Technical note: Validation of energy dispersive X-ray fluorescence for determination of indigestible markers in ruminant fecal and rumen fluid samples

Abstract

Determination of digestibility and passage rate is important for further understanding of nutrient utilization and thereby aids in improving nutrient utilization efficiency. Titanium dioxide and chromium ethylenediaminetetraacetic acid are commonly used as indigestible markers for determining passage rates of diets to aid in determination of digestibility. Analyzing Ti and Cr involves the use of procedures such as acid digestions, inductively coupled plasma spectroscopy, and atomic absorption. These commonly used methodologies involve hazardous chemicals, destruction of samples, and low sample throughput. The objective of this experiment was to develop and validate an accurate and precise method for measuring both Ti and Cr using energy dispersive X-ray fluorescence (ED-XRF). Energy dispersive X-ray fluorescence is an analytical technique used for analyzing elements in various sample types. The samples were added to the ED-XRF machine and irradiated with X-rays. The intensity of the X-rays emitted (termed fluorescent X-rays) was used for calculation of the concentration of the element. The method for Ti was constructed using fecal samples from cattle consuming three different diet types (finishing diet, dairy lactation diet, and grazing native range pasture). The Cr method was developed for rumen fluid analysis. We compared the machine-calculated concentrations of each element to the concentration calculated by a standard curve. For both the Ti and Cr, the standard curve-calculated value had a lower percent difference overall at 4.56% and 12.59%, respectively, compared to the machine percent difference of 8.35% and 16.38% for Ti and Cr, respectively. To determine accuracy and precision of the method, samples were spiked with various amounts of Ti or Cr and measured for their respective compounds with percent recovery and inter- and intra-assay CV-calculated thereafter. The average recovery for Ti across all diet types was 100.3%, and the recovery for Cr in rumen fluid was 95.7%. The average inter- and intra-assay CV for Ti, across all diet types, were 9.70% and 2.16%, respectively. For Cr, the average inter- and intra-assay CV were 5.42% and 8.45%, respectively. The ED-XRF method requires minimal additional chemicals, is cost-effective, and allows for sample preservation as well as a high throughput of samples. Our results indicate utilization of ED-XRF is an accurate and precise method for determination of Ti in feces and Cr in rumen fluid.

Utilization of energy dispersive X-ray fluorescence is a cost-effective and efficient method for analysis of titanium in fecal samples and chromium in rumen fluid.

Introduction

Indigestible markers, such as titanium dioxide (TiO₂) and chromium ethylenediaminetetraacetic acid (Cr-EDTA), are commonly utilized in digestibility studies to aid in calculation of nutrient digestibility and passage rates (Van Soest and Hall, 2020). Obtaining accurate estimates of digestibility and passage rate is important to address current concerns in animal nutrition research including improving nutrient utilization, decreasing overall production costs, and reducing environmental impacts (Harmon, 2020). Additionally, markers such as Cr-EDTA are crucial to providing estimates of rumen ­volume and liquid passage rates to improve understanding of volatile fatty acid supply to ruminants as estimates using volatile fatty acid concentrations are likely insufficient for evaluating ruminal fermentation (Hall et al., 2015).

TiO₂ has previously been validated as an acceptable digestibility marker (Titgemeyer et al., 2001) and has been used to determine ruminal digesta flow (Pina et al., 2009) and passage rate (Davis et al., 2014). The use of TiO₂ as a marker requires an accurate and precise laboratory method to quantitatively analyze the Ti concentration in digesta or feces. TiO₂ is typically analyzed using inductively coupled plasma (ICP), atomic absorption spectroscopy (AAS), or colorimetric procedures. These methods all require a complex acid digestion process and utilization of hazardous chemicals (Short et al., 1996; Barnett et al., 2016). Chromium-EDTA is commonly used for evaluation of liquid passage rate and liquid volume and is commonly analyzed using ICP or AAS (Shingfield et al., 2008; KrämerLund and Weisbjerg, 2013; Hall and Van Soest, 2019). Procedures utilizing ICP, AAS, or colorimetric procedures require the analyte to be in liquid form, which requires feed and feces to be wet ashed or dry ashed and dissolved in strong acids (Harrison et al., 1988; Short et al., 1996). Converting dry feed or digesta samples to liquid form adds additional steps, equipment, chemicals, error, sample dilution, and requires sample destruction. The use of strong acids also generates chemical waste requiring additional disposal and potential environmental impacts. While ICP has some benefits related to number of simultaneous analyses compared to AAS, ICP has a high operating cost and requires a high level of staff expertise (Wilschefski and Baxter, 2019). Further, AAS is generally limited in the elements available to be analyzed with very low sample throughput and a high sample volume required with the added limitation of analyzing for one element at a time (Hill and Fisher, 2017; Wilschefski and Baxter, 2019).

