The combination of simplex lattice design and chemometrics in the formulation of green tea leaves as transdermal matrix patch

Aim: This study was aimed to formulate a transdermal matrix patch using green tea leaf extract. Materials and methods: The transdermal matrix patch formulation was optimized by the simplex lattice design method. The correlation between responses was analyzed using chemometrics. The observed responses were: 1. the physical properties of the matrix patch, and 2. the percentage of dissolution efficiency of catechins, caffeine, and epigallocatechin gallate released from the patch. The determination of drug release kinetics was based on the curve-fitting analysis using zero-order, first-order, Higuchi, and Korsmeyer-Peppas models. Results: The results showed that the optimal formula was obtained using the mixture of HPMC K100, HPMC K4M, and PEG 400 at a ratio of 4.0: 4.5: 0.5. The principal component analysis (PCA) showed that %DE300 values of catechin caffeine and epigallocatechin gallate positively correlate. A similar condition was observed between the weight and thickness of the matrix. Drug release kinetics follows the Korsmeyer-Peppas model.


Introduction
Tea is a widely used beverage in worldwide (Michele et al. 2014;Setyawan et al. 2018a, b;Shiyan et al. 2019). Green tea has been reported to have antioxidant, antimutagenic, anticancer, antibacterial, antiobesity, antihypertensive, and antidiabetic properties (Batchelder et al. 2004;Geetha et al. 2004;Hsu 2005;Michele et al. 2014;Shiyan et al. 2020). Green tea contains many polyphenol compounds, such as catechins, which have been known to inhibit the process of initiation, promotion, and cancer progression (Bouzari et al. 2009). However, the oral bioavailability of catechin is very low (< 5%) (Baba et al. 2001). It also has a short elimination half-life, related to a fast systemic clearance (Hu et al. 2015). Epigallocatechin gallate (EGCG) also shows similar characteristics. Transdermal delivery, a method to administer drugs via the skin, can avoid problems related to the low oral bioavailability of many com-pounds (Chen et al. 2011). Therefore, it is considered to be a promising alternative delivery route of catechin-related compounds (Lambert et al. 2006).
The transdermal delivery has been around for a long time, administered either in cream, ointment, and patch dosage forms. The transdermal delivery system has been designed to provide a controlled and continuous drug delivery through the skin into the systemic circulation in a non-invasive manner (Prabhakar et al. 2013;Kumar et al. 2015). The conventional drug dosage form provides a fluctuating plasma drug concentration, which might lead to toxicity, underdose, and low efficacy conditions. The objective of drug delivery is to improve the efficacy, ensure safety, and improve patient compliance. Transdermal delivery has many advantages. These include: 1) bypassing the first-pass metabolism, 2) providing a non-invasive delivery manner, 3) facilitating a comfortable and practical use, 4) allowing a long duration of therapy, 5) providing an easy dosing and easy stopping whenever required, and 6) improving patient compliance (Prausnitz and Langer 2009;Sachan and Bajpai 2013).
Several studies on the transdermal formulation of green tea extract have been reported (Batchelder et al. 2004;Perva-Uzunalić et al. 2006). However, none focused on the drug release kinetics. Furthermore, there were no studies combining the simplex lattice design and chemometrics for formula optimization. The optimization study using the simplex lattice design (SLD) allows profiling of the mixture's effect on the parameters. This method is useful for setting formulas, optimizing formula variables, determining the number of trials, and keeping the total formula concentration consistency. The experimental responses observed in this study were the patch matrix's physical properties and the percentage of dissolution efficiency of catechins, caffeine, and epigallocatechin gallate. The relationship of each response was analyzed using chemometrics of principal component analysis (PCA).
This analysis extensively applied the statistical and mathematical approach, mainly the multivariate methods. PCA is a relatively simple, nonparametric method for extracting relevant information from the dataset, identifying patterns in data, and expressing the data to highlight their similarities and differences (Singh et al. 2013;Moraes et al. 2016). In this study, PCA is used to analyze the correlation between responses such as physical properties of patch matrix (weight and thickness) and percentage of dissolution efficiency (%DE) of catechins, caffeine, and epigallocatechin gallate.
Based on the description above, this recent study was aimed: 1) to optimize the transdermal patch matrix formulation, 2) to estimate the effect of excipient (HPMC K4M, HPMC K100, and PEG 400) on the physical properties of patch matrix and %DE of catechins, caffeine, and epigallocatechin gallate, 3) to analyze the correlations of each experimental response, and 4) to determine a mathematical equation model of catechins, caffeine, and EGCG release.

