SCIENCE CHINA Chemistry • ARTICLES •
May 2013 Vol.56 No.5: 664–671 doi: 10.1007/s11426-012-4765-9
Chemometrics-assisted excitation-emission fluorescence spectroscopy for simultaneous determination of ethoxyquin and tert-butylhydroquinone in biological fluid samples CHEN Yao, WU HaiLong*, WANG JianYao, ZHANG XiaoHua, LI Yong, ZHANG ShuRong & YU Ru-Qin State Key Laboratory of Chemo/Biosensing and Chemometrics, College of Chemistry and Chemical Engineering, Hunan University, Changsha 410082, China Received July 16, 2012; accepted August 7, 2012; published online October 17, 2012
A novel method applying simple, rapid, effective and inexpensive excitation-emission matrix (EEM) fluorescence spectroscopy coupled with second-order calibration method for simultaneous determination of ethoxyquin (EQ) and tert-butylhydroquinone (TBHQ) contents in biological fluid samples was developed. After a simple data preprocessing that was to insert zeros below the first-order Rayleigh scattering, the second-order calibration method based on the alternating normalization-weighed error (ANWE) algorithm was used to deal with EEM data. Via the introduced “second-order advantage”, the individual concentrations of the analytes of interest could be obtained even in the presence of uncalibrated interferences. The experimental concentration ranges for the analytes were as follows: EQ, from 4.58 to 20.6 g mL1 in plasma and from 6.87 to 20.6 g mL1 in urine; TBHQ, from 4.49 to 20.2 g mL1 in plasma and from 6.73 to 22.4 g mL1 in urine. The recoveries from spiked biological fluid samples were in the ranges of 92.8%–106.2% for EQ and 94.6%–107.2% for TBHQ. These results demonstrate that the three-dimensional EEM fluorescence with second-order calibration method is a powerful tool for obtaining both EQ and TBHQ quantitative results in plasma and urine samples, and could be applied to more complex matrices. ethoxyquin, tert-butylhydroquinone, biological fluid samples, fluorescence, alternating normalization-weighed error
1 Introduction Ethoxyquin (EQ) and tert-butylhydroquinone (TBHQ) are synthetic phenolic antioxidants that prolong the shelf life of foods by protecting them against deterioration caused by oxidation, such as fat rancidity and discoloration. They are used in various food products, mainly as animal feeds, unsaturated vegetable oils, many edible animal fats and meat products [1–6]. However, widespread studies, including long-term studies, have indicated that the superscalar use of *Corresponding author (email:
[email protected]) © Science China Press and Springer-Verlag Berlin Heidelberg 2012
synthetic phenolic antioxidants can result in potential health risks associated with their intake. EQ and TBHQ are suspected as being carcinogenic and responsible for a variety of unidentified health disorders [4, 7–16]. Most countries of the world have regulations for controlling the use of synthetic antioxidants in food applications, but the upper limits for EQ and TBHQ in food vary widely [4, 17–21]. The mechanism of toxicity of them is obviously ambiguous. And at the same time toxicological evaluation of synthetic phenolic antioxidants has been the subject of controversy in recent years [22]. To the best of our knowledge, limits are established according to the toxicity of substances. Understanding the quantity of EQ and TBHQ chem.scichina.com
www.springerlink.com
Chen Y, et al.