Energy dispersive X-ray fluorescence (ED-XRF) is an analytical technique used for analysis of elemental compositions. The technique is capable of both identifying and quantifying compounds in samples of solid, liquid, and powder forms. Some key benefits of ED-XRF are the lack of sample destruction, low volume of sample required, sensitivity of the measurement, and a high sample throughput. The premise of the technology is that a sample is irradiated with X-rays of specific energies (measured as electron volts) and each element releases characteristic X-rays termed “fluorescent X-rays”. These X-rays can be detected to determine what elements (e.g., Ti or Cr) are present based on the characteristic X-rays produced as well as the quantity of the element present based on the intensity of X-rays produced. One limitation for this methodology is the utilization of radiation which requires additional training, oversight, and safety protocols depending on institutional regulations. These concerns are minor in modern equipment and there is no expectation of radiation exposure. For livestock nutrition research specifically, this technique would provide the capability of analyzing samples for a wide range of elements utilizing only a single sample with simultaneous analysis (Anonymous, 2023). Energy dispersive X-ray fluorescence is lower in operation cost compared to ICP and AAS methods (Pyle et al., 1995) and requires a low input of labor (Anonymous, 2023).

Given the potential benefits of the ED-XRF method for indigestible markers in nutrition studies and the continuing and growing need for determining digestion kinetics, it is important to verify the appropriateness of this methodology for use in animal nutrition studies. As these markers have been previously validated for use as indigestible markers (Titgemeyer et al., 2001) and are commonly used in ruminant nutrition research, our goal was to improve upon the analysis of the concentration of the markers in rumen fluid and feces. Therefore, the objective of this experiment was to validate the use of ED-XRF for quantitative analysis of Ti in fecal samples from cattle consuming a finishing diet, lactation diet, or grazing native range and Cr in rumen fluid.

Materials and Methods

Animal procedures for this experiment were limited to fecal and rumen fluid collection and were approved by the Institutional Animal Care and Use Committee of Oklahoma State University (#22-39, #21-03, #20-34, and #19-6).

Internal vs. external calibration

Given the potential for elements such as carbon and nitrogen in the digesta samples to interfere with the analysis of Ti and Cr, we first sought to determine if an external calibration curve would improve the accuracy of the estimated element concentration compared to the internal calibration of the ED-XRF machine (EDX-7000; Shimadzu Scientific Instruments, Columbia, MD). Spiked samples were prepared for both Ti (feces) and Cr (rumen fluid). Fecal samples were collected from animals not exposed to TiO₂ consuming a finishing diet (feedlot), dairy lactation diet (dairy), and grazing native range (range) at Oklahoma State University’s (OSU) Willard Sparks Beef Research Center, the Ferguson Family Dairy Center, and Range Cattle Research Center, respectively, from a minimum of 10 animals at each location. The finishing diet fed to the steers for the feedlot samples contained (dry matter basis): 54.4% dry-rolled corn, 18% wet corn gluten feed (Sweet Bran; Cargill Inc., Dalhart, TX, USA), 12% prairie hay, 5% dry distillers grains with solubles, 5% vitamin and mineral supplement, 5% molasses, and 0.6% urea. The diet fed to the cows at the OSU dairy was consuming a diet described previously (Beck et al., 2022) and included 32.17% dry-rolled corn, 31.57% alfalfa hay, 9.01% whole cottonseed, 6.23% bermudagrass hay, 5.94% dry distillers grains, 4.83% soybean hulls, 4.24% SoyBest (Grain States Soya, Inc., West Point, NE, USA), 1.13% soybean meal, 1.24% MegaLac (Church & Dwight Co., Inc., Ewing Township, NJ), and 3.64% vitamin and minerals. Cows that were collected for the range samples were grazing native pasture in June of 2022 and additional bermudagrass hay and a supplement of dried distillers grain and dry-rolled corn.

