ARTICLE IN PRESS
LWT 40 (2007) 164–169 www.elsevier.com/locate/lwt
Cooked ham classification on the basis of brine injection level and pork breeding country Ernestina Casiraghi, Cristina Alamprese, Carlo Pompei Dipartimento di Scienze e Tecnologie Alimentari e Microbiologiche (DiSTAM), Universita` degli Studi di Milano, Via Celoria 2, 20133 Milano, Italy Received 20 September 2004; received in revised form 6 July 2005; accepted 14 July 2005
Abstract The aim of this study was to classify whole-leg cooked hams, made without polyphosphates, by linear discriminant analysis. Principal component analysis (PCA) was used for the selection of significant variables. Thirty-two variables were evaluated on 26 cooked hams prepared using different levels of brine injection and legs from pork bred in different countries (France or Denmark). Previously published data related to 20 hams were also used for classification. A chemometric model, based on ten variables, was obtained by using PCA. The variables were pH, moisture, protein, fat, NaCl, superficial wateriness, L* and a*/b* of biceps femoris muscle, modulus and elasticity index of semitendinosus muscle. Discriminant functions calculated using PCA-selected variables enable correct classification of the cooked hams according to the origin of the meat used and, when this is the same, according to the percentage of brine injected. r 2005 Swiss Society of Food Science and Technology. Published by Elsevier Ltd. All rights reserved. Keywords: Chemometric model; Cooked ham; Multivariate analysis
1. Introduction On average, cooked ham constitutes just over 26% by volume of the delicatessen products sold in Europe. Spain, France and Italy are the biggest consumers; in Italy cooked ham consumption was 4.9 kg/capita/year in 2002 (Associazione Industriali delle Carni). The final quality of the product depends both on the raw material used and on processing, which includes injection of brine, tumbling and cooking. Injection of brine ensures a uniform distribution of sodium chloride, nitrites and other possible ingredients (i.e. sugars, spices, polyphosphates, etc.). The brine injection level and the ingredients used are characteristic of each product and determine the cooked ham quality. In particular, products of higher quality are generally made without polyphosphates and with a low level of brine injection. Corresponding author. Tel. +39 0250316625; fax +39 0250316632. E-mail address:
[email protected] (C. Alamprese).
Tumbling is a mechanical operation which distributes the brine evenly inside the leg and causes the extraction of protein from muscle fibres (Pizza & Pedrielli, 2000). Cooking denatures the extracted proteins, welding the muscles and making the ham slice compact. It gives rise to the distinctive texture, the sensorial properties and a suitable hygienic quality of the final product. The aim of this work was to study the influence of brine injection level and pork meat origin on a set of chemical and physical variables usually used to describe quality of cooked ham. Many authors have suggested that the use of multivariate analysis to describe or classify cooked ham and sausages (Dellaglio, Casiraghi, & Pompei, 1996; de Pena, Cid, & Bello, 1998; Papadima, Arvanitoyannis, Bloukas, & Fournitzis, 1999). Therefore, multivariate analysis was applied to determine if samples could be distinguished on the basis of percentage of brine injection and country of origin of raw muscle. Twenty-six whole-leg cooked hams, with no added polyphosphates, were analysed; these were processed at the same plant but obtained from legs imported from different countries and
0023-6438/$30.00 r 2005 Swiss Society of Food Science and Technology. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.lwt.2005.07.007
ARTICLE IN PRESS E. Casiraghi et al. / LWT 40 (2007) 164–169
injected with different quantities of brine. Data relating to 20 other previously analysed samples, produced in the same plant, were also considered to obtain a greater data set for statistical analysis.
products were pasteurized by autoclaving at 105 1C for 10 min, cooled at 0 1C for about 24 h and stored at 0 1C. 2.2. Chemical analyses The producers sent samples to our laboratory under refrigerated conditions. Representative samples of each ham were obtained by mincing several slices (including surface fat) cut out from different sections of the ham. All hams of each group were analysed as follows:
2. Materials and methods 2.1. Samples A total of 55 de-boned whole-leg cooked hams, without polyphosphates, were considered in this research:
165
Twenty-six samples produced in the same factory over a 1-year period, differing for pork meat origin and brine injection level. Twenty cooked hams obtained from the same producer and previously analysed (Pompei & Spagnolello, 1997). Nine cooked hams of unknown pork breeding country and brine injection levels.
