1. INTRODUCTION:

Boswellia serrata contain Boswellic acids. It is highly lipophilic drug. The bioavailability of conventional formulation is less bioavailable as conventional formulation. Boswellia serrata contains six types of Boswellic acids (BAs) [11-keto-β-boswellic acid (KBA), acetyl-11-keto-β-boswellic acid (AKBA), β-boswellic acid (βBA), acetyl-β-boswellic acid (AβBA), α-boswellic acid (αBA), and acetyl-α-boswellic acid (AαBA)] as potent active agents for treatment of such inflammatory disorder by inhibiting 5-lipoxygenase (5-LO).^[1]Topical delivery of BSE through a skin could be considered as an alternative to the oral route, in order to shorten onset time, and to sustain the effect for longer periods. In this context, nanosized colloidal carriers have been the subject of widespread interest as having high solubilization capacity for lipophilic drugs.^[2]

More recently, there has been increasing focus on the utility of nanoemulsion as lipid-based formulations composed of isotropic mixtures of natural or synthetic oils with lipophilic or hydrophilic surfactants and cosurfactant/ cosolvent to form o/w nanoemulsion.^[3] Microemulsions/ Nanoemulsion have been recently proposed as carriers in pharmaceutics to achieve the sustained/controlled release of several drugs^[4]. Usually such formulations, are developed on trial and error approach by selecting component ratio from the NE region of pseudo-ternary phase diagram. By this conventional approach, it is possible to develop the formulation with specific characteristics; however, it is difficult to get the true optimum percentual composition. Chemometric methodologies currently provide effective tools for the study of pharmaceutical dosage forms. Nanoemulsion can be developed only for some set of the components ^[5]. Simplex lattice design initially was used to investigate the optimum ratio of components in order to determine existence of nanoemulsion. Then Partial least square regression analysis was applied to emphasize the role of nanoemulsion components. Successively, mixture design was used to select a set of microemulsions from which the permeation of BAs through a hydrophilic dialysis membrane was measured. The droplet size and rate of diffusion of the drug was then modelled as function of the mixture's percentual composition by PLS.

Nanoemulsions as the topical carrier offer significant advantages including low skin irritation, greater permeation ability due to nano droplet size and high drug-loading capacity^[6].The aim of the present work was to apply chemometric methodologies and multivariate analysis to develop these complex mixture systems that have not been investigated using such procedures. Nanoemulsion consisting of three biocompatible substances (Isopropyl Myristate, Tween 80, Transcutol P) were studied in order to assess their suitability as carriers for the topical application of BAs for inflammatory studies.

2. MATERIALS AND METHODS:

2.1. Material: The dry extract of Boswellia serrata (B. Serratta) was gifted by Pharmanza Herbal Pvt. Ltd., Gujarat, India. Isopropyl myristate, (IPM) was procured from SD Fine chemicals (Mumbai, India), Tween 80 was purchased from Sigma Aldrich (St. Louis, MO). Transcutol P provided as gratis sample from Gattefosse Corporation, Mumbai, India. HPLC grade methanol, and phosphoric acid, Acetonitrile were purchased from SD Fine chemicals (Mumbai, India). All other chemicals, buffer solution components and solvents were of analytical grade. Water was obtained from Milli Q water purification system (Millipore, MA).

2.2. Equilibrium solubility study:

The equilibrium solubility study was performed by adding an excess amount of B. serrata extract in 2ml of oils - isopropyl myristate, surfactants -Tween 80, and cosurfactant Transcutol P in 5-mL capacity stoppered vials separately vortexed using a Cyclo mixer [CM 101, REMI (INDIA)]. The vials were then kept at 25±0.5°C in an orbital shaker (CSI-24 BL, Remi Laboratories, and Ahmadabad, India) for 72 h to reach equilibrium. Following attainment of equilibrium, the supersaturated samples were centrifuged at 2,000 rpm for 15 min to separate undissolved amount. The obtained supernatants were quantified followed by filtered through a 0.45-μm membrane filter (Membrane Technologies, (Mumbai, India) using a HPLC (Shimadzu, Tokyo, Japan).

