Evaluation of X-ray fluorescence spectroscopy as a method for the rapid and direct determination of sodium in cheese

J. Dairy Sci. 98:5040–5051 http://dx.doi.org/10.3168/jds.2014-9055 © American Dairy Science Association®, 2015.

Evaluation of X-ray fluorescence spectroscopy as a method for the rapid and direct determination of sodium in cheese J. A. Stankey, C. Akbulut, J. E. Romero, and S. Govindasamy-Lucey1 Wisconsin Center for Dairy Research, University of Wisconsin, Madison 53706

ABSTRACT

Cheese manufacturers indirectly determine Na in cheese by analysis of Cl using the Volhard method, assuming that all Cl came from NaCl. This method overestimates the actual Na content in cheeses when Na replacers (e.g., KCl) are used. A direct and rapid method for Na detection is needed. X-ray fluorescence spectroscopy (XRF), a mineral analysis technique used in the mining industry, was investigated as an alternative method of Na detection in cheese. An XRF method for the detection of Na in cheese was developed and compared with inductively coupled plasma optical emission spectroscopy (ICP-OES; the reference method for Na in cheese) and Cl analyzer. Sodium quantification was performed by multi-point calibration with cheese standards spiked with NaCl ranging from 0 to 4% Na (wt/wt). The Na concentration of each of the cheese standards (discs: 30 mm × 7 mm) was quantified by the 3 methods. A single laboratory method validation was performed; linearity, precision, limit of detection, and limit of quantification were determined. An additional calibration graph was created using cheese standards made from natural or process cheeses manufactured with different ratios of Na:K. Both Na and K calibration curves were linear for the cheese standards. Sodium was quantified in a variety of commercial cheese samples. The Na data obtained by XRF were in agreement with those from ICP-OES and Cl analyzer for most commercial natural cheeses. The XRF method did not accurately determine Na concentration for several process cheese samples, compared with ICP-OES, likely due to the use of unknown types of Na-based emulsifying salts (ES). When a calibration curve was created for process cheese with the specific types of ES used for this cheese, Na content was successfully predicted in the samples. For natural cheeses, the limit of detection and limit of quantification for Na that can be determined with an acceptable level of repeatability, precision, and trueness was 82 and 246 Received November 2, 2014. Accepted April 13, 2015. 1 Corresponding author: [email protected]

mg/100 g of cheese, respectively. Calibration graphs should be created with standards that reflect the concentration range, ratio, and salt type present in the cheeses. This XRF method can be successfully used for the rapid and direct measurement of Na content in a wide variety of natural cheeses. Commercial process cheese manufacturers use proprietary blends of ES. We did find that the XRF technique worked for process cheese when the calibration graphs were created with the specific types of ES actually used. Key words: sodium, X-ray fluorescence spectroscopy, cheese INTRODUCTION

The Na content of cheese is typically determined coulometrically using a Cl analyzer (Johnson and Olson, 1985), which is rapid and easy to operate. However, this method indirectly determines Na by measuring the amount of Cl ions present in the cheese. This is a major problem when determining the Na content of cheeses containing potassium chloride (KCl), which is often used as a Na replacer. Additionally, the Cl analyzer is not used for process cheeses, which are manufactured with various Na-based emulsifying salts (ES), such as sodium citrate or sodium phosphate. Thus, when Na replacers are used in cheese, Na must be determined via an alternative method. In these cases, it is preferable to directly measure the Na content. Direct measurement of Na content in foods is commonly done using atomic absorption spectroscopy or inductively coupled plasma optical emission spectroscopy (ICP-OES). The ICP-OES and atomic absorption spectroscopy techniques require costly ultra-pure reagents, are laborious, and require complex sample preparation (e.g., ashing, digestion; Dolan and Capar, 2002; Ehling et al., 2010). Thus, a more practical chemical method is needed that is accurate (direct measurement of Na), quicker, and cheaper for use in routine quality control, which would allow cheese plants to easily monitor changes in salt content during the manufacturing process. The X-ray fluorescence analysis (XRF) method has previously been used to determine the concentration

5040

5041

DETERMINATION OF SODIUM IN CHEESE

of elements in a diverse number of applications including metal, cement, oil, polymer, plastics, mining, and minerals (Herbert and Street, 1974; Potts et al., 1984; Brouwer, 2003; West et al., 2010). X-ray fluorescence is also a rapid, precise, nondestructive, and a potential alternative method for direct Na determination in foods (Jastrzebska et al., 2003; Pashkova, 2009; Rinaldoni et al., 2009; Smagunova and Pashkova, 2013). Several studies have used XRF for the measurement of individual elements in milk and dairy products such as infant formula powders (Chan and Palmer, 2013; Fernandes et al., 2014). These XRF spectrometers have 2 basic components: the excitation source and the detection system. First, the sample is irradiated with X-rays and fluorescent X-rays are excited in the sample (Brouwer, 2003). These secondary X-rays are then measured for quantitative analysis of its elemental composition in the detection system (Brouwer, 2003). When an atom is hit with the incoming photon, an electron from the innermost orbital is expelled, which causes the atom to be in an excited, but unstable, state. To regain its stability, the atom must fill this gap by transferring an electron from the higher-energy outer orbital to the lower-energy inner orbital, which emits excess energy as an X-ray photon (Brouwer, 2003). Each element has discrete energy levels (e.g., dependent on the energy of the electron) so that the emitted radiation is characteristic for the particular element (e.g., “fingerprint”). A spectrum is created with distinct peaks for each element present in the sample, and the peak areas determine intensity; calibration is needed to relate intensity to actual concentrations. Sodium is the lightest element capable of being detected using energy dispersive XRF (Brouwer, 2003). X-ray fluorescence analysis was used for analysis of various elements in dairy foods; determination of P in melted and cottage cheeses (Jastrzebska et al., 2003), trace elements (i.e., Ba, Cu, Cr, Ni, and V) in Bryndza sheep cheese (Suhaj and Korenovska, 2008), and Na concentrations in several types of milks, dairy products, and infant formulas (Pashkova, 2009). These studies prepared cheese samples either by oven-drying, grinding, and pressing into pellets (Suhaj and Korenovska, 2008; Pashkova, 2009) or by microwave digestion of cheese (Jastrzebska et al., 2003) before XRF analysis. Some studies have successfully used XRF for the measurement of the mineral content of some dairy samples without any pretreatment; for example, Rinaldoni et al. (2009) successfully used XRF for the determination of Ca, K, P, Fe, and Zn in undiluted skim milks, ultrafiltered milks, and yogurts diluted with distilled water. This suggests that the XRF technique might be capable of Na analysis in cheeses without any pretreatment (e.g., no drying, ashing, or wet digestion).