These ham products were classified into six groups (products A–E, unknown) as outlined in Table 1. Hams were produced as follows: frozen hind legs, after selection by weight, were defrosted and mechanically de-boned. The average pH of the legs ranged from 5.8 to 6.0 for pork bred in France and from 5.6 to 5.9 for pork of Danish origin. The injection of brine was done using multineedle syringes. The brine composition for the samples of French origin was 8.5 g/100 g NaCl, 3.2 g/ 100 g dextrose, 0.13 g/100 g nitrites; for those of Danish origin it was 7.9 g/100 g NaCl, 3.0 g/100 g dextrose, 0.13 g/100 g nitrites. After brine injection, the samples were tumbled for 48 h. Hams were moulded, pressed and cooked for about 60–75 min for each kilogram of product. During cooking the core temperature reached 67 1C. Cooked hams were then cooled at 0 1C for about 24 h and, after de-moulding, were trimmed and packaged in vacuum plastic film-aluminium bags. Finally,
pH was measured by inserting a Xerolyt electrode (Ingold Messtechnic AG, Urdolf, Switzerland) in the minced sample. Moisture, ash, total nitrogen, fat and sodium chloride were determined according to the AOAC (1996) methods 950.46, 920.153, 928.08, 960.39 and 935.47, respectively. The nitrogen to protein conversion factor used was 6.25. The sample was prepared for NaCl titration by suspending 1.5 g of minced sample in 40 ml of distilled water at 55 1C and gently shaking for 45 min. Soluble nitrogen was determined, after protein precipitation with trichloroacetic acid, according to a method by Careri et al. (1993). All chemical analyses were performed in duplicate.
2.3. Physical analyses
Colour was determined, immediately after slicing, on a 3 mm thick ham slice; each of the following four muscles was evaluated on all hams of each group: biceps femoris (BF), quadriceps femoris (QF), semimembranosus (SM), semitendinosus (ST). L*, a* and b* indexes were determined according to CIE Hunter scale, using a reflectance colorimeter Croma Meter II (Minolta, Japan). The measuring aperture diameter was 10 mm and C/21 was the illuminant/viewing geometry. Results represent the average of seven determinations, each from different points of the
Table 1 Cooked ham samples Product
Number of samples
Pork breeding country
Brine injection level (%)
I. Produced in 1-year period A B C
10 8 8
France Denmark Denmark
25 30 35
II. Previously analysed D E
10 10
France Denmark
30 38
9
Unknown
Unknown
III. Unknown
ARTICLE IN PRESS E. Casiraghi et al. / LWT 40 (2007) 164–169
166
muscle surface, except for muscle QF, which was of a smaller size (four readings). Rheological indexes were evaluated on the two largest muscles, BF and ST of all hams, using an Instron Universal Testing Machine 4301 (Instron Ltd., High Wycombe, England) supported by Series IX Automated Material Testing System software (Instron Corp., 1990). A constant cross-head speed of 20 mm/ min and a piston 80 mm in diameter were used for uniaxial single and cyclic compression tests. For each test a minimum of five replicates were carried out on 25 mm diameter, 15 mm thick samples. Before testing, samples were conditioned at 5 1C for 60 min. Modulus (N/cm2), an index of product hardness, was calculated on the linear part of the single compression curve as the ratio between stress and relative deformation. Elasticity indexes were evaluated by performing four cyclic compressions between 0.5 and 30 N and were defined as the ratio between displacements recorded during the fourth and the first compression (C4/C1) and the second and the first compression (C2/C1). Superficial wateriness, i.e. the tendency of a cooked ham slice to release liquid, was measured on a 3 mm thick slice, cut out from the central part of all hams. The ham slice was placed between two sheets of blotting paper, which had been previously dried at 105 1C in a vacuum oven. A steady pressure of 5 g/cm2 was applied for 20 min at 20 1C, then the sheets of paper were removed, weighed and put in an oven at 105 1C until constant weight was reached. Results are the average of three replicates and are expressed as milligram of water released per square centimetre of the slice.
