A novel and enhanced approach for the assessment of the total carotenoid content of foods based on multipoint spectroscopic measurements

Depósito de investigación de la Universidad de Sevilla https://idus.us.es/ “This is an Accepted Manuscript of an article published by Elsevier in Food Chemistry on 15 June 2011, available at: https://doi.org/10.1016/j.foodchem.2010.11.167 .” 1 1 A NOVEL AND ENHANCED APPROACH FOR THE ASSESSMENT OF THE 2 TOTAL CAROTENOID CONTENT OF FOODS BASED ON MULTIPOINT 3 SPECTROSCOPIC MEASUREMENTS 4 5 MELÉNDEZ-MARTÍNEZ, A.J.a; AYALA, F.b; ECHÁVARRI, J.F.b.; NEGUERUELA, 6 A.I.c; ESCUDERO-GILETE, M.L.a; GONZÁLEZ-MIRET, M.L.a; VICARIO, I.M.a; 7 HEREDIA, F.J.* a 8 9 10 a Food Colour & Quality Lab., Dept. Nutrition & Food Science. Universidad de Sevilla. 11 Facultad de Farmacia. 41012-Sevilla, Spain 12 b Lab. of Colour La Rioja . Universidad de La Rioja. 26006-Logroño. Spain 13 c Dept. Applied Physics. Universidad de Zaragoza. Facultad de Veterinaria. 50013- 14 Zaragoza. Spain 15 16 17 18 * Author for correspondence: 19 Francisco J. Heredia 20 Food Colour & Quality Lab., Dept. Nutrition & Food Science. Universidad de Sevilla. 21 Facultad de Farmacia. 41012-Sevilla, Spain 22 Tel.: +34 954556495 Fax: +34 954557017 23 e-mail: [email protected] 24 2 25 ABSTRACT 26 We have devised a more sensible approach to estimate the carotenoid content of 27 orange juices, which can be regarded as a model system of food with intricate 28 carotenoid pattern. For this purpose spectroscopic information at several wavelengths 29 and spectra of the juices and not from their carotenoid extracts were considered, such 30 that more accurate and rapid quantitative assessments can be achieved. The wavelengths 31 proposed on the basis of the characteristic vector method were 420, 455, 515, 545 and 32 610 nm or 420, 445, 510, 545 and 605, depending on the measurement conditions. The 33 correlations between the carotenoid content and the reflectances at these wavelengths 34 were very good (R = 0.94 and 0.90, respectively). Additionally, it was demonstrated 35 that the colour of the juices could be assessed with very good accuracy considering 36 them. Due to its simplicity and rapidity, this method is intended to facilitate the quality 37 control of the carotenoid content of foodstuffs in the industry and/or in the field. 38 39 KEYWORDS: carotenoids; characteristic vector; citrus; colour; foods; orange juice; 40 quality control; tristimulus colorimetry 41 3 42 INTRODUCTION 43 The carotenoids are isoprenoid compounds biosynthesized by plants, fungi, 44 algae and bacteria that animals have to incorporate through the diet (Fraser & Bramley, 45 2004). They are natural pigments that provide colour to many structures (flowers, fruits, 46 seeds, tubers, roots, feathers, egg yolk, etc.) and therefore play key roles in the 47 communication between animals, the pollination, the dispersal of seeds, and the 48 acceptability of foods, among others. Pertaining to the involvement of carotenoids in 49 acceptability of foods it is important to consider that some aroma compounds (like 50 - -damascenone, etc.) are cleavage products of carotenoids 51 (Baldermann, Naim & Fleischmann, 2005; Carmona, Zalacain, Salinas & Alonso, 2006; 52 Silva Ferreira, Monteiro, Oliveira & Guedes de Pinho, 2008; Fleischmann & Zorn, 53 2008). Nevertheless, they also play essential roles in the photosynthesis, contributing to 54 the harvesting of light and, what it is more important, protecting the photosynthetic 55 organisms from the deleterious effects derived from photooxidation (Telfer, Pascal & 56 Gall, 2008). Additionally they are beneficial for animals and not only because some act 57 as vitamin A precursors. In this regard, it is well-known that some carotenoids exhibit 58 antioxidant activity (Burton, 1989; Demmig-Adams & Adams III, 2002; Astley, 59 Hughes, Wright, Elliott & Southon, 2004) and evidence is accumulating that they can be 60 beneficial for the prevention of serious human disorders (cancer, cardiovascular disease, 61 age-related macular degeneration, etc.) (Krinsky, 1989; Olson, 1999; Giovanucci, 62 2002).