The objective of this study was to develop, and validate, a method for the rapid, precise, and direct determination of Na in natural and process cheeses using XRF technology. Calibrations were created using a variety of cheese matrices containing different Na levels. These calibration cheese matrices were created by experimentally manufacturing a variety of unsalted cheeses (i.e., Cheddar, Mozzarella, Gouda, pizza, and nonfat) and spiking them with increasing levels of Na. The XRF equipment parameters, current, voltage, and runtime, were studied to improve the intensity of the Na measurements. Different sample preparation methods (shredded versus sliced, as well as the dimensions of the slices) were also evaluated. The method was validated in terms of linearity, repeatability, reproducibility, and trueness. The reference method, ICP-OES (Poitevin et al., 2009), was used to validate the XRF technology; the Cl analyzer method was used for comparison purposes. MATERIALS AND METHODS XRF Equipment

The Na contents of experimental cheese standards and commercial cheese samples were measured on an XRF instrument (X-Supreme8000, Oxford Instruments, Oxford, UK) and operating software (X-Supreme, version 1.99 build 45). The detector was purged with high-purity He gas (Airgas, Madison, WI). Preliminary experiments were carried out to determine XRF operating parameters, such as, voltage, current, and runtime (which influence peak intensity). The equipment was run at the following voltage and current levels: 4 kV at 750 μA for 180 s. Background corrections were applied for Na, K, and Cl. The software calculates the background corrections using the Lucas-Tooth and Price intensity-based X-ray correction model. The Lucas-Tooth and Price algorithm assumes that chemical variations within the sample (i.e., matrix effects) proportionally influence the fluorescent intensity of the element of interest (e.g., Na; Potts et al., 1984). The correction model uses terms that correct for absorption and enhancement effects from other elements (e.g., elements of interest are excited by incoming radiation but can also be excited by the fluorescence of other elements; Brouwer, 2003). The analysis volume of the X-Supreme8000 instrument was ~174 mm3, which was approximately an area of 62 mm2 by height of 0.7 mm. Thus, samples needed to be ≥0.7 mm thick and have a diameter of ≥8.8 mm. The samples were introduced into plastic cups (dimensions: 41 mm diameter × 28 mm height; Plastic Secondary Safety Window, Oxford Instruments). The sample cup consisted of 3 parts: film (4-μm-thick Film Journal of Dairy Science Vol. 98 No. 8, 2015

5042

STANKEY ET AL.

Poly-4, Oxford Instruments), inner cup, and outer cup. The plastic cups were reusable and the film was disposable. The outer cup was placed in a platform ring and film was laid across the platform. The inner cup was pushed down into position so that the film was a taut surface for the sample. The sample cup containing the prepared sample was placed in the XRF chamber, which has a 10-sample autosampler. After the chamber door was closed, the autosampler moved the sample to the analysis position where the sample was continually rotated while being irradiated. The continuous rotation of the sample helps to eliminate any nonhomogeneity effects from the surface of the sample (Brouwer, 2003). The accuracy of the XRF technique greatly depends on the uniformity and amount of the material the Xrays have to pass through. In this study, several sample properties, such as the format, thickness, and cheese composition, were evaluated to determine their influence on Na measurements. This particular piece of equipment is simple to operate, provides rapid (live) results, and is robust for potential quality analysis in a manufacturing plant type of setting. Preparation of Cheese Samples and Standards for XRF Calibration

Cheese standards, containing a wide range of Na concentrations, were prepared using natural cheeses (Cheddar, Gouda, pizza, Mozzarella, and nonfat), which were made using standard manufacturing protocols by licensed Wisconsin cheesemakers at the University of Wisconsin–Madison dairy processing plant. Because Na naturally occurs in milk, it was not possible to create a blank Na cheese matrix. Therefore, the performance of the method was evaluated using the standard addition method. The Na quantification was performed utilizing a multi-point calibration with unsalted cheeses. Freshly pressed blocks of cheese (≤1 d old) were obtained after milling, before dry salting. The cheese was finely shredded in a food processor (Cuisinart Inc., Greenwich, CT). Different amounts of NaCl (analytical grade, Fisher Scientific, Fair Lawn, NJ) were stirred into the shredded cheese with a spoon for 1 min (e.g., direct salting) to prepare cheese standards with different salt contents (i.e., 0 to 4.0 g of NaCl/100 g of cheese). The salted cheeses were then pressed in tri-cornered polypropylene beakers (250 mL, Fisher Scientific) using a hydraulic press (670 kPa for 1 h at 22°C). The pressed cheese blocks were sealed in resealable plastic bags and stored for 1 wk at 4°C to allow for Na equilibration. Using this dry-salting method, a maximum amount of salt (i.e., 4.0 g/100 g of cheese) could be added. Above this level, cheese texture became crumbly and difficult to work with. The XRF readings for these Journal of Dairy Science Vol. 98 No. 8, 2015

high-salt (i.e., ≥4.0 g/100 g of cheese) standards had much larger standard deviations due to the crumbly nature of the cheese sample. To address this issue (i.e., to have a calibration with >6% Na), several process cheese standards were made. Laboratory-scale process cheeses were prepared using the method described by Brickley et al. (2008). These high-salt cheese standards were made by adding different amounts of a single type of Na-based ES, trisodium citrate (analytical grade, Fisher Scientific), to unsalted pizza cheese base (i.e., 1.0, 2.0, 3.0, 5.0, and 7.0 g of trisodium citrate/100 g of cheese). The molten cheeses were then poured into tricornered polypropylene beakers, covered with plastic wrap, and sealed in plastic bags that were stored for 1 wk at 4°C to allow for Na equilibration. Cheese Standards with Varying Na and K Concentrations. A separate calibration plot for various concentrations of K was created by replacing some NaCl with KCl (i.e., 460, 510, 560, 620, 660, 680, and 750 mg of Na/100 g of cheese and 340, 310, 250, 230, 200, 140, and 105 mg of K/100 g of cheese, respectively) in an unsalted Cheddar cheese sample. This calibration plot was validated with a Cheddar cheese that was manufactured in the University of Wisconsin-Madison dairy plant, containing K-based salt replacers. Process Cheese Standards with K-Based Emulsifying Salts. A regular-salt Cheddar cheese (1.8% salt) was manufactured at the University of Wisconsin-Madison dairy plant and then used as the cheese base for the creation of various process cheese standards. Two ES, dipotassium phosphate (DPP; BK Giulini, Ludwigshafen, Germany) and disodium phosphate (DSP; BK Giulini), were blended in varying concentrations up to 2.0% (wt/wt) DSP and up to 1.75% (wt/wt) DPP. The process cheese standards (n = 8) were prepared using the small-scale process cheese method described by Brickley et al. (2008). The cheeses were sealed in plastic bags and stored at 4°C for 1 wk. These cheeses were sampled and analyzed by ICP-OES to quantify the actual amount of Na and K in order to create a calibration for measuring Na in the presence of K for commercial process cheeses (n = 8). Samples and Sample Preparation. To optimize the intensity of the Na peak in the samples, cheeses were introduced into the XRF equipment in different formats, as follows: (a) Discs (30 mm): cheese discs (30 mm diameter × 7 mm height) were prepared from the pressed block or commercial samples using a Hobart deli slicer (model 410, Hobart Manufacturing Co., Troy, OH) and a cork borer. The XRF equipment requires that samples have a diameter between 18.5 and 40 mm (the diameter of the

