Level of free 3-MCPD and 3-MCPD esters in glycerol model systems
The level of free 3-MCPD in glycerol model systems differed and was affected by the type of sweetener used (Table 2). The highest 3-MCPD level was reached for the model with the addition of steviol sweetener (114.8 μg kg−1). The lowest 3-MCPD level (51.9 μg kg−1) was found in the model with stevia leaves. The model with the addition of sucrose also demonstrated a high level of free 3-MCPD (99.6 μg kg−1), while its level in the models with the use of polyols was noticeably lower, from 57.3 μg kg−1 for erythritol to 77.4 μg kg−1 for maltitol. A post hoc test revealed three homogeneous subsets: (1) stevia leaves, control, erythritol, and xylitol; (2) maltitol; (3) sucrose and steviol sweetener. 3-MCPD esters were not detected in any of the models, due to the absence of added fat.
Level of free 3-MCPD and 3-MCPD esters in shortbread model systems
The level of free 3-MCPD in the model with the addition of triolein was roughly comparable to those received for the glycerol models (the same order of magnitude, Table 2), while for the models containing steviol sweetener and sucrose, the presence of fat resulted in a decrease of free 3-MCPD content of about 40%. The lowest level of free 3-MCPD was observed for control samples (55.5 μg kg−1), but the model system with the addition of erythritol still exhibited the lowest content of free 3-MCPD among sweetener model systems (62.4 μg kg−1), whereas the addition of stevia leaves induced the highest result (75.3 μg kg−1). The latter result was significantly different from the rest of the models. Significant differences were observed between the models of control and sucrose, and control and stevia leaves. The level of 3-MCPD was comparable in the model systems with the addition of xylitol, and maltitol. For the other samples, statistical analysis did not show any significant differences.
The levels of 3-MCPD esters in the shortbread model system expressed as 3-MCPD were substantially higher (10–16 times) compared to free 3-MCPD. The maximum content was obtained in the model containing sucrose (1112.4 μg kg−1) and in the model with steviol sweetener (1036.2 μg kg−1). The lowest ester content was discovered once again in the control model and the model containing erythritol (552 and 676.4 μg kg−1, respectively). Models including xylitol, maltitol, and stevia leaves behaved similarly and did not differ significantly (post hoc Tukey test results).
Effect of sweetener on free 3-MCPD content
Comparing the results obtained for both glycerol and shortbread systems, it seemed that some of the sweeteners influenced greatly the level of free 3-MCPD. This phenomenon was particularly noticeable for the model involving sucrose and polyols, especially erythritol. Other polyols, xylitol and maltitol, also had an impact on 3-MCPD formation; however, the effect was not as noticeable as for erythritol. An interesting occurrence was observed for steviol sweetener and stevia leaves, which, in some cases, enhanced 3-MCPD level in the investigated model systems.
In general, a possible explanation for the impact of sweetener on free or bound 3-MCPD formation arose from the mechanism of 3-MCPD formation. The higher level of 3-MCPD in glycerol model systems with the addition of sucrose was consistent with the previous studies [12, 13] and was simply caused by the decomposition of sugar with the generation of organic acids (lactic acid, levulinic acid, and formic acid ) that lowered pH. Low pH values are a valuable factor contributing to formation of 3-MCPD from chloride ions and glycerol. However, a comparable high content of 3-MCPD was also detected in models with the use of steviol sweetener. For this reason, the pH value of the investigated sweeteners was measured after thermal processing (Fig. 1a). As can be observed, the lowest pH values were obtained for steviol sweetener (4.88), while the highest was for erythritol (5.95). The rest of the sweeteners were characterised by pH values in the range 5.49–5.89. Among polyols, pH values decreased with the increase of the molecular mass and the number of carbon atoms. The only exception in the pH values was the stevia leaves; however, the amount of stevia leaves used in the model systems was lower.
In the case of the shortbread model systems, a similar correlation was observed, including the outliers for stevia leaves (Fig. 1b). For both systems, when excluding the values for stevia leaves, the coefficient of determination between 3-MCPD content and pH of sweeteners was equal to 92 and 85% for the glycerol and shortbread models, respectively. This confirmed that pH value had a strong impact on free 3-MCPD generation in the model systems.
Higher pH values were noticed for polyols, which can be attributed to their thermal stability. Polyols, in contrast to sucrose, do not decompose at higher temperatures and do not produce organic acids. Erythritol has excellent thermal stability even above 180 °C; xylitol is stable at temperatures up to its boiling point (216 °C) . Heating of maltitol, which is hydrogenated disaccharide, leads to its cleavage into sorbitol and glucose , which can be further decomposed, producing organic acids. Therefore, a lowered pH value is achieved that is in line with our observations. pH of steviol sweetener resulted from the presence of maltodextrin (polysaccharide), which was a basic ingredient of this sweetener. Excessive heating of such starch-derived sweeteners gives rise to a partial release a free glucose, anhydrosugars, and fructose-containing sugars  that can be further decomposed with the production of organic acids and lowering the pH level.
An interesting occurrence is a high level of 3-MCPD obtained in the shortbread model system incorporating stevia leaves. Data in the literature on the thermal stability of stevia during baking are generally conflicting and unclear. On one hand, it has been reported that the leaves, as well as the pure stevioside extracts, could be used in their natural state or cooked, and are thermostable at temperatures up to 200 °C. On the other hand, at temperatures exceeding 140 °C, forced decomposition was seen which resulted in total decomposition at 200 °C. Degradation products of stevioside included, e.g., traces of steviolbioside and glucose, which could be attributed to the rupture of the C19 ester bond in stevioside [37, 38]. Hence, it can be assumed that the traces of glucose could be attributed to the 3-MCPD formation through the lower pH level. Nonetheless, the amount of stevia leaves extract added to the model system was low, so the amount of glucose that could be released upon heating would be low as well. Furthermore, pH values of the stevia leaf extract were similar to maltitol; however, the corresponding free 3-MCPD levels varied.