High performance liquid chromatography analysis:

Quantitative determination of six different types of Boswellic acids was performed by a validated HPLC method. A Shimadzu-model HPLC equipped with quaternary LC-10A VP pump, variable wavelength programmable UV/ VIS detector, SPD-10AVP column oven (Shimadzu), SCL 10AVP system controller (Shimadzu), Rheodyne injector fitted with a 20-μl loop was used and the data were recorded and evaluated using Class-VP 5.032 software. Chromatographic separation was achieved on a reversed-phase C-18 column, LiChrospher®100 (5 μm, 250×4.6 mm inner diameter) using a mobile phase A 200ml water + 0.01 % Phosphoric acid + 800ml acetonitrile and mobile B 100% Acetonitrile at a flow rate of 1 ml/min with UV detection at 210nm and 250 nm. The mobile phase was filtered through 0.22-μm nylon filter prior to use. The peak area correlated linearly with Boswellic acids concentration in the range of 5–50 µg/ml with the lowest detection limit at 0.5µg/ml, and the average correlation coefficient was 0.9976.

2.3. Construction of phase diagrams and formulation of BAs-loaded NE:

2.3.1. Construction of pseudo ternary phase diagrams:

10 ml of Oil: Smix mixture of water added drop by drop. The sample was checked by visual observation as well as % transmittance. If the sample was an isotropic and clear solution, or was cloudy or showed the phase separation, it was not a NE. the boundary point between NE and non-NE was determined and corresponding ratio was recorded during titration. The NE system pseudo-ternary phase diagram was constructed by labelling the recorded boundary points in a ternary plot depended on Prosim Software.

2.3.2. Formulations of Nanoemulsion:

Series of nanoemulsion mixtures were selected based on the design matrix from the Simplex lattice mixture design and formulation were prepared as follows: BSE was dissolved in IPM and then Smix containing Tween 80: Transcutol P was added, resulting mixture was homogenized with high speed homogenizer [IKA Pvt. Ltd, Germany] at 6000RPM for 10-15 min to solubilize the drug. The drop wise addition of water was done to resulting mixture till nanoemulsion was formed. The systems obtained were observed upto 48 hrsfor any signs of instability according to the transparency criterion classified into micro/nanoemulsion or thermodynamically unstable system respectively ^[7].

Table: 1. Actual and transformed values as per simplex lattice design.

Mixtures	Actual Formulation Components/ transformed components value
Mixtures	A	B	C
NE-1	8/1	8/0	5/0
NE-2	2/0	36/1	5/0
NE-3	2/0	8/0	20/1
NE-4	4/0.5	18/0.5	5/0
NE-5	4/0.5	8/0	10/0.5
NE-6	2/0	18/0.5	10/0.5
NE-7	3/0.33	12/0.33	0.33/7

Where, A = % oil, B = % Smix, C = % water as independent variables.

2.4. Evaluation of nanoemulsion mixtures:

2.4.1. Droplet size analysis:

The mean droplet size and zeta potential of the various nanoemulsions was determined by photon correlation spectroscopy using a Zetasizer 1000 HS (Malvern Instruments, Worcestershire, UK). Sample was extemporaneously diluted in Milli-Q water (Millipore Corp., USA) and injected in the apparatus. Each sample was analyzed twice, each analysis consisted of five replicates.

2.4.2.In-vitro drug diffusion study:

In vitro diffusion studies of nanoemulsions were performed using a modified Franz diffusion cell attached with thermostatic water bath at 25°C±1°C. A dialysis membrane (MWCO: 12,000 Da, Himedia) with 1.98 cm²effective surface area of membrane viz. a pore size of 0.45µm was used. Each one gm of drug loaded nanoemulsion was placed in the donor compartment. The receiver compartment was filled with phosphate buffer (pH 7.4) as dialysis medium at 25°C±1°C and 400rpm on magnetic stirrer. Aliquots (25µl) were periodically withdrawn, at the same time sample was simultaneously refilled with fresh acceptor phase fresh acceptor phase, from the receiver compartment through a side tube and analyzed using HPLC.

