*2.6. Experimental Design*

A three-factor, three-level Box-Behnken experiment design was applied to evaluate the effects of different parameters on drug release rate. Briefly, the design is equal to the three replicated centre points and the set of points lying at the midpoint of each surface of the three-dimensional cube that defines the region of interest of each parameter. The three independent variables (*X*1, *X*2 or *X*3) were the concentration of CA in the sub-coating aqueous dispersion (*X*1), the sub-coating weight gain based on the dry uncoated pellet mass (*X*2), and the ADEC coating weight gain based on the sub coated pellets mass (*X*3). Each variable was coded to be in the range of −1, 0, 1, which represented different variable levels. Levels of the factors and constraints for the in vitro drug release based on preliminary pharmacokinetic study are listed in Table 1. The design required 15 experimental formulations. The independent variables of each formulation and their responses are listed in Table 2. The response surface model generated by the design is given as Equation (1):

$$Y = a\_0 + a\_{1 \times 1} + a\_2 X\_2 + a\_3 X\_3 + a\_4 X\_1 X\_2 + a\_5 X\_2 X\_3 + a\_6 X\_1 X\_3 + a\_7 X\_1^2 + a\_8 X\_2^2 + a\_9 X\_3^2 \tag{1}$$

where *Y* is the response parameter, *X*1, *X*2, and *X*3 are the independent parameters, *a*0 is the intercept, *a*1 − *a*3 are the main effect coefficients, *a*4 − *a*9 are coefficients of parameters with interaction or quadratic effects. Statistical analysis of the model was performed in Design-Expert software (V.8.0.6, Stat-Ease Inc., Minneapolis, MN, USA). The regression models of *Y*1, *Y*2, and *Y*3 were evaluated in terms of statistically significant coefficients using analysis of variance (ANOVA) and *r*2 values. Only coefficients with *p* values less than 0.05 were constructed in the models. In addition, response surface plots were

performed to visualize the effect of parameters and their interactions on the responses. Design space, which was determined from the common region of successful operating ranges for the responses, was established following the obtained response surface to clarify the optimal formulation.


**Table 1.** The factors and responses of the Box-Behnken design.

**Table 2.** Independent variables and observed responses of the Box-Behnken design.


#### *2.7. In Vitro Release of LXP and CA*

A dissolution test was carried out at 37 ◦C in 900 mL water, using a dissolution apparatus (78X-6A, Huanghai medicine inspecting institute, China) with the basket rotation speed of 100 rpm, which is specified in China Pharmacopoeia. The prepared sustained-release pellets containing 90 mg of anhydrous LXP were added to the dissolution apparatus. At pre-determined intervals, 5 mL of the sample was withdrawn and replaced with fresh medium. Then the samples were analyzed by HPLC.

In order to better understand the impact of CA on the drug release rate, simultaneous release profiles of CA and LXP in formulations with different CA concentrations were conducted. Additionally, the impact of dissolution media on the release of CA and LXP were investigated by performing the dissolution tests in the following media: pH 1.0 HCl, pH 4.5 and 6.8 phosphate buffers, and water. The contents of CA and LXP in the formulations were determined by HPLC.

### *2.8. Release Mechanism Studies*

The in vitro release mechanisms of LXP were analyzed by seven kinetics models. As shown in Table 3, *Qt* is the release amount of LXP at time *t*, *Q*0 is the initial amount of LXP in the pellets, *k*0 is the zero order release constant and *k*1 is the first order release constant, *kH* is the Higuchi dissolution constant, *n* is exponent constant characterizing different release mechanisms, *a* is a time scale parameter and *b* is a shape parameter that characterizes the curves of the release profiles. The dissolution data of LXP were fitted to these models by linear or non-linear least-squares fitting methods. The correlation coefficients calculated by regression analysis were used to evaluate the goodness of fit for each model.


**Table 3.** Models for drug release.