Materials and Instrumentation
Dried green tea (Camellia sinensis L.) Kuntze was harvested from Mitra Kerinci Farm in West Sumatra, Indonesia. HPMC K100, HPMC K4M, and PEG 400 (all are of pharmaceutical grade) were obtained as gifts from Colorcon Indonesia. Epigallocatechin gallate (EGCG), catechin, and caffeine (all are of analytical grade) were purchased from Sigma-Aldrich, Singapore.

System suitability test (SST)
System suitability test (SST) was performed by injecting a mixture of standard solutions of catechin, caffeine, and EGCG with a concentration of 10 µg/mL. The parameters such as retention time, peak area, peak height, theoretical plate, tailing factor, resolution, and the high equivalent of the theoretical plate (HETP) were generated from each chromatogram. The selectivity was presented based on the resolution (Rs) values.

Linearity, the limit of detection (LoD), and limit of quantitation (LoQ)
A minimum of 5 concentrations series of the solution was prepared from the standard solutions of caffeine, catechins, and EGCG and then subjected to HPLC measurement. The responses were used to observe the value of the slope, intercept, and linear regression of the relationship curve between the content (x-axis) and the peak area of the chromatogram (y-axis). LoD and LoQ values were calculated based on the Signal to Noise ratio (S/N) of 3:1 and 10:1, respectively (Indrati et al. 2018;Shiyan et al. 2019).

Accuracy and precision
The accuracy was presented by the recovery percentage by analyzing three different series of analyte concentrations for six replications. Precision was determined as the standard deviation or relative standard deviation (RSD) values. In the critical method, it is generally accepted that the RSD must be less than the RSD determined by the Association of Official Analytical Chemists (AOAC).

Formula optimization
Formula optimization was performed by the simplex lattice design method (SLD) with the factors of HPMC K4M, HPMC K100, and PEG 400. This process was performed by using Design Expert ver. 7 software. Matrix weight, matrix thickness, dissolution efficiency values (%DE 300 ) of catechins, caffeine, and EGCG were used as the evaluated responses. The optimal formula was determined based on the criteria: 1) the smallest value of weight and thickness of the matrix; 2) the largest %DE 300 values of gallic acid, catechin, caffeine, and EGCG. Parameter of %DE 300 was calculated based on the area under the curve of the released compound from the matrix during 300 minutes. The determination of the released compound was performed by the RP-HPLC method.

Chemometrics
A chemometrics model is used to determine the correlation between observational responses. PCA models were implemented to evaluate the changes of the variable on each run. This process was performed by using Minitab software. PCA's output is a scree plot, score plot, loading plot, and a bi-plot graph describing the correlation and its effects on the principle components (PC).

Release kinetic study
Determination of the compound's release kinetics was carried out by the curve fitting analyses of the observation release curve based on the mathematical approaches, i.e., the zero-order, the first-order, Higuchi, and Korsmeyer-Peppas models (Costa and Lobo 2003).

Method validation
The chromatogram is presented in Figure 1. The system suitability test (SST) result showed good results, as indicated by the analytical parameters provided in Table 1. The coefficient of variation (CV) for all parameter was less than 2%, CV of Resolution (Rs) of < 2%, tailing factor (< 2), and theoretical plates (N > 2000) (Snyder et al. 1997;Indrati et al. 2018;Setyawan et al. 2020).
Linearity, the limit of detection (LoD), and limit of quantitation (LoQ).

