HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text (PDF) Interpretive Summary Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via Google Scholar Google Scholar Articles by Soyeurt, H. Articles by Dardenne, P. Search for Related Content PubMed PubMed Citation Articles by Soyeurt, H. Articles by Dardenne, P.

Potential estimation of major mineral contents in cow milk using mid-infrared spectrometry

H. Soyeurt^*,1, D. Bruwier^*, J.-M. Romnee, N. Gengler^*,, C. Bertozzi, D. Veselko^# and P. Dardenne

^* Gembloux Agricultural University, Animal Science Unit, B-5030 Gembloux, Belgium
Walloon Agricultural Research Centre, Quality Department, B-5030 Gembloux, Belgium
National Fund for Scientific Research, B-1000 Brussels, Belgium
Walloon Breeding Association, B-5590 Ciney, Belgium
^# Milk Committee, B-4651 Battice, Belgium

¹ Corresponding author: soyeurt.h{at}fsagx.ac.be

	ABSTRACT

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSIONS ACKNOWLEDGMENTS REFERENCES

 
Milk and dairy products are a major source of minerals, particularly calcium, involved in several metabolic functions in humans. Currently, several dairy products are enriched with calcium to prevent osteoporosis. The development of an inexpensive and fast quantitative analysis for minerals is required to offer dairy farmers an opportunity to improve the added value of the produced milk. The aim of this study was to develop 5 equations to measure Ca, K, Mg, Na, and P contents directly in bovine milk using mid-infrared (MIR) spectrometry. A total of 1,543 milk samples were collected between March 2005 and May 2006 from 478 cows during the Walloon milk recording and analyzed by MIR spectrometry. Using a principal component approach, 62 milk samples were selected by their spectral variability and separated in 2 calibration sets. Five outliers were detected and deleted. The mineral contents of the selected samples were measured by inductively coupled plasma atomic emission spectrometry. Using partial least squares combined with a repeatability file, 5 calibration equations were built to estimate the contents of Ca, K, Mg, Na, and P in milk. To assess the accuracy of the developed equations, a full cross-validation and an external validation were performed. The cross-validation coefficients of determination (R²cv) were 0.80, 0.70, and 0.79 for Ca, Na, and P, respectively (n = 57), and 0.23 and 0.50 for K and Mg, respectively (n = 31). Only Ca, Na, and P equations showed sufficient R²cv for a potential application. These equations were validated using 30 new milk samples. The validation coefficients of determination were 0.97, 0.14, and 0.88 for Ca, Na, and P, respectively, suggesting the potential to use the Ca and P calibration equations. The last 30 samples were added to the initial milk samples and the calibration equations were rebuilt. The R²cv for Ca, K, Mg, Na, and P were 0.87, 0.36, 0.65, 0.65, and 0.85, respectively, confirming the potential utilization of the Ca and P equations. Even if new samples should be added in the calibration set, the first results of this study showed the feasibility to quantify the calcium and phosphorus directly in bovine milk using MIR spectrometry.

Key Words: milk • mineral • mid-infrared spectrometry • calcium

	INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSIONS ACKNOWLEDGMENTS REFERENCES

 
Milk and dairy products are the main source of calcium (around 1,200 mg/L of milk) in Western diets (Devriese et al., 2006). In Belgium, cheese is the major source of calcium (27%) followed by raw milk (14%). Globally, dairy products contribute 54% of the daily dietary intake of calcium (Devriese et al., 2006). In the European Union, this contribution varies between 45 and 70% (Guéguen and Pointillart, 2000). According to Gerrior and Bente (2002), milk provides more than 70% of the calcium available in the US food supply.

Consumption of the multifunctional nutriment calcium is important for humans because calcium plays a role in several metabolic functions such as blood coagulation and muscular contractions. The average recommended dietary calcium intake is about 900 mg/d for adults (19 to 59 yr old) and 1,200 mg/d for adolescents and the elderly (Guéguen and Pointillart, 2000; Devriese et al., 2006). However, consumption of Ca by women and men is too low. In Belgium, in 2004, the average consumption of dietary calcium was 838 mg/d for men and 716 mg/d for women (e.g., Heaney, 2000; Devriese et al., 2006). According to Huth et al. (2006), the low intake of calcium by Americans and the large difference between recommended and typical dietary Ca intakes are recognized as major public health problems. In fact, low consumption of Ca has several effects on human health such as osteoporosis, which is more frequent in women than in men (e.g., Guéguen and Pointillart, 2000; Lanou et al., 2005; Devriese et al., 2006; Huth et al., 2006), arterial hypertension, colon cancer (Guéguen and Pointillart, 2000; Huth et al., 2006), and regulation of body weight and body fat (Huth et al., 2006).