2.5. Experimental Design:

A Simplex lattice mixture design was used for the search of the microemulsion area of existence. Mixture design was used for planning the mixtures to be tested for building reliable regression models. From Cornell reports, a mixture experiment is defined as an experiment where the response is assumed to depend only on the relative proportions of the components present in the mixture and not on the amount of the mixture itself ^[8]. This is what was expected for the nanoemulsions under investigation. In this perspective, in order to be able to calculate a model non-linear in the composition variables, we chose the Simplex lattice design, augmented with the introduction of three interior experiments. Such a design can be used to calculate up to a complete third degree model and provides additional information for validating the model itself. Moreover, this design is characterized by the need and proper distribution of the information throughout the experimental region and is especially suitable for detecting the design space in the interior of the triangular region.

2.5.1. Application of Multivariate data techniques for optimization of BAs loaded NE:

Application of Multivariate data techniques, called design of experiments (DOE) and partial least square regression (PLSR) involve the concept of ‘mixture designs’ for changing mixture composition and explaining how such changes will affect the properties of the mixture ^[9]. Percentual composition of nanoemulsion mixtures were optimized using Simplex lattice mixture design represented as three-components by an equilateral triangle in two-dimensional space. Seven formulations were selected from each vertex (A, B and C), at the halfway point between vertices (AB, BC and AC), and last one at the center point (ABC) ^[10]. The percentual composition of oil, Smix and water were selected as independent variables. The mean droplet size and cumulative permeation of drug from nanoemulsions were chosen responses. For the formulation design, selected responses for seven formulations were used to fit polynomial equations which can help to predict the properties of all possible formulations within the design space. Polynomial equations were generated and model graphs in the form of contour plots were constructed through Design Expert 7.0.1 Software (Stat- Ease, Inc., Minneapolis, MN).Comparison of the observed and predicted responses were critically made and percentage prediction error was also calculated with respect to the observed responses. Mean droplet size and permeation rate of BAs from developed nanoemulsions were measured and the results of observed and predicted responses are shown in Table 2.

Equilateral triangle representing simplex lattice design for three components.

Further optimization was performed using the desirability function ^[11]. In brief, for a response to be minimized, the desirability function was defined as:

Eq. (1)

Table 2.Mean droplet size and cumulative permeation of seven different formulations as per simplex lattice design.

Mixtures	Response: Y₁ Droplet size[nm]		Response: Y₂ Qn permeation [µg/cm²]		% Prediction error	% Prediction error
Mixtures	Experimental result	Predicted result	Experimental result	Predicted result	R₁	R₂
NE-1	23.38±1.98	22.01±2.83	20.95 ±2.54	22.01±2.83	5.85	5.05
NE-2	21.8 ±2.35	22.51± 2.32	20.44 ±1.28	22.51± 2.32	3.25	10.01
NE-3	29.31±3.45	21.59± 3.02	19.48 ±1.54	21.59± 3.02	0.26	10.83
NE-4	12.89±1.25	23.54± 3.80	19.17 ±2.45	23.54± 3.80	8.26	2.27
NE-5	18.48±2.84	22.91± 2.43	23.24±4.01	22.91± 2.43	2.39	1.41
NE-6	15.98±3.21	23.32± 2.74	25.15±1.47	23.32± 2.74	4.59	7.27
NE-7	25.33±2.24	21.99± 2.50	20.48±2.78	21.99± 2.50	13.1	7.37

Data of Mean droplet size (Y1) and Cumulative permeation (Y2) were shown as mean± SEM (n= 3).

Table 3. Mixtures of the experimental design and corresponding observed responses as per PLS regression.

A	B	C	AB	BC	AC	A²	B²	C²	Kp (10-2)	Mean Droplet size
8	36	5	288	180	40	64	1296	25	3.13	23.38
2	36	5	72	180	10	4	1296	25	7.06	21.8
2	8	20	16	160	40	4	64	400	6.72	29.31
4	18	5	72	90	20	16	324	25	4.74	12.89
4	8	10	32	80	40	16	64	100	3.58	18.48
2	18	10	36	180	20	4	324	100	3.5	15.98
3	12	7	36	84	21	9	144	49	5.02	25.33

where, A = % oil, B = % Smix, C = % water as independent variables, and kp = permeability coefficient and mean droplet size were dependent variables for PLSR modeling

For a response to be maximized, the desirability function was defined as:

The limits were selected for Y1: Ymax = 29.31nm (largest tolerable droplet size) and Ymin = 12.89nm (lowest tolerable droplet size); Y2: Ymax = 25.15 µg/cm²(Desired cumulative permeation) and Ymin = 19.17 µg/cm²(lowest tolerable permeation). Then the overall desirability (D) was calculated as follows:

Eq. (3)

Graphs of these properties in the form of contour plot were constructed using Design Expert 7.0.1 Software (Stat- Ease, Inc., Minneapolis, MN) the same as the design. The responses for seven formulations were fitted to a special cubic polynomial model.