2.4. Statistical analyses Systat 5.03 software for Windows (Systat Inc., Evanston, IL, 1993) was used for calculation of
Pearson’s correlation matrix, for one-way variance analysis and for linear discriminant analysis (LDA). Principal component analysis (PCA) was performed using Unscrambler 7.5 software (Camo Asa, Trondheim, Norway).
3. Results and discussion 3.1. Physico-chemical analyses Table 2 reports the average value and the standard deviations of pH determination, chemical analyses and superficial wateriness for products A–C. The moisture level and the sodium chloride content resulted substantially from the brine injection level. The average figure increased significantly (Po0:001) in product C compared to product A. Product B was indistinguishable from both A and C. Both moisture and sodium chloride showed low values of standard deviation for all three types of product, indicating that brine injection is a well applied and standardized process. pH, protein, soluble nitrogen, fat and ash were not statistically different as a function of brine injection level and pork breeding country. Standard deviation value of the fat content was high (from 2.8 to 3.4 g/100 g), indicating a high variability within each product, which is due to the different amount of surface fat of each cooked ham. The trend in the values of superficial wateriness was consistent with the data relating to moisture and thus with the brine injection level and with pork breeding country. Products A–C all proved to be statistically different (Po0:05). In Table 3, average values and standard deviations for the modulus and the elasticity indexes are shown; these were obtained separately on muscles BF and ST for each
Table 2 Mean7standard deviation of pH, chemical parameters and superficial wateriness of the three products Product A (n ¼ 10) pH Moisture (g/100 g) Protein (g/100 g) Protein (g/100 g d.m.) Soluble nitrogen (g/100 g) Fat (g/100 g) Fat (g/100 g d.m.) Ash (g/100 g) NaCl (g/100 g) Superficial wateriness (mg/cm2)
6.0470.26 66.172.1a 18.471.2a 54.676.7a 2.170.4a 13.373.4a 38.877.5a 2.770.2a 2.170.1a 7.270.7a
a
Product B (n ¼ 8) a
5.9870.24 68.672.0ab 17.671.3a 56.577.4a 2.470.3a 11.373.2a 35.778.0a 2.870.2a 2.270.1ab 9.071.4b
n ¼ number of hams analysed in each group. d.m. ¼ dry matter. a,b,c Means with different letters indicate significant differences (Po0:05) among products.
Product C (n ¼ 8) 5.9470.25a 70.672.7b 17.570.8a 60.378.7a 2.170.5a 10.372.8a 35.077.4a 2.970.2a 2.270.1b 10.470.7c
ARTICLE IN PRESS E. Casiraghi et al. / LWT 40 (2007) 164–169
167
values. QF is not distinguishable from the other muscles in any product. For variable b* some significant differences were found (Po0:05) between different muscles within the same product, although these were not consistent. The a*/b* ratio showed a significant difference (Po0:05) between muscle ST and the other muscles in B and C products. In the case of product A, muscle ST proved to be significantly different (Po0:05) only from muscle BF. The relative lack of homogeneity of the colour parameters in the four muscles was already observed by Pompei and Spagnolello (1997). Variance analysis performed on the same muscle among A–C products did not reveal significant differences for any colour parameter or muscle considered.