Milk is a good source of other major minerals such as magnesium, phosphorus, and potassium. In the United States, dairy products provide about 16% of Mg and 32% of P and Na dietary intakes. Like calcium, phosphorus and magnesium are involved in bone health and human development. The combination of sodium, magnesium, calcium, and other milk components such as vitamins, protein, and essential fatty acids has a beneficial effect on blood pressure regulation (Huth et al., 2006).

Currently, several dairy products are enriched in calcium to prevent osteoporosis. However, according to Huth et al. (2006), the combination of minerals in milk seems to be more effective than minerals taken alone. Improving the mineral content in milk could be an interesting opportunity for dairy farmers to add value to the milk produced. To do this, the farmers could use the natural sources of variation influencing the contents of mineral in milk. For instance, Mouillet et al. (1975) and Cerbulis and Farrell (1976) showed breed differences for the mineral profile in bovine milk. Davis et al. (2001) observed within-breed differences of calcium for constant protein content in milk, suggesting a potential animal selection effect on mineral contents. The use of genetic variation (within or across breeds) requires a large data set. Currently, one of the faster methodologies—inductively coupled plasma atomic emission spectrometry (ICP-AES)—is too expensive to permit routine analysis of milk samples collected, for instance, during the regular milk recording managed by different breed organizations. However, the regular milk recording permits the rapid collection of a large amount of data. The current method used to measure the contents of fat, protein, lactose, and urea during regular milk recording is mid-infrared (MIR) spectrometry. Using this technology to measure the contents of mineral in milk could permit the development of selection and management tools for dairy farmers to improve the nutritional quality of milk. Thus, the objective of this research was to study the feasibility of developing calibration equations permitting the quantification of Ca, K, Mg, Na, and P directly in bovine milk using MIR spectrometry.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSIONS ACKNOWLEDGMENTS REFERENCES

 
Sampling and Recording Spectra Files
The development of calibration equations requires the analysis of samples showing a large variability of milk composition. Therefore, different known sources of variation for milk composition were taken into account during the sampling (e.g., geographical situation, season of sampling, breed). Following standard procedures (International Committee for Animal Recording, 2004), 2 x 1,543 milk samples of 50 mL (50% morning milk and 50% evening milk) were collected for 1 yr (March 2005 to May 2006) during the milk recording of the Walloon region of Belgium organized by the Walloon Breeding Association (Ciney, Belgium). These samples were collected from 478 cows belonging to 8 herds and representing 6 dairy breeds (dual-purpose Belgian Blue, Holstein-Friesian, Jersey, Montbeliarde, Normande, and Meuse-Rhine-Yssel type Red and White). One sample was analyzed directly after sampling by MIR spectrometry using a MilkoScan FT6000 spectrometer (Foss, Hillerød, Denmark) by the Milk Committee (Battice, Belgium). All of the generated spectra were recorded in a database. The second sample was conserved at –26°C pending chemical analysis.

Selection of Samples
The measurement of mineral contents for all collected samples was not conceivable because of the cost of chemical analysis. The selection of samples was difficult because any indicator traits of mineral contents in milk were available. To maximize the milk composition variability of the samples used in the calibration set, 100 samples were selected from their spectral variability using a principal components approach.

Reference Method
The reference mineral contents used for the calibration process were measured using ICP-AES (Ultima, Jobin-Yvon, Longjumeau, France). The dispersive system was a Czerny-Turner monochromator; the focal distance was 1 m. A concentric nebulizer composed of Meinhard glass and a cyclonic room was used. The spectral data ranged from 120 to 800 nm. Humidification of argon was used as well as a peristaltic pump. The integration parameters were as follows: nebulizer gas flow = 0.75 min^–1; pressure of nebulizer = 3 bar; Rf power = 1,100 W; fixed time of rising = 60 s; fast speed rinsing pump; time of transfer = 15 s; time of stabilization = 45 s; fast speed of transfer pump; synchronization time = 0 s; normal speed of pump = 20 m/s. The spectral lines for the detection of Ca, K, Mg, Na, and P were 318, 766, 279, 590, and 178 nm, respectively.