2.5.2. Partial least square regression analysis:

PLS regression analyses a set of components that performs a simultaneous decomposition of variable (X) and response (Y) with the constraint explaining maximum possible covariance between them. The partial least squares regression approach was selected ^[12-14]. New set of orthogonal variables are obtained as linear combinations of the original ones. These variables are calculated in such a way to avoid collinearity present in the independent block and correlating them with the choosen responses ^[15]. The three components oil, Smix and water were selected as independent variables and permeability coefficient and droplet size were choosen as dependent variables. The correlation matrix with independent and dependent variable with its labels is shown in Table 3. The selection of the best set of independent variables was performed using a step-wise algorithm ^[16]. A model that shows lowest predicted sum of squares (PRESS) is used for prediction purposes. The PRESS was evaluated through a leave-one-out cross validation procedure ^[5], an option available in most of the softwares that can handle chemometric analysis. EXCEL stat software was used to carry out PLSR in this investigation.

3. RESULTS AND DISCUSSION:

3.2. Construction of Pseudo ternary phase diagram:

Pseudo ternary phase diagrams were constructed in the absence of BSE to identify nanoemulsion region depicts an important tool to study the phase behavior of nanoemulsion. It can be represented in a triangular format (triangle) which has three coordinates. (1) Oil phase, (2) Surfactant/co-surfactant phase and (3) aqueous phase at fixed weight ratios (Smix ratios 1:1, 2:1, 3:1, 4:1, 1:2, 1:3).The different blend of surfactant and cosurfactants (Km) were chosen in increasing concentration of surfactant with respect to co-surfactant and increasing concentration of co-surfactant with respect to surfactant for detailed study of the phase diagrams [2]. The concentrations of components were recorded in order to complete the pseudoternary phase diagrams, and then the contents of oil, Smix and water at appropriate weight ratios were selected based on these results.

An o/w nanoemulsion region was found towards the water-rich apex of the phase diagram. As the surfactant concentration was increased in the Smix ratio 1:1 (Fig. 1(a)), a higher nanoemulsion region was observed. The probable reasons are reduction of the interfacial tension, increased the fluidity of the interface. Kawakami K and coworkers reported that greater penetration of the oil phase in the hydrophobic region of the surfactant monomers ^[17]. From the phase diagrams in fig. 1 (e) and (f), formulation with Km 1:2, 1:3 showed very narrow area of nanoemulsion formation compared with others. So, further study in direction of increasing concentration of cosurfactant was not carried out. Phase diagram in Fig. 1 (a) and (c) with Km 1:1 and 3:1 has largest area of nanoemulsion formation followed by Km 4:1 (Fig. 1 (d)). Beyond this point, there is decrease in the area of NE formation. Various blends were selected from the nanoemulsion region to obtain B. serrata loaded nanoemulsions. From pseudo ternary phase diagrams, the formulations in which the amount of oil phase completely solubilized the drug and which could accommodate the optimum quantity of Smix and distilled water were selected for the optimization study.

3.3. Demonstration of Multivariate analysis techniques:

Simplex lattice mixture design was used to optimize the composition of nanoemulsions. The mixture components and response variables were related using polynomial equation with statistical analysis though Design-Expert® software. The approximations of response values (Y₁, Y₂) based on the special cubic model was most suitable with lowest PRESS value. The value of the coefficients exhibit the effect of these variables on the response.

Figure: 2shows contour plots for R1-Droplet size, R2-percutaneous absorption and overlay plot of superimposing the contour plots of the two responses.