Formula optimization
The transdermal patch matrix was prepared using a solvent casting technique. The excipient composition was processed by the simplex lattice design (SLD) method with 13 runs ( Table 2). The experimental results showed that matrix weight and thickness were 0.48-1.16 g and 0.10-0.33 mm, respectively. The %DE 300 of catechin, caffeine, and EGCG were 15.22-47.83%, 21.92-53.47%, and 1.88-8.71%, respectively. The results showed that the %DE 300 is all less than 90% over a period of 300 minutes. It can be caused by several things, such as the influence of dissolution media. The osmolarity of the dissolution medium contributed to auto-oxidation. Increasing the ionic strength at pH > 5 with the addition of sodium chloride will increase the rate of degradation of catechins and EGCG. EGCG is stable in acidic conditions (pH < 4). At a higher pH, it will trigger the EGCG epimerization process to become gallocatechin gallate (GCG). In addition, EGCG and catechin derivatives were able to interact with HPMC to form complex compounds that produce sediment or aggregates through hydrophobic interactions and hydrogen bonds. This complexation increased the stability of EGCG and catechins, but on the other hand, reduced the release of catechins and EGCG. ANOVA analysis showed that all treatments generated a significant difference (p < 0.05). A mathematical equation describing the relationship between these components of the experimental response (matrix weight and matrix thickness) has followed a linear model (equations 1 and 2).
Those equations demonstrate that all components play important roles in the increase in matrix weight and thickness. The PEG 400 (C) represents the most prominent role in increasing the weight of the matrix. PEG 400 might act as a plasticizer. Therefore its addition induced greater mobility of the polymer chains by replacing polymer-polymer interactions with polymer-plasticizer interactions. PEG 400 has non-volatile properties. The higher level of PEG 400 used in the matrix would make the matrix weight and thickness increase because PEG did not evaporate during the drying process. ANOVA analysis showed that all treatments produced significant differences in the %DE 300 values (p < 0.05). The special-cubic equation models appropriately describe the relationship between these components to the experimental response of %DE 300 values of catechin, caffeine, and EGCG (equations 3, 4, and 5).
The HPMC K4M (B) represents the most prominent role in increasing the level of %DE 300 of catechin, caffeine, and EGCG. The higher level of HPMC K4M used in the matrix would decrease the matrix viscosity and the diffusion layer to be quickly diffused out.
The optimum formula selection was determined by several criteria, namely the minimum weight and thickness of the matrix, the maximum of %DE 300 of catechin, caffeine, and EGCG. The numerical approach showed that the optimal formula was at a ratio of HPMC K100 (A), HPMC K4M (B), PEG 400 (C) at 4.0: 4.5: 0.5. The contour plot diagram of the optimum formula is presented in Figure 3. The optimal formula estimated matrix weight, matrix thickness, %DE 300 of catechin, %DE 300 of caffeine, and %DE 300 of EGCG at 0.67 g, 0.22 mm, 48.44, 49.68, 8.99%, respectively.

Chemometrics responses analysis
Principal component analysis (PCA) was carried out to determine the correlation of each response. PCA is an impor- , which are in linear combinations of the original variables describing each object (X 1 , X 2 ,……X n ). If there are five variables or responses, there will be five principal components (PC). There are five parts of PCA, namely data, score, loading, and residual. The score of PCs is also called as hidden or latent variable. Samples that have the same scores of PC can be understood as the same object. The PCA process generated five factors or PC. Based on the scree plot graph in Figure 4a, only four PC were able to produce the data variation of 99.20%. The score plot graph in Figure 4b shows that the formula is divided into several quadrants with different distances from one another. Some formulas show similar response characters based on the proximity value of the PC value. For example, formula 8, 9, and 10 had proximity values of PC 1 and PC 2 . It is illustrated that formula 8, 9, 10 had proximity of response character. The loading plot shows the strength of each variable affecting the PC. The angle between vectors describes how these variables correlate with one another. If two vectors form a narrow-angle, it indicates a positive correlation between the two variables, and if the vectors form an angle ≥ 90 o , then they are not correlated or negatively correlated. The loading plot graph in Figure 4c showed that %DE of catechin, caffeine, and epigallocatechin gallate have a positive correlation. A similar correlation was also present between the weight and thickness of the matrix. In contrast, the matrix patch's physical properties have a negative correlation to the % DE of the compounds. PCA biplot graph in Figure 4d is merged a usual score plot with a loading plot (Rohman and Putri 2019).