(Voutilainen, Nurmi, Mursu & Rissanen, 2006; Trumbo & Ellwood, 2006; 63 Lidebjer, Leanderson, Ernerudh & Jonasson, 2007). 64 Due to all the facts enumerated above, the need to assess the carotenoids 65 occurring in foodstuffs is indisputable. Although scientists can choose from a wide 66 choice of HPLC methods to determine them, more rapid methods are needed in some 4 67 situations, such as quality control in the industry and/or the field. In this regard, it is 68 important to note that the chromatographic analysis of carotenoids involves typically 69 several steps (extraction, saponification, chromatographic analysis), so they are not 70 appropriate for almost real-time assessments. Even though the estimation of the total 71 carotenoid content of a source by spectrophotometry is still commonplace in some 72 laboratories, such procedure presents drawbacks. The most important one is without any 73 doubt that the accuracy of these measurements is far from desirable, since the extinction 74 coefficients for carotenoids are not as accurate as they should be and sometimes even 75 arbitrary coefficients are used (Britton & Young, 1993). On the other hand, only the 76 spectroscopic information at one point is considered, despite the fact that the visible 77 spectra of most carotenoids have three maxima within an interval of some 40 nm. 78 Additionally, the pigments have to be extracted from the food, which inevitably requires 79 some time and poses a drawback if the levels of carotenoids have to be determined on 80 site. 81 In this paper we describe a novel approach that simplifies and improves the 82 determination of the total carotenoid content of foods by considering spectroscopic 83 information of the food itself (not from an extract) at several wavelengths. Additionally, 84 this procedure allows to accurately assess the colour of the product, which is one of the 85 main attributes related to its acceptability. 86 87 EXPERIMENTAL PROCEDURES 88 Samples 89 Eighty-two commercially available orange juices (31 ultrafrozen and 51 90 thermally-treated juices) were analyzed. The samples were selected such that they 91 covered all the possible colours of orange juice that can be found in the Spanish market. 5 92 The thermally-treated samples were purchased from retailers based in Seville (Spain) 93 and the deep-frozen ones were kindly provided by Zumos Vitafresh. 94 recommendations concerning the storage of the orange juices were followed until their 95 analysis. The deep-frozen orange juices were thawed at ambient temperature. 96 The samples were considered independently in order to obtain the maximum 97 possible variability and make the correlations between colour or spectroscopic 98 information and pigments more robust since the set of samples varied widely in these 99 100 parameters. 101 Carotenoid analysis 102 The analysis of carotenoids was carried out according to the routine protocol 103 used in our laboratory. This methodology is fully explained elsewhere (Meléndez- 104 Martínez, Vicario & Heredia, 2007a). In brief the carotenoids were extracted with a 105 mixture hexane/methanol/acetone (50:25:25, v/v/v, containing 0.1% butylated 106 hydroxytoluene) and subsequently subjected for 1 hour to an alkaline hydrolysis with 107 ethanolic KOH (10% w/v) under dim light and at room temperature. After washing four 108 times with water the coloured extracts were dried out on a rotary evaporator and re- 109 dissolved in 1 ml of acetone:methanol (1:2, v/v, containing 0.1% butylated 110 hydroxytoluene) for the HPLC analyses. 111 The HPLC analysis were performed on an Agilent 1100 system fitted with a 112 quaternary pump, a column temperature control module, a PDA detector and an 113 autosampler l from the vials (Agilent, Palo Alto, CA, United States). 114 The carotenoids were separated on an YMC C30 column (250 × 4.6 mm ) (YMC, 115 Wilmington, NC, USA). Water, methanol (MeOH) and methyl-ter-butyl ether (MTBE) 6 116 were used in the mobile phase according to a gradient elution. The flow rate was 117 1ml/min and the peaks were monitored at 430 nm. 118 Most of the carotenoids detected were identified by comparison of their UV/Vis 119 spectroscopic and chromatographic characteristics with those of reference standards 120 isolated from appropriate sources or semi-synthesized. These standards were 121 stereomutated to help identify geometrical isomers. More detailed information 122 concerning the identification of the carotenoids occurring in orange juices is provided in 123 previous papers. 