Rumen fluid samples were collected from eight ruminally cannulated Holstein steers with no exposure to Cr-EDTA. Samples were collected from multiple locations in the ventral sac of the rumen using a stainless-steel sampling probe with 0.297 mm openings in the mesh (Bar Diamond Inc., Parma, ID, USA). Fifty milliliters of samples from all steers were pooled together into a glass beaker and stirred using a stir bar and stir plate to create a composite sample devoid of Cr-EDTA before being aliquoted into 50-mL conical tubes. Aliquots were centrifuged (2,500 × g for 10 min at 4 °C) to obtain the supernatant which was transferred to a new 50-mL conical tube and frozen at −20 °C until Cr validation procedures. Feces from cattle of each diet were spiked to achieve Ti concentrations of 1.049 (High), 0.450 (Med), and 0.150 (Low) g/kg by adding the amount of TiO₂ needed to achieve the respective Ti concentration. Rumen fluid was spiked to achieve Cr concentrations of 0.5 (High), 0.25 (Med-High), 0.10 (Med-Low), and 0.005 (Low) g/L. The estimated concentrations calculated from the internal calibration of ED-XRF machine were compared with the calculated concentrations of the spike samples from the external calibration curve (see Standard preparation section below). Additionally, these values were compared to the known concentration of the spike by calculating a percent difference using the following equation:

$${\displaystyle \begin{array}{l}{\displaystyle Percentdifference=[\left(\frac{\begin{array}{c}(calculatedconcentration-actualconcentration)\end{array}}{\begin{array}{c}(calculatedconcentration+actualconcentration)\end{array}}\right)÷2]\ast 100}\end{array}}$$

Standard preparation

Titanium

As ED-XRF technology is sensitive to matrix effects (Bowers, 2019) standards were prepared in fecal samples collected from animals consuming a feedlot, dairy, or range diet as described in the previous section_. Samples were dried in a forced air oven at 105 °C for 24 h. After drying, all fecal samples were ground and compiled by location of collection. Standards were made to achieve the following concentrations: 0.05, 0.1, 0.5, 1, 2, 3, 4, and 5 g/kg TiO₂. A blank sample (0 g/kg TiO₂) was also included in the set of standards. The 5 g/kg TiO₂ standard was prepared by adding blank feces and TiO₂ to a 2-quart plastic container with subsequent hand mixing for 3 min to ensure homogenous dispersion of the TiO₂. The remaining standards were prepared similarly by mixing blank feces with a higher concentration standard. Using the molecular weights of TiO₂ and Ti, the corresponding concentration of Ti was determined using molecular weight of TiO₂ (79.866 g/mol) and the atomic weight of Ti (47.867 g/mol). Details of standard preparation can be found in Table 1.

Table 1.

Preparation of standards for the external calibration curve for analysis of titanium in feces. For this experiment, standards were prepared in dried feces that were collected from dairy cows consuming a lactation ration, beef steers consuming a finishing ration, and cows grazing native range

Standard ID	Concentration, g/kg TiO2	Concentration, g/kg Ti	Amount of standard, g	Feces, g
A	4.00	2.40	40 of stock1	10.00
B	3.00	1.80	37.5 of A	12.50
C	2.00	1.20	33.33 of B	16.67
D	1.00	0.60	25 of C	25.00
E	0.50	0.30	25 of D	25.00
F	0.10	0.06	10 of E	40.00
G	0.05	0.03	25 of F	25.00

¹Stock fecal sample prepared to achieve a concentration of 5 g/kg TiO₂ (× 3.00 g/kg Ti) by combining 0.25 g TiO₂ with 49.75 g of feces.