The polynomial equations comprise the coefficients for intercept, first-order main effects, interaction term, a positive sign of coefficient indicate a synergistic effect while negative term indicates an antagonistic effect upon the response. After generating the polynomial equations through MLRA (Multiple linear regression analysis) relating the dependent and independent variables, mixture components were optimized for the responses Y₁ and Y₂. When the ratio of Smix was close to 1:1 (w/w), droplet size of nanoemulsion was lowest, also the permeation of Boswellic acids from nanoemulsion was high, because the small droplet size of nanoemulsion facilitates penetration of the drug for topical absorption. Topical absorption is influenced by nano droplet sized in vehicle and the partition coefficient. The solubility of drug in vehicle is also increased because of IPM and Transcutol P are present in nanoemulsion which contributes as penetration enhancer.

The result of polynomial equation for selected responces were obtained through Design expert software as shown in Eq. 4 and Eq. 5.

Droplet Size Y₁ = +34.66 *X₁ +16.95 *X₂ +11.81 *X₃ +0.61* X₁ X₂ +15.63* X₁X₃+ 68.35 *X₂ X₃- 400.6 *X₁X₂X₃ . (r2= 0.998) …… …………… Eq. 4

Cumulative Permeation [µg/cm²] Y₂ = +20.19* X₁+23.35* X₂ +23.60*X₃-2.55 *X₁X₂ -5.15 *X₁X₃ -10.81 *X₂X₃+28.2 *X₁X₂X_3.(r2=0.998)……………… Eq. 5

Eqs.(4) and (5) may be used to calculate the predicted values for other formulations in the design space. Table 2containsthe predicted values of droplet size and cumulative permeation with % prediction error of seven formulation with respect to observed responces. The equation has good prediction ability.

Regression coefficients for measured responses were shown in Table 4. Adequate precision value was greater (should be >4) indicated special cubic model be use for further navigation. Model graphs of the droplet size and the cumulative amount of Boswellic acids permeated were constructed in the form of contour plots (Fig. 2-A.1,2-A.2) and the optimized formulation was chosen by superimposing the contour plots of the two responses revealed design space, shown in overlay plot.(Fig. 2-A.3).

3.4. Optimization by desirability function

The droplet size (Y1) and cumulative permeation (Y2) responses were transformed into the desirability scales d1, d2, respectively. Y1 had to be minimized, while Y2 had to be maximized. The overall objective function (D) was calculated by Eq. (3) and the result is shown in Table 5.

The model was also fitted with a special cubic polynomial model and optimized by the “total subset” variable selection method after calculating with step width of 0.1, the maximum desirability function (1.00) value was obtained at X1: 0, X2: 0.45, X3: 0.55. To confirm the model adequacy for prediction, one batch of formulation under the optimum composition were prepared. As the results shown in Table III demonstrate, the model was validated due to close agreement between the predicted and experimental results, and the appropriate amount of components chosen for the optimal formulation of BSE-loaded NE were : IPM : 2%, Tween 80:Transcutol P (Smix) : 18% and Water : 10%.

3.5. Partial least square regression analysis:

Randall D. Tobias reported that the X- independent variables and Y-dependent variables are chosen so that the relationship between successive pairs of scores is as strong as possible. In principle, this is like a robust form of redundancy analysis, seeking directions in the factor space that are associated with high variation in the responses but biasing them toward directions that are accurately predicted.

The expansion of the Simplex lattice mixture design was constrained according to pharmaceutical requirements, as result of nanoemulsion are expressed by particular ranges of the components as pseudo-ternary phase diagram.

Figure 3. Standardized coefficients of variable Kp (Permeability coefficients) and droplet size.

Figure: 4 shows Plot of predicted vs experimental Kp (Fig. 4-a) and droplet size (Fig 4-b) , where Fig. 4-c shows Biplot chart obtained from final PLS model

Table: 6Experimental, predicted and calculated permeability coefficients and droplet size by means of the best PLS model

Mixtures	Kp exp.	Kp pred.	Kp calc.	DLS exp.	DLS pred.	DLS calc.
NE-1	3.13	4.29	4.68	23.38	21.03	19.93
NE-2	7.06	6.58	5.49	21.8	22.88	20.36
NE-3	6.72	5.59	5.71	29.31	26.48	24.58
NE-4	4.74	3.88	4.04	12.89	18.45	19.54
NE-5	3.58	2.66	4.95	18.48	19.63	20.31
NE-6	3.5	3.80	3.71	15.98	13.31	13.03
NE-7	5.02	5.42	5.70	25.33	21.28	22.28

The equations obtained through partial least square regression used to calculate selected responses as shown in Eq. 6 and Eq. 7 obtained through XL State software.