124 The absolute concentrations of the pigments were worked out by external 125 calibration and the total contents of carotenoids as the sum of the content of the 126 individual pigments. 127 Colour measurement 128 A more detailed description of the methodology followed for the objective 129 colour measurements can be found anywhere else (Meléndez-Martínez, Vicario & 130 Heredia, 2004). In brief, visible reflectance measurements (380-770 nm) were recorded 131 at 1 nm intervals on a CAS 140 B spectroradiometer (Instrument Systems, Munich, 132 Germany) equipped with a Tamron zoom mod. SP 23A (Tamron USA, Inc., Commack, 133 NY, USA), a Top 100 telescope optical probe (Instrument Systems, Munich, Germany) 134 and an external incandescent lamp. The apparatus takes three consecutive measurements 135 at 1 second intervals in order to compensate for possible variations in the conditions of 136 the measurements, such that the colour parameters obtained are averages of that 137 triplicate. 138 The within-laboratory repeatability (within-day precision) was developed 139 according to ISO-5725-2:1994 (ISO, 1994). It was ascertained by analyzing ten times 140 within the same day the colour of an orange juice. The repeatability is expressed as the 7 141 coefficient of variation (CV) or relative standard deviation (percentage of SD), being 142 lower than 3.6 % for all the CIELAB colorimetric variables (0.66% for L*, 3.57% for 143 a*, 2.40% for b*, 2.41% for C*ab, and 0.29% for hab). 144 The Illuminant D65 and the 10° Observer were taken as references and the colour 145 coordinates of the uniform colour space CIELAB were considered (CIE, 1978). 146 A transparent plastic cuvette was used for the measurements. The blank 147 measurements were made with the cuvette filled with distilled water and placed against 148 a reference BaSO4 pressed plate (USRS-99-010, Labsphere Inc. North Sutton, NH, 149 USA). The samples were measured against a white and a black background because it 150 has been reported that the colour parameters obtained by using one or another are 151 differently correlated with the visual perception of the colour and other parameters 152 (Meléndez-Martínez et al., 2004). 153 Mathematical method of the characteristic vector method 154 The analysis of the characteristic vectors has been carried out following the 155 adaptation of the method made by Simonds (Simonds, 1963), using software developed 156 by Laboratorio de Color de la Rioja. 157 The fundamentals of the characteristic vector analysis (Lebart, Morineau & 158 Fenelon, 1985) can be summarized as follows: the response data (visible reflectance 159 spectrum) are available for r = 301 levels of the variable (wavelength) and can be 160 plotted as a curve. Consequently, for each sample the r values of make up a one-row 161 r-column vector of response data. For n samples the vectors can be laid out as a matrix 162 of r columns and n rows. In this regard, a set of p characteristic vectors (p < r), can be 163 found such that, when they are properly added to the mean response vector, they will 164 approximate to any of the primary response vectors. The outcome is that the 165 components of the generic vectors, that is, the reflectance spectra, can be expressed as: 8 780 380 780 380 780 380 780 380 780 380 M 1 V 1, M 2 V 2, ..... M p V p, Eq. [1] 166 In this equation, refers to the spectral reflectance of the sample, to the component 167 of the mean vector, V to the components of the characteristic ith vector and Mi to the 168 specific coefficients, named scalar multiples, of every reconstructed vector. 