Chromium

Standards were individually prepared in ultra-pure water (Milli-Q, Millipore Sigma, Burlington, MA, USA) to achieve the following concentrations of Cr: 0.001, 0.005, 0.01, 0.025, 0.05, 0.10, 0.25, 0.50, 1.0, 1.5, and 2.0 g/L of Cr supplied as Cr-EDTA. Standards were prepared in ultra-pure water as opposed to rumen fluid due to the volatile nature of rumen fluid; additionally, there was no difference in the baseline intensity values of water compared to rumen fluid with no added Cr. This wide range of concentrations was initially used to determine if there was a limit of linearity, and the calibration curve was compared to a limited range of concentrations that is more practical for most experiments (0.005 to 0.50 g/L). A blank ultra-pure water sample (0 g/L Cr) was also included in the set of standards. A stock solution of Cr-EDTA (2.77 g/L Cr) was prepared as previously described (Binnerts et al., 1968) and used for the standard preparation. Briefly, chromium trichloride was dissolved in ultra-pure water and a solution of disodium-EDTA was added. This solution was gently boiled for 1 h with excess EDTA neutralized using 1 M calcium chloride. The solution was allowed to cool, and the pH was adjusted to 6.7 using 10 N sodium hydroxide before being diluted with ultra-pure water. Using a 10-mL volumetric flask, 2 mL of ultra-pure water was added followed by the appropriate volume of the stock Cr-EDTA solution (Table 2). Standards were then diluted to 10 mL in their respective flask using ultra-pure water. Standards were transferred to a 15-mL conical tube and vortexed.

Table 2.

Preparation of chromium-EDTA (Cr-EDTA) standards for the external calibration curve. Standards were prepared by adding the denoted volume of a stock solution of Cr-EDTA¹ to a 10 mL volumetric flask and diluting to volume with deionized water.

Concentration Cr, g/L	Volume of stock1 added, µL
0.001	3.6
0.005	18.1
0.010	36.1
0.025	90.3
0.050	180.5
0.100	361.0
0.250	903.0
0.500	1805.0
1.000	3610.0
1.500	5415.0
2.000	7220.0

¹Concentration of stock solution was 2.77 g/L Cr (Binnerts et al., 1968).

Procedure validations

Titanium

One to two grams of standard/sample were placed into a double open-ended X-cell sample cup (catalog number 219-85000-55; Shimadzu Scientific Instruments) lined on one end with polypropylene film (3520 Polypropylene 0.2 mil thick,7.3 cm; SPEX SamplePrep, Metuchen, NJ). In duplicate, standards and/or samples were loaded into the EDX-7000 for analysis. Ti was measured using the TiKa line and carbon as a balance. The X-ray energy value was set to 4.5 kiloelectron volts (keV) with an analysis range of 4.3 to 4.7 keV. A collimator size of 10 mm was utilized, and the reference system group was sysmetal. The voltage was set to 50 kV, and the current was set to 100 microamperes (µA). Due to nitrogen in the atmosphere potentially interfering with the analysis, helium was used to flush the system prior to analysis. The intensity, defined as counts per second per µA of electrical current (cps/µA), was recorded as well as Ti% calculated by the internal calibration. An external calibration curve of known Ti concentration (g/kg) versus intensity was generated (GraphPad Prism 9.4.1, GraphPad Software, San Diego, CA; Figure 1).

Figure 1.

Standard curves for Ti analysis across diet type: feedlot (filled circle), range (filled square), and dairy (filled triangle). Equations—feedlot: y = 23.1603x + 1.0386, range: y = 13.8396x + 1.2160, and dairy: y = 18.2595x + 2.5794.