Permeability coefficient: Kp (10^-2)= 4.30-0.10*A 1.09E-03*B+4.72E-02*C -1.55E-03*AB+3.78E-03*AC-1.95E-03*BC-9.08E-03*A²+4.64E 05*B²+2.22E-03*C2

(RMSE 1.25)………………………………………Eq. 6

Droplet size (DLS) = 3.42-5.79E-02*A + 5.37E 02*B + 0.34*C + 5.73E 03*AB + 4.16E 02*AC+0.10*BC+1.27E-02*A²+ 2.08E-03*B²+1.70E-02*C²

(RMSE 2.12)………………………………. Eq. 7

RMSE: Residual mean sum of square error)

A limited area shows the combinations of the variables that will result into formation of microemulsion.

The classical multiple regression analysis is unsuitable to evolve a model which incorporates interaction and polynomial terms since the number of trials are few in the nanoemulsion region. PLSR can be used to handle this issue. The PLS models often used for prediction purpose rather than interpretation. The independent and the dependent variables with original data matrix was expanded to incorporate interaction and polynomial terms before performing PLSR.

The graph of the experimental and predicted responses for both dependent variables are shown in (Fig. 4-A.1, 4-A.2). Biplot chart generated through modelling of data illustrated in fig. 6 indicates partial least square analysis. Fairly linear relationship was observed between calculated and predicted and observed responses. Residual mean sum of square value (RMSEV) from PLSR was nearer to zero (8.76) as compared to simplex lattice model (34.69).and hence it is concluded that PLSR proved to be an effective tool to predict responses. The use of PLS and its active cross-validation makes the results more stable and reliable.

The results of the calculation using first derivative variables which explained 73.8% of the variance in prediction with one latent variable, contained two variables, B and B², whose OLS type coefficients are reported in Eq.6 The predictive ability of the model is shown as a plot of predicted against experimental permeabilities in Fig. 4, while the numeric data are reported in Table 6.

At different equilibria involving BAs and nanoemulsion components can affect the permeation process: transport of the free drug, diffusion of drug aggregates, interaction of the drug with the components of the nanoemulsion interphase, micellization of the drug present in the continuous phase and, primarily, drug partition in the internal phase of nanoemulsion. In the present work, the resulting model, obtained with the first set of variables, indicates that the oil and cosurfactant plays a very important role in the permeation of the drug through the dialysis membrane.

CONCLUSION:

The optimized formulation revealed higher formulation rate and extent of drug permeation. Also higher concentration of drug in nanoemulsion vehicle resulted in high concentration gradient, which might be the main permeation mechanism of drugs into the membrane. These findings confirm the reservoir effect of the internal phase of components of the nanoemulsion, and demonstrate the influence of micellization of the drug on its releasing behaviour. Both calculated models showed good predictive abilities of the BAs permeability from nanoemulsions through a hydrophylic membrane. Therefore, they could be used to select proper mixture compositions to achieve a desired drug release performance. The use of chemometric tools is also favored by FDA in this era of quality by design and process analytical technology.

ACKNOWLEDGEMENT:

The PhD research work was supported by Research grant (GUJCOST/MRP/362) from GUJCOST, Gandhinagar, Gujarat, India. Author like to acknowledge Gattefosse, and Abitec Corporation, India for providing the gift samples of Lipid excipients. The authors are thankful to Dr. Atindra Shukla - Shah Schulman Centre for Surface science and Nanotechnology, Dharmsinh Desai University, Nadiad, Gujarat, India for providing research facilities. The authors are thankful to thanks to Vrunda Suthar, and Mukesh Sharma for guiding during research work.