Sci China Chem
in the biological fluid samples may reinforce the mechanism of toxicity of them, and the exact usage of them. Because of the unsolved problems, we should explore a quick, inexpensive and effective method for EQ and TBHQ residue quantification in biological fluid samples. At present, various analytical methods have been reported for the determination of synthetic phenolic antioxidants in vegetable oils as well as in different types of food, including micellar electrokinetic capillary chromatography with electrochemical detection [21], gas chromatography (GC) [23], high-performance liquid chromatography (HPLC) [1, 2, 18], gas chromatography-mass spectrometry (GC/MS) [24], liquid chromatography-ion trap mass spectrometry (LC/ITMS) [25] and liquid chromatography/tandem mass spectrometry (LC/MS/MS) [22, 26]. Unfortunately, those methods may inherently suffer from the main disadvantages associated with the need of time-consuming and complicated pre-processing. Few measures have been used for simultaneous determination of EQ and TBHQ in vegetable oils [24]. And up-to-now there has been no attempt on determination of EQ and TBHQ in biological fluid samples following literature reports. Because of the demand for sensitive and effective quantification methods of EQ and TBHQ, there is much interest in exploring alternative analytical techniques. Excitationemission fluorescence matrices (EEMs) produce three-way data which convey certain advantages: the measurements are carried out on a single instrument, the signals are selective and sensitive, and the obtained models are trilinear. Contemporaneously, the second-order calibration methods have been noted in the last two decades owing to the benefit called second-order advantage [27, 28]. While decomposing the three-way data array, this advantage would help the researchers obtain qualitative and quantitative information of interested species in their co-existence in an invasive manner without disturbing the equilibrium of the original interaction system. Thus, in this context we turn our attention to the spectrofluorimetry coupled with second-order calibration method base on the alternating normalization-weighted error (ANWE) algorithm [29] for the analysis of EQ and TBHQ in biological fluid samples. The three-dimensional fluorescence data usually comprises the Rayleigh scatterings that do not conform to the trilinear model, which cannot be normally resolved by second-order calibration method. In this paper, after inserting zeros close to and outside the first-order Rayleigh scatter lines [30], the EEM data was decomposed by ANWE algorithm. From the resolved results, the EEM data combined with ANWE show up as a considerably useful combination for the simultaneous quantitative determination of EQ and TBHQ in biological fluid samples. The methodology may help researchers deeply penetrate into their mechanism of toxicity and help official regulations formulate the uniform limit for EQ and TBHQ in food.
665
May (2013) Vol.56 No.5
2 Experimental 2.1
Reagents and chemicals
Analytical-reagent grade chemicals were employed in all experiments. Iso-propanol and n-hexane were purchased from Chemical Reagent Research Institute (Tianjin, China). Stock solutions (257 g mL1 and 202 g mL1, respectively) of ethoxyquin (Sigma, USA) and tert-butylhydroquinone (Sigma, USA) were prepared by dissolving each compound in iso-propanol/n-hexane (4:1, by vol) and stored at 4 °C for a maximum 2 of days. From these concentrated solutions, 2.06 g mL1 and 2.02 g mL1 working solutions were prepared, respectively. The plasma was taken from the Blood Center in Changsha. The morning urine was supplied by normal people. 2.2
Preparation of analytical samples
All analytical solutions were prepared in a 10 mL centrifuge tube by dissolving in iso-propanol/n-hexane (4:1, by vol), and their concentrations were shown in Table 1. The biological fluid samples spiked with or without EQ and TBHQ should be ultrasonically treated for 15 min and then centrifuged at 10000 r min1 for 10 min. 2.3
Apparatus
F-4500 fluorescence spectrophotometer from Hitachi (Japan) equipped with a Xenon lamp. Fluorescence measurements Table 1 No. C1 C2 C3 C4 C5 C6 C7 C8 C9 C10 PP11 PP12 PP13 PP14 PP15 PP16 PU17 PU18 PU19 PU20 PU21 PU22
Contents of all analytical samples EQ (g mL1) 0 0.114 0.206 0.183 0.160 0.0229 0.114 0.0916 0.0687 0.0458 0.183 0.114 0.0916 0.0687 0.0458 0.137 0.0687 0.114 0.0916 0.206 0.160 0.137
TBHQ (g mL1) Plasma (L) 0.112 0 0 0 0.0224 0 0.0449 0 0.0673 0 0.0898 0 0.112 0 0.202 0 0.157 0 0.179 0 0.0449 100 0.112 100 0.202 100 0.157 100 0.179 100 0.135 100 0.0673 0 0.0898 0 0.112 0 0.157 0 0.202 0 0.224 0
Urine (L) 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 100 100 100 100 100 100
666
Chen Y, et al.