Another question was whether the polyols could interact with other shortbread batter components, such as amino acids contained in flour. It is commonly known that reducing sugars or monomers that are generated from polymetric sugars can react with amino compounds in the Maillard reaction. This mechanism can also explain why glucose promotes 3-MCPD generation, i.e., via the removal of amino acids by the Maillard reaction. Polyols do not act in the Maillard reaction [35, 36]. When glucose or starch-derived sweeteners are absent, amino compounds can inhibit 3-MCPD formation . This also explains the decrease of 3-MCPD formation in the models with the addition of polyols.
Effect of sweetener on esterified 3-MCPD content
Regarding 3-MCPD bound in the form of esters, its formation from triglycerides and chloride ions is significantly more complex. As has been mentioned in the Introduction, most proposed mechanisms of 3-MCPD ester formation take place in acidic media; hence, it was expected that the formation of 3-MCPD esters depended on pH value. Another mechanism, suggested by Zhang et al. , involves reactions with free radicals. Therefore, in this case, 3-MCPD ester formation should be related to radical scavengers present in the model system. In conclusion, two hypotheses explaining the influence of sweetener on 3-MCPD ester formation can be assumed: (1) the impact of pH generated during thermal processing of sweeteners; (2) the effect of sweeteners as antioxidants that can inhibit lipid decomposition.
To examine a hypothetical effect of pH resulting from the use of various sweeteners on 3-MCPD esters formation, a mixture of sweetener and flour was heated in the same conditions as model systems and then dissolved in water. Received values of pH for each sweetener are presented in Fig. 1c. As can be seen, the lowest pH values were obtained for sucrose (5.12), while the highest was for erythritol (5.98). The high determination coefficient (93%) suggested that pH value altered the 3-MCPD generation in model systems. As pH decreased, more 3-MCPD ester was formed, which can be seen as proof that an acidic environment promotes the reaction between lipids and chloride ions.
The second hypothesis explaining the inhibitory effect of polyols on the formation of 3-MCPD esters is based on the potential radical mechanism of 3-MCPD ester generation that was proposed by Zhang et al. . The authors observed that four antioxidants (L-ascorbyl palmitate, α-tocopherol, lipophilic tea polyphenols, and rosemary extract) could decrease 3-MCPD esters in palm oil during deodorization. Consequently, we hypothesised that compounds with antioxidant capacity should contribute to the scavenging of free radicals and the inhibition of 3-MCPD ester generation.
In the case of polyols, their antioxidant properties have already been reported by Faraji et al. , who revealed that polyols could function as antioxidants in fish oil. Hartog et al.  demonstrated that in vitro erythritol was an excellent radical scavenger. Nevertheless, our study did not confirm these findings: none of the six examined sweeteners, either raw or after thermal processing, demonstrated any antioxidant properties (data not shown, due to lack of any significant results).
The only source of compounds with antioxidant properties could, therefore, be compounds formed during the Maillard reaction. It was previously confirmed that reductones and melanoidins formed in browning reactions present antioxidative activity based on reducing power and metal chelating capability . Nonetheless, as has been mentioned earlier, heating of sugar alcohols does not lead to the browning process that occurs when sugars are heated. Therefore, they should not form compounds with antioxidant potential. However, to explore this relation, we decided to assess the antioxidant capacity of a mixture of flour and sweeteners that had previously been subjected to heat treatment at 200 °C.
Figure 1d shows the relationship between the antioxidant capacities of compounds produced during heating and the amount of esters formed. As can be seen from the figure, the highest potential antioxidant capacity was obtained for the model incorporating stevia leaves (1.03 mol Trolox kg−1) and sucrose (0.88 mol Trolox kg−1), whereas the smallest was for the control (0.53 mol Trolox kg−1) and the model involving erythritol (0.59 mol Trolox kg−1). Slightly less antioxidant potential than those obtained for sugar was found for steviol sweetener containing maltodextrin, which could also partly take part in the Maillard reaction. However, the antioxidant capacity of Maillard products was about ten times lower than, e.g., the antioxidant capacity of spices . The relationship between the amount of produced esters and the antioxidant potential of model systems was, therefore, the opposite of the hypothesis that antioxidants would inhibit the formation of esters of 3-MCPD.
As has been suggested before, in the case of free MCPD, the increasing effect between the amount of produced ester and the antioxidant capacity of the samples can be explained by the capture by the Maillard reaction of amino acids that could  increase the pH favouring retention of forming esters and  inhibit the formation of 3-MCPD esters. Nonetheless, the latter hypothesis was formed in the case of free 3-MCPD, and no previous research on the relation between amino acids and 3-MCPD esters formation has been performed.
Another possible factor could be that some of the compounds of the Maillard reaction can promote fat oxidation . However, it must be emphasised that the Maillard reaction usually occurs at higher temperature and requires more energy than the oxidation of fats , so compounds of the Maillard reactions might not have a direct inhibitory effect on lipid oxidation. In addition, this effect has not been explored enough in recent literature. Another problem is that all processes occurring upon heat treatment of fat-rich food are correlated, such that the products of each reaction can modify the other .
The most effective formation of 3-MCPD esters occurs at about 230 °C ; this temperature is close to the process of caramelisation, which itself generates an acidic condition. In conclusion, pH seems to have a dominant effect over other factors contributing to 3-MCPD ester generation.