5043

DETERMINATION OF SODIUM IN CHEESE

sample cup). We decided to cut the cheeses to a diameter of 30 mm. For process cheese singles, individual cheese slices were stacked so that they had a height of approximately 7 mm. Five 30-mm-diameter samples were taken from each stack. Process cheese food was cooled in a freezer at −20°C for 10 min before sampling to firm up the cheese and allow for easier and more uniform slicing. The cheese discs were stored in resealable plastic bags at 4°C until analysis. (b) Shredded cheese: cheese samples were shredded using a food processor (model DLC-2011CHB, Cuisinart Inc.) for 10 to 15 s to produce uniform cheese particles. The shredded cheese (5 or 10 g) samples (n = 5) were added to the prepared sample cups and measured by the XRF. This experiment was repeated but the shredded cheese (5 or 10 g) samples were pressed to remove any air gaps using a nylon plug (54-ZX1223, Oxford Instruments), which fit tightly inside the inner cup. In addition to the cheese standards, various commercial samples of natural (n = 7) and process (n = 3) cheese were purchased locally and sampled in disc form (30 mm diameter × 7 mm height). XRF Method Validation

Single laboratory validation (SLV) was carried out because our laboratory was the only laboratory with access to this XRF instrument; the SLV procedure was performed according to AOAC guidelines (AOAC International, 2012). Measurements were carried out using blind duplicates of commercially purchased natural cheeses (n = 13) with varying chemical composition and Na contents. Blocks of cheese were divided in half, repackaged, labeled with random 2- or 3-digit numbers, and used to prepare the samples (slicing and sampling) for XRF analysis. At least 12 samples were taken from each cheese, sealed in a resealable plastic bag, and stored at 4°C until analysis. Duplicate samples from each bag were measured on the XRF equipment using the same calibration method (calibration range was from 250 to 1,600 mg of Na/100 g of cheese). Measurements taken with the XRF equipment were carried out 4 times (instrument sessions) spanning 2 wk. Inter- and intraday parameters were referred to as repeatability (r) and within-laboratory reproducibility (R), respectively (Horwitz and Albert, 2006). Parameters including r, R, relative standard deviation (RSDr and RSDR), relative repeatability (rrel), and relative within-laboratory reproducibility (Rrel) values were calculated according to standard protocols (IDF, 1991). The Horitz ratio

(HorRatR), which indicates reliability of these values (Horwitz and Albert, 2006), was determined using the equation HorRatR = RSDR ÷ PRSDR = RSDR ÷ 2C−0.15,

[1]

where PRSDR is the predicted relative within-laboratory standard deviation (Masotti et al., 2012). The PRSDR is equivalent to 2C−0.15, where C is the mass fraction of the Na concentration. The HorRatR is a unitless value used to express the acceptability of a chemical method of analysis with respect to precision (Horwitz and Albert, 2006). The expected range for a HorRatR value in a SLV study is between 0.3 and 1.5, ideally <1.0 (Poitevin et al., 2009). The XRF software reports raw data as intensity counts per second (cps; the area under the curve) versus energy (keV) and converts intensity to Na concentration using the Lucas–Tooth and Price matrix correction algorithm. The limit of detection (LOD) and the limit of quantification (LOQ), expressed as milligrams per 100 g of cheese, were estimated using the standard deviation (SD) from the analysis of at least 4 reagent blanks (unsalted or unbrined cheese standards made without any additional salt), run on a single day, and measured over 3 separate days (instrument sessions). The SD for intensity of the background noise for these blanks was used to calculate the LOD and LOQ. Specifically, LOD was calculated as 3 × SD and the LOQ was calculated as 3 × LOD. ICP-OES Analysis

Analysis of Na and K was done using the standard AOAC method 984.27 (Poitevin et al., 2009) for ICPOES (Vista-MPX Simultaneous ICP-OES, Varian Inc., Palo Alto, CA). Readings were taken at 589.6 and 769.9 nm for Na and K, respectively (Mozafar et al., 1990; Govindasamy-Lucey et al., 2007). The same samples used for XRF analysis were saved for ICP-OES testing. The values obtained from ICP-OES analysis for Na and K were used to create the calibration curves in the XRF software. Chloride Analyzer

Sodium contents of all cheese standards and commercial samples were also measured indirectly by potentiometric Cl titration method (M926 Chloride Analyzer, Nelson & Jameson Inc., Marshfield, WI; ISO, 2006; Johnson and Olson, 1985). These values were used for comparison with the results from the XRF analysis method. Journal of Dairy Science Vol. 98 No. 8, 2015

5044

STANKEY ET AL.

Statistical Analysis

Unless otherwise specified, 3 replicates were performed for each sample. An ANOVA was carried out with PROC GLM in SAS (version 9.3; SAS Institute Inc., Cary, NC). Scheffé’s multiple-comparison test was carried out to evaluate differences in the treatment means at a significance level of P < 0.05. RESULTS AND DISCUSSION XRF Operating Parameters: Effect of Current, Voltage, and Runtime on Na Peak Intensity