Chromium

Prior to beginning the assay, all standards and samples were allowed to come to room temperature to minimize temperature variation during analysis. After vortexting, 1.0 mL of each standard or sample was pipetted into a double open-ended X-cell sample cup (catalog number 219-85000-55; Shimadzu Scientific Instruments) lined on one end with polypropylene X-ray film (3520 Polypropylene 0.2 mil thick, 7.3 cm; SPEX SamplePrep). Standards and samples were then analyzed, in duplicate, using the EDX-7000. Samples were analyzed using the same procedures as detailed above with exception of using the CrKa line, and the energy value was set to 5.42 keV. The analysis range for Cr was 5.22 to 5.62 keV. The intensity was recorded as well as the Cr% calculated by internal calibration. An external calibration curve of Cr concentration (g/L) versus intensity (cps/µA) was plotted using intensity values obtained from the standards (GraphPad Prism 9.4.1, GraphPad Software; Figure 2), and each sample’s concentration of Cr was calculated using the average intensity value for that sample.

Figure 2.

Extended standard curve for Cr analysis with a maximized figure showing the standard curve in the range of 0.001 to 0.5 g/L Cr.

Limits of blank and detection

For fecal Ti, the limits of blank and detection were determined for each of the three diets as previously described as well as for Cr in rumen fluid (Armbruster and Pry, 2008). Briefly, the limit of the blank (LoB) was determined by analyzing the intensity of the blank sample over 20 replicates to calculate the intensity mean (mean_blank) and standard deviation (SD_blank). The LoB was calculated using the following equation:

$${\displaystyle \begin{array}{l}{\displaystyle LoB\mathbf{=}\mathbf{m}\mathbf{e}\mathbf{a}{\mathbf{n}}_{blank}\mathbf{+}\mathbf{1}\mathbf{.645}\left(\mathbf{S}{\mathbf{D}}_{blank}\right)}\end{array}}$$

The limit of detection (LoD) was determined by analyzing the intensity of the lowest concentration standard 20 times to calculate the standard deviation. The LoD was calculated using the following equation:

$${\displaystyle \begin{array}{l}{\displaystyle LoD=LoB+1.645\left(S{D}_{lowstandard}\right)}\end{array}}$$

After determining the LoB and LoD for the intensity values, the LoB and LoD (g Ti/kg) were calculated using the respective standard curve.

Accuracy and precision

Percent recovery of the known amount of Ti and Cr added to spiked samples was utilized as a measure of accuracy. Spiked samples were generated as described above. Each spiked sample was analyzed 20 times on three separate days (total of 60 replicates per spiked sample) and percent recovery was calculated using the following equation:

$${\displaystyle \begin{array}{l}{\displaystyle \mathrm{\%}Recovery=\left(\frac{\mathrm{g}\mathrm{m}\mathrm{a}\mathrm{r}\mathrm{k}\mathrm{e}\mathrm{r}\mathrm{m}\mathrm{e}\mathrm{a}\mathrm{s}\mathrm{u}\mathrm{r}\mathrm{e}\mathrm{d}}{\mathrm{g}\mathrm{m}\mathrm{a}\mathrm{r}\mathrm{k}\mathrm{e}\mathrm{r}\mathrm{a}\mathrm{d}\mathrm{d}\mathrm{e}\mathrm{d}}\right)\times 100}\end{array}}$$

Intra- and inter-assay CV were calculated to determine precision of the assay.

Statistical analysis

Means for the recoveries of each marker were determined using Means procedure of SAS 9.4 (SAS Institute Inc., Cary, NC). For the Ti analysis, a standard curve (six replicates per standard) was used to calculate the slope for each diet type. To determine the effect of fecal sample source on the calculated slopes, the fixed effect of fecal sample source (i.e., feedlot, dairy, or range) was analyzed (PROC MIXED, SAS Inst. Inc.). The differences between marker concentration calculated by the external calibration curve and the ED-XRF machine-calculated concentration were assessed using a pairwise t-test with PROC TTEST of SAS 9.4 (SAS Inst. Inc.). All analyses were considered significant when P < 0.05.

