*3.5. All Parameters Measured on the Eight FDOE Samples*

The application of the Wilcoxon rank-sum (Mann–Whitney U test) made it possible to reach some general considerations on the contributions that the individual three factors made to the formation of acrylamide in the experimental biscuit matrix. For this reason, the RIB was excluded from this test. Table 4 shows that a significant reduction in the acrylamide concentration in the experimental biscuits was obtained by decreasing the amount of AB in the recipe, using 1.5 g instead of 4.5 g, while also reducing the baking TP (150 ◦C in the initial phase and in the middle phase instead of 160 ◦C in both of them). However, as for the two SRT (2 and 3 min), no significant effect was shown with respect to acrylamide concentration.


**Table 4.** Wilcoxon rank-sum (Mann–Whitney U test).

Results of the Wilcoxon rank-sum (Mann–Whitney U test) in the samples set are reported as *p*-value. *p* < 0.001, very strong significant effect (bold and underlined); *p* < 0.01, strong significant effect (bold); *p* < 0.05, moderate significant effect (standard); *p* < 0.1, weak significant effect (italics); *p* > 0.1, n.s.: not significant. SRT: steam release time; TP: temperature program; AB: ammonium bicarbonate percentage.

The concentration of acrylamide increased in parallel with the increase in both the TP and AB due to the crucial role that these two factors play in triggering and quickening the Maillard reaction and the non-enzymic browning in general, while SRT seemed to play a less relevant role.

The statistical analysis was also extended to the other parameters analyzed. In addition to acrylamide concentration in biscuits, Table 3 shows the data concerning the values of the sample sizes, the CIELab color coordinates, and *aw*. The Wilcoxon rank-sum (Mann– Whitney U test) was applied to evaluate the influence of TP, SRT, and AB on the studied parameters (Table 4). As for the size parameters, length and width did not show any significant effect, whereas SRT significantly affected thickness (*p* = 0.0574). Indeed, by enhancing the SRT, the biscuit thickness significantly increased.

The small differences deriving from the manual preparation of the forms did not cause any substantial changes in length and width during the baking process. Conversely, the SRT influenced the thickness of the samples. This phenomenon could be explained by making different hypotheses: (i) the steam flow could have reduced the water loss from the biscuit because it decreases the vapor pressure between the surrounding environment and the biscuit surface [33], (ii) when the water evaporates, it stretches the gluten net where it is physically trapped, since, when the water tries to leave the dough, the latter is dragged upwards due to the passage of the water [33].

Contrarily, TP and AB did not have any significant effect on biscuit thickness. Temperature variation did not affect the rate of decomposition of the leavening agents, which occurs at temperatures far lower than 100 ◦C [34]. Concerning the AB concentration, the results show that the partial replacement of this leavening agent with sodium bicarbonate-cream of tartar was successful, and it did not cause any changes in the structure of the biscuits [15].

Regarding the color parameters, the wide difference in the RIB values compared with the experimental samples of the FDOE was clear. In the RIB samples, the values of redness (a\*) and yellowness (b\*) increased consistently with the increase in acrylamide concentration, to the detriment of values of brightness (L\*), which were at the same time reduced.

Having excluded RIB from the statistical comparison to carry out a finer evaluation of the combination of the FDOE factors, other aspects also emerged. L\* was significantly affected (*p* = 0.0016) by AB level, causing higher brightness when AB increased. This influence was also revealed by other authors [23,35], also considering different elaborations of the CIELab color parameters: for example, the E-value [36] and the browning index (BI) [22], both dependent on the L\* values. The a\* coordinate showed a significant effect for TP (*p* = 0.0238), giving a higher tendency to redness as temperature was increased. A weaker effect was shown by a\* due to AB (*p* = 0.0659), while AB heavily affected the b\* coordinate (*p* = 0.0008), thus giving a more intense yellowness as higher temperatures and lower AB levels were applied.

The higher concentration of AB significantly increased L\* and decreased b\* in the FDOE samples. This twofold effect was related to the ammonia release [11]. Contrarily, the lower concentration of AB entailed a higher concentration of sodium ions (from sodium bicarbonate), which yielded a darker biscuit surface color during baking [37,38].

On the other hand, the effect of TP on the behavior of color parameters was more predictable. Higher TP values significantly affected a\* and b\*, increasing both values. Taking into consideration that a\* refers to the value on the green–red axis and b\* refers to the value on the blue–yellow axis, it can be assumed that the more intense the thermal damage produced, the more evident the browning phenomenon on the surface.

Another way to compare colors of different samples is the evaluation of the color distance (ΔE). ΔE is a dimensionless parameter that arises from the combination of the L\*, a\*, and b\* values when pairs of samples are considered. ΔE of sample pairs leads to determining whether or not there is a difference in the colors perceived by the human eye according to specific thresholds [39], i.e., ΔE < 0.2 indicates an imperceptible difference between colors; 0.2 < ΔE < 0.5, a very small difference between colors; 0.5 < ΔE < 1.5, a small difference between colors; 2 < ΔE < 3, a barely distinguishable difference between colors; 3 < ΔE < 6, a very distinguishable difference between colors; 6 < ΔE < 12, a large color difference; and ΔE >12, completely different colors.

Table 5 shows the data obtained when applying these thresholds to the ΔE of sample pairs in the present work. This diversity of colors could easily be distinguished by the naked eye for most of the FDOE samples when compared to the RIB, while with regards to samples 5 and 6, the "difference" was even larger. In order to further investigate this point, specific sensory analyses should be run to understand if this discrepancy may lead consumers to reject the product.

**Table 5.** Color distance (ΔE) calculated for all possible pairs of samples. Samples 1–8 belong to the FDOE; RIB is the reference biscuit.


The last parameter considered was *aw*. The different levels of TP and AB did not have significant effects on the *aw* in biscuits, while this *aw* was affected only by SRT (Table 4). Indeed, samples 1, 3, 5, and 7, obtained with a 3 min RST, showed higher *aw* values than those obtained with a 2 min application only. For this reason, a careful assessment of shelflife in combination with the product's sensory properties should be taken into consideration. However, *aw* values of the FDOE biscuits were around 0.3, which represents a reference value for this type of food matrix.