Preliminary work was carried out to enhance the Na peak intensity by changing the current, voltage, and runtime. Figure 1a shows an XRF spectrum scan obtained for a commercial sample of low-moisture, partskim (LMPS) Mozzarella. Sodium, being one of the lightest elements, was detected first (at photon energy ~1.6 × 10−19 kJ) and the intensity was also low (~5 cps), compared with the other elements. A portion of Figure 1a was enhanced to show the region of interest (ROI), where the Na peak was detected (Figure 1b). The ROI for the Na peak was between 1.53 and 1.76 × 10−19 kJ. Further experiments were carried out in which the current, voltage, and runtime were changed to try to increase the peak area and overall Na intensity at the ROI (Figure 2). The area under the curve increased with increasing voltage (Figure 2a) and decreased with increasing current (Figure 2b). The runtime was increased from 90 to 180 s, which increased the overall number of readings (cps) taken, which improved the signal to noise ratio (data not shown). Subsequent measurements for the calibrations of Na in cheeses, and for the SLV procedure, were carried out using the following conditions: current of 750 μA and a voltage of 4 kV for a 180-s runtime. The XRF equipment used for these experiments had an autosampler feature (but without an internal temperature controller), which allowed 10 samples to be loaded and then analyzed consecutively. Cheese samples exhibited oiling off and possible moisture loss after 10 consecutive runs, which suggests that the samples were being heated within the instrument chamber (initial cheese temperature: 5°C; cheese temperature after 10th measurement of the same sample: ~35°C). A single full-fat Cheddar cheese sample was measured 10 consecutive times (run time: 90 s; total residence time: 15 min) to determine SD between the repeated intensity readings. A layer of oil (oiling off) was observed on the sample film after the final reading and Na intensity decreased with increasing residence time in the autosampler (results not shown). Similarly, Journal of Dairy Science Vol. 98 No. 8, 2015

Figure 1. An example of (a) a typical X-ray fluorescence spectroscopy (XRF) spectrum scan showing all the different elements for a commercial low-moisture, part-skim Mozzarella cheese (2% Na, wt/ wt) sample, and (b) an enlarged portion of the spectrum scan to show the Na peak. Test was run at 5 kV at 600 μA for 90 s on a cheese disc (30 mm diameter × 7 mm height) at ambient temperature. cps = counts per second.

Smagunova and Pashkova (2013) found discoloration in both skim and whole milk powder-pellets and oiling off in whole milk powder-pellets after successive measurements and long measurement times (>20 min per sample; sample chamber temperature was ~38°C). Smagunova and Pashkova (2013) reported that the melting of fat contributes to scattering (enhancement of the XRF intensity) effects when measuring elements like Na due to primary and secondary radiation. Oiling off was not observed when the experiment was repeated with a reduced-fat Cheddar cheese sample, and for this cheese type Na intensity was not affected by 10 consecutive runs (P > 0.05). To prevent excessive oiling off and to ensure consistency between samples, use of

DETERMINATION OF SODIUM IN CHEESE

Figure 2. Changes in the intensities and peak areas of the Na peak (region of interest: 1.53 to 1.76 10−19 kJ) with (a) changing voltage levels (4 to 9 kV) at constant current (300 μA) and (b) changing current (400 to 600 μA) and voltage levels (5 to 7 kV). Both experiments ran for 90 s using a commercial low-moisture, part-skim Mozzarella cheese (2% Na, wt/wt) disc (30 mm diameter × 7 mm height) at ambient temperature. cps = counts per second.

the autosampler was discontinued and only one sample was analyzed at a time (run time: 180 s; total residence time: 4 min; cheese temperature after analysis: ~10°C). Effect of Cheese Form, Thickness, and Composition on Na Peak Intensity

The fluorescence emitted by a material and its subsequent detection is greatly dependent on the sample density, thickness, and composition (SzczerbowskaBoruchowska, 2012). We explored the effect of different formats and dimensions of samples. Cheese Forms. A low-fat (6% fat) Cheddar cheese was prepared in 2 different forms: sliced into discs or shredded using a food processor. Sodium intensity was

5045

higher when samples were in the sliced disc form (66 cps) compared with when the same samples were shredded and introduced in the XRF cup (34 cps). When the cheese was shredded and placed in the sample cup, air gaps remained between the individual shredded particles, which influenced the intensity. Na intensity was improved when a nylon plug (54-ZX1223, Oxford Instruments) was used to manually compress the shredded cheese. Despite varying the mass of shredded cheese used, the intensity remained significantly lower (5 g = 46.9 cps, 10 g = 46.6 cps) than when the samples were in a disc form (P < 0.05). Thus, the mass of shredded cheese sample used did not have a significant effect on the intensity of the Na peak. Cheese Thickness. More X-ray radiation will generally be absorbed as the sample thickness increases because more radiation is unable to leave the sample (Brouwer, 2003). A commercial Colby cheese sample was sliced to 3 different thickness using a deli slicer (3, 7, or 17.5 mm) and cut into 30-mm-diameter discs. Samples (n = 5) were run using the predetermined operating conditions (4 kV at 750 μA for 180 s) and replicated 3 times. The area under the Na peak varied with sample thickness. When compared with the actual Na concentration (744 mg of Na/100 g of cheese, determined by ICP-OES), the XRF concentrations using the thicker slices (7 and 17.5 mm) were closer to the actual concentration (697 and 657 mg of Na/100 g of cheese, respectively). Thinner (3 mm) slices overestimated the Na concentration (825 mg of Na/100 g of cheese) compared with the actual Na concentration (P < 0.05). Thicker samples (i.e., >5 mm) are usually preferred for analysis because more radiation will be reflected back and reach the detector than thinner samples (Brouwer, 2003). Cheese Composition. It is desirable to create a single method using the XRF equipment that can measure the Na concentration in a wide range of cheeses, regardless of manufacturer, cheese type, or age. To determine whether compositional differences (e.g., moisture, fat, or protein contents; Table 1) in the cheese matrix affected the accuracy of the XRF equipment in determining changes in Na concentration, we created 6 unique calibration curves using a variety of cheese types (i.e., Cheddar, Gouda, pizza, Mozzarella, nonfat, and process cheese), which varied in moisture, fat, and protein contents. The 6 calibration curves ranged from 0 to 4% NaCl (wt/wt); other mineral components were not modified. The Na concentrations for the 6 calibration curves were combined into a master linear best-fit calibration curve, shown in Figure 3, along with their corresponding regression equations and coefficient of determination values. The master curve was linear (R2 ≥ 0.99), which suggests that the compositional differJournal of Dairy Science Vol. 98 No. 8, 2015

5046

STANKEY ET AL.