Some liquids can be analyzed without the mineralization stage (which increases the cost of analysis, the time needed, and, frequently, the risk of sample contamination). Nobrega et al. (1997) tested the direct analysis of milk without mineralization using a dilution in an aqueous solution at pH 8 containing a tertiary amine mixed at 5 and 10% to prevent the precipitation of proteins. Murcia et al. (1999) performed direct analysis of diluted milk samples (1:50) by addition of Fluka Triton X-100 (polyethylene glycol tert-octylphenyl ether; Sigma-Aldrich, Birnem, Belgium) as a tensioactive dispersant to improve the repeatability of measurements. The results obtained with and without mineralization stage were similar (Murcia et al., 1999).

Before the development of calibration equations, the first step of this study was to validate the hypothesis that the measurements of minerals by ICP-AES with and without a previous mineralization stage should be similar. Four milk samples were chosen based on their extreme fat content (2.38, 2.40, 7.52, and 7.62 mg/dL of milk) because fat is a critical point in milk analysis by ICP-AES. These selected samples were then analyzed by ICP-AES with and without the mineralization stage. Each sample was analyzed 3 times for each treatment. For the dispersive analysis, each sample was diluted at 1:100 and Triton X-100 was added at 1:1000. The diluent was ultra-pure water obtained with a MilliQ Gradient instrument (Millipore, Billerica, MA). A nitroperchloric mineralization by microwave was used. A total of 0.5 g of sample was introduced in a Teflon vessel. Then, 1 mL of H₂O₂ at 30% (Merck, Brussels, Belgium) and 6 mL of HNO₃ at 65% (Merck) were added. The vessel was hermetically sealed and placed in the microwave. Two microwaves (MLS 1200 mega, Milestone Microwave Laboratory Systems, Bergamo, Italy) were used. The first contained a rotor with 6 places and the second had a rotor with 10 places. The process of mineralization using the microwave containing a rotor with 6 places was 1 min at 250 W; 1 min at 0 W; 5 min at 250 W; 5 min at 400 W; 5 min at 650 W; and 10 min of ventilation. The process of mineralization using the microwave containing a rotor with 10 places was: 2 min at 250 W; 2 min at 0 W; 5 min at 250 W; 5 min at 450 W; 5 min at 650 W; 5 min at 450 W; and 10 min of ventilation. The dispersed samples were analyzed first followed by the mineralized samples to prevent potential precipitation of proteins. After mineralization, the samples were filtered using Whatman filter paper (595 , diameter 125 mm, Scheider and Schuell, Dassel, Germany). For the filtration of samples for the quantification of P, a paper poor in P was used (Whatman, 512 , diameter 150 mm, Scheicher and Schuell).

Estimation of Uncertainty
The uncertainty of measurement comprised different sources of uncertainty: 1) the uncertainty of repeatability among independent analyses executed on the same sample (U_repeat); 2) the uncertainty about the recovery percentage of mineral element estimated using a certified reference milk (CRM) sample (U_recup); 3) the uncertainty on the weighing of sample (U_weight). The calibration certificate of the precision balance provided this information; 4) the uncertainty of the dilutions of the sample (U_dilution). The manufacturer of the vessel mentioned this characteristic; 5) the uncertainty of the ICP-AES instrument (U_ICP). The instrument gave this uncertainty from repeated analysis (3 times) of the sample milk sample; 6) the uncertainty of the concentration of the stock solution [standard solution: CertiPur (Merck) for the standard needed for the calibration of the ICP-AES instrument (U_stock)]. The manufacturer provided this information; 7) the uncertainty of the dilutions of the standard needed to establish the calibration of the ICP-AES instrument (U_standard); and finally, 8) the uncertainty about the purity of the CRM sample used (U_CRM). The provider gave this characteristic.