Sci China Chem
among all samples were carried out using 1.00 cm quartz cells, slit widths, 5 nm, wavelength excitation range, from 200 to 420 nm (each 2 nm) and wavelength emission range, from 290 to 560 nm (each 3 nm), scan rate, 12000 nm/min. All the programs used in the paper were written in-house in the Matlab environment and run on a personal computer E4600 processor with 1 GB RAM under Windows XP operating system. 2.4
Chemometric methodology
2.4.1 Trilinear model for second-order calibration The excitation-emission matrix data of all analytical samples can be stacked to compose a three-way data set. The size of three-way data array X is I×J×K (I, number of excitation wavelengths; J, number of emission wavelengths; K, number of samples). The trilinear model for second-order calibration is given by N
xijk ain b jn ckn eijk n 1
(i =1, 2,…, I; j=1, 2,…, J; k=1, 2,…, K) (1) where xijk , ain , b jn , ckn and eijk are the typical elements of X (I×J×K), A (I×N), B (J×N), C (K×N) and E (I×J×K), respectively. N denotes the number of factors, which should be considered as the total number of detectable species, including the analytes and the background as well as uncalibrated interferences. Matrix A is the relative excitation profile of the N species; matrix B is the relative emission profile of the N species; matrix C is the relative concentrations of the N species in K samples; E is an I×J×K three-way residual array. 2.4.2 ANWE algorithm The alternating normalization-weighted error (ANWE) algorithm is designed by our laboratory, which has been used successfully for resolution of the three-way data array in analytical chemistry [29, 31, 32]. It uses the least-squares principle to alternatively minimize three different objective functions. The iterative procedure and calibration for ANWE are similar to parallel factor analysis (PARAFAC) [33, 34]. Compared with other algorithms, ANWE can obtain reliable content results for the analytes of interest, hold fast convergence, be of insensitivity to over-estimated component number and successfully solve highly collinear problems. 2.5 Simultaneous determination of EQ and TBHQ in biological fluid samples Samples were prepared for detection into fluorescence spectrophotometer as described in Table 1. The spectra of iso-propanol/n-hexane (4:1, by vol) blank solution, plasma blank solution with 100 times dilution and urine blank solution with 100 times dilution were recorded in triplicate ex-
May (2013) Vol.56 No.5
periments during the whole research process. The first ten samples (C1-C10) were built as a calibration set, and the rest ones (PP11-PU22) were constructed as two prediction sets (one is in plasma, another is in urine). Fluorescence EEMs were collected for each of these and translated into a suitable format. Following this, the three-way data were set up, and then inserted zeros outside the data area to remove the first-order Rayleigh scatter lines. The new three-way data, after subtracting the corresponding solvent blank, was analyzed by ANWE algorithm. 2.6 Pretreatment of excitation-emission fluorescence matrices Figures 1(a) and (b) show the three-dimensional plot corresponding to the EEM for the predicting samples PP16 (one of the spiked plasma samples) and PU22 (one of the spiked urine samples) in wide spectral excitation and emission ranges, showing the presence of both Rayleigh and Raman scatterings. To avoid the presence of signals that are uncorrelated with the target concentrations of the studied analytes, EEMs, as a general rule, should be truncated. In such situation, however, the spectra of EQ and TBHQ lie close to the Raleigh scattering, the truncated scope of their fluorescence data may cause certain mathematical difficulties in decomposition for losing massive information of analytes. Thus, we employ a method [30] that is inserting zeros below the first-order Rayleigh scattering shown in Figures 1(c) and (d), which can effectively eliminate the effect of Rayleigh scattering and retain more useful information. That is, in added plasma samples, excitation from 244 to 420 nm at 2 nm intervals (I = 89) and emission from 314 to 536 nm at 3 nm intervals (J = 75), making a total of 6675 spectral points, as well as excitation from 246 to 420 nm at 2 nm intervals (I = 88) and emission from 322 to 560 nm at 3 nm intervals (J = 80), making a total of 7040 spectral points in added urine samples. Then the pretreated matrix data of all analytical samples (see Table 1) can be stacked to shape into a cube. There are two cubes. The size of the first one X1 is 89×75×16 in plasma, and the other X2 is 88×80× 16 in urine. Whether in plasma or in urine, sixteen samples are prepared for the simultaneous determination of EQ and TBHQ. The first ten samples constitute a calibration set, and the rest samples built as a prediction set.
3 3.1
Results and discussion Plasma sample
To begin with, the selection of the number of spectral components in cube should be ascertained. The core consistency test [35] drops to a very low value when using five spectral components to model the cube, suggesting that N = 4 is a sensible choice. ANWE algorithm was applied to cubes of data formed by the EEMs for the 10 calibration samples
Chen Y, et al.
Sci China Chem
May (2013) Vol.56 No.5
667
Figure 1 Three-dimensional plots of the excitation-emission matrix fluorescence: (a) the sample PP16 and (b) the sample PU22, showing the presence of Rayleigh and Raman scatterings. (c) and (d) correspond to (a) and (b), respectively. But their fluorescence spectra were inserted zeros below the first order Rayleigh scattering.
