Is authentication of the geographic origin of poultry meat and dried beef improved by combining multiple trace element and oxygen isotope analysis?
B.M. Franke a,b,c, R. Hadorn b, J.O. Bosset b, G. Gremaud c, M. Kreuzer a,∗
a ETH Zurich, Institute of Animal Science, Universitaetstrasse 2, 8092 Zurich, Switzerland
b Agroscope Liebefeld-Posieux Research Station, CH-3003 Berne-Liebefeld, Switzerland
c Swiss Federal Office of Public Health, CH-3003 Berne-Liebefeld, Switzerland
Abstract
Data available on contents of up to 72 different trace elements and the oxygen isotope ratio of 78 poultry breast and 71 dried beef samples were analysed to determine whether the accuracy of the prediction of the geographic origin is improved by combining promising methods. Validation was performed by determining the origin of a smaller sub-group by using a statistical model established from the data of the second, larger, sub-group. As expected, also the combined data proved useful for the determination of the geographic origin of meat samples. However, combining data did not clearly reduce the percentage of incorrectly classified individual samples compared to the two approaches applied separately. In poultry, cross-validation and validation resulted in 83 and 50% correct classifications, respectively. The corresponding values in dried beef were 73 and 43%. In conclusion, compared to element signature data alone, combining both methods did not improve predictions of origin.
Keywords: Poultry; Dried beef; Geographic origin; Authenticity
1. Introduction
Legislation in the EU and in Switzerland demands a better traceability of food. Therefore, also meat and meat products have to be labeled with their country of origin (2000/13/EG;
Eidgenössisches Departement des Inneren, 2005). The development of appropriate tools to authenticate these origins independent from paper traceability increases consumers’
safety and contributes to the protection of producers against potential frauds and misrepresentations as well.
In the last years several studies were conducted with the aim to determine the geographic origin of meat (Hegerding, Seidler, Danneel, Gessler & Nowak, 2002; Piasentier, Valusso, Camin & Versini, 2003; Boner & Förstel, 2004; Renou, Bielicki, Deponge, Gachon, Micol &
Ritz, 2004; Shintu, Caldarelli & Franke, 2007). In a larger research programme, various analytical tools initially considered to be of high potential for this purpose (Franke, Gremaud, Hadorn & Kreuzer, 2005) have been tested. Especially various minerals and trace elements (Franke, Haldimann, Gremaud, Bosset, Hadorn & Kreuzer, 2007a; Franke et al., 2007b) and the oxygen isotope ratio (δ18O) (Franke et al., 2008) have been shown to be quite useful for geographic authentication of poultry meat and dried beef. Still classifications to countries of origin of the raw meat or country of processing (dried beef) is
∗ Corresponding author. Tel.: +41 44 632 5972; fax: +41 44 632 1128; E-mail address:
not possible in a totally correct way, especially in case of nearby origins and when meat is processed at the same plant. Characteristic profiles of elements may be related to the environment of the animals, e.g through specific geologic profiles or through (anthropogenic) pollution (Franke et al., 2007a) while distance from the ocean and other external factors are responsible for variations in δ18O (Boner & Förstel, 2004; Franke et al., 2008). The applicability of both methods is limited, either by incontrollable feed supplementation with, sometimes essential, elements or when processing such as drying modifies the isotope ratio. Initially present differences in oxygen isotope ratio between European and South American meat (Boner & Förstel, 2001, 2004; Förstel & Lickfett, 2002) thus may be masked (Franke et al., 2008).
The power to discriminate between samples of different geographic origin might be increased by combining techniques based on different principles and the limitations outlined above could be overcome. In the present study, therefore, available data of elements (Franke et al., 2007a, 2007b) and δ18O (Franke et al., 2008) were combined for overall statistical analyses comparing the accuracy of geographic authentication with that found when applying the methods independently.
2. Materials and methods 2.1. Samples
A total of 78 frozen poultry breasts had been obtained from Brazil (n = 14), France (n = 13), Germany (n = 15), Hungary (n = 16), and Switzerland (n = 20) between February 2004 and December 2005. This basically included two sample sets, the first comprising 22 samples from these countries (Phase I), the second 56 samples (Phase II). The authenticity of all samples had been certified with valid custom documents, specifying place and date of slaughter. The samples were vacuum-sealed in vapour-impermeable plastic bags and kept frozen at -25 °C until being analysed.
