer to total drug (X1; d,IR/d,T) and ratio of Eudragit RL PO to drug in the SR layer (X3; Eud/d,SR) on drug release at pH 1.2 at 2 h (Y2; Rel@2h,pH1.2), and c the effects of ratio of drug in the IR layer to total drug (X1; d,IR/d,T) and ratio of HPMC to drug in the SR layer (X2; HPMC/d,SR) on drug release at pH 6.8 at 12 h (Y3; Rel@12h,pH6.8)

Tak et al. Table VI. Predicted and Observed Response Values for the Loxoprofen Tablets Prepared Under Optimized Conditions Responses

Predicted value

Observed value

Biasness (%)a

Y1: Drug release at distilled water at 30 min (%) Y2: Drug release at pH 1.2 at 2 h (%) Y3: Drug release at pH 6.8 at 12 h (%)

53.36 44.87 94.13

51.05 47.44 100.79

4.33% −5.73% −7.08%

a

(Predicted value-measured value) × 100 / predicted value

6.8 at 12 h (Y3) increased, initially pretty sharply while slowly later on, as the ratio of drug in the IR layer to total drug (X1) increased, while it decreased with an increase in the ratio of HPMC to drug in the SR layer. The sharp rate of initial drug release can be ascribed to quick early release of the drug from the IR layer; the steady and sustained release thereafter could be a result of slow drug diffusion from the SR layer after it imbibes water and swells gradually. Optimization Using Desirability Function After generating the model polynomial equations to relate the dependent and independent variables, the process was optimized for all three responses simultaneously using desirability function. Multiple responses including Y1, Y2, and Y3 were transformed into individual desirability scale d1, d2, and d3, respectively. Factors were set within the range. Constraints were set for all responses. Y1 and Y2 were set as 50 and 45%, respectively, while Y3 was set to be maximized. Equal importance was given to all responses. The global desirability value was calculated by combining all the individual desirability functions as the geometric mean using an extensive grid and feasibility search over the domain. The suggested optimized formulation was 0.46, 2.5, and 0.69 for X1, X2, and X3, respectively, with the corresponding desirability (D) value of 0.836 (Table V, Fig. S3). This factor level combination predicted the response as Y1 = 53.36%, Y2 = 44.87%, and Y3 = 94.13%. To confirm the model adequacy for the prediction, three batches of the optimized formulations were prepared, in vitro assay was performed, and all responses were evaluated for each

formulation (Table VI and Table S4). The optimized loxoprofen IR/SR bilayer tablet had 92.5 ± 5.0 N of hardness, 449.9 ± 1.3 mg of weight variation, 51.05% of drug release in distilled water at 30 min, 47.44% in pH 1.2 at 2 h, and 100.79% in pH 6.8 at 12 h, respectively. It can be concluded that the experimental values were in close agreement with predicted values, indicating the success of the design for evaluation and optimization of the formulation. Evaluation of the Optimized Formulation After evaluation of obtained results, a dissolution test was performed for the optimized formulation for comparison with expected results. The optimized formulation F16 showed results in close agreement with expected dissolution percentage from Design-Expert suggesting the success of the BBD combined with desirability function for the evaluation and optimization of loxoprofen IR/SR tablets. This formulation showed 2% higher loxoprofen release at pH 1.2 and 1% higher in distilled water than expected results with almost 100% drug release at pH 6.8 suggesting suitability of F16 for achievement of our ultimate goal (Fig. 4). Stability Test of Formulations Content uniformity test and dissolution test of the optimized formulation were performed for 8 weeks. Final optimized formulation showed better content u n i f o r m i t y c om p a r e d w i t h c o m m e r c i a l pr o d u c t s (Table VII). Commercial product showed irregular content of loxoprofen with a dramatic decrease upon

Fig. 4. Dissolution profiles of the optimized formulation (F16) at a pH 1.2 (2 h) and 6.8 (10 h), and b distilled water. Values are mean ± standard deviation (SD, n = 6)

Bilayered Tablets of Loxoprofen Using Box–Behnken Design Table VII. Stability Test (Content Uniformity) of the Optimized Formulation and the Commercial Product Time (weeks)

Optimized formulation

Commercial product

0 2 4 8

99.6 ± 3.5 98.5 ± 1.1 97.6 ± 6.3 91.7 ± 4.0

97.4 ± 5.1 98.2 ± 7.3 74.1 ± 0.1 71.8 ± 7.4

storage. In contrast, the optimized IR/SR formulation showed a slight decrease in loxoprofen content, lower than 10%. Likewise, changes in dissolution profiles in distilled water showed a similar tendency, wherein the optimized formulation was better compared to the commercial product (Fig. 5). Commercial products showed a dramatic change from 2 to 8 weeks with an irregular tendency and large standard deviations.

CONCLUSION IR/SR bilayer tablet formulation of loxoprofen was successfully prepared using BBD combined with desirability function for optimization. The results suggested that the optimized formulation showed appropriate weight and hardness with sustained dissolution profiles in different pH conditions and better stability compared to the commercial product. Therefore, the optimized bilayer matrix tablet of loxoprofen can be considered as a suitable substitute for the current commercial product. ACKNOWLEDGMENTS This research was supported by the Yeungnam University research grants in 2015.