The CRM sample used was a powdered skimmed milk (CRM-063R) of the Community Bureau of Reference (Brussels, Belgium). In accordance with Skoog et al. (1997), the total uncertainty of the chemical method (U_total) was calculated following this expression:

Calibration Procedure
The calibration equations were established using 3 different calibration sets. The first calibration set contained the 40 samples showing the highest spectral variability from the 100 samples initially selected. Eight samples were not correctly conserved and were not analyzed by ICP-AES. Following the same approach, a second set including 30 samples was created from the 60 remaining samples. The third calibration set contained 30 samples selected using the Ca calibration equation developed based on the samples contained in the first and the second calibration sets (more details in the Results and Discussion section). A total of 92 samples were analyzed by ICP-AES; 5 outliers were detected and deleted. From the reference mineral contents and the corresponding spectral data, 5 calibration equations for the prediction of Ca, K, Mg, Na, and P contents were developed using a program for multivariate calibration (WINISI III; http://www.winisi.com/; Foss, Hillerød, Denmark). The use of partial least squares regressions was preferred in this study over other methods such as multiple linear or principal components regressions because the use of partial least squares regressions limits the presence of noise in the calibration equations when a limited number of samples are used. In fact, the PLS method compresses all spectral data (Martens and Jensen, 1982; Frank et al., 1984) and simultaneously maximizes the variability of the dependent variable (Martens and Naes, 1987).

To linearize the spectra, the initial spectral data expressed in transmittance were converted into absorbance using the following formula: absorbance = log(transmittance^–1).

To improve the repeatability of the calibration equations across instruments, a repeatability file containing milk spectra analyzed by 4 different MilkoScan FT6000 (Foss) was used: 3 located at the Milk Committee (Battice, Belgium) and 1 located at Convis (Ettelbruck, Luxembourg). The approach was similar to the one described by Westerhaus (1990). Some statistical parameters of the developed calibration equations were calculated to assess the accuracy of the prediction: the mean and the standard deviation (SD) of the reference mineral contents measured by ICP-AES, the standard error of calibration (SEC), and the calibration coefficient of determination (R²_C).

Validation
The developed calibration equations were validated using 2 different methods: an internal validation using a full cross-validation and an external validation using milk samples that were not used to build the considered equations.

A full cross-validation was applied to determine the most appropriate number of factors used and also to assess the robustness of the developed equations. Cross-validation uses the same samples for validation and calibration processes. Full cross-validation leaves out one sample and then performs a calibration with the remaining samples (Williams, 2007). This procedure is repeated until every sample has been predicted once. Finally, the validation errors are combined into a standard error of cross-validation (SECV). Consequently, some additional statistical parameters were calculated to assess the accuracy of calibration equations: the SECV and the cross-validation coefficient of determination (R²_CV). The ratio of SECV to SD (RPD) was also calculated to assess the efficiency of the calibration (Williams, 2007).

External validation was based on the comparison between the mineral contents predicted by MIR spectrometry and those measured by the reference chemical analysis (ICP-AES) for milk samples that were not used for the calibration procedure. The prediction values were obtained by applying the developed calibration equations on collected spectra using the calibration program (WINISI III; http://www.winisi.com/; Foss). The validation coefficient of determination (R²v) was calculated.

	RESULTS AND DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSIONS ACKNOWLEDGMENTS REFERENCES

 
Reference Method
Table 1 depicts the results obtained for the analysis by ICP-AES of the 4 selected samples with and without the mineralization pretreatment. This experiment showed that these 2 treatments (with and without mineralization) differed significantly only for K. Indeed, the P-value obtained by the test of equality of means and presented in Table 1 was <0.001 for K. The coefficient of variation (CV) observed from the measurement of Na by direct milk dispersion in the ICP-AES spectrometer was greater than that calculated for the measurement of Na after mineralization. Thus, the direct analysis of milk sample by ICP-AES was not appropriate to measure the Na content. However, in accordance with the CV values, the direct analysis of milk by ICP-AES was more efficient for the measurement of the other studied minerals (Table 1).

The average differences for the mineral contents measured by the 2 considered methods are also presented in Table 1. The difference ranged between 7.8 and 15.2%. As expected based on the previous results, the greatest difference was observed for Na. Globally, the values given by the direct analysis of milk sample were greater than those measured after a mineralization stage.