Concerning the dried beef, totally 74 samples (thereof 21 being collected in Phase I), having been prepared either from M. biceps femoris or from M. semitendinosus, were either directly collected from the production places (samples processed in Switzerland) or purchased from producers in Australia (n = 8), Austria (n = 6), Canada (n = 8), and USA (n = 5) between May 2004 and February 2006. The Austrian samples had been produced from Brazilian raw meat, whereas all other non-Swiss samples had raw meat from the country of processing as its origin. Part of the Swiss samples had been produced in the Swiss canton of Valais using Swiss raw meat (n = 15) and the other part in the canton of Grisons using Swiss (n = 16) and Brazilian raw meat (n = 16). All dried beef samples were produced by curing and various sequences of air drying and pressing. Former studies (Franke et al., 2007) showed final dry matter contents of 46±5 %, equivalent to about 40 % water loss. Slight variations in recipes (e.g. amount of salt, kind of herbs) and technology (salt application, curing, drying, pressing, use of moulds, storing, packaging, etc.) may have occurred within the same type of product, depending on the producer. The samples were vacuum-sealed in vapour-impermeable plastic bags and kept at -5 °C until being analysed.
The individual poultry and dried beef samples were independent from each other by being either obtained from different producers or obtained from different production batches when originating from the same producer. No detailed information on conditions of animal fattening, meat handling and processing were available, as would be also the case in potential future routine control of declared geographic origins.
2.2. Chemical analysis
Samples had been analysed for δ18O as described in detail by Franke et al (2008) and for element content as outlined in Franke et al. (2007a, 2007b). Briefly, water for oxygen isotope analysis, extracted from approx. 10 g of meat, had been obtained, using still frozen samples to avoid drip loss and thus possible shifts in δ18O, by azeotropic distillation with toluene (130 °C, 18 h) using the Bidwell-Sterling apparatus being built as outlined by Schäfer (1967). In these water samples, the oxygen isotope ratio was determined with the help of an IRMS (Delta-Plus XL, Finnigan, Bremen, Germany). For element analysis, samples of dried beef (0.4 to 0.5 g) and poultry (0.8 to 1 g) were prepared by micro-wave assisted digestion with nitric acid under pressure. The digested material was analysed using a sector field ICP-MS (Element 2, Finnigan MAT, Bremen, Germany) using a CertiPUR® Rhodium ICP Standard (Merck, Darmstadt, Germany).
2.3. Statistical analysis
Poultry samples were grouped according to their geographic origin, while dried beef samples were grouped considering both raw meat origin and place of processing. The results of the reference material, analysed for elements in both sampling phases, were compared using the Wilcoxon-Mann-Whitney test (probability > 0.01) to be able to exclude all those elements from statistical analysis where concentrations varied with time (cf.
Franke et al., 2007a). Also elements with results remaining below the detection limit in at least one of the two sampling phases were excluded from further statistical analysis.
Analysis of variance (ANOVA) was performed with each single variable and, in case of significance (p < 0.05), Bonferroni-adjusted pairwise comparison was performed in order to identify the origins being separated from the others by individual variables. Different from Franke et al. (2007a), this was done on the combined dataset of both sampling phases. For δ18O, results were taken from Franke et al. (2008). Additionally, δ18O and element analyses were combined and Linear Discriminant Analysis (LDA) in stepwise backward elimination (probability f to enter / to remove = 0.15) was performed. By this way, (i) cross-validation, jackknifed-type, classification matrices were built using all samples (cross-validation) and (ii) the origin of the samples from Phase 1 was predicted using the data of Phase 2 samples (validation). All statistical analyses were performed by Systat (version 11, Systat Software Inc, Richmond, California, USA).
3. Results and discussion 3.1. General
According to the results of the comparison based on the Wilcoxon-Mann-Whitney, the elements 45Sc, 53Cr, 67Zn, 68Zn, 104Pd, 128Te, 141Pr, 151Eu, 153Eu, 161Dy, 163Dy, 169Tm, 171Yb,
172Yb, 203Tl, 209Bi, 238U were excluded. Reasons for that are described in Franke et al.
(2007a). These elements are mostly rare in nature and of low concentration in meat.
3.2. Poultry meat
In the univariate mode of analysis, poultry meat groups from any country of origin could be distinguished from that of other countries by using at least one element (Table 1).
that case just 75As was useful. Possible reasons for the discriminating power of the elements As, Rb, Sr and Tl found to be significantly different among countries have been discussed earlier (Franke et al., 2007b).