C1–C10, together with the plasma samples PP11–PP16 shown in Table 1. Figure 2, where the components have been labeled according to the order assigned by the model in the cube, displays the actual spectral profiles and the resolved spectral profiles from the decomposition of the excitation-emission matrix fluorescence data array obtained for both the calibration and predicted samples employing ANWE. Figures 2(a) and (b) stand for excitation profile and emission profile, respectively. As can be seen, the overlapping is very strong in the useful spectral regions, posing great challenges on the used quantitative technique. And seen from Figure 2(c) that shows the concentration distribution of EQ, TBHQ and interferences, the background of plasma interferes with EQ and TBHQ. In such situation, the resolved spectra coincide well with
the fluorescence spectra actually measured in EQ and TBHQ individually. And in Table 2, the average predicted recoveries of EQ and TBHQ gained from ANWE are 99.1±2.5% and 100.6±5.3%, respectively. With the aid of second-order advantage, the spectral profiles of EQ and TBHQ can be extracted. Those results (see Table 2) obtained for spiked plasma samples clearly illustrated that the ANWE algorithm, which can be used for improving selectivity by mathematical means, has the capacity of simultaneous quantitative analysis of EQ and TBHQ in complex plasma matrix. And those results fully illustrate the meaning and usefulness of the “second-order advantage”. This useful property avoids the requirement of either interference removal, as in standard working curve (zeroth-order calibration), or the construction of a large and diverse calibration
668
Chen Y, et al.
Sci China Chem
May (2013) Vol.56 No.5
Figure 2 Normalized excitation (a) and emission (b) profiles, including the resolved fluorescence spectra and the actual fluorescence spectra of EQ and TBHQ, and relative concentration (c) profile, which were obtained from plasma samples by ANWE method with four factors. Table 2 Results obtained when applying ANWE algorithm analysis to plasma samples spiked with ethoxyquin (EQ) and tert-butylhydroquinone (TBHQ) EQ (g mL1) actual predicted a) PP11 0.183 0.170(92.8) PP12 0.114 0.116(101.8) PP13 0.0916 0.0903(98.6) PP14 0.0687 0.0690(100.4) PP15 0.0458 0.0469(102.4) PP16 0.137 0.135(98.5) Average recovery (%) 99.1±2.5 Sample
TBHQ (g mL1) actual predicted a) 0.0449 0.0425(94.6) 0.112 0.108(96.4) 0.202 0.213(105.4) 0.157 0.165(105.1) 0.179 0.192(107.2) 0.135 0.128(94.8) 100.6±5.3
a) Recovery in parentheses.
set as in multivariate calibration (first-order calibration). 3.2
Urine sample
The analysis of urine samples is similar to the plasma samples, using the ANWE method as well. The estimated number of components is also 4 after calculated by the core consistency test. The excitation and emission spectra resolved by ANWE algorithm associated with the actual spectra are shown in the Figures 3(a) and (b), respectively. Figure 3(c) that is a relative concentration profile. EQ and TBHQ are
subject to seriously background interferences. But that does not affect the resolution. The resolved spectra coincide well with the actual fluorescence spectra. The prediction results for the urine samples are listed in Table 3. For recovery analysis, six real samples spiked with urine were prepared with the six concentration levels of EQ and TBHQ as described in experimental section. As seen from Table 3, the recovery results are satisfactory. In the urine samples, the average predicted recoveries of EQ and TBHQ gained from ANWE are 99.9±2.0% and 100.9±1.3%, respectively. This is also a remarkable achievement, because of successful prediction for interested analytes in the samples of the complexity of biological fluid samples. The method is reliable and guarantees repeatability of results and the process preparation of the sample is not complicated. In marked contrast, in edible vegetable oil the recovery values of GC-MS [24] were EQ: 75.6%–90.7%; TBHQ: 93.8%–112%, but those in our method were EQ: 92.8%– 106.2%; TBHQ: 94.6%–107.2%. Obviously the recoveries of our method are better. The regression coefficient of EQ and TBHQ in this experiment was 0.9997 and 0.9990, respectively. This illustrates that our method has good linearity of EQ and TBHQ. The plasma concentration ranges of EQ and TBHQ are from 4.58 to 20.6 g mL1 and from 4.49
Chen Y, et al.