169 This analysis is to be applied to n experimental curves. The model computes the 170 Mi coefficients necessary for the reconstruction of each of them, as well as the 171 percentage of the variability among the family of homologous response curves 172 explained for each characteristic vector. The formulae to work out the tristimulus 173 values, in accordance to the 174 recommendations, would be: X 10 k L x Y 10 k L y Z 10 k L z Eq.[2] 175 where k is a normalizing factor, L the spectral emission of the reference illuminant, 176 x , y , z , the colour-matching functions corresponding to the observer consider as 177 reference, and the interval of the spectral measurements. 178 If Eq. [1] is substituted into Eq. [2], the following expressions would be obtained 179 for the tristimulus values (in this case X, although analogous expressions are obtained 180 for Y and Z): 181 1 8 1 X 10 k M1V1, M 2V2, M pVp, L x 182 Expressed otherwise: X10 k L x M1k V1, L x M p k Vp, L x Eq.[3] 183 as the Mi coefficients are not dependent on the wavelength for any single reconstructed 184 curve. 780 380 780 380 15 301 Likewise, the following equations were obtained for the measurements made with black 302 background: 303 304 X10 = 0.125891 + 6.09084 420 + 12.8171 445 + 2.57889 510 545 + 58.9423 305 605 Eq. [10] 306 307 Y10 = - 420 445 510 545 + 37.1951 308 605 Eq. [11] 309 310 Z10 420 445 510 - 545 + 0.0107494 311 605 Eq. [12] 312 313 In the light of the results obtained and considering Figure 1b it can be claimed 314 that the reflectance spectra of the orange juices studied can be reconstructed by 315 considering wavelengths at which the absorption of light is very high (420, 445 and 455 316 nm), intermediate (510, 515 and 545 nm) and low (605 and 610 nm). Since the original 317 reflectance spectra obtained for the juices were clearly different as a function of the 318 background used for the measurements, some of the wavelengths obtained were 319 different depending on the background used for the colour measurements. 320 The CIELAB colour coordinates corresponding to the 82 orange juice samples 321 were calculated from the tristimulus values obtained considering the equations 7-12. To 322 compare the colour coordinates computed from the original spectra (summarized in 323 Table 1) with the ones calculated from the reconstructed spectra their colour differences 324 were calculated. Notwithstanding that the CIEDE2000 colour difference formula is 325 recommended for industrial applications for small colour differences under reference 16 326 conditions (CIE, 2004), we opted for using the CIELAB colour difference formula 327 (CIE, 2004) because visual colour differences were not involved. 328 Considering the measurements made with white background, the highest 329 0.99 330 CIELAB units for a*, 1.69 units for b* and 0.17 units for L*. Regarding the 331 measurements made with black background, the highest differences were 0.6, 1.65, and 332 0.23 CIELAB units for a*, b* and L*, respectively. In this sense it was observed that the 333 highest colour differences ( Eab) were 1.70 and 1.60 CIELAB units for white and black 334 background, respectively. In any case, it was observed that the values of Eab were 335 lower than 0.5 CIELAB units in most of the cases (81% of the cases considering the 336 white background and 74% of the cases considering the black background) (Figure 2). 337 These results indicated that the reflectance spectra of diverse orange juices can be 338 3 3 8 339 3 3 9 reconstructed very accurately using as few as 5 wavelengths. 340 Relationships spectroscopic measurements carotenoid content 341 Even though the instrumental colour measurements by Tristimulus Colorimetry 342 cannot provide detailed information regarding the carotenoid profile of complex sources 343 of carotenoids they are very useful for other purposes. These measurements have 344 inherent advantages (ultra-rapidity, versatility, non-destructiveness, possibility of 345 automatization, portability, etc.) that make them very appropriate to obtain almost real- 346 time information and therefore amenable for quality control purposes. Thus, the 347 applicability of objective colour assessments to estimate the carotenoid content of 348 diverse foods has been the goal of several studies conducted in the last years (Mínguez- 349 Mosquera, Rejano-Navarro, Gandul-Rojas, Sánchez-Gómez & Garrido-Fernández, 17 350 1991; Meléndez-Martínez, Vicario & Heredia, 2003; Konopka, Kozirok & Rotkiewicz, 18 351 2004; Ruiz, Reich, Bureau, Renard & Audergon, 2008; Ornelas-Paz, Yahia & Gardea, 352 2008; Moyano, Meléndez-Martínez, Alba & Heredia, 2008a; Moyano, Meléndez- 353 Martínez, Alba & Heredia, 2008b). Likewise, they have proved to be useful for the 354 estimation of the vitamin A activity of diverse orange juices (Meléndez-Martínez, 355 Vicario & Heredia, 2007b). In any case, some of these methods required expensive 356 apparatus which are not always affordable for the quality control departments of the 357 industry or for producers. Likewise, the interpretation of the data involves being 358 familiar with the objective measurement of colour by tristimulus colorimetry, which is 359 not as intuitive for technicians as just taking absorbance/reflectance readings at only a 360 few wavelengths. In this sense we have proposed as few as five wavelengths from 361 which the whole visible spectra (380-770 nm) of orange juices can be reconstructed. 362 These wavelengths are expected to be appropriate for the rapid estimation of the total 363 carotenoid content of any orange juice, as the miscellaneous set of samples analyzed in 364 this study differed greatly in colour and in their carotenoid content (Table 1). Thus, 365 considering quantitative data, it was observed that the total carotenoid content of the 366 orange juices analyzed ranged roughly from 2 to 35 mg/l. Important qualitative changes 367 in the carotenoid profiles of the samples were also noticed. In fact, in the light of the 368 data we have reported in previous surveys (Melendez-Martinez et al., 2008) it can be 369 inferred that clear qualitative changes take place in the carotenoid pattern of orange 370 juices as a result of the treatments they undergo in the industry and the length and 371 conditions of storage. At this point it is also important to note that the carotenoid 372 content of oranges (and of all vegetable products) depends on other factors, like 373 genotype, stage of ripeness, the climatic conditions of the area of production, etc. 374 (Mouly, Gaydou, Lapierre & Corsetti, 1999; Lee & Castle, 2001), which adds further 375 variability to our set of samples. 19 376 To ascertain whether the wavelengths selected by the mathematical model could 377 be really used to estimate the carotenoid content of the diverse orange juice surveyed, 378 multiple regression analyses were carried out. For this purpose, the total contents of 379 carotenoids were considered as dependent variable and the reflectances at the selected 380 wavelengths as independent variables. The results are summarized in Table 2. The 381 regression coefficients were high regardless of the background used for the 382 measurements, although they were higher when the reflectances obtained for the white 383 background were considered (R = 0.94 and 0.90 for white and black measurements, 384 respectively). Likewise, high significance (p<0.01) correlations between the total 385 content of carotenoids and the reflectances at 515, 545, and 610 nm (for white 386 background) and at 420, 510, and 545 nm (for black background). These results 387 indicated that it is possible to obtain a good rapid quantitative assessment of the 388 carotenoid fraction of orange juices from the reflectances values obtained at 5 389 wavelengths. In this regard, the B coefficients that enable to work out the total 390 carotenoid content from the reflectances at those wavelengths were computed (Table 2). 