cooked ham. Considering the different muscles of each product, both the elasticity indexes (C2/Cl and C4/Cl) showed significant differences (Po0:001) between BF and ST. The elasticity indexes are a measure of product cohesiveness: the higher are the values, the more cohesive is the ham. In more detail, C2/C1 is a measure of the cohesiveness of the ham referred to a first bite, while C4/C1 is more related to a chewing action. As regards to modulus values, the most significant difference (Po0:001) between the two muscles considered was observed in product A. For both BF and ST muscles, elasticity indexes and modulus did not show any significant differences among the three cooked ham products as a function of brine injection level and pork breeding country. The colour parameters in Table 4 show the average figures for each type of product. The variance analysis calculated for the four muscles belonging to the same ham type showed significant differences for parameters L* and a*, between muscle ST and muscles BF and SM (Po0:01): in all products ST had lower L* and higher a*
3.2. Classification of the cooked hams A data matrix formed by the 26 samples of products A–C and by 20 out of the 32 variables measured was used for PCA. This is a technique used to interpret large data matrices, and based on identifying the most
Table 3 Rheological parameters (mean7standard deviation) of biceps femoris (BF) and semitendinosus (ST) muscles in the three products Product A (n ¼ 10)
Elasticity index C2/C1 Elasticity index C4/C1 Modulus (N/cm2)
Product B (n ¼ 8)
Product C (n ¼ 8)
BF
ST
BF
ST
BF
ST
0.4970.05a 0.4570.05a 20.4472.21a
0.6070.07b 0.5570.09b 25.0474.56b
0.5170.07a 0.4770.07a 23.2475.09a
0.6170.05b 0.5770.05b 24.8374.33a
0.4770.03a 0.4270.03a 20.2773.00a
0.6370.05b 0.5970.06b 22.9973.57b
n ¼ number of hams analysed in each group. a,b Means with different letters indicate significant differences (Po0:05) between muscles of the same product.
Table 4 Colour parameters (mean7standard deviation) of the four muscles in the three products L*
a*
b*
a*/b*
Product A (n ¼ 10) BF QF SM ST
64.472.5b 62.473.3ab 64.573.7b 60.672.4a
13.171.1a 13.871.7ab 12.871.8a 14.671.1b
6.270.7ab 6.670.6b 6.070.3a 6.270.3ab
2.170.1a 2.170.4ab 2.270.4ab 2.470.2b
Product B (n ¼ 8) BF QF SM ST
65.072.5c 62.372.8b 64.772.4bc 61.272.6a
12.471.3a 13.172.1ab 11.971.5a 13.970.9b
6.070.4ab 6.270.5b 6.170.3ab 5.870.4a
2.170.3a 2.170.3a 2.070.3a 2.470.2b
Product C (n ¼ 8) BF QF SM ST
64.570.8b 63.071.6ab 65.471.6b 60.971.9a
12.770.8a 13.070.9ab 12.171.0a 14.171.1b
6.170.4a 6.670.6b 6.170.6a 6.171.2a
2.170.2a 2.070.3a 2.070.2a 2.470.3b
n ¼ number of hams analysed in each group. BF, biceps femoris; QF, quadriceps femoris; SM, semimembranosus; ST, semitendinosus. a,b,c Means with different letters indicate significant differences (Po0:05) between muscles of the same product.