Optimum formula verification
This verification was performed to ensure that the results predicted by the model did not differ significantly from the results of the observations. The number of experiments was replicated three times, and the data results are presented in Table 3. Verification results showed that there was no significant difference between the model-based prediction and the observation of each response (p > 0.05).

Release kinetics determination
The release profiles in Figure 5a were concluded that the kinetics of catechin, caffeine, and EGCG followed the Korsmeyer-Peppas (equation 6).

M/Mt~ = Kt n (6)
Where M/Mt~ is a fraction of drug released at time t, k is the release rate constant, and n is the release exponent. The assessment is based on the value of the correlation coefficient (r) between the profile of the observed curve against the predicted curve. The value of r ≤ 1 indicates a good correlation in both analyses. Based on the data results in Table 5, the release rate (k) of catechin, caffeine, and EGCG were 0.10-0.33 mg/hour, 0.23-0.51 mg/ hour, and 0.01-0.08 mg/hour, respectively. The catechin, caffeine, and EGCG diffusion exponent value (n) were 0.29-1.11, 0.17-0.37, and 0.39-0.74, respectively. Based on the diffusion exponent value (n), the mechanism of catechin release (Fig. 5b) followed the Fickian diffusion (run 1, 2, 9, and 11), where the diffusion rate is less than relaxation. Some formulas follow non-Fickian diffusion (run 3, 4, 5, 6, 7, 8, 10, and 12) that the diffusion and the relaxation rate is balanced. Only run 13 followed the relaxation mechanism. Caffeine release (Fig. 5b) followed a Fickian diffusion, in which the rate of diffusion is lower than the relaxation. Five formulas followed the Fickian diffusion (run 4, 8, 11, 12, and 13), while the remaining eight formulas followed non-Fickian diffusion. EGCG release (Fig. 5b) followed Fickian diffusion, where the rate of the diffusion is slower than the relaxation. Four formulas followed the Fickian diffusion (run 4, 8, 11, and 12), and the remaining nine formulas followed non-Fickian diffusion. The difference in this mechanism is related to the polymer composition. It has different physical and chemical properties, resulting in the polymer's water penetration Figure 5. Dissolution profiles of caffeine, catechin, and EGCG from matrix patch transdermal for 300 minutes (a), Korsmeyer-Peppas fitting curve results (red lines) of catechin, caffeine, and EGCG drug release of the optimum formula (blue pattern) was performed by Solver (b).  leading to erosion. The dissolution medium's penetration caused the diffusion mechanism into the pores of the matrix produced by HPMC (hydrophilic polymer) and dissolving the compound. An increase in the volume of the medium caused the matrix expansion, followed by drug diffusion.

Conclusions
The results showed the optimal formula was obtained by a combination of HPMC K100: HPMC K4M: PEG 400 (4.0: 4.5: 0.5). The HPMC K4M represents the most prominent role in increasing the level of %DE 300 of catechin, caffeine, and EGCG. The optimal formula produced weight, matrix thickness, %DE of catechin, caffeine, and EGCG were 0.67 g, 0.22 mm, 48.44, 49.68, 8.99%, respectively, with a desirability value of 0.830. The PCA showed that %DE 300 catechin, %DE 300 caffeine, and %DE 300 epigallocatechin gallate have a positive correlation, as well as between the weight and thickness of the matrix. The drug release kinetics followed the Korsmeyer-Peppas model.