(Ackman, 2005), the high AV obtaine

(Ackman, 2005), the high AV obtained, AV > 20 (Ackman, 2005),
denotes an advanced state of rancidity.
In Table 1, the effective removal of primary oxidation products
during the bleaching process is shown. For some bleached oils, processed at the highest clay amount and temperature (5 wt% and
130C), no significant value of PV was observed. Nevertheless, carrying out the bleaching process using low concentrations of clay,
1–3 wt%, at low process temperatures, 90–110C, did not reduce
the PV of the oil; on the contrary these conditions even enhanced
the formation of peroxides due to the exposure of the oil to high
temperatures. Considering the secondary oxidation products,Table
1shows an AV reduction in all the bleached oils. This fact denotes
the capacity of the acid-activated earths to adsorb these types of
compounds (Rossi et al., 2003; Sathivel, 2010). From the AV results
shown inTable 1, it could be clearly observed, within the experimental conditions assayed, that the adsorption of secondary oxidation products is more effective at higher temperatures and
concentration of activated earth. On the other hand, a direct correspondence between the Rancimat induction period, IP, and any of
the oxidation indices, PV, AV and totox, measured in the bleached
oil samples was not observed (Table 1).
The initialL

, a

, and b

values of the DNSO were, respectively,
71.46, 16.39, and 96.75 (Table 1). These CIELAB coordinates denote
a dark brown color, attributable to pigments such as carotenoids
(Indrasena and Barrow, 2010). The bleaching process increased
the lightness of the oil, reaching L

values up to 83.61. Besides,
the activated earth adsorption process effectively reduceda

value
indicating that a decrease in red pigment occurred. As a consequence, the oil samples became lighter and slightly more transparent. Nevertheless, the b

value of the bleached oils generally
suffered a minor reduction taking values in the range of 78.97–
99.56, which denotes yellowish color. Chroma values of the
bleached oils, which represent the intensity of color, were reduced,
while hue-angle value increased with the concentration of the activated earth. The hue-angle values were in the range of 79.53–90.76
which implies yellow–orange color for the bleached oils.
3.2. Statistical modeling
The experimental data of each measured variable were fitted to
a complete quadratic model. The polynomial coefficients for the
surface response model were calculated by multiple regressions,
and the results are expressed inTable 2for the oxidation and in
Table 3for the color parameters. Ap-value of associated probability was also calculated for each term of the regression model.
Selecting a confidence level of 95%, ap-value greater than 0.05
was not considered to be statistically significant. Similar statistical
procedures have already been employed to optimize numerous
processes involved in the up-grading of fish by-products such as
extraction of fish oil from herring by-products (Aidos et al.,
2003), enzymatic hydrolysis of sardine wastes (Dumay et al.,
2006) and hydraulic pressing of sardine discards (Pérez-Gálvez
et al., 2009).
Table 2andTable 3show that all the measured variables (FFA,
PV, AV, totox, IP,a

, b

, L

, chroma and hue-angle) are highly dependent on the linear effect of clay percentage, with an associated
probability p< 0.0001. Regarding temperature, its linear effect
was statistically significant for FFA, PV, AV, totox and IP, with
p< 0.001. On the other hand, the time was the input variable having the lowest influence on the measured variables, with linear effects being statistically significant (p< 0.005) in the cases of FFA,
AV, totox andL

. Quadratic effects were found to be significant only
for clay percentage in the cases of L

, b

, chroma and hue-angle;
and time in the cases of FFA, PV, AV andL

. The p-values for the
remaining effects indicated that the interaction between experimental factors was not statistically significant (p> 0.05).