ARTICLE IN PRESS E. Casiraghi et al. / LWT 40 (2007) 164–169
168
important directions of variability in a multivariate data space. In this way, a small number of components, determined as a linear combination of the measured variables, are used to replace the original variables measured in the experiment, thus reducing their number (Naes, Baardseth, Helgesen, & Isaksson, 1996; Resurreccion, 1988). Variables not considered in the data matrix were related to colour parameters: only L* and a*/b* of muscles BF and ST were taken into account. Besides, a correlation matrix between the measured variables enabled the identification of closely correlated parameters that, providing the same information on data variability, could be ruled out. This excluded from the data matrix protein, fat on dry matter, ash, L* of muscle ST, a*/b* of muscle ST, elasticity index C4/Cl of muscle BF, and elasticity index C2/C1 for both muscles ST and BF. By considering the contribution of each variable to the total explained variance, i.e. the variance explained by the 20 variables considered together, it was possible to eliminate variables that contributed little towards variance explanation (soluble nitrogen and modulus of muscle BF). It was thus possible to draw up a model with ten variables. The same variables were identified in a previous work (Pompei & Spagnolello, 1997) as the most important descriptive indexes for cooked ham. Table 5 shows, for each of the ten selected variables, the percentages of variance explained in relation to the first component and to both first and second component. Total variance of the first two components was 56.4%. The parameters moisture, proteins on dry matter, fat, salt and superficial wateriness gave a greater contribution to the definition of the first component, whereas the parameters L* of muscle 1, a*/b* of muscle BF, modulus for muscle ST and elasticity index C4/C1 of muscle ST explained more of the second component variability. From the bi-plot of the first two principal components (Fig. 1), it can be noted that those parameters (moisture,
protein on dry matter, superficial wateriness and salt) which have high values in the products with a high brine percentage fall in the right part of the plot, while parameters which have high levels in products with a lower brine injection (fat, modulus of muscle ST) are on the left. We can therefore affirm that the first component describes the hams mainly according to brine injection percentage. The second component takes into account the parameters of colour, consistency and pH, related to raw meat characteristics. From the bi-plot it can be seen that the three ham types are not clearly distinguished: the 30% brine-injected product is not well differentiated both from the 25% brine injection product on the left of the graph, and from the 35% brine-injected product to the right of the graph. To get a wider data set for further statistical evaluation, it was decided to test the same chemometric model on samples analysed both in this work and in a previous study (Pompei & Spagnolello, 1997). To this aim the samples analysed in this research (products A–C) were inserted as ‘evaluation set’ in the previously constructed model based on 20 samples considered as ‘training set’ (products D and E). From Fig. 2 it can be noted that evaluation set and training set overlap on the PC1–PC2 plane, thus proving that the chemometric model is valid and the ten variables utilized to construct it describe whole-leg cooked ham. The 26 samples used in this study (products A–C) and the 20 samples (products D and E) analysed in the previous research, whose pork breeding country and brine level were known, were subsequently evaluated using LDA. This is a multivariate technique aimed at determining which set of variables best discriminates one group of objects from another (Resurreccion, 1988). Initially, LDA was performed on all 46 samples, which
Table 5 Percentage of variance explained by the ten selected variables Variable
PC1
PC1+PC2
pH Moisture Protein on dry matter Fat NaCl Superficial wateriness L* muscle BF a*/b* muscle BF Modulus muscle ST C4/C1 muscle ST All variables
19.2 84.7 57.7 72.6 41.9 48.1 0.5 0.6 12.7 0.7 33.9
31.1 92.3 80.3 82.7 55.5 58.5 66.9 33.0 46.5 17.1 56.4
BF, biceps femoris; ST, semitendinosus.
Fig. 1. Bi-plot on the first two component plane of the three types of ham (A, ~; B, K; C, ’) and of the ten selected variables (n): pH (1), moisture (2), proteins on dry matter (3), fat (4), NaCl (5), superficial wateriness (6), L* muscle BF (7), a*/b* muscle BF (8), modulus muscle ST (9) and elasticity index C4/C1 muscle ST (10).
