A statistical study on the development of metronidazole-alginate-chitosan nanocomposite formulations using Full factorial designs

The purpose of this study was to investigate the effect of chitosan (CS) and Alginate (Alg) polymers concentrations and CaCl2 concentration on metronidazole (MET) drug loading (LE), size particles and zeta potential. Nanocomposites were prepared by ionotropic pregelation method. A (2 *3 *2) *3= 36 full factorial design (FFD) was used to predict statistical equation and responses. The MET-CS-AlgNPs nanocomposites were characterized by X-ray diffraction, Fourier-transform infrared spectroscopy, thermal gravimetric analysis, scanning electron microscope and in vitro drug release studies. All data indicated the presence of drug into MET-CS-AlgNPs nanocomposites. The release profile of MET-CS-AlgNPs nanocomposites was found to be sustained


Introduction
In pharmacy, the term of optimization can be defined as the formulation and the process to obtain the best combination of product. In optimization steps, one generally experiments by a chain of reasonable steps, changing the variables at a time until an acceptable system is produced. Different types of optimization techniques were used in the literature; like Experimental Design (ED), Formulation by Design (FBD) and Quality by Design (QBD) [1][2][3].
Previously, every novel pharmaceutical formulation (PF) was designed by studying the influence of variables on the characteristics of dosage form, through changing one single variable at a time while keeping other variables constant (PF). The disadvantage of this technique does not guarantee to give the true optimum process and the product obtained may be suboptimal.
Experimental Design (ED) is a technique that is used specifically to avoid the disadvantage of PF method by examining various problems that are encountered during research, development and production (ED). It is obvious that if experiments are performed randomly, the results obtained will also be random. Therefore, it is essential to plan the experiments in method so that the relevant information will be obtained. Experimental designs (ED) is a method where the independent variables are initially screened to establish and they are more effective to the product [4][5][6]. The second step is the 'optimization'; in this step, a number of decisions are to be made, what plan variables will be changed to create an optimal design, and what requirements should be met. This step is the most important step in optimization design, and designers may spend much time on modeling during the optimization process [7]. The third step is called mixture designs; it contains changing mixture content and exploring how these changes will effect on the mixture. ED is a statistical method that describes a set of combinations of variables. The experimental design is chosen depending on several factors; such as number of factors, their levels, possible interactions and the order of the model.
The same procedure was used to form MET-CS-AlgNPs nanocomposites using only 100 mg MET mixed with Alg solution.

Methodology
First, we presented the modeling of the responses (loading efficiency, particle size and zeta potentials). Secondly, we build up the FFD to perform the experiments. Then, we used the multiple regressions to develop the loading efficiency, particle size and zeta potentials models responses. Finally, the analysis of concentration variance was used to analyze the experimental data to predict the effects and contribution of parameters on responses.

Modeling of different responses
The loading efficiency percent is the first response which was taken as a parameter and defined as the amount of total entrapped drug divided by the total nanoparticles weight. The second and third responses taken in this work was particle size and zeta potentials. Table 1 shows three independent parameters and their levels

Multiple Regression method
Multiple Regressions is a statistical method that is used to estimate the correlation between dependent and independent variables. The term of correlation coefficient (R²) indicates how well data fit the multiple regression model. It provides a measure of how well observed outcomes are replicated by the model, as the proportion of total variation of outcomes explained by the model. An R² close to 1 indicates that the regression model perfectly fits the data.

The metronidazole loading efficiency
The high speed centrifugation instrument was used in this work to determine the where Tp is the total MET used to prepare the nanocomposites, and Tf is the free MET in the supernatant.

Particle size, and zeta potential of nanocomposites
Particle size and zeta potential of nanocomposites were analyzed through dynamic light scattering (DLS) with Zetasizer Nano S (Malvern, UK) at The Arab Pharmaceutical Manufacturing. The analysis was performed in triplicate at a temperature of 25 o C.

Controlled release study of the MET from the nanocomposites
In-vitro release of MET from nanocomposites is determined by primary method in HCl with pH 1.2, using a Perkin Elmer UV-Vis spectrophotometer with λmax of 323 nm. A suitable amount of each nanocomposite was added to 2 mL of the media. The cumulative amount of drug released into the solution was measured at preset time intervals at corresponding λmax.
The percentage release of the MET into the release media was obtained by: The concentration which corresponding to 100% release was obtained by adding a known amount of nanocomposites in 2 mL HCl in addition to using sonicate and heat the nanocomposites at 37 o C.