Table 1
Univariate differentiation of poultry breast meat of various originsa
Country Brazil France Germany Hungary Switzerland Variable
n 14 13 15 16 20
Brazil
—
X X X 0
X 0 X X
X X 0 X
X X 0 X
85Rb
86,88Sr
205Tl δ18O France X
X
0 —
X 0 X
0 X X
0 X X
86,88Sr
205Tl δ18O Germany 0
X X
X 0
X —
0 X X
0 X X
86,88Sr
205Tl δ18O Hungary X
0 X X
X X 0 X
X X 0 X
—
X 0 0 0
75As
205Tl
86,88Sr δ18O Switzerland 0
X X
X 0 X
X 0 X
0 0
0 —
205Tl
86, 88Sr δ18O
a X = differentiation possible; 0 = no differentiation possible. Results for δ18O taken from Franke et al. (2008).
The cross-validation classification matrix gave a mean classification rate of 83 %, based on the elements 82Se, 85Rb, 88Sr, 155Gd, 205Tl, 23Na, 51V, and on δ18O (Table 2). Just samples from Brazil could be classified completely correct, but one French sample was misclassified as being Brazilian. German samples were classified as to be correct with the exception of one sample which was classified as being French. Between Hungarian and Swiss samples there were some misclassifications in both directions, but all samples of these two groups were classified as belonging to those. For validation, the elements 75As,
82Se, 85Rb, 88Sr, 95Mo, 142Nd, 205Tl, 23Na, 44Ca, 51V and δ18O were selected by the model.
The mean rate of correct classification when validating the method was 50 %. All French and Brazilian samples were classified correctly, while Germany samples were completely misclassified as French. No Swiss sample was identified as being from there, but most samples were classified as being Hungarian. There were other misclassifications in other groups, too. Using element signature alone had given average correct classifications of 73 and 50 % of the samples for cross-validation and validation, respectively (Franke et al., 2007a). This means that no real improvement of the accuracy of authentication was achieved when including δ18O results into the dataset, although the variable had been selected by the model for both types of validation.
Table 2
Cross-validation classification and validation matrix of poultry breast meat samples of various originsa
Predicted origin
Actual origin Brazil France Germany Hungary Switzerland
% correct
Cross-validation
Brazil 14 0 0 0 0 100
France 1 9 2 1 0 69
Germany 0 1 14 0 0 93
Hungary 0 0 0 13 3 81
Switzerland 0 0 0 5 15 75
Overall 15 10 16 19 18 83
Validation
Brazil 4 0 0 0 0 100
France 0 2 0 0 0 100
Germany 0 3 0 0 0 0
Hungary 0 1 0 5 0 83
Switzerland 1 1 0 5 0 0
Overall 5 7 0 10 0 50
a Figures represent sample numbers.
3.3. Dried beef meat
Analysis of variance gave several single variables to differentiate between nearly all origin groups of dried beef (Table 3). Only Canadian and Austrian samples could not be distinguished by any of these individual variables and, within dried beef of Swiss raw meat origin, sites of processing (Grisons or Valais). Reasons for the inability to differ between the latter two origins are probably that the raw meat may originate from the entire country for both processing sites and is processed under similar conditions even in different Swiss regions. However, it remains unclear why Canadian and Austrian dried beef could not be distinguished. Reasons explaining the discriminating power of the elements Ca, Cu, Ga, Ni, Pd, Rb, Tl, V and Zn were discussed previously (Franke et al., 2007b).
The LDA carried out across all dried beef samples selected 10B, 42Ca, 111Cd, 63Cu, 57Fe,
7Li, 55Mn, 95Mo, 85Rb, 88Sr, 51V, and δ18O as the most discriminating variables. The mean classification rate of all samples in the cross-validation classification matrix was 73 % (Table 4). Canadian samples were identified completely correctly, but one US-sample was misclassified as being Canadian. For dried beef processed in Europe, it was not possible to differ between that processed in different countries but with the raw beef originating from the same country (Brazilian beef processed either in Austria or in the Swiss canton of Grisons). The same holds true for different raw meat origin processed at the same site (either Brazilian or Swiss beef processed in the canton of Grisons), maybe also because differences in the shift in the oxygen isotope ratio due to the drying process. However, all samples of Swiss raw meat and processed in Switzerland were determined as being processed in Switzerland. For validation of the geographic origin of dried beef 10B, 137Ba,
42Ca, 57Fe, 69Ga, 7Li, 24Mg, 55Mn, 85Rb, 88Sr, 51V and δ18O were found to be most discriminating. The mean classification rate in the validation step was 43 %. Again all Canadian samples were correctly classified. Here it was possible to differ between
differentiation between samples of different raw meat origin and same processing place as well as different Swiss processing sites still was not possible. Using element signature alone had provided average correct classifications of 70 and 48 % of the samples for cross-validation and validation, respectively (Franke et al., 2007a). Therefore, as found for poultry meat, including δ18O did not improve authentication in dried beef despite δ18O had been always included by the model.