Results and Discussion

Titanium

The difference between concentrations calculated from the external calibration curve and the concentration given by the machine was evaluated for all feces collected from cattle consuming all diets and spiked concentrations. The concentration from the external calibration curve and the ­concentration from the internal calibration were different for all diets and spiked concentrations (P < 0.001; Table 3) except for the 1.049 g/kg range spiked sample (P =0.47). Additionally, the percent error, on average, was less for the external calibration curve-calculated concentration (4.56%) compared to the internal calibration calculated (8.35%) indicating the external calibration curve-calculated concentration is more accurate. For most samples, the Ti concentration calculated from the standard curve produced concentrations more similar to the expected concentration. The lack of difference between the Ti concentration from the standard curve and machine-calculated concentration for the High concentration from the range sample may be due in part to the high standard error. Given these results, it is recommended when utilizing an ED-XRF machine for Ti analysis, to utilize an external calibration curve.

Table 3.

Comparison of titanium (Ti) concentrations of spiked fecal samples from dairy cows consuming a lactation ration, beef steers consuming a finishing ration, and cows grazing native range and containing high (1.049 g/kg), medium (0.450 g/kg), or low (0.150 g/kg) concentrations of Ti as measured using an external calibration curve or the internal calibration of an energy dispersive X-ray fluorescence analyzer. The percent difference of the resulting concentration as extrapolated from an external calibration curve or internal calibration to determine which calibration method was more accurate

	Ti concentration, g/kg				Percent difference
	External calibration curve	Internal calibration	SEM1	P-value	External calibration curve	Internal calibration
High, 1.049 g/kg
Feedlot	0.974	1.205	0.0035	<0.001	5.01	9.02
Range	1.133	1.025	0.1471	0.47	5.07	1.55
Dairy	1.051	1.021	0.0010	<0.001	0.13	1.81
Medium, 0.450 g/kg
Feedlot	0.447	0.529	0.0016	<0.001	0.45	10.48
Range	0.493	0.390	0.0024	<0.001	5.99	9.76
Dairy	0.403	0.369	0.0003	<0.001	7.48	13.64
Low, 0.150 g/kg
Feedlot	0.167	0.173	0.0007	<0.001	7.02	9.27
Range	0.147	0.126	0.0013	<0.001	1.35	11.94
Dairy	0.171	0.134	0.0098	<0.001	8.54	7.66

¹Standard error of the mean (n = 60).

The external calibration curve range chosen for Ti analysis was linear across all three diets with R² > 0.996 (Figure 1). The slopes for the standard curves differed among diets (P < 0.05), and the intercept for dairy samples was greater compared to the feedlot and range samples (Table 4). Given the results above regarding the improvement when using an external calibration curve compared to internal calibration and the difference in the slopes and intercepts of external calibrations depending on source of feces, it is highly recommended that researchers collect feces from the individual trials prior to dosing Ti to generate external calibration curves.

Table 4.

Recovery of titanium (Ti) in spiked fecal samples from dairy cows consuming a lactation ration, beef steers consuming a finishing ration, and cows grazing native range and containing high (1.049 g/kg), medium (0.450 g/kg), or low (0.150 g/kg) concentrations of Ti as measured using an external calibration curve. The intra- and inter-assay coefficient of variation (CV) measured in the spiked fecal samples

	Recovery, % ± Standard Deviation	Intra-assay CV, %	Inter-assay CV, %
High, 1.049 g/kg
Feedlot	92.3 ± 4.59	5.22	4.76
Range	100.6 ± 0.73	5.11	0.62
Dairy	100.2 ± 1.47	7.14	1.38
Medium, 0.450 g/kg
Feedlot	93.2 ± 10.57	6.53	2.45
Range	99.3 ± 1.57	8.58	1.15
Dairy	92.0 ± 3.22	16.23	3.00
Low, 0.150 g/kg
Feedlot	113.2 ± 1.94	8.45	1.36
Range	97.7 ± 4.54	11.88	2.05
Dairy	114.3 ± 4.23	18.12	2.66

The LoD for Ti measured in feces from cattle consuming native range was extremely low and below the lowest point included in the standard curve. The feedlot and dairy feces LoD were above that of the lowest standard concentration (Table 5). Therefore, animals should be dosed to achieve fecal Ti concentrations greater than 0.073 g/kg. The dose required to achieve concentrations above the limit of detection should be evaluated for each experiment as it will depend on body weight, feed intake, digestibility, dosing scheme, and timing of collection (Ellis et al., 1994; Moffet and Gunter, 2020). Overall, the LoD should be taken into consideration prior to animal dosage to ensure analytical capabilities will meet the needs of the experiment.