Declaration of Interest:

There is no conflict of interest

REFERENCES:

1. Gerbeth K, Meins J, Kirste S, Momm F, Schubert-Zsilavecz M, Abdel-Tawab M., Determination of major boswellic acids in plasma by high-pressure liquid chromatography/mass spectrometry. J Pharm Biomed Anal, 2011; 56: 998–1005. DOI:10.1016/J.FITOTE.2012.10.009

2. Adnan Azeem, Mohammad Rizwan, Roop K. Khar, and Sushama Talegaonkar, Nanoemulsion Components Screening and Selection: a Technical Note AAPS PharmSciTech, 2009;10:69-76. DOI:10.1208/S12249-008-9178-X

3. A.M.S. Villar, B.C. Naveros, A.C.C. Campmany, M.A. Trenchs, M. Aróztegui, B. Rocabert, L.H. Bellowa, Design and optimization of self nanoemulsifying drug delivery systems (SNEDDS) for enhanced dissolution of gemfibrozil, Int. J.Pharm. 2012;431:161–175.

4. Miiller, B.W. and Kleinebudde, P., Untersuchungen an sogenannten Mikroemulsionssystemen. Pharm. Ind., 1988;50: 1301-1306.

5. Gould, P.L., Optimization methods for the development of dosage forms. Int. J. Pharm. Tech. Prod. Mfr., 1984;5:19-24.

6. Sonneville-Aubrun, O., Simonnet, J.-T., L’Alloret, F., Nanoemulsions: a new vehicle for skincare products. Adv. Colloid Interface Sci. 2004;108-109:145–149

7. Lagues, M., Ober, R. and Taupin, C., Study of structure and electrical conductivity in microemulsions: evidence for percolation mechanism and phase inversion. J. Phys. Lett., 1978;39: 487-490.

8. Cornell, J.A., Experiments with mixtures, 2nd Edn., Wiley, New York, 1991, p. 13.

9. Eriksson, L., Johansson, E., Wikstrom, C., Mixture design: design generation, PLS analysis, and model usage. Chemomet. Intell. Lab. Syst. 1998;43: 1–24.

10. Weiwei Zhu, Aihua Yu, Weihong Wang, Ruiqian Dong, Formulation design of microemulsion for dermal delivery of penciclovir. Int J Pharm, 2008;360:184-190.

11. Liu, Y., Zhang, P., Feng, N., Zhang, X., Wu, S., Zhao, J. Optimization and in situ-intestinal absorption of self-microemulsifying drug delivery system of oridonin. Int. J. Pharm. 2009;365: 136–142.

12. Wold, H., Soft modeling: The basic design and some extensions. In Joereskog, K.G. and Wold, H. (Eds), Systems under Indirect Observation: Causility-Structure-Prediction. II. North-Holland, Amsterdam, 1982, 1-54.

13. Geladi, P. and Kowalski, B.R., Partial least-squares regression: a tutorial. Anal, Chim. Aeta 1986;185:1-17.

14. Manne, R., Analysis of two partial least-squares algorithms for multivariate calibration, Chemometrics lntell. Lab. Systems, 1987;2:187-197.

15. Pattarino F, Marengo E. Gasco M.R. Carpignano R., Experimental design and partial least squares in the study of complex mixtures: microemulsions as drug carriers. Int J Pharm. 1993;91: 157-165

16. Marengo, E. and Todeschini, R., A fast method for the calculation of partial least-squares coefficients. Chemometrics Intel. Lab. Systems, 1991;12:117-120.

17. Kawakami K, Yoshikawa T, Moroto Y, Kanaoka E, Takahashi K, Nishihara Y., Microemulsion formulation for enhanced absorption of poorly soluble drugs I. Prescription design J Control Release, 2002;81: 65-74.

Received on 21.10.2014 Modified on 30.10.2014

Research J. Pharm. and Tech. 7(12): Dec. 2014; Page 1446-1455

Coefficients	Y1	Y2	Coefficients	Y1	Y2
Std. Dev.	4.99	4.20	R-squared	0.9223	0.9163
Mean	22.09	21.68	Adjusted R-squared	0.9114	0.9187
C.V. %	22.60	19.38	Predicted R squared	0.9012	0. 9001
PRESS	28862.80	20434.79	Adequate Precision	4.848	6.032

Mixture components	Predicted Value from Overlay graph by Design expert	Experi-mental Value	% Bias
Oil X1	0	2%	----
Smix X2	0.44	18%	----
Water X3	0.55	10%	----
Droplet size (nm)	30.99 ± 3.99	23.38 ± 1.19	32.5
Cumulative Permeation (µg/cm2)	20.81 ± 2.74	23.98 ± 3.52	-13.21