Instrumentation
Powder X-ray diffraction (XRD) patterns were used to determine the crystal structure of the samples in the range of 2-70 degrees on an XRD-6000 diffractometer

X-ray diffraction (XRD)for MET-CS-AlgNPs nanocomposites
From the literature, the XRD diffract gram of CS have crystalline properties with intense peak at 2θ equals to 19.7º.At the same time, the XRD diffract gram of Alg have a semi-crystalline properties with a peak at 2θ equals to 13.6º [36].  14. 5 39.1 C

FTIR spectroscopic analysis of CS-AlgNPs and MET-CS-AlgNPs
FTIR spectra of MET, CS-AlgNPs and MET-CS-AlgNPs are presented in Figure 2.
For CS-AlgNPs, a band at 3296 cm -1 was observed due to O-H and N-H stretching.
Absorptions due to vibration asymmetry CH2 and symmetry CH2 were located at 2930 cm -1 and 2850 cm -1 , respectively. The strong band near 1589 cm -1 corresponded to vibration C=O, vibration C-N and bending N-H (amide I). A symmetric stretching band of the COOgroup was centered near 1420 cm -1 [37].
For MET-CS-AlgNPs, some bands were downshifting, for example 3296 cm -1 to 3283 cm -1 , from 1589 cm -1 to 1585 cm -1 and from 1408 cm -1 to 1413 cm -1 ; this can be explained by the interaction between MET and CS-AlgNPs. between 137 and 288 o C and had a mineral residue of 0.9%. [38], which is due to the vaporization of volatile components [39].
The CS-AlgNPs show three main thermal stages, the first stage of the decomposition process occurred between 60 and 200 o C, which is due to vaporization of volatile components, such as water molecule immobilize between chitosan chains during the coating [40]. Focusing on the structure of CS and Alg, it can be seen that H2O molecules can be bounded by hydroxyl group [41]. The second responsible weight loss occurred between 200 and 520 o C due to the release of water bounded to the functional groups of CS and Alg polymers, which was not completely removed in the first step of dehydration, also degradation of both polymers.

Figure 4 Scanning electron microscopy (SEM) micrographs (100000× magnification) of CS-AlgNPs (A) and MET-CS-AlgNPs (B)
The CS-AlgNPs and metronidazole loaded CS-AlgNPs were morphologically characterized using SEM as shown in Figure 4. The micrographs of CS-AlgNPs at Figure 4A show that the nanoparticle has a smooth surface and spherical shape as found in the literature [42]. Figure 4B shows that metronidazole loaded CS-AlgNPs nanocomposites had also spherical shape.

Multiple linear regression analysis using full quadratic
The linear (CS, Alg, and CaCl2), linear-square (CS*CS, Alg*Alg, and CaCl2*CaCl2), linear-interaction equations (CS*Alg, CS*CaCl2and Alg*CaCl2) have been fitted using Minitab software for LE, size and zeta potential variables. The equations can be given in terms of the coded values of the independent variables as the following (Table 3):   The Figure 5B represents the effect of different parameters toward zeta size. The results indicate that all the effect is statistically significant. The effect of A (Alg) has the highest standardized effect on the zeta size followed by B, C, BC, BB, AB and AC. Hence, the term BB should not be considered for the empirical relation. All the significance of factors can be shown at the normal plot ( Figure 6B).

Normal probability plot
Normal probability is used commonly in the drug formulation. For example, Ghadiri, M [43], and Javaid [44] recommend use of these plots for estimating the goodness of fit of a hypothesized distribution. If the sample is actually distributed as hypothesized, one would expect the plot of the ordered observations at Y axis versus the order statistic means at X axis to be approximately linear. Thus the correlation coefficient which measures the degree of linear association between two random variables is an appropriate test statistics. A correlation coefficient of 1.0 shows a perfect correlation between two variables.
In our work, the residual analysis was employed to study the random behavior of the residuals, and normal probability was prepared for residual errors of LE, zeta size and zeta potential responses, as shown in Figure7

Residuals versus fitted value
The residual analysis plot consists of residuals and fitted value on the y and x axis, respectively. The plot is used to detect three things, its non-linearity, unequal error variances, and outliers.
At Figure 8 A-C, the residuals "bounce randomly" around the zero line; this indicates that the relationship is linear and reasonable. In addition, the residuals roughly form a "horizontal band" around the zero line, which indicates that the variances of the error terms are equal. Finally, there is no one residual "stands out" from the basic random pattern of residuals and this means that there are no outliers.