391 When the correlations colour coordinates-carotenoid content were considered, 392 some interesting information was obtained. Taking into account the colour coordinates 393 computed from the original spectra it was observed that a* was the only coordinate 394 having high significance (p<0.001), for measurements made with the white and the 395 black backgrounds. When the set of scalar colour coordinates (L*, a*, b*) was included 396 in the multiple regression model to consider complete colorimetric information 397 (Meléndez-Martínez et al., 2007b), a* was also the only coordinate having high 398 significance. Analogous simple and multiple regression analyses were carried out 399 considering the L*, a* and b* coordinates obtained from the reconstructed spectra. The 400 results of these regression analyses were in agreement with those obtained taking into 20 401 account the colour parameters obtained from the original spectra (Table 3), which also 402 confirms the goodness of the 5 wavelength selected by the mathematical model. With 403 regard to the regression coefficients obtained when the total carotenoid content was 404 considered as dependent variable and the set of scalar coordinates as independent 405 variables, several interesting remarks can be made. On one hand, it was seen that the 406 coefficients obtained for the colour coordinates computed from the original spectra 407 (R=0.82 and R=0.81 for the white and the black backgrounds, respectively) were almost 408 identical to those obtained from the coordinates computed from the reconstructed 409 spectra (R=0.81 and R=0.81 for the white and the black backgrounds, respectively). 410 This observation is and additional proof of the goodness of the characteristic vector 411 method to simplify the colour measurement without losing much accuracy. On the other 412 hand, it is important to note that these coefficients are clearly lower than those obtained 413 when the correlations between the reflectances at the wavelengths selected and the 414 carotenoid content were evaluated (Table 2). These data seem to indicate that the 415 consideration of reflectance values at a number of key wavelengths may be more useful 416 to assess the pigment content of carotenoid-containing foods than their colour 417 coordinates. In any case, the frequency histograms of the differences between the total 418 carotenoid content values obtained by HPLC and those calculated from colour 419 coordinates and the selected wavelengths (Figure 3) indicate that both types of 420 4 2 0 421 4 2 1 parameters can be used to estimate with good accuracy the total pigment content. 422 ACKNOWLEDGMENTS 423 The authors thank the company Zumos Vitafresh (Almonte, southern Spain) for 424 4 2 4 425 425 21 sup plyi ng the dee pfroz en ora nge juic es anal yze d in this stud y. 22 426 REFERENCES 427 428 Astley, S. B., Hughes, D.A., Wright, A.J.A., Elliott, R.M., & Southon, S. (2004). DNA 429 damage and susceptibility to oxidative damage in lymphocytes: Effects of 430 carotenoids in vitro and in vivo. British Journal of Nutrition, 91(1), 53-61. 431 Ayala, F., Echávarri, J.F., Juárez, M.C., & Negueruela, A.I. (1993). Propuesta de un 432 nuevo método de medida del color de vinos tintos y rosados. Óptica Pura y 433 Aplicada, 26(2), 533-541. 434 Baldermann, S., Naim, M., & Fleischmann, P. (2005). Enzymatic carotenoid 435 degradation and aroma formation in nectarines (Prunus persica). Food Research 436 International, 38, 833-836. 437 Britton, G. (1983). The biochemistry of natural pigments. Cambridge, United Kingdom: 438 Cambridge University Press. 439 Britton, G. (1995). Structure and properties of carotenoids in relation to function. The 440 FASEB Journal, 9, 1551-1558. 441 Britton, G., & Young, A. (1993). Methods for isolation and analysis of carotenoids. In 442 A. Young & G. Britton, Carotenoids in photosynthesis (pp. 409-458). London: 443 Chapman & Hall. 444 Burton, G. W. (1989). Antioxidant action of carotenoids. Journal of Nutrition, 119, 109- 445 111. 446 Carmona, M., Zalacain, A., Salinas, M.R., & Alonso, G.L. (2006). Generation of saffron 447 volatiles by thermal carotenoid degradation. Journal of Agricultural and Food 448 Chemistry, 54, 6825-6834. 23 449 CIE (1978). Recommendations on Uniform Color Spaces, Color-Difference Equations, 450 Psychometric Color Terms, CIE Publication No. 15 (E-1.3.1) 1971, Supplement 451 2. Vienna: Bureau Central de la CIE. 452 CIE (2004). Technical Report Colorimetry. Vienna, Austria: Commission Internationale 453 de l'Eclairage Central Bureau. 454 Demmig-Adams, B., & Adams III, W.W. (2002). Antioxidants in photosynthesis and 455 human nutrition. Science, 298, 2149-2154. 