Table 5.

External calibration curve data from standards made in feces dairy cows consuming a lactation ration, beef steers consuming a finishing ration, and cows grazing native range

Item	Feedlot	Range	Dairy
Calibration curve slope1	23.16	13.84	18.26
Calibration curve intercept2	1.039	1.216	2.579
R 2	0.9967	0.9982	0.9990
Limit of blank, g/kg	0.0510	0.0062	0.0417
Limit of detection, g/kg	0.0596	0.0276	0.0729

¹All slopes differ P ≤ 0.001.

²All intercepts differ P ≤ 0.001.

For all diet types and spike concentrations, recoveries ranged from 92.3% to 114.3% (Table 4) indicating the accuracy of this method is acceptable. The recoveries reported here are similar to previously reported results in chicken feces using a dry-ashing and concentrated acid digestion method (Short et al., 1996), as well as methods that modify Short et al. (1996) in cattle feces (Titgemeyer et al., 2001) and horse feces (Fowler et al., 2022). Thus, utilization of the ED-XRF for analysis of Ti will likely produce accurate results.

The inter-assay CV was <5% for all diet types and spiked concentrations (Table 4), and the intra-assay CV for the feedlot and range samples was <15% for all spiked concentrations. However, for the dairy samples, the CV was 16.23 and 18.12 for the 0.450 (Medium) and 0.150 (Low) g/kg Ti spiked concentrations, respectively. It is generally accepted that the CV for bioanalytical assays should be less than 15% (Tiwari and Tiwari, 2010). Given the cutoff of 15% CV, this procedure displays acceptable precision between assays, and for the feedlot and range diet types were precise within assay. Precision between assay was decreased for the dairy samples. The LoD for the dairy samples was also numerically increased compared to the feedlot and range diets indicating the assay may not be as sensitive for the dairy samples resulting in an increase in variation and decreased sensitivity. This may be due to the diet composition of typical dairy lactation diets being more variable in nutrient type than typical finishing and grazing diets (Mahanna, 2015; National Academies of Sciences and Medicine, 2016). The dairy diet fed is an intermediate mixture of concentrates and forages compared to finishing diets and grazed native range. It is possible that the complex nature of mixed rations used in dairies could complicate the analysis or there might be a greater concentration of minerals (e.g., calcium) included in the lactation ration compared to the finishing or grazing system. This may lead to a variation in fecal composition and increase overall variability in the assay. These data suggest that ED-XRF is a precise method for analysis for Ti; however, matrix effects and interference should be evaluated for individual experiments.

In a separate experiment, fecal samples were collected from cows dosed with TiO₂ and analyzed for Ti using the method described here. The samples (n = 104) had an inter- and intra-assay CV of 3.7 and 1.1%, respectively. One sample had an intensity CV > 5%. No samples had an intensity CV > 10%. Thus, analyzing fecal samples for Ti from cattle dosed with TiO₂ via ED-XRF produced precise results and is an acceptable method for Ti analysis.

Chromium

The Cr concentrations calculated from the external calibration curve were different than those calculated from the ED-XRF machine (Table 6; P < 0.001), and on average, the external calibration curve-calculated concentration had a lower percent difference (12.59%) compared to the internal calibration-calculated value (16.38%; Table 6). The internal calibration-calculated concentration had a lower percent error for both the 0.50 (High) and 0.25 (Medium-High) g/L spiked samples; however, the external calibration curve-calculated concentration resulted in more consistent results overall. For the lower concentration spikes, the ED-XRF machine overestimated the concentration of Cr, particularly for the 0.005 (Low) g/L, while producing a similar estimate to the known concentration for the higher concentration spikes. The standard curve-calculated Cr concentrations were consistently similar to the expected concentration. Therefore, utilizing a standard curve to calculate unknown samples may provide a more consistent and accurate result, especially at lower concentrations.