Table 1. Composition of the experimental cheeses (before salt addition) Cheese (no salt added) Cheddar Gouda Mozzarella Nonfat Pizza5

Moisture1 (%)

Protein2 (%)

Fat3 (%)

Na4 (mg/100 g)

40.9c 44.2b 44.7b 58.9a 43.3bc

23.5b 23.5b 30.3a 34.6a 31.0a

31.1a 29.4b 20.9c 1.21d 21.7e

30.1a 25.7a 38.5a 38.6a 30.6a

a-e

Means within a column with different superscripts differ (P < 0.05). Determined as 100 – total solids. 2 Total percentage N × 6.35 by Kjeldahl. 3 Determined by Mojonnier. 4 Determined by Cl analyzer. 5 Used as a base to make the laboratory-scale process cheese. 1

ences between cheese types did not affect the ability of the XRF to successfully determine Na. These findings are similar to other studies looking at the effect of fat content in milk powders on the determination of several minerals using XRF (Gunicheva, 2010; Smagunova and Pashkova, 2013). Based on this result, any natural cheese type (regardless of gross composition) can be successfully analyzed using any of the 6 calibration methods created with an individual cheese type. XRF Method Validation

Linearity Range and Calibration Curve. Linearity (R2 ≥ 0.99) was observed in all the calibration

Figure 3. Calibration curve for Na concentration established using a variety of cheeses: Gouda (), Mozzarella (‫)چ‬, Cheddar (), process (‫)ڐ‬, pizza (), and nonfat () spiked with NaCl and ranging in total Na content from 30 to 1,800 mg of Na/100 g of cheese. The Na content was calculated from X-ray fluorescence spectroscopy (XRF) intensity versus Na content determined by inductively coupled plasma-optical emission spectroscopy (ICP-OES). Solid line (and regression equation) represents the linear regression curve for all types of cheese tested (i.e., master calibration curve). Values are means of 3 replicates; error bars indicate SD. Journal of Dairy Science Vol. 98 No. 8, 2015

curves when background corrections were applied (Figure 3). The range of Na in the calibration curves (100 to 1,500 mg of Na/100 g of cheese) encompassed the typical Na contents found in most commercial natural cheeses (300 to 700 mg of Na/100 g of cheese) as well as commercial process cheeses (930 to 1,600 mg/100 g; Agarwal et al., 2011). One calibration curve was selected that had the largest range of Na concentrations to verify the trueness of the method. The regression equation for the calibration curve was y = 0.9856x + 21.569. The coefficient of determination calculated from the weighted linear regression analysis was 0.99 (linear over the range). Sensitivity (LOD and LOQ). Dairy products contain at least a trace amount of Na; thus, the creation of a true blank was not possible. Using the conditions described in the established method, the LOD, calculated as 3 × SD of an unsalted cheese sample, was equal to 82 mg of Na/100 g of cheese. The LOQ for an established method (i.e., the smallest amount of Na that can be quantitatively determined in cheese with suitable precision and accuracy; AOAC International, 2012) was calculated as 3 × LOD, which was equal to 246 mg of Na/100 g of cheese. The procedure for XRF as described in this study was therefore suitable for the analysis of low-Na cheeses (i.e., containing ≤280 mg/100 g of cheese). Trueness. The trueness (that is, how close a result is to the true value) was determined on a wide range of commercial cheese (both natural and process) samples (n = 10) for comparison with the ICP-OES and Cl analyzer methods (Table 2). The Na data obtained by XRF analysis were in close agreement with ICP-OES for most of the commercial cheeses except for Muenster (brand B), low-Na Cheddar cheese (brand C), and the process cheeses: deli deluxe reduced fat pasteurized process American cheese slices (brand A), deluxe pasteurized process American cheese product slices (brand B), and pasteurized prepared American singles cheese

5047

DETERMINATION OF SODIUM IN CHEESE

Table 2. Comparison of Na (mg/100 g) in commercial cheeses measured by inductively coupled plasma-optical emission spectroscopy (ICPOES), X-ray fluorescence spectroscopy (XRF), and Cl analyzer Na2 (mg/100 g of cheese) Cheese type Brick Colby Deli deluxe reduced fat pasteurized process American cheese slices Deluxe pasteurized process American cheese product slices Low-moisture, part-skim Mozzarella Medium aged Cheddar Monterey Jack Muenster Pasteurized prepared American singles cheese product Low-Na Cheddar cheese

Cl analyzer

SEM

P-value

Ca3

K3

630a 657a 544c

23.44 23.94 10.25

NS NS <0.0001

561 545 1741

56 48 123

1,509a

684b

46.13

<0.01

1192

72

563a 540a 638a 581b 1,018a

733a 633a 682a 679a 604b

633a 642a 679a 643a 527b

43.24 27.18 7.93 7.94 24.50

NS NS NS <0.01 <0.02

635 652 582 591 1202

47 42 39 68 310

262a

304b

306b

3.28

<0.001

771

97

Brand1

ICP-OES

XRF

B B A

568a 569a 1,448a

620a 652a 1,366b

B

1,392a

B B B B A C

a-c

Means within a row with different superscripts differ in Na concentration (P < 0.05). Letters A, B, and C refer to the different manufacturers/distributors of the cheeses. 2 Mean values of 3 determinations. 3 Measured by ICP-OES method. 1

product (brand A). The discrepancy between ICP-OES and XRF data for the Muenster cheese may have been due to a sampling issue, as Muenster cheese is typically brine-salted (our calibration method was created by direct-salting of curd). The commercial cheese samples were prepared for XRF analysis by slicing blocks purchased from the grocery store into discs. Samples for ICP-OES analyses were shredded first and then ashed, which eliminates any possible variation in Na due to the sample location from within a block of brine-salted cheese. The Cl analyzer accurately predicted Na content of most of the natural cheeses (made without any K-salts) when values were compared with ICP-OES with the exception of the Muenster (brand B) and low-Na Cheddar cheese (brand C). To verify the need for the standards to be an accurate representation of our samples, an additional set of XRF calibration standards were formulated by varying both K and Na concentrations. Initially, aqueous solutions were prepared that contained NaCl and KCl at different levels (up to 50% substitution) to determine whether the XRF method was capable of accurately measuring the changes in Na intensity when K was present at much higher quantities. The intensities for each of the Na-K blend standards were plotted against the known concentrations. Linearity (R2 > 0.98) was observed for both Na and K in the blend standard solutions, which demonstrated that XRF could accurately predict Na when K was present in increasing quantities in solutions (results not shown). Cheese standards were then prepared from an unsalted Cheddar with different

ratios of Na:K as described in the methods section. Due to the different composition of the cheese matrix (i.e., increased K concentration), an additional correction factor (i.e., Ca) was applied to the regression model for these standards. Linearity for Na (R2 = 0.98) and K (R2 = 0.99) were both observed in the Cheddar cheese calibrations made with varying Na to K ratios (Figure 4a). Trueness was assessed on experimental Cheddar cheeses manufactured with different K-salt based replacers. Data obtained from XRF was compared against ICP-OES and Cl analyzer methods (Table 3). This XRF calibration method allowed for closer (better) prediction (compared with ICP-OES) of the Na content (P > 0.05) in the presence of high levels of K. The XRF method was more accurate when compared with the Cl analyzer, which was unable to predict (P < 0.05) Na levels accurately in cheeses manufactured with K-salt replacers (Table 3). Precision Parameters (Repeatability and Reproducibility). The amount of Na determined in duplicate analyses of each of the 13 commercial natural cheeses on 4 different days is presented in Table 4. The RSDr varied between 0.9 and 13.3%, and the RSDR ranged from 0.9 to 13.3%. For the 15 natural cheese samples, the calculated HorRatR value ranged from 0.2 to 2.9 (mean = 0.74). The HorRatR value is a performance parameter, which indicates the acceptability of a chemical method of analysis with respect to precision. Generally, a HorRatR <1 is considered acceptable for SLV studies (Masotti et al., 2012). Some other studies have suggested that the extrapolations of HorRatR Journal of Dairy Science Vol. 98 No. 8, 2015

5048

STANKEY ET AL.

