*Data Analysis*

Table 1 shows the experimental treatments in the CCD design of RSM applied in optimizing the removal of mycotoxins from PKC using chitosan in Table 2, the estimated regression coefficients and the corresponding *R<sup>2</sup>* values, *p*-value and model lack of fit are presented. Each response was evaluated as a function of main (linear), quadratic and interaction effects of pH (*x*1), time (*x*2) and temperature (*x*3). As revealed by ANOVA, the coefficients of multiple determination (*R<sup>2</sup>*) for the responses ranged between 0.89 and 0.98, thus illustrating that the quadratic polynomial models obtained adequately represented the experimental data (Table 2). The models sufficiently related the studied independent variables to the simultaneous removal of eight mycotoxins by CTS, indicating a perfect fit (*p* < 0.05) with the second-order response surface equations. A good model fit is indicated by the *R<sup>2</sup>* values being at least 0.80 [27] Apparently, chitosan showed poor adsorption for DON, HT-2 and T-2 as the *R<sup>2</sup>* values for these responses were all less than 0.80 (i.e., 0.62, 0.55 and 0.71 respectively) and their *p*-values were likewise not significant (*p* > 0.05). These low *R<sup>2</sup>* values indicate that the predicted responses differed considerably from the experimental values and the models were not adequate. This observation may be due to the fact that trichothecenes, being non-ionizable molecules having a bulky epoxy group, have poor adsorption with plane surfaces. As a consequence, they adsorb on very few adsorbent agents [28].


**Table 1.** Experimental design (coded) of the central composite design (CCD).


**Table 2.** Regression coefficient, *<sup>R</sup>2*, *p*-value and lack of fit test for the reduced response surface models.

\* Non-Significant (*p* > 0.05).

Judging from the *F* and *p* values for the main, quadratic and interaction effects of each independent variable (pH, time and temperature) as seen in Table 3, most of them showed significant effects (*p* < 0.05) for the removal of mycotoxins. Compared to the quadratic and interaction effects, the main linear effects were more significant (*p* < 0.05) for mycotoxin removal. As all factors had significant (*p* < 0.05) effects on the mycotoxin removal (Table 3), they should therefore be retained as critical parameters in the final reduced model for fitting with the experimental data. Furthermore, the interaction effects of all the independent variables were significant (*p* > 0.05) only in the removal of AFB1. Except in the case of ZEA, the interaction effects of pH with the other two factors was significant (*p* < 0.05) in the removal of the mycotoxins. The reason may be due to the structure and the differences in the solubility of ZEA, owing to its polarity. The resorcinol moiety of ZEA is fused to a 14-atom macrocyclic lactone ring, with a double bond in the trans isomeric form with ketone and methyl groups [29], and its deprotonated form (ZEN) exists at pH > 7.62 [30,31]. Therefore, at acidic pH, the main binding mechanism will likely be hydrophobic interactions. Therefore, the removal of ZEA depended on temperature during equilibrium time. On the other hand, the results showed that the interaction of pH and temperature was significant (*p* < 0.05) for response variables, such as AFB1, AFG1, AFG2, OTA and FB1, with the effect being most significant for AFB1 and OTA reduction as their *F* values were 97.42 and 70.21 respectively. The interaction effect of pH and time was significant (*p* < 0.05) for removal of AFB1, AFB2, AFG1 and OTA (Table 3). Likewise, the interaction of pH and time had the highest influence on the removal of AFB1 due to its rather high *F*-value of 127.30. Therefore, it can be observed that pH was more influential for mycotoxin removal than the other variables studied. An acidic pH causes the release of more H<sup>+</sup> ions that may react with the adsorbent or adsorbate, thereby affecting results. Being a polycationic polymer, the surface of chitosan would become strongly positively charged under more acidic conditions, due to protonation of amino groups. This would cause increased electrostatic reactions between it and the negatively-charged mycotoxin molecules, leading to better adsorption. However, with gradual increase in pH, the adsorbent (chitosan) surface carries more negative charges, causing repulsion between it and the mycotoxin molecules. This eventually results in reduced adsorption capacity [20].


**Table 3.** Significant Probability (*p*-value and *F*-value) of the independent variable effects in the reduced response surface models.

a Non-Significant (*p* > 0.05).

The interaction effects of the processing variables on mycotoxin removal, shown to be significant from ANOVA results (Table 3), are explained visually with three-dimensional (3-D) response surface plots in Figure 1a–m. It can be seen from Figure 1a–c that the removal of AFB1 was significantly affected (*p* > 0.05) by the interaction of all the independent variables. The interaction effect x1x3 was also significant (*p* < 0.05) for almost all the target mycotoxins except ZEA, AFB2 and FB2, thus indicating an overall positive effect on mycotoxin reduction. Maximum removal of AFB1 and AFG2 were seen in the middle of pH (*x*1) with two other factors (*<sup>x</sup>*2, *x*3) for AFB1 (Figure 1a–c) and pH (*x*1) and temperature (*x*3) for AFG2 (Figure 1g). Likewise, maximum removal of ZEA was illustrated in the middle of *x*2 and *x*3 in Figure 1j.

  **H PLW**

 

**HS PHW**

 

 **%)\$**  **%)**

**+S**

 

**H PLW**

**Figure 1.** Response surface plots showing the interaction effect of independent variables on the reduction of 8 mycotoxins (**a**–**c**) AFB1, (**d**) AFB2, (**e**–**f**) AFG1, (**g**) AFG2, (**h**–**i**) OTA), (**j**) ZEA), (**k**–**l**) FB1, (**m**) FB2.