Sci China Chem
May (2013) Vol.56 No.5
669
Figure 3 Normalized excitation (a) and emission (b) profiles, including the resolved fluorescence spectra and the actual fluorescence spectra of EQ and TBHQ, and relative concentration (c) profile, which were obtained from urine samples by ANWE method with four factors. Table 3 Results obtained when applying ANWE algorithm analysis to urine samples spiked with ethoxyquin (EQ) and tert-butylhydroquinone (TBHQ) EQ (g mL1) actual predicted a) PU17 0.0687 0.0730(106.2) PU18 0.114 0.113(99.1) PU19 0.0916 0.0914(99.8) PU20 0.206 0.205(99.5) PU21 0.160 0.155(96.9) PU22 0.137 0.134(97.8) Average recovery (%) 99.9±2.0 Sample
TBHQ (g mL1) actual predicted a) 0.0673 0.0687(102.1) 0.0898 0.0919(102.3) 0.112 0.110(98.2) 0.157 0.157(100.0) 0.202 0.202(100.0) 0.224 0.224(100.0) 100.9±1.3
a) Recovery in parentheses.
to 20.2 g mL1, which are from 6.87 to 20.6 g mL1 and from 6.73 to 22.4 g mL1 in urine, respectively. To the best of our knowledge, those concentration ranges meet the requirements of the food testing. As outlined before, under plasma and urine conditions, the prediction results are desirable. It is show that EEM fluorescence spectroscopy coupled with second-order calibration is an effective and relia-
ble method. And in the context of three-way fluorescence analysis, the ANWE algorithm is being successfully used for data processing, because it achieves decomposition of three-dimensional arrays in a unique manner, allowing relative concentrations and spectral profiles of individual sample components to be extracted directly. The term second-order advantage has been coined to describe this property, which holds an immense potentiality in the analysis of complex samples. 3.3 Figures of merit The study based on ANWE calibration method also furnishes interesting figures of merit. The recoveries in predicted concentrations have been reported in Table 2 and Table 3. Other analytical figures of merit, including sensitivity (SEN) [36], selectivity (SEL) [37], limit of detection (LOD) and limit of quantification (LOQ) [38, 39], are collected in Table 4. The LOD values gained from ANWE are very low, which is enough to study EQ and TBHQ’s mechanism of toxicity. It is shown that the overall prediction
670 Table 4
Chen Y, et al.
Sci China Chem
1
Analytical figures of merit in both of plasma and urine
Figure of merit SEN (mL g1) SEL LOD (ng mL1) LOQ (ng mL1)
Plasma EQ TBHQ 23196.0 1889.8 0.8 0.2 0.2 4.5 0.6 13.7
May (2013) Vol.56 No.5
Urine EQ 12042.0 0.4 1.3 4.1
TBHQ 1453.9 0.2 13.8 41.7
abilities of them are satisfactory, whereas the figures of merit of EQ are better than TBHQ. The reason may be that the fluorescence signal intensity of background is more interference with TBHQ compared with EQ. As can be seen, the second-order calibration method based on ANWE can yield desirable outcomes for the simultaneous measurement of EQ and TBHQ in the biological fluid samples. Converting the limits of detection shown in Table 4 to plasma levels imply values of 20 ng mL1 for EQ, and 450 ng mL1 for TBHQ, but those in urine are 130 ng mL1 for EQ, and 4170 ng mL1 for TBHQ. The converted limits of detection for EQ and TBHQ in plasma or urine are also very low. In a word, the results should be deemed satisfactory in view of the complex matrices. The figures of merit once again demonstrated that EEM fluorescence spectroscopy coupled with ANWE algorithm is an effective and inexpensive method for the qualitative and quantitative characterization of EQ and TBHQ in biological fluid samples.
4 Conclusions A new analytical method for simultaneous determination of EQ and TBHQ in biological fluid samples was developed in this paper. We used the inserting zeros below the first-order Rayleigh scattering to handle the excitation-emission matrix fluorescence data first. And then, the satisfactory results were obtained by the alternating normalization-weighed error (ANWE) method. The results indicated that the excitation-emission matrix fluorescence data coupled with second-order calibration method based on ANWE algorithm was a useful and practical method for the qualitative and quantitative characterization of EQ and TBHQ in biological fluid samples due to its celerity, efficiency, reproducibility, small sample volume and ease of pretreating the complex matrix. The second-order advantage was adequately exploited. The second-order calibration method may help the chemists to resolve the complex matrices and seek out the spectra of individual interest components with the simple pretreatment, even without any physical or chemical separation. And the present work may provide a new avenue for the deeper investigation on the toxicity of EQ and TBHQ in human body. The authors gratefully acknowledge the National Natural Science Foundation of China (21175041) and the National Basic Research Program (2012CB910602) for financial support.