For the 2 methods, the repeatability can be considered as good because the CV values were 3% except for the measurement of Na performed by ICP-AES without previous mineralization (Table 1).

This inequality between the concentrations obtained by ICP-AES with and without mineralization was in agreement with a previous study conducted by Roulez (2001). The results obtained by that author showed the highest contents of Ca, K, and Na for the direct method of analysis compared with the mineralization pretreatment. The use of nitric acid could precipitate a part of the proteins and modify the spatial repartition of mineral in the solution. The use of Triton X-100 was not investigated. In contrast to the results of Roulez (2001), Murcia et al. (1999) used this tensioactive dispersant and showed similar results with and without mineralization. In this study, the measurement of Na by the two proposed methods did not provide similar results.

Based on the obtained results and the fact that the use of the ICP-AES with a previous mineralization stage increases the possibility of sample loss, direct analysis of milk sample by ICP-AES without mineralization was used to quantify the reference mineral contents needed for the development of calibration equations.

Sample Selection
Figure 1 illustrates the spectral variability of the 1,543 collected milk samples. All spectra were plotted and the variation is clearly observable. For convenience, the standard deviation for each spectral data points (1,060 in total for the spectra generated by the MilkoScan FT6000) is also represented. For financial reasons, it was not possible to analyze all collected samples. Principal components analysis was applied to select the most interesting samples to elaborate the first and second calibration sets. Thirty-one principal components described 99.98% of the spectral variability. As explained in the Materials and Methods section, a total of 100 milk samples were selected and 57 of the selected samples showing the largest spectral variability were analyzed by ICP-AES.

View larger version (15K): [in this window] [in a new window]   Figure 1. Spectral variability of the 1,543 collected milk samples. The gray background represents the spectral variability of the collected samples (all spectra are plotted). The continuous black line presents the standard deviation for the 1,060 spectral data points.

View larger version (10K): [in this window] [in a new window]   Figure 2. Reference calcium contents measured by inductively coupled plasma atomic emission spectroscopy from the 57 samples initially selected.

View larger version (11K): [in this window] [in a new window]   Figure 3. Calcium contents predicted by mid-infrared spectrometry using the calibration equation of calcium developed from 57 samples.

Uncertainty
Table 3 shows the uncertainty of the method used to measure the reference mineral contents. The highest uncertainty came from the recuperation percentage of mineral element in the CRM sample. The second highest contributors to uncertainty were related to the weighing of sample and to the purity of the CRM sample used. The total uncertainty ranged between 3.95 and 5.55% (Table 3). These values were globally <5% and were, therefore, acceptable.

View larger version (18K): [in this window] [in a new window]   Figure 4. Relation between the reference contents of calcium and phosphorus measured by inductively coupled plasma atomic emission spectroscopy and prediction of Ca and P by mid-infrared spectrometry using the second calibration equation established for Ca and P for the samples selected for the third calibration set.

According to Williams (2007), calibration equations showing R² values between 0.83 and 0.90 can be used with caution for most applications, including research. In conclusion, based on the obtained statistical parameters, this study showed the feasibility of quantifying the contents of Ca and P directly in milk using MIR spectrometry. This application could have interesting potential interest for dairy farmers. For instance, Harmon (1994) suggested that the variation of minerals in milk could be studied to detect the presence of mastitis.