Table 3. Comparison of Na (mg/100 g) of experimental Cheddar cheeses made with varying concentrations of K (mg/100 g) measured by inductively coupled plasma-optical emission spectroscopy (ICP-OES), X-ray fluorescence spectroscopy (XRF), and Cl analyzer Na2 (mg/100 g of cheese) Sample High-K Cheddar Low-K Cheddar

K1 357 70

ICP-OES b

433 643a

XRF b

468 697a

Cl analyzer a

597 631a

SEM

P-value

5.10 15.8

<0.01 NS

a,b

Means within a row with different superscripts differ in Na concentration (P < 0.05). Measured by ICP-OES method. 2 Mean values of 3 determinations. 1

values to a SLV study are expected to be from 0.3 to 1.5 (Poitevin et al., 2009). Consistent deviations from the ratio on the low side (values <0.3) may indicate unreported averaging or excellent training and operator

Figure 4. Calibration curves for Na () and K () established using (a) natural cheeses spiked with NaCl and KCl or (b) process cheeses made with disodium phosphate and dipotassium phosphate emulsifying salts. The Na or K content was calculated from X-ray fluorescence spectroscopy (XRF) intensity versus the Na or K content determined by inductively coupled plasma optical emission spectroscopy (ICP-OES). Solid lines represent the linear regression curves. Values are means of 4 replicates; error bars indicate SD. Journal of Dairy Science Vol. 98 No. 8, 2015

experience, whereas consistent deviations on the high side (values >1.5) may indicate inhomogeneity of the test samples, need for further method optimization, more training, operating below the limit of determination, or an unsatisfactory method. The HorRatR values were ≥1.0 for the following cheeses: Asiago (brand G), LMPS Mozzarella (brand F), Swiss (brand H), and Parmesan (brand B) cheeses (as denoted by the italicized font in Table 4). The salt contents of the 2 LMPS Mozzarella cheeses tested were very different (417 and 828 mg/100 g, Table 4). The LMPS Mozzarella cheeses samples originated from 2 different manufacturers. Commercial cheeses of the same variety have a wide range of salt levels (Agarwal et al., 2011). Additionally, the LMPS Mozzarella, Asiago, Swiss, and Parmesan cheeses were all brine salted. Brine salting results in a Na gradient in cheeses (Guinee, 2004). Because the sample preparation for the XRF method involved using a disc of the following dimensions being cut (30 mm diameter × 7 mm height) from somewhere in the block, the within-sample variation could have been due to an existing salt gradient within the cheese block. This within-sample (block) variation can significantly affect the repeatability, reproducibility, and subsequently the calculated HorRatR parameters. When the brine-salted cheeses were excluded from Table 4, the HorRatR range decreased to 0.2 and 0.8 (average = 0.46). Brine-salted cheeses for XRF analysis should therefore be ground (like the preparation step for ICP-OES digestion) to have homogeneous samples and eliminate salt-gradient effects. Calibration curves should be made specifically for brine-salted cheeses thus ensuring that the sample taken for analysis is homogeneous and fully representative of the cheese block. A suggestion is grinding a portion of the cheese block (as is done in Cl analysis sample preparation), weighing a predetermined amount of the ground cheese into a sample cup, and removing most of the surface irregularities (air gaps) with a nylon plunger. It should be noted that the calibration method must be created using the same sample preparation procedure, as will be used for cheese analysis.

5049

DETERMINATION OF SODIUM IN CHEESE

Table 4. Statistical analysis of data corrected for outliers for determining precision parameters1 for the within-laboratory study on 13 commercial natural cheese samples for the determination of Na in cheese by X-ray fluorescence spectroscopy (XRF) Cheese type 3

Asiago Brick Colby LMPS4 Mozzarella3 LMPS4 Mozzarella Mild aged Cheddar Monterey Jack Muenster Parmesan Parmesan3 Provolone Sharp Cheddar Swiss3

Brand2

Na content (mg/100 g)

r (mg/100 g)

RSDr (%)

rrel (%)

R (mg/100 g)

RSDR (%)

Rrel (%)

PRSDR (%)

HorRatR

G B H F H H H B I B J B H

1,323 788 782 417 828 648 692 800 653 698 662 701 325

131 28.7 20.8 155 70.6 21.6 30.8 20.1 36.5 97.3 37.1 22.8 68.0

3.5 1.3 0.9 13.3 3.0 1.2 1.6 0.9 2.0 5.0 2.0 1.2 7.5

9.9 3.7 2.7 37 8.5 3.3 4.4 2.5 5.6 14 5.6 3.3 21

150 28.7 22.2 155 74.7 21.6 30.8 20.1 55.3 106 37.1 22.8 80.5

4.0 1.3 1.0 13.3 3.2 1.2 1.6 0.9 3.0 5.4 2.0 1.1 8.9

11 3.7 2.8 37 9.0 3.3 4.4 2.5 8.5 15 5.6 3.3 25

3.8 4.1 4.1 4.5 4.1 4.3 4.2 4.1 4.2 4.2 4.2 4.2 4.7

1.0 0.3 0.2 2.9 0.8 0.3 0.4 0.2 0.7 1.3 0.5 0.3 1.9

1

r = repeatability; RSDr = relative standard deviation of repeatability; rrel = relative repeatability; R = within-laboratory reproducibility; RSDR = relative standard deviation of within-laboratory reproducibility; Rrel = relative within-laboratory reproducibility; PRSDR = predicted RSDR; HorRatR = Horwitz ratio (RSDR/PRSDR). 2 Letters B, F, G, H, I, and J refer to the different manufacturers/distributors of the cheeses. 3 Italicized rows indicate data sets with HorRatR ≥ 1.0. 4 Low-moisture, part-skim.