ARTICLE IN PRESS E. Casiraghi et al. / LWT 40 (2007) 164–169
169
class correctly classified 96.1% of samples. Only one of the 35% brine-injected samples was classified as belonging to 30% brine-injected hams. Subsequently, data from nine samples (not reported) whose pork breeding country and brine injection levels were unknown were inserted in the model, and LDA was again performed to identify raw material origin. The model classified all the samples inserted as French. LDA was subsequently used to classify them according to brine injection level. The samples were classified as 30% brine injected. The producer later confirmed that the raw hams were of French origin and that samples had been injected with 30% brine, thus allowing to ascertain that the model had classified them correctly. Fig. 2. Plot of the samples used for the validation of a previously constructed chemometric model: training set, ~; evaluation set, &. Table 6 Sample assignment to the different levels of brine injection according to the discriminant analysis results Brine injection (%)
38
35
30
25
n
38 35 30 25
9 (90%) 1 1 0
0 7 (88%) 5 0
1 0 10 (56%) 1
0 0 2 9 (90%)
10 8 18 10
Note: Percentages in brackets are referred to samples correctly classified. n ¼ number of hams in each group.
were assigned to different groups depending on the percentage of brine injection. The discriminant functions that were calculated correctly classified 35 samples out of 46, i.e. 76%. Table 6 reports the assignment of samples to the different groups according to the discriminant analysis results, showing the percentage of samples correctly classified for each class. Only 56% of the 30% brine-injected samples were correctly classified; it should be noted that this group contained cooked hams obtained from raw materials of two different countries. The samples that were most correctly classified (90%) belonged to products A and E, injected with 25% and 38% brine, respectively. In order to determine the influence of pork breeding country on the characteristics of the finished product, LDA was subsequently carried out assigning the samples to two classes: raw materials from Denmark and from France. LDA results correctly classified 42 samples out of 46 and, more specifically, 92.3% of Danish samples (24 out of 26) and 90% of French samples (18 out of 20). Finally, within each class—French origin and Danish origin —the samples were classified according to brine injection levels. LDA on class of French origin correctly classified 100% of samples, whereas LDA on Danish
4. Conclusions This study validated a chemometric model for the description of cooked ham based on ten chemical and physical variables, which was developed in a previous study (Pompei & Spagnolello, 1997). Moreover, these ten variables allow one to correctly classify, within the examined data set, commercial cooked hams according to the pork breeding country and, when this is the same, according to the brine injection percentage. In conclusion, since the assessment of meat origin and brine injection level of cooked ham is the result of the interaction of different characteristics and cannot be evaluated by any single characteristic, a practical application of this study is the possibility to use a multivariate approach to verify meat traceability. References AOAC. (1996). Official methods of analysis. Gaithersburg, MD: Association of Official Analytical Chemists. Careri, M., Mangia, A., Barbieri, G., Bolzoni, L., Virgili, R., & Parolari, G. (1993). Sensory property relationships to chemical data of Italiantype dry-cured ham. Journal of Food Science, 58, 968–972. Dellaglio, S., Casiraghi, E., & Pompei, C. (1996). Chemical, physical and sensory attributes for the characterization of an Italian drycured sausage. Meat Science, 42, 25–35. de Pena, P. M., Cid, C. M., & Bello, J. (1998). A method for identification of frozen meat used for production of cooked ham. Meat Science, 48, 257–264. Naes, T., Baardseth, P., Helgesen, H., & Isaksson, T. (1996). Multivariate techniques in the analysis of meat quality. Meat Science, 43, S135–S149. Papadima, S. N., Arvanitoyannis, I., Bloukas, J. G., & Fournitzis, G. C. (1999). Chemometric model for describing Greek traditional sausages. Meat Science, 51, 271–277. Pizza, A., & Pedrielli, R. (2000). Effetto delle tecniche di zangolatura e di cottura sulla resa e sull0 accettabilita` del prosciutto cotto ottenuto da cosce di diversa qualita`. Industria Conserve, 75, 171–182. Pompei, C., & Spagnolello, A. (1997). Chemometric model for describing cooked ham. Italian Journal of Food Science, 9, 3–15. Resurreccion, A. V. A. (1988). Applications of multivariate methods in food quality evaluation. Food Technology, 42, 128–136.