Interpretation of Calibration Equations
The correlations between Ca and P contents and spectral data were also estimated to detect the most related spectral zones to mineral predictions by MIR. The greatest correlations for Ca were located between 1,454 and 1,458 cm^–1 (R = 0.74) and between 2,831 and 2,970 cm^–1 (R = 0.73). The greatest correlations for P were located between 1,200 and 1,277 cm^–1 (R = 0.77), between 2,841 and 2,974 cm^–1 with a maximum at 2,974 cm^–1 (R = 0.77), and between 1,442 and 1,469 cm^–1 (R = 0.71). The spectral data located at 2,927 cm^–1 and 2,858 cm^–1 are related to milk fat. Consequently, the MIR predictions of Ca and P are linked to the content of fat, which was expected because the correlations between Ca or P and fat were positive (Table 5). The greatest correlations observed between 1,743 and 1,747 cm^–1 were related to the carbonyl group of ester bonds present in milk fat and organic acid. The spectral zone between 1,446 and 1,470 cm^–1 is characterized by CH₂ and CH₃ groups present in milk fat and proteins (Coates, 2000). The high correlation located at 1,242 cm^–1 is related to the P=O bond present in phospholipids (Bertrand and Dufour, 2006). The spectral zones known between Ca and the carboxylate group of casein at 1,410 cm^–1 and 1,575 cm^–1 (Byler and Farrell, 1989) did not show high correlation (R = 0.46 and R = 0.41, respectively). Upreti and Metzger (2006) studied the feasibility of developing calibration equations to measure organic P and bound Ca in Cheddar cheese. These authors based on the previous work of Fernandez et al. (2003) limited the development of calibration equations to spectral data between 1,050 and 900 cm^–1. In this region, Upreti and Metzger (2006) observed a region from 956 to 946 cm^–1 related to the concentration of organic P. The region located around 980 cm^–1 observed by those authors seemed to be related to bound Ca. In our study, these regions were not the most interesting for the measurement of Ca and P by MIR spectrometry.

	CONCLUSIONS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSIONS ACKNOWLEDGMENTS REFERENCES

 
This study showed the feasibility of predicting the contents of Ca and P by MIR spectrometry. The results obtained for the calibration equations that predict the Mg content indicated the need to increase the number of reference samples analyzed by ICP-AES used for the calibration procedure. The method used to measure mineral content (ICP-AES without mineralization) was inappropriate to determine the Na concentration in milk. As the calibration equations are strongly influenced by the reference values used, this inability could partly explain the poor prediction by MIR observed for this mineral. Finally, the prediction of K by MIR was not possible.

As MIR spectrometry is used routinely by the milk labs to estimate the contents of the major milk components such as the percentage of fat and protein (used for milk payment and milk recording), implementation of the developed equations to estimate the concentrations of Ca and P in milk could be interesting for dairy farmers to give added value to the produced milk and to improve cheese properties.

	ACKNOWLEDGMENTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSIONS ACKNOWLEDGMENTS REFERENCES

 
The authors acknowledge the financial support provided through grants 2.4507.02F (2), F.4552.05, and 2.4623.08 of the National Fund for Scientific Research (Brussels, Belgium). Nicolas Gengler, research associate at the National Fund for Scientific Research (Brussels, Belgium) acknowledges its support.

Received for publication September 20, 2008. Accepted for publication January 20, 2009.

	REFERENCES

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSIONS ACKNOWLEDGMENTS REFERENCES

 


Bertrand, D., and E. Dufour. 2006. La Spectroscopie Infrarouge et Ses Applications Analytiques. 2nd ed. Lavoisier Tec et Doc, Paris, France.

Byler, D. M. and Farrell, H. M.. 1989. Infrared spectroscopic evidence for calcium ion interaction with carboxylate groups of casein. J. Dairy Sci. 72:1719–1723.[Abstract/Free Full Text]

Cerbulis, J. and Farrell, H. M.. 1976. Composition of the milks of dairy cattle. II. Ash, calcium, magnesium, and phosphorus. J. Dairy Sci. 59:589–593.[Abstract/Free Full Text]

Coates, J. 2000. Interpretation of infrared spectra, a practical approach. Pages 10815–10837 in Encyclopedia of Analytical Chemistry. R. A. Meyers, ed. John Wiley & Sons, New York, NY.

Davis, S. R., Farr, V. C., Knowles, S. O., Lee, J., Kolver, E. S. and Auldist, M. J.. 2001. Sources of variation in milk calcium content. Aust. J. Dairy Technol. 56:156.

Devriese, S., I. Huybrechts, M. Moreau, and H. Van Oyen. 2006. Enquête de consommation alimentaire Belge A – 2004. Institut Scientifique de Santé Publique. ISP/EPI Reports. http://www.iph.fgov.be/epidemio/epifr/foodfr/table04.htm. Accessed Sep. 2, 2008.

Fernandez, C., Ausar, S. F., Badini, R. G., Castagna, L. F., Bianco, I. D. and Beltramo, D. M.. 2003. An FTIR spectroscopy study of the interaction between alpha s-casein-bound phosphosyryl group and chitosan. Int. Dairy J. 13:897–901.[CrossRef]

Frank, I. E., Feikema, J., Constantine, N. and Kowalski, B. R.. 1984. Prediction of product quality from spectral data using the partial least-squares method. J. Chem. Inf. Comput. Sci. 24:20–24.