The rrel and Rrel values ranged from 2.5 to 37% and 2.5 to 37%, respectively, which represent expected variability between results when a sample was analyzed in duplicate; these values were higher for the brine-salted cheeses (as denoted by the italicized font in Table 4). When the values from brine-salted cheeses are excluded, then rrel and Rrel values ranged from 2.5 to 8.5% and 2.5 to 9.0%, respectively. XRF Analysis of Process Cheeses

The XRF method was unable to accurately predict the Na content of 2 of the commercial process cheese samples: deli deluxe reduced fat pasteurized process American cheese slices (brand A) and pasteurized prepared American singles cheese product (brand A; Table 2). The incorrect estimation of Na concentration by XRF is due to an enhancement effect because the intensity of the X-rays of the element of interest (i.e., Na) depends not only on the concentration of Na in the sample, but also on the overall mineral composition (Brouwer, 2003). Variations in the matrix mineral composition (e.g., addition of different ES in process cheese manufacture) result in changes in the mean absorption coefficients of both the primary radiation source and the fluorescence radiation of the element of interest. We did not observe these problems in estimating Na concentration in natural cheeses because it is a simpler mineral system; in our standards the only elements with varying concentrations were Na and Cl. The absorption effects from the additional elements (e.g., found in ES) may result in either a decrease or an increase in the

measured Na intensity, depending on whether the elemental composition changes diminish or augment the mass absorption coefficient (Brouwer, 2003). We hypothesized that the XRF underestimated Na concentration because the XRF mineral composition of these standards did not accurately reflect those of the samples being analyzed. The deli deluxe reduced fat pasteurized process American cheese slices (brand A) and pasteurized prepared American singles cheese product (brand A) both contained higher K concentrations (123 and 310 mg/100 g of cheese, respectively as determined by ICP-OES; Table 2) than the calibration standards we used for this XRF method (range: 30 to 60 mg/100 g of cheese). Despite using K or Ca (which were also present in higher concentrations in these process cheeses) as background correction factors, the calculated Na values were still less (P < 0.05) than the reference ICP-OES values. As expected, the Cl analyzer underestimated the Na content of the process cheeses (Table 2), which was due to the indirect method for Na calculation used by the Cl analyzer instrument. Because process cheeses are manufactured with ES that contain Na but do not contain Cl (e.g., Na-phosphate or Na-citrate), Na cannot be accurately calculated using the Cl analyzer method (Table 2). Commercial process cheeses are manufactured with a wide range (blends) of ES; they sometimes also contain K-based salt replacers. A calibration plot was created using process cheese standards made with varying concentrations of Na- and K-based ES. A linear relationship (R2 ≥ 0.99) was obtained between known Na or K Journal of Dairy Science Vol. 98 No. 8, 2015

5050

STANKEY ET AL.

Table 5. Comparison of Na content (mg/100 g) of commercial process cheeses measured by inductively coupled plasma optical emission spectroscopy (ICP-OES), X-ray fluorescence spectroscopy (XRF), and Cl analyzer Na2 (mg/100 g of cheese) Cheese type

Brand1

American pasteurized process cheese product singles Deli deluxe sharp Cheddar pasteurized process cheese slices Pasteurized prepared American cheese product Pasteurized prepared American singles cheese product Pasteurized prepared American singles cheese product Pasteurized process American cheese food singles Pasteurized process American cheese product slices Swiss style pasteurized process cheese product singles

D A A A A E D D

ICP-OES 1,326a 1,527a 1,331a 1,033a 1,036a 1,267a 1,162a 1,249a

XRF 965b 1,185b 1,034b 1,134b 1,280b 1,024b 1,296b 1,241a

Cl analyzer 825c 920c 692c 635c 595c 808c 694c 761b

SEM

P-value

K3

43.4 64.1 59.1 19.0 39.9 38.8 18.7 40.1

<0.05 <0.05 <0.05 <0.01 <0.01 <0.05 <0.01 <0.05

342 124 340 320 300 235 156 206

a–c

Means within a row with different superscripts differ in Na concentration (P < 0.05). Letters A, E, and D refer to the different manufacturers/distributors of the cheeses. 2 Mean values of 3 determinations. 3 Measured by ICP-OES method. 1

concentration (obtained by ICP-OES) and calculated concentrations from XRF technique when the background corrections were applied for these experimental process cheese samples (Figure 4b). Trueness was assessed on a wide range of commercial process cheese samples by comparison of the XRF method against the ICP-OES and Cl analyzer methods (Table 5). Our experimentally created process cheese calibration was only able to accurately (P > 0.05 compared with Na result from ICP-OES) calculate Na in 1 out of the 8 commercial process cheese samples, that is, Swiss-style pasteurized process cheese product singles (brand D; Table 5). Two explanations for this are possible. First, the specific types of ES that were used to make these commercial process cheese samples likely affected the XRF results. Second, the process cheese standards for our calibration curves were made using only 2 specific types of ES (i.e., DSP and DPP). During the manufacture of commercial process cheeses, blends of often 3 or more different types of ES are used. Our sample calibration standards did not accurately reflect the matrix chemical composition of all of the different commercial process cheese samples. When ES are added during the manufacture, a range of different type of salt complexes can be formed, involving Cacasein crosslinking and new forms of phosphate salts [e.g., Ca-pyrophosphate complex (Lucey et al., 2011)]. In addition, some ES salts remain undissolved in process cheese and some other soluble forms of calcium phosphate may be present. Thus, all of these complex salt interactions in the process cheese would likely create interferences for the XRF technique unless custom calibration curves can be created. It is likely that the accuracy of the XRF would improve if calibration standards were a true representation of the samples being

Journal of Dairy Science Vol. 98 No. 8, 2015

measured, as occurred in our experimental process cheese work (Table 5). This method shows promise for the direct and rapid measurement of process cheeses at a manufacturing plant where standards can be used for calibration that are representative of the process cheese samples being manufactured. A process cheese manufacturer would be unlikely to encounter the problems we faced because they would create calibration standards using their manufactured cheeses as the base material. As expected, the Cl analyzer significantly (P < 0.05) underestimated Na concentration in process cheeses when compared with the ICP-OES reference values. The XRF method was compared with the Cl analyzer method (Table 5). As expected, the Cl analyzer was unable to accurately predict Na levels in any process cheeses containing higher K levels or when non-Cl-based Na-ES (e.g., disodium phosphate, trisodium phosphate, sodium metaphosphates, or trisodium citrate) are used in process cheese manufacture. CONCLUSIONS

The XRF technique was demonstrated to be a rapid and reliable method for the routine determination of Na in natural cheeses. No complex sample preparation was required; a slice of cheese could easily be used to accurately measure Na content. However, for brinedsalted cheeses, we suggest that the cheeses be ground so as to present homogeneous samples and eliminate salt-gradient effects with discs. Although the XRF method worked for our experimental process cheeses, cheese plants would need to create unique calibration curves specific to the types (blends) of ES used in process cheese. When we made process cheeses with known