2
3
4
5 6
7
8
9 10
11 12
13
14
15
16
17
18 19
20
21 22
23
Karovičová J, Šimko P. Determination of synthetic phenolic antioxidants in food by high-performance liquid chromatography. J Chromatogr A, 2000, 882(1-2): 271–281 Yankah VV, Ushio H, Ohshima T, Koizumi C. Quantitative determination of butylated hydroxyanisole, butylated hydroxytoluene, and tert-butyl hydroquinone in oils, foods, and biological fluids by high-performance liquid chromatography with fluorometric detection. Lipids, 1998, 33(11): 1139–1145 Blaszczyk A, Skolimowski J. Apoptosis and cytotoxicity caused by ethoxyquin salts in human lymphocytes in vitro. Food Chem, 2007, 105(3): 1159–1163 Dolatabadi JEN, Kashanian S. A review on DNA interaction with synthetic phenolic food additives. Food Res Int, 2010, 43(5): 1223– 1230 Melton SL. Methodology for following lipid oxidation in muscle foods. Food Technol, 1983, 37(7): 105–111 Bohne VJB, Lundebye AK, Hamre K. Accumulation and depuration of the synthetic antioxidant ethoxyquin in the muscle of Atlantic salmon. Food Chem Toxicol, 2008, 46(5): 1834–1843 Błaszczyk A, Skolimowski J. Apootosis and cytotoxicity caused by ethoxyquin and two of its salts. Cell Mol Biol Lett, 2005, 10(1): 15–21 Alanko K, Jolanki R, Estlander T, Kanerva L. Occupational “multivitamin allergy” caused by the antioxidant ethoxyquin. Contact Dermatitis, 1998, 39(5): 263–264 Dzanis DA. Safety of ethoxyquin in dog foods. J Nutr, 1991, 121(Suppl. 11): S163–S164 Rubel DM, Freeman S. Allergic contact dermatitis to ethoxyquin in a farmer handling chicken feeds. Australas J Dermatol, 1998, 39(2): 89–91 Blaszczyk A. DNA damage induced by ethoxyquin in human peripheral lymphocytes. Toxicol Lett, 2006, 163(1): 77–83 Manson M, Green J, Driver H. Ethoxyquin alone induces preneoplastic changes in rat kidney whilst preventing induction of such lesions in liver by aflatoxin B1. Carcinog, 1987, 8(5): 723–728 Hernández M, Reyes J, Gómez-Lojero C, Sayavedra M, Meléndez E. Inhibition of the renal uptake of paminohippurate and tetraethylammonium by the antioxidant ethoxyquin in the rat. Food Chem Toxicol, 1993, 31(5): 363–367 Reyes JL, Hernández ME, Meléndez E, Gómez-Lojero C. Inhibitory effect of the antioxidant ethoxyquin on electron transport in the mitochondrial respiratory chain. Biochem Pharmacol, 1995, 49(3): 283– 289 Okubo T, Nagai F, Ushiyama K, Kano I. Contribution of oxygen radicals to DNA cleavage by quinone compounds derived from phenolic antioxidants, tert-butylhydroquinone and 2,5-di-tert-butylhydroquinone. Toxicol Lett, 1997, 90(1): 11–18 Nagai F, Okubo T, Ushiyama K, Satoh K, Kano I. Formation of hydroxydeoxyguanosine in calf thymus DNA treated with tertbutylhydroquinone, a major metabolite of butylated hydroxyanisole. Toxicol Lett, 1996, 89(2): 163–167 Ding T, Shen CY, Jiang Y, Yao YG, Tao J, Zhao HM, Shen WJ, Xu JZ, Yao RR. Determination of Ethoxyquin Residues in Fruits by HPLC-MS/MS. J Chin Mass Spectrom Soc, 2009, 30(5): 307–310 Lu YJ, Cheng J, Jiang JS. Determination of Tertiary Butylhydroquinone in Foods by HPLC. Mod Sci Instrum, 2008, 5: 58–60 André C, Castanheira I, Cruz J, Paseiro P, Sanches-Silva A. Analytical strategies to evaluate antioxidants in food: a review. Trends Food Sci Technol, 2010, 21(5): 229–246 Noel Robledo S, Alicia Zón M, Daniel Ceballos C, Fernández H. Qualitative and quantitative electroanalysis of synthetic phenolic antioxidant mixtures in edible oils based on their acid-base properties. Food Chem, 2011, 127(3): 1361–1369 Hao PP, Sun WL, Huang W, Ni JR. Metabolic pathways of tertiary butylhydroquinone in rats. J Toxicol, 2007, 21(1): 30–32 Guan YQ, Chu QC, Fu L, Ye JN. Determination of phenolic antioxidants by micellar electrokinetic capillary chromatography with electrochemical detection. Food Chem, 2006, 94(1): 157–162. Yang MH, Lin HJ, Choong YM. A rapid gas chromatographic meth-
Chen Y, et al.
