The preparation of akara was carried out by a baiana de acarajé according to traditional practices (Rogério et al., 2014). A volume of 30 L of CPO (a mixture of liquid and solid phases) obtained by industrial extraction at Nazaré city (Brazil) and conditioned in tinplate cans was acquired at São Joaquim Fair, in Salvador city (Brazil). The oil was completely homogenized at 45 °C. A volume of 20 mL of the fresh oil was filtered in a glass wool filter, protected with nitrogen gas and stored in an amber bottle at −20 °C until analysis (Jorge and Gonçalves, 1998).

The frying was always performed outdoors. The total frying time was 5 h a day and the frying process was carried out over 5 consecutive days. The deep-frying began with 5 L of CPO, which were heated in an enameled pan (10 L capacity; 46 and 26 cm external and internal diameter, respectively; 13 cm high) for 12 min in the presence of a white onion (~60g). Five units of raw mass (uniform balls of 95 g each) were successively added and deep-fried in the CPO (475g/L of oil). The frying temperature ranged from 143 to 188 °C in the beginning (day 1) and 159–178 °C in the end (day 5). At the end of each day the oil was decanted, filtered hot and stored at room temperature in an enameled pan with a cover until the next day. The same procedure was repeated over the following four days. The total frying time was 25 h.

Replenishment with oil was performed according to the baiana’s practices. Fresh oil was added on days 1 and 2, whereas blends of fresh and used oil were added during the following days (Table 1).

The results obtained for the fresh oil are listed in Table 2. The fatty acid composition and levels of carotenoids were typical of an Elaeis guineensis palm oil, presenting a proportion of saturated and unsaturated fatty acids of almost 50:50% and carotenoid contents of nearly 600 mg/kg oil. The contents of individual fatty acids and total carotenoids were within the ranges found in the Codex Alimentarius (2017) and those previously reported for Brazilian crude palm oils (Almeida et al., 2013). The PV also corresponded to normal values for crude oils according to Codex Alimentarius (2017), which recommends a maximum of 15 meq O₂/kg oil. The levels of polar compounds were relatively high, but common for crude palm oil (Almeida et al., 2013). The analysis of the polar components by HPSEC showed that the polar fraction was mainly constituted by hydrolytic degradation products, diacylglycerols and free fatty acids. This fact is frequent in crude oils obtained from unhealthy or damaged oleaginous fruits (Sambanthamurthi et al., 2000). The oxidative stability index determined by Rancimat was also within the range reported for crude palm oil (Almeida et al., 2013).

Polar compounds. The total content of polar compounds (TPC) has been widely used to evaluate the quality of frying oils. They correspond to the products formed through the major reactions that take place during frying. The TPC is comprised of polymerized triacylglycerols, oxidized triacylglycerols, diacylglycerols and free fatty acids, i.e. compounds formed by thermal, oxidative and hydrolytic degradation, respectively (Dobarganes et al., 2000).

The changes in TPC contents observed in the oil during frying are shown in Table 3.

The TPC content increased linearly with the frying time (R² = 0.96), as observed in numerous frying studies (Olivero-David et al., 2017). However, no significant increment was found from the first to the second day of frying, i.e. between the oils used during 5 and 10 h, respectively. This fact can be attributed to the addition of fresh oil. Not only does the addition of fresh oil cause a dilution effect on the degradation products, but also it exerts some protection because of the reposition of antioxidants (Romero et al., 1998). After 15-h frying, the CPO exceeded the limit of 25% established in most of recommendations and regulations in different countries (Firestone, 2007). This fact was concomitant with the addition of a blend of fresh and used oil from the third day of the study, which emphasizes the importance of replenishing with fresh oil.

Compared to the standard method, lower levels of TPC were provided by the two quick tests, Testo 270 and Fri-check, which ranged from 9.5 to 22.2% and from 5.2 to 16.7%, respectively. These considerable differences with the standard method could be related to the unusually elevated levels of hydrolysis products found in this kind of oil. It is well known that polymerization and oxidation are the predominant reactions under frying conditions. Hydrolysis also occurs, but at a much lower level (Sebedio et al., 1996). The calibration of rapid tests, including those based on physical properties, is normally performed using common refined vegetable oils that are generally obtained from oleaginous seeds. The levels of hydrolysis products in such oils are minimal compared to those found in palm oil, and more concretely in crude oils like the one of the present study. As a result, rapid tests are calibrated to measure those changes imparted by the most abundant compounds formed during frying, i.e. polymers and oxidized triacylglycerols. As an example, the contribution of the hydrolysis products to the oil’s viscosity should definitely be much lower than that produced by the polymerization compounds, which could explain the low values provided by the Fri-check test.

Despite the differences in the levels of TPC given by Testo and Fri-check, both tests showed strong linear correlations with the standard method (r = 0.860 for Testo 270 and r = 0.947 for Fri-check). This fact indicates that the two tests responded well to the physicochemical changes occurring in the oil during frying. In a previous study, Testo 270 also showed linear correlation with the standard method, but the correlation presented by Fri-check was poor (Almeida et al., 2017). The main difference between the previous study and the present work is that a great variety of frying crude palm oils (collected at each point of sale of akara) with different solid contents were evaluated at that time. Therefore, the low correlation found could have been related to the heterogeneity of the samples analyzed.