Table 3
Univariate differentiation of dried beef of various raw meat origins and places of processinga Country AU-AU BR-AT BR-GR CH-GR CH-VS CN-CN US-US Variable
n 8 6 16 16 15 8 5
AU-AU
—
X 0 0 X 0 X
X X X 0 X X
X X X 0 0 X
X X 0 0 0 X
X X 0 X X X
0 0 0 X 0 X
10B
111Cd
69Ga
7Li
85Rb δ18O BR-AT X
X 0 X
—
X X 0 0
X 0 X 0
0 0 X 0
0 0 0 0
0 X 0 X
42Ca
7Li
85Rb δ18O BR-GR X
0 X 0 X X
0 0 0 X 0 0
—
0 X 0 0 X 0
0 0 0 0 X 0
0 0 0 X 0 0
X 0 0 X X X
111Cd
57Fe
69Ga
7Li
85Rb δ18O CH-GR 0
X 0 0 X
0 0 0 X 0
X 0 0 X 0
—
0 0 0 0 0
0 0 0 X 0
0 0 X 0 X
57Fe
69Ga
7Li
85Rb δ18O CH-VS 0
0 X
0 X 0
0 X 0
0 0
0 —
X X
X 0 X
7Li
85Rb δ18O CN-CN X
X X
0 0 0
X 0 0
0 X 0
X X
0 —
X X
7Li
85Rb δ18O US-US 0
X X 0 X X
0 X X 0 X X
X X X 0 X X
X X X 0 X X
X X X 0 X X
X X X X X 0
—
10B
7Li
55Mn
85Rb 51V δ18O
First acronym, country of origin; second acronym, site of processing. AU Australia; AT, Austria; CN, Canada;
BR, Brazil; GR, Canton of Grisons, Switzerland; US, United States of America; VS, Canton of Valais, Switzerland.
aX = differentiation possible; 0 = no differentiation possible. Results for δ18O taken from Franke et al. (2008).
Table 4
Cross-validation classification and validation matrix of dried beef samples of various raw meat origins and places of processing a
Predicted origin
Actual origin AU-AU BR-AT BR-GR CH-GR CH-VS CN-CN US-US % correct Cross-validation
AU-AU 7 0 0 1 0 0 0 88
BR-AT 0 5 0 0 1 0 0 83
BR-GR 1 1 11 0 2 0 1 69
CH-GR 0 0 2 10 4 0 0 63
CH-VS 0 0 0 6 9 0 0 60
CN-CN 0 0 0 0 0 8 0 100
US-US 0 0 0 0 0 1 4 80
Overall 8 6 13 17 16 9 5 73
Validation
AU-AU 0 0 0 4 0 0 0 0
BR-AT 0 1 0 1 0 0 0 50
BR-GR 0 0 2 0 0 1 1 50
CH-GR 0 0 1 3 0 0 0 75
CH-VS 0 0 1 1 1 0 0 33
CN-CN 0 0 0 0 0 2 0 100
US-US 0 0 0 2 0 0 0 0
Overall 0 1 4 11 1 3 1 43
First acronym, country of origin; second acronym, site of processing. AU Australia; AT, Austria; CN, Canada;
BR, Brazil; GR, Canton of Grisons, Switzerland; US, United States of America; VS, Canton of Valais, Switzerland.
a Figures represent sample numbers.
4. Conclusions
As with element signature and δ18O alone, a differentiation of the origins of poultry meat and dried beef was possible but just with some limitations. Nearly all origins showed significant group differences in individual elements and/or in δ18O, but it was not possible to determine clear cut-off-values to differ between certain origins, which makes a application of the method for official controls difficult, if not impossible. The quality of prediction by the multivariate design with the aim to identify ‘unknown’ samples (cross- validation and validation) was not improved by combining the two approaches found promising earlier as individual tools. This means that the search for other, more complementary, approaches is still open in order to have a sound geographic authentication of the origin of meat.