24
25
26
27 28 29
30
31
Sci China Chem
od for direct determination of BHA, BHT and TBHQ in edible oils and fats. Food Res Int, 2002, 35(7): 627–633 Guo L, Xie MY, Yan AP, Wan YQ, Wu YM. Simultaneous determination of five synthetic antioxidants in edible vegetable oil by GC-MS. Anal Bioanal Chem, 2006, 386(6): 1881–1887 Huang W, Gu YC, Niu H. Determination of tertiary-butylhydroquinone and its metabolites in rat serum by liquid chromatographyion trap mass spectrometry. Lipids, 2008, 43(3): 281–288 Kmellar B, Pareja L, Ferrer C, Fodor P, Fernández-Alba AR. Study of the effects of operational parameters on multiresidue pesticide analysis by LC-MS/MS. Talanta, 2011, 84(2): 262–273 Kowalski BR, Seasholtz MB. Recent developments in multivariate calibration. J Chemom, 1991, 5(3): 129–145 Booksh KS, Kowalski BR. Theory of analytical chemistry. Anal Chem, 1994, 66(15): 782A–791A Xia AL, Wu HL, Zhu SH, Han QJ, Zhang Y, Yu RQ. Determination of psoralen in human plasma using excitation-emission matrix fluorescence coupled to second-order calibration. Anal Sci, 2008, 24(9): 1171–1176 Thygesen LG, Rinnan A, Barsberg S, Moller JKS. Stabilizing the PARAFAC decomposition of fluorescence spectra by insertion of zeros outside the data area. Chemom Intell Lab Syst, 2004, 71(2): 97–106 Li YN, Wu HL, Nie JF, Li SF, Yu YJ, Zhang SR, Yu RQ. Interference-free determination of abscisic acid and gibberellin in plant samples using excitation-emission matrix fluorescence based on oxidation derivatization coupled with second-order calibration
May (2013) Vol.56 No.5
32
33
34
35
36
37
38
39
671
methods. Anal Methods, 2009, 1: 115–122 Wang XM, Wu HL, Nie JF, Li YN, Yu YJ, Yu RQ. Rapid determination of thiabendazole in orange extract using excitationemission matrix fluorescence and second-order calibration based on alternating trilinear decomposition/alternating normalization-weighted error algorithms. Sci China Chem, 2008, 51(8): 729–735 Harshman RA. Foundations of the PARAFAC procedure: Models and conditions for an explanatory multimodal factor analysis. UCLA Working Papers in Phonetics, 1970, 16: 1–84 Carroll JD, Chang JJ. Analysis of individual differences in multidimensional scaling via an n-way generalization of ‘Eckart-Young’ decomposition. Psychometrika, 1970, 35(3): 283–319 Bro R, Kiers HAL. A new efficient method for determining the number of components in PARAFAC models. J Chemom, 2003, 17(5): 274–286 Olivieri AC, Faber NKM. A closed-form expression for computing the sensitivity in second-order bilinear calibration. J Chemom, 2005, 19(11-12): 583–592 Olivieri AC. Computing sensitivity and selectivity in parallel factor analysis and related multiway techniques: the need for further developments in net analyte signal theory. Anal Chem, 2005, 77(15): 4936–4946 Boqué R, Ferré J, Faber NKM, Rius FX. Limit of detection estimator for second-order bilinear calibration. Anal Chim Acta, 2002, 451(2): 313–321 Lorber A. Error propagation and figures of merit for quantification by solving matrix equations. Anal Chem, 1986, 58(6): 1167–1172