DETERMINATION OF SODIUM IN CHEESE

types of ES, we were able to accurately determine the Na in the cheeses because we created calibration standards with these ES. The sensitivity of the technique allowed for the rapid (<5 min) determination of Na in natural cheeses with an LOQ of 245 mg of Na/100 g of cheese, which would make it suitable for the analyses of even low Na cheeses (e.g., ≤280 mg of Na/100 g of cheese). The technique was relatively easy to use once a standard method and calibration graph were developed. The good linearity and satisfactory performance in terms of recovery, trueness, and precision factor support the adoption of this method for in-plant monitoring of Na in natural cheeses where Na replacers are being used. REFERENCES Agarwal, S., D. McCoy, W. Graves, P. D. Gerard, and S. Clark. 2011. Sodium content in retail Cheddar, Mozzarella, and process cheeses varies considerably in the United States. J. Dairy Sci. 94:1605– 1615. AOAC International. 2012. AOAC guidelines for single laboratory validation of chemical methods for dietary supplements and botanicals. Pages 1−32 in Appendix K. Official Methods of Analysis of AOAC International. G. W. Latimer, ed. AOAC International, Gaithersburg, MD. Brickley, C. A., S. Govindasamy-Lucey, J. J. Jaeggi, M. E. Johnson, P. L. H. McSweeney, and J. A. Lucey. 2008. Influence of emulsifying salts on the textural properties of nonfat process cheese made from direct acid cheese bases. J. Dairy Sci. 91:39–48. Brouwer, P. 2003. Theory of XRF: Getting Acquainted with the Principles. PANalytical B.V., Almelo, the Netherlands. Chan, J. C., and P. T. Palmer. 2013. Determination of calcium in powdered milk via X-ray fluorescence using external standard and standard addition based methods. J. Chem. Educ. 90:1218–1221. Dolan, S. P., and S. G. Capar. 2002. Multi-element analysis of food by microwave digestion and inductively coupled plasma-atomic emission spectrometry. J. Food Compos. Anal. 15:593–615. Ehling, S., S. Tefera, R. Earl, and S. Cole. 2010. Comparison of analytical methods to determine sodium content of low-sodium foods. J. AOAC Int. 93:628–637. Fernandes, T. A. P., J. A. A. Brito, and L. M. L. Gonçalves. 2014. Analysis of micronutrients and heavy metals in Portuguese infant milk powders by wavelength dispersive X-ray fluorescence spectrometry (WDXRF). Food Anal. Methods 7:3–8. Govindasamy-Lucey, S., J. J. Jaeggi, M. E. Johnson, T. Wang, and J. A. Lucey. 2007. Use of cold microfiltration retentates produced with polymeric membranes for standardization of milks for manufacture of pizza cheese. J. Dairy Sci. 90:4552–4568. Guinee, T. P. 2004. Salting and the role of salt in cheese. Int. J. Dairy Technol. 57:99–109. Gunicheva, T. N. 2010. Advisability of X-ray fluorescence analysis of dry residue of cow milk applied to monitor environment. X-Ray Spectrom. 39:22–27.

5051

Herbert, A. J., and K. Street. 1974. Non-dispersive soft X-ray fluorescence spectrometer for quantitative analysis of the major elements in rocks and minerals. Anal. Chem. 46:203–207. Horwitz, W., and R. Albert. 2006. The Horitz ratio (HorRat): A useful index of method performance with respect to precision. J. AOAC Int. 89:1095–1109. IDF. 1991. Milk and milk products. Precision characteristics of analytical methods—Outline of collaborative study procedure. IDF Standard No 135B. International Dairy Federation, Brussels, Belgium. ISO. 2006. Cheese and processed cheese products—Determination of chloride content—Potentiometric titration method. ISO 5943:2006 (IDF 88:2006). International Standards Organization, Geneva, Switzerland. Jastrzebska, A., B. Brudka, T. Szymanski, and E. Szyk. 2003. Determination of phosphorus in food samples by X-ray fluorescence spectrometry and standard spectrophotometric method. Food Chem. 83:463–467. Johnson, M. E., and N. F. Olson. 1985. A comparison of available methods for determining salt levels in cheese. J. Dairy Sci. 68:1020–1024. Lucey, J. A., A. Maurer-Rothmann, and S. Kaliappan. 2011. Functionality of ingredients: Emulsifying salts. Pages 110−132 in Processed Cheese and Analogues. A. Tamime, ed. Wiley-Blackwell, Oxford, UK. Masotti, F., D. Erba, I. De Noni, and L. Pellegrino. 2012. Rapid determination of sodium in milk and milk products by capillary zone electrophoresis. J. Dairy Sci. 95:2872–2881. Mozafar, A., J. J. Oertli, D. Tuor, and T. Roesch. 1990. Calibration of ICP to reduce the interference of Na with the determination of K. Commun. Soil Sci. Plan. 21:1607–1617. Pashkova, G. V. 2009. X-ray fluorescence determination of element contents in milk and dairy products. Food Anal. Methods 2:303– 310. Poitevin, E., M. Nicolas, L. Graveleau, J. Richoz, D. Andrey, and F. Monard. 2009. Improvement of AOAC Official Method 984.27 for the determination of nine nutritional elements in food products by inductively coupled plasma-atomic emission spectroscopy after microwave digestion: Single laboratory validation and ring trial. J. AOAC Int. 92:1484–1518. Potts, P. J., P. C. Webb, and J. S. Watson. 1984. Energy-dispersive X-ray fluorescence analysis of silicate rocks for major and trace elements. X-Ray Spectrom. 13:2–15. Rinaldoni, A. N., M. E. Campderros, A. Perez Padilla, E. Perino, and J. E. Fernandez. 2009. Analytical determinations of minerals content by XRF, ICP and EEA in ultrafiltered milk and yogurt. Lat. Am. Appl. Res. 39:113–118. Smagunova, A. N., and G. V. Pashkova. 2013. Choice of optimal conditions for X-ray fluorescence analysis of milk products with varying fat content. X-Ray Spectrom. 42:546–551. Suhaj, M., and M. Korenovska. 2008. Correlation and distribution of elemental markers of origin in the production of Bryndza sheep cheese. Food Chem. 107:551–557. Szczerbowska-Boruchowska, M. 2012. Sample thickness considerations for quantitative X-ray fluorescence analysis of the soft and skeletal tissues of the human body—Theoretical evaluation and experimental validation. X-Ray Spectrom. 41:328–337. West, M., A. T. Ellis, P. J. Potts, C. Streli, C. Vanhoof, D. Wegrzynek, and P. Wobrauschek. 2010. Atomic spectrometry update–X-ray fluorescence spectrometry. J. Anal. At. Spectrom. 25:1503–1545.

Journal of Dairy Science Vol. 98 No. 8, 2015