The composition of polar compounds was analyzed by HPSEC in selected samples and the results are listed in Table 4. Although no significant differences were found for the content of TPC between the oils used for 5- and 10-h frying, the proportion of the different components was slightly different. As expected, the levels of polymerization and oxidation compounds were slightly greater for the 10-h sample. In agreement with other studies, no significant differences were found for the hydrolysis compounds, i.e. DG and FFA, as a consequence of frying (Sebedio et al., 1996; Olivero-David et al., 2017). Given that the levels of hydrolysis products were high in the fresh oil, the limit of 25% TPC was reached when the thermal and oxidative degradations were not elevated compared to refined oils obtained from oleaginous seeds. This fact is of interest from the nutritional point of view because DG and FFA are the same compounds produced in the digestion of fats and therefore they do not affect the nutritional value of oils.

Fatty acid composition. As expected, the fatty acid composition determined by GC (Table 5) showed considerable losses in unsaturated fatty acids during frying. Significant losses were even detected at 5-h frying. At the end of the study the losses in oleic and linoleic acids were 13% and 26%, respectively. While the relative losses were higher for the polyunsaturated fatty acid, oleic acid exhibited a greater decrease in absolute amounts (5.6 g/100 g of oil). This fact can be attributed to the much higher content of oleic acid and especially to the interactions between the thermoxidative reactions of oleic and linoleic acids (Morales et al., 2012).

Carotenoids. A steady drop in carotenoids took place simultaneously to the formation of degradation compounds. The level of carotenoids was practically exhausted at the end of the study, falling from 584.9 (± 5.0) mg·kg⁻¹ to 21.5 (± 1.2) mg·kg⁻¹, which is in agreement with a previous report (Correia et al., 2017). Therefore, due to the losses in these components with antioxidant properties, the oil became partially unprotected after 25-h frying.

Oxidative stability index. The oxidative stability index (OSI) determined by the Rancimat test is a widely used measurement of the resistance to oxidation of oils under accelerated conditions, i.e. high temperature and high oxygen availability. The OSI does not exactly provide information about the frying performance of oils because of the different conditions and reaction mechanisms involved in the test. However, it is a useful measure to evaluate different oils in a short period of time and to establish a relative order of oxidative stability. Unless oxygen is not a limiting factor, the stability order given by the test should be practically the same as the one obtained at room or moderate temperature. Therefore, the OSI can be used to get to know the oxidative behavior of oils or fat-containing foods during storage (Velasco and Dobarganes, 2002). In addition, the OSI is an indirect measurement of the antioxidant/pro-oxidant balance. In fact, it can be defined as the time required for the complete depletion of antioxidants at the test conditions, i.e. when the oil becomes unprotected and oxidation is accelerated. As expected, the OSI of the oil in the present study decreased progressively during frying as a consequence of the accumulation of thermo oxidative compounds, which accelerated the oil degradation process, and the gradual loss in antioxidants outlined above (Figure 1). The OSI decreased by 62.5% at the end of the frying study.

¹H-NMR spectroscopic measurements. ¹H-NMR spectroscopy can be applied to quantify the fatty acid chains of triacylglycerols and so estimate the fatty acid composition (Barison et al., 2010). Figure 2 illustrates the spectra obtained in the present study. Signals A and B correspond to hydrogens bond to β carbonyl carbons and to hydrogens bond to methylene carbons located between two double bonds (linoleate), respectively. Therefore, signals A and B are used to determine saturated and polyunsaturated fatty acids, respectively. In addition to the fatty acid composition, the ¹H-NMR spectra also provide information on degradation products. For instance, 1,2-diacylglycerols can be estimated by hydrogens at the sn-3 and sn-2 positions, which give signals C and D, respectively. Also, core aldehydes can be determined through signal E, which corresponds to the hydrogen of formyl groups.

The results for the fatty acid composition of the oil samples estimated by ¹H-NMR are listed in Table 6. A significant increase in signal A was observed at the end of the frying process, indicating an increase in saturated compounds. Although this signal is normally used to show an increase in saturated fatty acids (Barison et al., 2010), the formation of methylene hydrogen does not necessarily mean the formation of single saturated fatty acids. It is known that the loss in double bounds occurs concomitantly with the formation of functional groups containing oxygen or in condensation reactions to form polymerization compounds. In this regard, the results by Frankel et al., (1981) have shown that thermally decomposed methyl oleate and methyl linoleate produced about 77% and 67% saturated compounds, respectively.

As expected, a progressive decline in linoleic acid was observed by ¹H-NMR (Table 6), which is in agreement with the results obtained by GC. In this regard, a decrease in signal C with frying can be observed in Figure 2. Oleic acid also decreased in a similar way to that observed by GC.