456 Fleischmann, P., & Zorn, H. (2008). Enzymic pathways for formation of carotenoid 457 cleavage products. In G. Britton, S. Liaaen-Jensen, & H. Pfander, Carotenoids. 458 Volume 4: Natural Functions (pp. 341-366). Basel, Switzerland: Birkhäuser. 459 Fraser, P. D., & Bramley, P.M. (2004). The biosynthesis and nutritional uses of 460 carotenoids. Progress in Lipid Research, 43, 228-265. 461 García-Parrilla, M. C., Ayala, F., Echávarri, J.F., Troncoso, A.Mª., Negueruela, A.I., & 462 Heredia, F.J. (1998). Measurement of wine vinegars' color: application of the 463 characteristic vector method. Journal of Agricultural and Food Chemistry, 46, 464 4238-4241. 465 Giovanucci, E. (2002). Lycopene and prostate cancer risk. Methodological 466 considerations in the epidemiologic literature. Pure and Applied Chemistry, 467 74(8), 1427-1434. 468 ISO (1994) Accuracy (trueness and precision) of measurement methods and results Part 469 2: Basic method for the determination of repeatability and reproducibility of a 470 standard measurement method. ISO-5725-2:1994 24 471 Kön, S., Kolbe, H., Korger, M., Köpsel, C., Mayer, B., Auweter, H., Lüddecke, E., 472 Bettermann, H., & Martin, H.-D. (2008). Aggregation and interface behaviour of 473 carotenoids. In G. Britton, S. Liaaen-Jensen, & H. Pfander, Carotenoids. Volume 474 4: Natural functions (pp. 53-98). Basel, Switzerland: Birkhäuser. 475 Konopka, I., Kozirok, W., & Rotkiewicz, D. (2004). Lipids and carotenoids of wheat 476 grain and flour and attempt of correlating them with digital image analysis of 477 kernel surface and cross-sections. Food Research International, 37, 429-438. 478 Krinsky, N. I. (1989). Carotenoids and cancer in animal models. Journal of Nutrition, 479 119, 123-126. 480 Lebart, L. A., Morineau, A., & Fenelon, J.P. (1985). Tratamiento estadístico. 481 Barcelona, Spain: Marcombo. 482 Lee, H. S., & Castle, W.S. (2001). Seasonal changes of carotenoid pigments and color 483 in Hamlin, Earlygold, and Budd Blood orange juices. Journal of Agricultural 484 and Food Chemistry, 49, 877-882. 485 Lidebjer, C., Leanderson, P., Ernerudh, J., & Jonasson, L. (2007). Low plasma levels of 486 oxygenated carotenoids in patients with coronary artery disease. Nutrition, 487 Metabolism and Cardiovascular Diseases, 17(6), 448-456. 488 Marx, M., Schieber, A., & Carle, R. (2000). Quantitative determination of carotene 489 stereoisomers in carrot juices and vitamin supplemented (ATBC) drinks. Food 490 Chemistry, 70, 403-408. Table 3. Summary of the multiple regression analyses carried out to evaluate the correlations between the total carotenoid content and the colour coordinates obtained from the original and the reconstructed spectra From original spectra From reconstructed spectra Background Variables B coefficient Std. error of estimate B coefficient Std. error of estimate White L* 0.1271 a* 2.1603** 5.6257 0.07059 2.07541** 5.6932 Black a* 2.524167** 5.7294 2.48255** 5.8057 * Significant at p<0.01; ** significant at p<0.001 b* -0.1072 -0.08703* Intercept -12.6542 -8.81539 L* -0.021504 -0.05097 b* -0.161832* -0.14881 Intercept 9.030370 10.44026 Figure 1. Typical visible spectra: a. Absorption spectrum of an orange juice carotenoid extract 400 450 500 550 600 650 700 750 b. Reflection spectra of an orange juice obtained with white and black backgrounds 400 450 500 550 600 650 700 750 70 60 50 40 30 20 10 0 Eab < 0.1 0.1-0.5 0.5 - 1.0 1.0 - 1.5 1.5 - 2.0 > 2.0 60 Eab 50 40 30 20 10 0 < 0.1 0.1-0.5 0.5 - 1.0 1.0 - 1.5 1.5 - 2.0 > 2.0 Figure 2. Frequency histogram of colour differences corresponding to original and reconstructed spectra obtained for white (upper figure) and black background (lower figure) Figure 3. Frequency histogram of carotenoid content differences between original data and reconstructed spectra data obtained for white and black background. (ct: Total carotenoid content; cr: Carotenoid content calculated from reconstructed spectra): a) carotenoid content calculated with L*a*b* regression equation; b) carotenoid content calculated with reflectances regression equation. a) 30 28 26 24 22 20 18 16 14 12 10 8 6 4 2 0 2 4 6 8 10 12 14 16 18 b) 36 34 32 30 28 26 24 22 20 18 16 14 12 10 8 6 4 2 0 2 4 6 8 10 12 14 16 18 20