REFERENCES

Fig. 5. Dissolution profiles at different time points of a the commercial product and b the optimized formulation (F16) in distilled water following storage. Values are mean ± standard deviation (SD, n = 6)

1. Tanaka Y, Nishikawa Y, Hayashi R. Species differences in metabolism of sodium 2-[4-(2-oxocyclopentylmethyl)phenyl]propionate dehydrate (loxoprofen sodium), a new anti-inflammatory agent. Chem Pharm Bull. 1983;31(10):3656–64. 2. Mu R, Bao CD, Chen ZW, Zheng Y, Wang GC, Zhao DB, et al. Efficacy and safety of loxoprofen hydrogel patch versus loxoprofen tablet in patients with knee osteoarthritis: a randomized controlled non-inferiority trial. Clin Rheumatol. 2014. doi:10.1007/s10067-014-2701-4. 3. Saito M, Kawatani M, Kinoshita Y, Satoh K, Miyagawa I. Effectiveness of an anti-inflammatory drug, loxoprofen, for patients with nocturia. Int J Urol. 2005;12(8):779–82. 4. Sakamoto C, Soen S. Efficacy and safety of the selective cyclooxygenase-2 inhibitor celecoxib in the treatment of rheumatoid arthritis and osteoarthritis in Japan. Digestion. 2011;83(1–2):108–23. 5. Ohta S, Komai O, Hanakawa H. Comparative study of the clinical efficacy of the selective cyclooxygenase-2 inhibitor celecoxib compared with loxoprofen in patients with frozen shoulder. Mod Rheumatol. 2014;24(1):144–9. 6. Yamashita Y, Sano N, Shimohira D, Danjo A, Goto M. A parallel-group comparison study of celecoxib with loxoprofen sodium in third mandibular molar extraction patients. Int J Oral Maxillofac Surg. 2014;43(12):1509–13. 7. Poudel BK, Marasini N, Tran TH, Choi HG, Yong CS, Kim JO. Formulation, characterization and optimization of valsartan selfmicroemulsifying drug delivery system using statistical design of experiment. Chem Pharm Bull. 2012;60(11):1409–18. 8. Lionberger RA, Lee SL, Lee LM, Raw A, Yu LX. Quality by design: concepts for ANDAs. AAPS J. 2008;10(2):268–76. 9. Gupta B, Poudel BK, Tran TH, Pradhan R, Cho HJ, Jeong JH, et al. Modulation of pharmacokinetic and cytotoxicity profile of imatinib base by employing optimized nanostructured lipid carriers. Pharm Res. 2015;32:2912–27. 10. Gupta B, Poudel BK, Pathak S, Tak JW, Lee HH, Jeong JH, et al. Effects of formulation variables on the particle size and drug encapsulation of imatinib-loaded solid lipid nanoparticles. AAPS PharmSciTech. 2015. doi:10.1208/ s12249-015-0384-z. 11. Meka VS, Thing LK, Gorajana A, Kolapalli VR. Formulation and optimization of gastric floating drug delivery system using central composite design and its biopharmaceutical evaluation. Pak J Pharm Sci. 2015;28(4):1373–87.

Tak et al. 12. Ferreira SLC, Bruns RE, Ferreira HS, Matos GD, David JM, Brandao GC, et al. Box-Behnken design: an alternative for the optimization of analytical methods. Anal Chim Acta. 2007;597(2):179–86. 13. Jhee OH, Lee MH, Shaw LM, Lee SE, Park KH, Kang JS. Pharmacokinetics and bioequivalence study of two brands of loxoprofen tablets in healthy volunteers. Arzneimittelforschung. 2007;57(8):542–6. 14. Wu F, Jin T. Polymer-based sustained release dosage forms for protein drugs, challenges and recent advances. AAPS PharmSciTech. 2008;9(4):1218–29. 15. Box GEP, Behnken DW. Some new three level designs for the study of quantitative variables. Technometrics. 1960;2(4):455–75. 16. Marasini N, Yan YD, Poudel BK, Choi HG, Yong CS, Kim JO. Development and optimization of self-nanoemulsifying drug delivery system with enhanced bioavailability by Box-Behnken design and desirability function. J Pharm Sci. 2012;101(12):4584–96.

17. Derringer G, Suich R. Simultaneous optimization of several response variables. J Qual Technol. 1980;12:214–9. 18. Chopra S, Patil GV, Motwani SK. Release modulating hydrophilic matrix systems of losartan potassium: optimization of formulation using statistical experimental design. Eur J Pharm Biopharm. 2007;66:73–82. 19. Kramar A, Turk S, Vrecer F. Statistical optimization of diclofenac sustained release pellets coated with polymethacrylic films. Int J Pharm. 2003;256:43–52. 20. Vasavid P, Chaiwatanarat T, Pusuwan P, Sritara C, Roysri K, Namwongprom S, et al. Normal solid gastric emptying values measured by scintigraphy using Asian-style meal: a multicenter study in healthy volunteers. J Neurogastroenterol Motil. 2014;20(3):371–8. 21. Kumar P, Ganure AL, Subudhi BB, Shukla S. Preparation and characterization of pH-sensitive methyl methacrylate-g-starch/ hydroxypropylated starch hydrogels: in vitro and in vivo study on release of esomeprazole magnesium. Drug Deliv Transl Res. 2015;5(3):243–56.