Gerrior, S., and L. Bente. 2002. Nutrient content of the US food supply, 1909–1999: A summary report. Center for Nutrition Policy and Promotion, USDA, Washington, DC.

Guéguen, L. and Pointillart, A.. 2000. The bioavailability of dietary calcium. J. Am. Coll. Nutr. 19:119S–136S.[Abstract/Free Full Text]

Harmon, R. J. 1994. Physiology of mastitis and factors affecting somatic cell counts. J. Dairy Sci. 77:2103–2112.[Abstract]

Heaney, R. P. 2000. Calcium, dairy products and osteoporosis. J. Am. Coll. Nutr. 19:83S–99S.[Abstract/Free Full Text]

Huth, P. J., DiRienzo, D. B. and Miller, G. D.. 2006. Major scientific advances with dairy foods in nutrition and health. J. Dairy Sci. 89:1207–1221.[Abstract/Free Full Text]

International Committee for Animal Recording. 2004. Section 2. ICAR rules, standards and guidelines for dairy production recording. http://www.icar.org/docs/Rules%20and%20regulations/Guidelines/Guidelines_2005_final_low_resolution.pdf Accessed Nov. 17, 2005.

Lanou, A. J., Berkow, S. E. and Barnard, N. D.. 2005. Calcium, dairy products, and bone health in children and young adults: A reevaluation of the evidence. Pediatrics 115:736–743.[Abstract/Free Full Text]

Martens, H., and S. A. Jensen. 1982. Partial least squares regression: A new two-stage NIR calibration method. Pages 607–647 in Proc. 7th World Cereal and Bread Congress, Prague. K. Holas, ed. Elsevier, Amsterdam, the Netherlands.

Martens, H., and T. Naes. 1987. Multivariate Calibration by Data Compression. Wiley, Chichester, UK.

Mouillet, L., Luquet, F. M. and Casalis, J.. 1975. Contribution à l’étude des variations de la teneur en sels minéraux du lait de vache dans différentes regions françaises. Lait 55:683–694.[CrossRef]

Murcia, M. A., Vera, A., Martinez-Tome, M., Munoz, A., Hernandez-Cordoba, M. and Ortiz-Gonzalez, R.. 1999. Fast determination of the Ca, Mg, K, Na and Zn contents in milk and non-dairy imitation milk using ICP-AES without mineralization stage. Lebens. Wiss. Technol. 32:175–179.[CrossRef]

Nobrega, J. A., Gelinas, Y., Krushevska, A. and Barnes, R. M.. 1997. Direct determination of major and trace elements in milk by inductively coupled plasma atomic emission and mass spectrometry. J. Anal. At. Spectrom. 12:1243–1246.[CrossRef]

Roulez, D. 2001. Contribution à la mise en service d’un spectromètre d’émission atomique couplé à un plasma induit : Études de diverses matrices. MS thesis. Haute Ecole Charleroi-Europe, Charleroi, Belgium.

Skoog, D. A., D. M. West, and F. J. Holler. 1997. Chimie Analytique. 7th ed. De Boeck Université, Brussels, Belgium.

Upreti, P. and Metzger, L. E.. 2006. Utilization of Fourier transform infrared spectroscopy for measurement of organic phosphorus and bound calcium in Cheddar cheese. J. Dairy Sci. 89:1926–1937.[Abstract/Free Full Text]

Westerhaus, M. O. 1990. Improving repeatability of NIR calibrations across instruments. Pages 671–674 in Proc. 3rd Int. Conf. Near Infrared Spectroscopy, Brussels, Belgium. Agricultural Research Centre, Gembloux, Belgium.

Williams, P. 2007. Near-infrared technology—Getting the best out of light. PDK Grain, Namaimo, Canada.

This Article Abstract Full Text (PDF) Interpretive Summary Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via Google Scholar Google Scholar Articles by Soyeurt, H. Articles by Dardenne, P. Search for Related Content PubMed PubMed Citation Articles by Soyeurt, H. Articles by Dardenne, P.