Figure 9 Residuals versus observation order for LE (A), zeta size (B) and zeta potential (C).
The use of the residuals versus observation order is to verify the assumption that the As can be seen in Figure 11 A-1, the minimum zeta size has been collected by using high concentration of Alg and the lowest concentration of CS. From Figure 11 A-2, the particle size below 50 nm can be prepared by using 50 mg of CS and 400 mg of Alg. As can be seen in Figure 11B-1, the particle size below 120 nm can be prepared by using 400 mg of Alg and 60 mg of CaCl2 ( Figure 11B-2). In the case of CS and CaCl2 variables at Figure 11C, the lower size than 120 nm can be collected by using CS concentration of 200 ng and CaCl2 range between 30 and 60 mg.    Figure 12 A shows the combined effect of Alg and CS concentrations. As can be seen, the zeta potential is bigger than-5.0mV and it was obtained at concentration of Alg which is bigger than 300 mg and Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 21 January 2020 doi:10.20944/preprints202001.0234.v1 CS concentration between 50-75 mg and 160-200 mg; whereas zeta potential is lower than -12.5 mV which was obtained at concentration of Alg below than 300 mg and CS concentration between 60-185 mg. Figure 12B shows  Figure 12C shows the 3D surface and contour plots represent a rising ridge surface.
As the color gets darker, the zeta potential response become at -4 mV, this is occur at high concentration of CaCl2 (55-60 mg) and CS concentrations below 50 mg and higher than 200 mg. The zeta potential response at -4 mV can be collected at low concentration of CaCl2 (below 30 mV) and CS concentration between 75-175 mg. All factors seem to affect the LE, zeta size and zeta potential because the line is not horizontal. At Figure 13A, Alg with concentration of 200 mg show higher LE comparing with 400 mg Alg. A CaCl2 with 30 mg has a higher LE mean than 60 mg.

Main Effects Plot for LE, zeta size and zeta potential
The CS also affects the LE. CS with 200 mg had a higher LE mean than 100 and 50 mg.
It is evident from Figure 13B that zeta size is minimum at the highest level of Alg and CaCl2 and the lowest level of CS.
From Figure 13C, . Figure 15 Interaction effects of factors on the zeta size As Figure 15 shows, there is an interaction between the Alg*CS ( Figure 15A) and CS*CaCl2 ( Figure 15C). The Figure 15A shows that there is a significant interaction between factors Alg and CS. The green and red lines (200 and 125 mg CS respectively) show that the mean size response decreases when the factor Alg level is low. While in Figure 15C, the green, red and blue lines which are corresponding to 60, 45 and 30 mg CaCl2 respectively shows that the size mean response decreases when the factor CS level is low.

Figure 16 Interaction effects of factors on the zeta potential
At interaction plot for Figure 16A and B, the lines are not parallel; this interaction effect indicates that the relationship between Alg concentration and zeta potential depends on the value of CS ( Figure 16A) and CaCl2 ( Figure 16B). For example, if we use Alg with concentration 200 mg, then CaCl2 at 30 mg is associated with the -20 mV ( Figure 16B). However, if we use Alg with concentration 200 mg, then CS with 50 and 125 mg is associated with the -10 mV ( Figure 16A).

Validation test
The comparison of experimental results with predicted values was shown in Table 6.
From the table, the theoretical values for response were close to experimentally obtained values. This result indicates that the mathematical models can be successfully be used to predict the LE, zeta size and zeta potential values for any combination of the Alg, CS and CaCl2 within the range of the performed experimentation.  respectively. The polysaccharides are highly hydrolyzed, so MET can be diffused out quickly and easily [45]. In addition, at the second stage, the release profile of MET presented with slowly sustained release from the nanocomposites.  nanocomposites. In addition, the MAC21 nanocomposites follow the first kinetic model (Table 7).

Conclusion
For the multiple linear regression analysis, mathematical models for LE, zeta size and zeta potential were developed using response surface methodology to formulate the input parameters Alg, CS and CaCl2 concentrations. Selected mathematical models showed that the developed response surface methodology models were statistically significant and suitable for all sanding conditions to have higher R² and R²-adjusted values. High correlation values were determined between the experimental data and predicted ones. The concentrations of Alg, CS and CaCl2with value 200 mg, 200 mg and 60 mg, respectively were determined as optimum conditions resulting in maximum LE;400 mg, 50 mg and 60 mg respectively, for minimum zeta size and finally 200 mg, 129 mg and 30 mg, respectively for -19 mV zeta potential. The verification experiment was carried out to check the validity of the developed mathematical model that predicted LE, zeta size and zeta potential within the range of 10% error limit.