Acknowledgements
We are grateful to Bell AG, Bischofberger AG, Fredag AG and Micarna SA, which provided the poultry samples as well as to Albert Spiess AG, Cher-Mignon SA, Metzgerei Beat
Bündner Fleischtrocknerei, Rapelli SA and Surselva Fleischwaren AG, which provided the European and Australian dried beef samples. We also like to thank the Swiss embassies of USA and Canada who assisted in organising the American and Canadian dried beef samples and the Swiss Federal Office of Public Health for its financial support.
References
Boner, M., & Förstel, H. (2001). Möglichkeiten der Herkunftsbestimmung von Lebensmitteln mit Hilfe der Stabil-Isotopen-Methode am Beispiel von Rindfleisch. Lebensmittelchemie, 55, 151.
Boner, M., & Förstel, H. (2004). Stable isotope variation as a tool to trace the authenticity of beef. Analytical and Bioanalytical Chemistry, 378(2), 301–310.
Eidgenössisches Departement des Inneren. (2005) Verordnung des EDI über die Kennzeichnung und Anpreisung von Lebensmitteln. http://www.admin.ch/ch/d/sr/8/817.022.21.de.pdf. Accessed on 28 August 2007
Förstel, H., & Lickfett, J. (2002). Mit isotopem Fingerabdruck den Lebensmitteln auf der Spur. Bio World, 1, 26–27.
Franke, B. M., Gremaud, G., Hadorn, R., & Kreuzer, M. (2005). Geographic Origin of meat – elements of an analytical approach to its authentication. European Food Research and Technology, 221(3-4), 493–
503.
Franke, B. M., Haldimann, M., Gremaud, G., Bosset, J. O., Hadorn, R., & Kreuzer, M. (2007a). Element signature analysis: its validation as a tool for geographic authentication of the origin of dried beef and poultry meat. European Food Research and Technology, online first, DOI 10.1007/s00217-007-0776-8.
Franke, B. M., Haldimann, M., Reimann, J., Baumer, B., Gremaud, G., Bosset, J. O., Hadorn, R., & Kreuzer, M. (2007b). Indications for the applicability of element signature analysis for the determination of the geographic origin of dried beef and poultry meat. European Food Research and Technology, 225(3-4), 501–509.
Franke, B. M., Koslitz, S., Micaux, F., Piantini, U., Maury, V., Pfammatter, E., Wunderli, S., Gremaud, G., Bosset, J. O., Hadorn, R., & Kreuzer, M. (2008). Tracing the geographic origin of poultry meat and dried beef with oxygen and strontium isotope ratios. European Food Research and Technology, 226(4), 761-769
Hegerding, L., Seidler, D., Danneel, H. J., Gessler, A., & Nowak, B. (2002). Sauerstoffisotopen-Verhältnis- Analyse zur Herkunftsbestimmung von Rindfleisch. Fleischwirtschaft, 82(4), 95–100.
Piasentier, E., Valusso, R., Camin, F., & Versini, G. (2003). Stable isotope ratio analysis for authentication of lamb meat. Meat Science, 64(3), 239–247.
Renou, J. P., Bielicki, G., Deponge, C., Gachon, P., Micol, D., & Ritz, P. (2004). Characterization of animal products according to geographic origin and feeding diet using nuclear magnetic resonance and isotope ratio mass spectrometry. Part II: Beef meat. Food Chemistry, 86(2), 251–256.
Schäfer,W. (1967). Wasserbestimmung. In: Acker, L., Bergner, K. G., Diemair, W., Heimann, W., Kirmeier, F., Schormüller, J., & Souci, S. (Eds.), Handbuch der Lebensmittelchemie, 2. Bd/Teil 2 Analytik der Lebensmittel Nachweis und Bestimmung von LM-Inhaltsstoffen. (pp. 22) Berlin, Heidelberg, New York:
Springer-Verlag.
Shintu, L., Caldarelli, S., & Franke, B. M. (2007). Pre-selection of potential molecular markers for the geographic origin of dried beef meat by HR-MAS NMR spectroscopy. Meat Science, 76(4), 700–707.