Table 6.

Comparison of chromium (Cr) concentrations of spiked rumen fluid samples containing high (0.50 g/L), medium-high (0.25 g/L), medium-low (0.10 g/L), or low (0.005 g/L) concentrations of Cr as measured using an external calibration curve or the internal calibration of an energy dispersive X-ray fluorescence analyzer. The percent difference of the resulting concentration as extrapolated from an external calibration curve or internal calibration to determine which calibration method was more accurate

	Cr concentration, g/L				Percent difference		Recovery, %	Intra-assay CV	Inter-assay CV
	External calibration curve	Internal calibration	SEM1	P-value	External calibration curve	Internal calibration
High, 0.50 g/L	0.474	0.492	0.0001	<0.001	3.59	1.08	94.8 ± 5.64	3.32	5.83
Medium-High, 0.25 g/L	0.233	0.246	0.0003	<0.001	4.75	1.08	114.9 ± 22.80	2.63	2.45
Medium-Low, 0.10 g/L	0.109	0.120	0.0001	<0.001	5.66	11.76	108.6 ± 18.08	4.69	15.30
Low, 0.005 g/L	0.003	0.013	0.0006	<0.001	36.36	51.61	64.6 ± 11.23	11.02	10.21

¹Standard error of the mean (n = 60).

The extended external calibration curve displayed an R² > 0.999 (Figure 2) as did the narrowed range standard curve (R² = 0.999). Additionally, the slope and intercept for the narrowed standard curve were numerically different (slope = 39.35, intercept = 0.3776) compared to the extended standard curve (slope =35.36, intercept =0.4735). The narrowed standard curve range is more similar to concentrations of previously reported measurements (Hall and Van Soest, 2019). While the standard curve is linear up to 2.0 g/L Cr, the small difference in the slopes between the narrowed and extended curve may result in variation; thus, a standard curve should be used in accordance with expected Cr concentrations.

The LoD for Cr samples was 0.0011 g/L indicating this method can analyze low concentrations. For all spiked ­samples, the recovery was >94.8% (Table 6) except for the Low spike sample averaged 64.6% indicating that the ED-XRF, though having a low limit of detection, may be less accurate at very low concentrations of Cr. Therefore, it is important to appropriately dose Cr to achieve concentrations that are easily detectable. The intra- and inter-assay CV for all spiked concentrations were <12%. Given the accepted limit for bioanalytical assay CV of <15% (Tiwari and Tiwari, 2010), this method for analysis of Cr in rumen fluid is acceptably precise.

Rumen fluid samples from steers consuming bermudagrass hay from a sample pilot project were dosed with Cr-EDTA ruminally with Cr analyzed using the method described here. The average CV of the intensity of the samples (n = 24) was 5.589%. Fourteen samples had an intensity CV > 5%, and two samples had an intensity CV > 10%. Given relatively high number of samples with CV > 5%, there might be an advantage to analyzing samples in triplicate.

Conclusion

The use of the ED-XRF machine produced accurate and precise results for determination of Ti and Cr in fecal samples of cattle consuming a finishing diet, lactation diet, or grazing native range and rumen fluid samples, respectively, when using an external calibration curve. When using this method for analysis of Ti in fecal samples, it is important to utilize an external calibration curve generated from feces containing no Ti from animals consuming a diet as similar as possible to those evaluated in the experiment to minimize matrix effects. This method is less hazardous than traditional methods and increases the efficiency of analysis through reduction of labor and cost. Further research to determine the applicability to other digestibility markers may be warranted.

Glossary

Abbreviations:

µA

microamperes

AAS

atomic absorption spectroscopy

cps/µA

counts per second per microampere

Cr-EDTA

chromium ethylenediaminetetraacetic acid

ED-XRF

energy dispersive X-ray fluorescence

ICP

inductively coupled plasma

keV

kiloelectron volt

kV

kilovolts

LoB

limit of blank

LoD

limit of detection

TiO2

titanium dioxide

Conflict of interest statement.The authors declare no conflict of interest.